diff --git a/.devops/full-cuda.Dockerfile b/.devops/full-cuda.Dockerfile
new file mode 100644
index 0000000000000..e5fcb37d6fe7a
--- /dev/null
+++ b/.devops/full-cuda.Dockerfile
@@ -0,0 +1,33 @@
+ARG UBUNTU_VERSION=22.04
+
+# This needs to generally match the container host's environment.
+ARG CUDA_VERSION=11.7.1
+
+# Target the CUDA build image
+ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
+
+FROM ${BASE_CUDA_DEV_CONTAINER} as build
+
+# Unless otherwise specified, we make a fat build.
+ARG CUDA_DOCKER_ARCH=all
+
+RUN apt-get update && \
+    apt-get install -y build-essential python3 python3-pip
+
+COPY requirements.txt requirements.txt
+
+RUN pip install --upgrade pip setuptools wheel \
+    && pip install -r requirements.txt
+
+WORKDIR /app
+
+COPY . .
+
+# Set nvcc architecture
+ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
+# Enable cuBLAS
+ENV LLAMA_CUBLAS=1
+
+RUN make
+
+ENTRYPOINT ["/app/.devops/tools.sh"]
diff --git a/.devops/main-cuda.Dockerfile b/.devops/main-cuda.Dockerfile
new file mode 100644
index 0000000000000..30c01196ab520
--- /dev/null
+++ b/.devops/main-cuda.Dockerfile
@@ -0,0 +1,32 @@
+ARG UBUNTU_VERSION=22.04
+# This needs to generally match the container host's environment.
+ARG CUDA_VERSION=11.7.1
+# Target the CUDA build image
+ARG BASE_CUDA_DEV_CONTAINER=nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
+# Target the CUDA runtime image
+ARG BASE_CUDA_RUN_CONTAINER=nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION}
+
+FROM ${BASE_CUDA_DEV_CONTAINER} as build
+
+# Unless otherwise specified, we make a fat build.
+ARG CUDA_DOCKER_ARCH=all
+
+RUN apt-get update && \
+    apt-get install -y build-essential
+
+WORKDIR /app
+
+COPY . .
+
+# Set nvcc architecture
+ENV CUDA_DOCKER_ARCH=${CUDA_DOCKER_ARCH}
+# Enable cuBLAS
+ENV LLAMA_CUBLAS=1
+
+RUN make
+
+FROM ${BASE_CUDA_RUN_CONTAINER} as runtime
+
+COPY --from=build /app/main /main
+
+ENTRYPOINT [ "/main" ]
diff --git a/.github/workflows/build.yml b/.github/workflows/build.yml
index aec43bd923bb5..f6a2dd6daa198 100644
--- a/.github/workflows/build.yml
+++ b/.github/workflows/build.yml
@@ -16,7 +16,10 @@ on:
     paths: ['**/CMakeLists.txt', '**/Makefile', '**/*.h', '**/*.hpp', '**/*.c', '**/*.cpp', '**/*.cu']
 
 env:
- BRANCH_NAME: ${{ github.head_ref || github.ref_name }}
+  BRANCH_NAME: ${{ github.head_ref || github.ref_name }}
+  GGML_NLOOP: 3
+  GGML_NITER: 1
+  GGML_N_THREADS: 1
 
 jobs:
   ubuntu-focal-make:
@@ -64,7 +67,7 @@ jobs:
         id: cmake_test
         run: |
           cd build
-          ctest --verbose
+          ctest --verbose --timeout 900
 
   ubuntu-latest-cmake-sanitizer:
     runs-on: ubuntu-latest
@@ -99,7 +102,7 @@ jobs:
         id: cmake_test
         run: |
           cd build
-          ctest --verbose
+          ctest --verbose --timeout 900
 
   macOS-latest-make:
     runs-on: macos-latest
@@ -137,19 +140,21 @@ jobs:
       - name: Build
         id: cmake_build
         run: |
+          sysctl -a
           mkdir build
           cd build
-          cmake -DLLAMA_AVX2=OFF ..
+          cmake -DLLAMA_AVX2=OFF -DLLAMA_FMA=OFF ..
           cmake --build . --config Release
 
       - name: Test
         id: cmake_test
         run: |
           cd build
-          ctest --verbose
+          ctest --verbose --timeout 900
 
   windows-latest-cmake:
     runs-on: windows-latest
+
     env:
       OPENBLAS_VERSION: 0.3.23
       OPENCL_VERSION: 2023.04.17
@@ -248,7 +253,7 @@ jobs:
         if: ${{ matrix.build != 'clblast' && (matrix.build != 'avx512' || env.HAS_AVX512F == '1') }} # Test AVX-512 only when possible
         run: |
           cd build
-          ctest -C Release --verbose
+          ctest -C Release --verbose --timeout 900
 
       - name: Get commit hash
         id: commit
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 4d44f91755c0a..6042f6a42fdc4 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -68,8 +68,9 @@ option(LLAMA_ACCELERATE                      "llama: enable Accelerate framework
 option(LLAMA_BLAS                            "llama: use BLAS"                                  OFF)
 set(LLAMA_BLAS_VENDOR "Generic" CACHE STRING "llama: BLAS library vendor")
 option(LLAMA_CUBLAS                          "llama: use cuBLAS"                                OFF)
+option(LLAMA_CUDA_FORCE_DMMV                 "llama: use dmmv instead of mmvq CUDA kernels"     OFF)
 set(LLAMA_CUDA_DMMV_X      "32" CACHE STRING "llama: x stride for dmmv CUDA kernels")
-set(LLAMA_CUDA_DMMV_Y       "1" CACHE STRING "llama: y block size for dmmv CUDA kernels")
+set(LLAMA_CUDA_MMV_Y        "1" CACHE STRING "llama: y block size for mmv CUDA kernels")
 option(LLAMA_CUDA_DMMV_F16                   "llama: use 16 bit floats for dmmv CUDA kernels"   OFF)
 set(LLAMA_CUDA_KQUANTS_ITER "2" CACHE STRING "llama: iters./thread per block for Q2_K/Q6_K")
 option(LLAMA_HIPBLAS                         "llama: use hipBLAS"                               OFF)
@@ -217,6 +218,9 @@ if (LLAMA_BLAS)
         message(STATUS "BLAS found, Includes: ${BLAS_INCLUDE_DIRS}")
         add_compile_options(${BLAS_LINKER_FLAGS})
         add_compile_definitions(GGML_USE_OPENBLAS)
+        if (${BLAS_INCLUDE_DIRS} MATCHES "mkl" AND (${LLAMA_BLAS_VENDOR} MATCHES "Generic" OR ${LLAMA_BLAS_VENDOR} MATCHES "Intel"))
+            add_compile_definitions(GGML_BLAS_USE_MKL)
+        endif()
         set(LLAMA_EXTRA_LIBS ${LLAMA_EXTRA_LIBS} ${BLAS_LIBRARIES})
         set(LLAMA_EXTRA_INCLUDES ${LLAMA_EXTRA_INCLUDES} ${BLAS_INCLUDE_DIRS})
 
@@ -247,8 +251,14 @@ if (LLAMA_CUBLAS)
         set(GGML_SOURCES_CUDA ggml-cuda.cu ggml-cuda.h)
 
         add_compile_definitions(GGML_USE_CUBLAS)
+        if (LLAMA_CUDA_FORCE_DMMV)
+            add_compile_definitions(GGML_CUDA_FORCE_DMMV)
+        endif()
         add_compile_definitions(GGML_CUDA_DMMV_X=${LLAMA_CUDA_DMMV_X})
-        add_compile_definitions(GGML_CUDA_DMMV_Y=${LLAMA_CUDA_DMMV_Y})
+        add_compile_definitions(GGML_CUDA_MMV_Y=${LLAMA_CUDA_MMV_Y})
+        if (DEFINED LLAMA_CUDA_DMMV_Y)
+            add_compile_definitions(GGML_CUDA_MMV_Y=${LLAMA_CUDA_DMMV_Y}) # for backwards compatibility
+        endif()
         if (LLAMA_CUDA_DMMV_F16)
             add_compile_definitions(GGML_CUDA_DMMV_F16)
         endif()
@@ -264,7 +274,7 @@ if (LLAMA_CUBLAS)
         if (LLAMA_CUDA_DMMV_F16)
             set(CMAKE_CUDA_ARCHITECTURES "61") # needed for f16 CUDA intrinsics
         else()
-            set(CMAKE_CUDA_ARCHITECTURES "52") # lowest CUDA 12 standard
+            set(CMAKE_CUDA_ARCHITECTURES "52;61") # lowest CUDA 12 standard + lowest for integer intrinsics
         endif()
     endif()
     message(STATUS "Using CUDA architectures: ${CMAKE_CUDA_ARCHITECTURES}")
diff --git a/Makefile b/Makefile
index d8f12fce74ec5..38d65aebc6c01 100644
--- a/Makefile
+++ b/Makefile
@@ -163,17 +163,27 @@ ifdef LLAMA_CUBLAS
 	LDFLAGS   += -lcublas -lculibos -lcudart -lcublasLt -lpthread -ldl -lrt -L/usr/local/cuda/lib64 -L/opt/cuda/lib64 -L$(CUDA_PATH)/targets/x86_64-linux/lib
 	OBJS      += ggml-cuda.o
 	NVCC      = nvcc
-	NVCCFLAGS = --forward-unknown-to-host-compiler -arch=native
+	NVCCFLAGS = --forward-unknown-to-host-compiler
+ifdef CUDA_DOCKER_ARCH
+	NVCCFLAGS += -Wno-deprecated-gpu-targets -arch=$(CUDA_DOCKER_ARCH)
+else
+	NVCCFLAGS += -arch=native
+endif # CUDA_DOCKER_ARCH
+ifdef LLAMA_CUDA_FORCE_DMMV
+	NVCCFLAGS += -DGGML_CUDA_FORCE_DMMV
+endif # LLAMA_CUDA_FORCE_DMMV
 ifdef LLAMA_CUDA_DMMV_X
 	NVCCFLAGS += -DGGML_CUDA_DMMV_X=$(LLAMA_CUDA_DMMV_X)
 else
 	NVCCFLAGS += -DGGML_CUDA_DMMV_X=32
 endif # LLAMA_CUDA_DMMV_X
-ifdef LLAMA_CUDA_DMMV_Y
-	NVCCFLAGS += -DGGML_CUDA_DMMV_Y=$(LLAMA_CUDA_DMMV_Y)
+ifdef LLAMA_CUDA_MMV_Y
+	NVCCFLAGS += -DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_MMV_Y)
+else ifdef LLAMA_CUDA_DMMV_Y
+	NVCCFLAGS += -DGGML_CUDA_MMV_Y=$(LLAMA_CUDA_DMMV_Y) # for backwards compatibility
 else
-	NVCCFLAGS += -DGGML_CUDA_DMMV_Y=1
-endif # LLAMA_CUDA_DMMV_Y
+	NVCCFLAGS += -DGGML_CUDA_MMV_Y=1
+endif # LLAMA_CUDA_MMV_Y
 ifdef LLAMA_CUDA_DMMV_F16
 	NVCCFLAGS += -DGGML_CUDA_DMMV_F16
 endif # LLAMA_CUDA_DMMV_F16
@@ -182,6 +192,7 @@ ifdef LLAMA_CUDA_KQUANTS_ITER
 else
 	NVCCFLAGS += -DK_QUANTS_PER_ITERATION=2
 endif
+
 ggml-cuda.o: ggml-cuda.cu ggml-cuda.h
 	$(NVCC) $(NVCCFLAGS) $(CXXFLAGS) -Wno-pedantic -c $< -o $@
 endif # LLAMA_CUBLAS
diff --git a/README.md b/README.md
index e890dc9c22e75..daa71c2b939da 100644
--- a/README.md
+++ b/README.md
@@ -11,6 +11,7 @@ Inference of [LLaMA](https://arxiv.org/abs/2302.13971) model in pure C/C++
 
 **Hot topics:**
 
+- Simple web chat example: https://github.com/ggerganov/llama.cpp/pull/1998
 - k-quants now support super-block size of 64: https://github.com/ggerganov/llama.cpp/pull/2001
 - New roadmap: https://github.com/users/ggerganov/projects/7
 - Azure CI brainstorming: https://github.com/ggerganov/llama.cpp/discussions/1985
@@ -85,7 +86,7 @@ as the main playground for developing new features for the [ggml](https://github
 - [X] [OpenBuddy 🐶 (Multilingual)](https://github.com/OpenBuddy/OpenBuddy)
 - [X] [Pygmalion 7B / Metharme 7B](#using-pygmalion-7b--metharme-7b)
 - [X] [WizardLM](https://github.com/nlpxucan/WizardLM)
-- [X] [Baichuan-7B](https://huggingface.co/baichuan-inc/baichuan-7B)
+- [X] [Baichuan-7B](https://huggingface.co/baichuan-inc/baichuan-7B) and its derivations (such as [baichuan-7b-sft](https://huggingface.co/hiyouga/baichuan-7b-sft))
 
 **Bindings:**
 
@@ -344,8 +345,9 @@ Building the program with BLAS support may lead to some performance improvements
 
   | Option                  | Legal values           | Default | Description |
   |-------------------------|------------------------|---------|-------------|
+  | LLAMA_CUDA_FORCE_DMMV   | Boolean                |   false | Force the use of dequantization + matrix vector multiplication kernels instead of using kernels that do matrix vector multiplication on quantized data. By default the decision is made based on compute capability (MMVQ for 7.0/Turing/RTX 2000 or higher). Does not affect k-quants. |
   | LLAMA_CUDA_DMMV_X       | Positive integer >= 32 |      32 | Number of values in x direction processed by the CUDA dequantization + matrix vector multiplication kernel per iteration. Increasing this value can improve performance on fast GPUs. Power of 2 heavily recommended. Does not affect k-quants. |
-  | LLAMA_CUDA_DMMV_Y       | Positive integer       |       1 | Block size in y direction for the CUDA dequantization + mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended. Does not affect k-quants. |
+  | LLAMA_CUDA_MMV_Y       | Positive integer       |       1 | Block size in y direction for the CUDA mul mat vec kernels. Increasing this value can improve performance on fast GPUs. Power of 2 recommended. Does not affect k-quants. |
   | LLAMA_CUDA_DMMV_F16     | Boolean                |   false | If enabled, use half-precision floating point arithmetic for the CUDA dequantization + mul mat vec kernels. Can improve performance on relatively recent GPUs. |
   | LLAMA_CUDA_KQUANTS_ITER | 1 or 2                 |       2 | Number of values processed per iteration and per CUDA thread for Q2_K and Q6_K quantization formats. Setting this value to 1 can improve performance for slow GPUs. |
 
@@ -693,7 +695,7 @@ export LD_LIBRARY_PATH=/vendor/lib64:$LD_LIBRARY_PATH
 
 For easy and swift re-execution, consider documenting this final part in a .sh script file. This will enable you to rerun the process with minimal hassle.
 
-Place your desired model into the `/llama.cpp/models/` directory and execute the `./main (...)` script.
+Place your desired model into the `~/llama.cpp/models/` directory and execute the `./main (...)` script.
 
 ### Docker
 
@@ -729,6 +731,38 @@ or with a light image:
 docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:light -m /models/7B/ggml-model-q4_0.bin -p "Building a website can be done in 10 simple steps:" -n 512
 ```
 
+### Docker With CUDA
+
+Assuming one has the [nvidia-container-toolkit](https://github.com/NVIDIA/nvidia-container-toolkit) properly installed on Linux, or is using a GPU enabled cloud, `cuBLAS` should be accessible inside the container.
+
+#### Building Locally
+
+```bash
+docker build -t local/llama.cpp:full-cuda -f .devops/full-cuda.Dockerfile .
+docker build -t local/llama.cpp:light-cuda -f .devops/main-cuda.Dockerfile .
+```
+
+You may want to pass in some different `ARGS`, depending on the CUDA environment supported by your container host, as well as the GPU architecture.
+
+The defaults are:
+
+- `CUDA_VERSION` set to `11.7.1`
+- `CUDA_DOCKER_ARCH` set to `all`
+
+The resulting images, are essentially the same as the non-CUDA images:
+
+1. `local/llama.cpp:full-cuda`: This image includes both the main executable file and the tools to convert LLaMA models into ggml and convert into 4-bit quantization.
+2. `local/llama.cpp:light-cuda`: This image only includes the main executable file.
+
+#### Usage
+
+After building locally, Usage is similar to the non-CUDA examples, but you'll need to add the `--gpus` flag. You will also want to use the `--n-gpu-layers` flag.
+
+```bash
+docker run --gpus all -v /path/to/models:/models local/llama.cpp:full-cuda --run -m /models/7B/ggml-model-q4_0.bin -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1
+docker run --gpus all -v /path/to/models:/models local/llama.cpp:light-cuda -m /models/7B/ggml-model-q4_0.bin -p "Building a website can be done in 10 simple steps:" -n 512 --n-gpu-layers 1
+```
+
 ### Contributing
 
 - Contributors can open PRs
@@ -749,5 +783,10 @@ docker run -v /path/to/models:/models ghcr.io/ggerganov/llama.cpp:light -m /mode
 
 ### Docs
 
-- [GGML tips & tricks](https://github.com/ggerganov/llama.cpp/wiki/GGML-Tips-&-Tricks)
+- [main](./examples/main/README.md)
+- [server](./examples/server/README.md)
+- [embd-input](./examples/embd-input/README.md)
+- [jeopardy](./examples/jeopardy/README.md)
+- [BLIS](./docs/BLIS.md)
 - [Performance troubleshooting](./docs/token_generation_performance_tips.md)
+- [GGML tips & tricks](https://github.com/ggerganov/llama.cpp/wiki/GGML-Tips-&-Tricks)
diff --git a/convert.py b/convert.py
index 14269277627b1..7a2705e5c506f 100644
--- a/convert.py
+++ b/convert.py
@@ -154,9 +154,15 @@ def guessed(model: 'LazyModel') -> 'Params':
         # try transformer naming first
         if "model.layers.0.self_attn.q_proj.weight" in model:
             n_layer=next(i for i in itertools.count() if f"model.layers.{i}.self_attn.q_proj.weight" not in model)
+        elif "model.layers.0.self_attn.W_pack.weight" in model:   # next: try baichuan naming
+            n_layer=next(i for i in itertools.count() if f"model.layers.{i}.self_attn.W_pack.weight" not in model)
         else:
             n_layer=next(i for i in itertools.count() if f"layers.{i}.attention.wq.weight" not in model)
 
+        if n_layer < 1:
+            raise Exception("failed to guess 'n_layer'. This model is unknown or unsupported.\n"
+                            "Suggestion: provide 'config.json' of the model in the same directory containing model files.")
+
         n_head=n_embd // 128 # guessed
 
         return Params(
@@ -822,6 +828,7 @@ def lazy_load_torch_file(outer_fp: IO[bytes], path: Path) -> ModelPlus:
 
 
 SAFETENSORS_DATA_TYPES: Dict[str, DataType] = {
+    'BF16': DT_BF16,
     'F16': DT_F16,
     'F32': DT_F32,
     'I32': DT_I32,
diff --git a/examples/alpaca.sh b/examples/alpaca.sh
index aef207f364797..8d2bae6918b62 100755
--- a/examples/alpaca.sh
+++ b/examples/alpaca.sh
@@ -7,7 +7,7 @@
 cd `dirname $0`
 cd ..
 
-./main -m ./models/ggml-alpaca-7b-q4.bin \
+./main -m ./models/alpaca.13b.ggmlv3.q8_0.bin \
        --color \
        -f ./prompts/alpaca.txt \
        --ctx_size 2048 \
diff --git a/examples/baby-llama/baby-llama.cpp b/examples/baby-llama/baby-llama.cpp
index 212f54d32cbad..4965881ecec22 100644
--- a/examples/baby-llama/baby-llama.cpp
+++ b/examples/baby-llama/baby-llama.cpp
@@ -31,6 +31,17 @@ float frand_normal(struct random_normal_distribution * rnd) {
     return ((r < rnd->min) ? (rnd->min) : (r > rnd->max) ? (rnd->max) : r);
 }
 
+void ggml_graph_compute_helper(std::vector<uint8_t> & buf, ggml_cgraph * graph, int n_threads) {
+    struct ggml_cplan plan = ggml_graph_plan(graph, n_threads);
+
+    if (plan.work_size > 0) {
+        buf.resize(plan.work_size);
+        plan.work_data = buf.data();
+    }
+
+    ggml_graph_compute(graph, &plan);
+}
+
 struct ggml_tensor * randomize_tensor(
         struct ggml_tensor * tensor,
         int ndims,
@@ -1569,6 +1580,8 @@ int main(int argc, char ** argv) {
     int n_tokens = model.hparams.n_ctx;
     int n_vocab  = model.hparams.n_vocab;
 
+    std::vector<uint8_t> work_buffer;
+
     for (int ex=0; ex<n_examples; ++ex) {
         struct ggml_init_params params = {
             /*.mem_size   =*/ compute_size,
@@ -1586,7 +1599,6 @@ int main(int argc, char ** argv) {
         int n_past = 0;
 
         ggml_cgraph gf = {};
-        gf.n_threads = 1;
 
         get_example_targets_batch(ctx0, 64*ex+0,  tokens_input, targets);
 
@@ -1595,7 +1607,7 @@ int main(int argc, char ** argv) {
         struct ggml_tensor * e = square_error_loss(ctx0, targets, logits);
 
         ggml_build_forward_expand(&gf, e);
-        ggml_graph_compute(ctx0, &gf);
+        ggml_graph_compute_helper(work_buffer, &gf, /*n_threads*/ 1);
 
         float error_before_opt = ggml_get_f32_1d(e, 0);
 
@@ -1611,7 +1623,7 @@ int main(int argc, char ** argv) {
         ggml_opt(ctx0, opt_params_lbfgs, e);
         //
         ggml_build_forward_expand(&gf, e);
-        ggml_graph_compute(ctx0, &gf);
+        ggml_graph_compute_helper(work_buffer, &gf, /*n_threads*/ 1);
 
         float error_after_opt = ggml_get_f32_1d(e, 0);
 
@@ -1659,13 +1671,12 @@ int main(int argc, char ** argv) {
             struct ggml_context * ctx0 = ggml_init(params);
 
             ggml_cgraph gf = {};
-            gf.n_threads = 1;
 
             int n_past = 0;
             struct ggml_tensor * logits = forward(&model, &kv_self, ctx0, &gf, tokens_input, sample_ctx, n_past);
 
             ggml_build_forward_expand(&gf, logits);
-            ggml_graph_compute(ctx0, &gf);
+            ggml_graph_compute_helper(work_buffer, &gf, /*n_threads*/ 1);
 
             struct ggml_tensor * best_samples = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, sample_ctx);
             struct ggml_tensor * probs        = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_vocab, sample_ctx);
@@ -1687,10 +1698,11 @@ int main(int argc, char ** argv) {
     }
 
     print_matrix(model.tok_embeddings);
-
     printf("done\n");
+
     // ggml_free(kv_self.ctx);
     // ggml_free(model_lora.ctx);
     ggml_free(model.ctx);
+
     return 0;
 }
diff --git a/examples/benchmark/benchmark-matmult.cpp b/examples/benchmark/benchmark-matmult.cpp
index 39d15caeb7779..f7215f43bb31c 100644
--- a/examples/benchmark/benchmark-matmult.cpp
+++ b/examples/benchmark/benchmark-matmult.cpp
@@ -20,6 +20,17 @@
 #pragma warning(disable: 4244 4267) // possible loss of data
 #endif
 
+void ggml_graph_compute_helper(std::vector<uint8_t> & buf, ggml_cgraph * graph, int n_threads) {
+    struct ggml_cplan plan = ggml_graph_plan(graph, n_threads);
+
+    if (plan.work_size > 0) {
+        buf.resize(plan.work_size);
+        plan.work_data = buf.data();
+    }
+
+    ggml_graph_compute(graph, &plan);
+}
+
 float tensor_sum_elements(const ggml_tensor * tensor) {
     float sum = 0;
     if (tensor->type==GGML_TYPE_F32) {
@@ -159,13 +170,14 @@ int main(int argc, char ** argv)  {
     // printf("Creating compute graph\n");
     struct ggml_cgraph gf = ggml_build_forward(m11xm2);
 
-    gf.n_threads=benchmark_params.n_threads;
-    printf("cgraph->n_threads=%i\n",gf.n_threads);
+    printf("n_threads=%i\n", benchmark_params.n_threads);
 
     TENSOR_DUMP(m11);
     TENSOR_DUMP(m2);
 
-    ggml_graph_compute(ctx, &gf);
+    std::vector<uint8_t> work_buffer;
+
+    ggml_graph_compute_helper(work_buffer, &gf, benchmark_params.n_threads);
 
     TENSOR_DUMP(gf.nodes[0]);
 
@@ -187,7 +199,6 @@ int main(int argc, char ** argv)  {
 
     // printf("Creating compute graph\n");
     struct ggml_cgraph gf31 = ggml_build_forward(q31);
-    gf31.n_threads=benchmark_params.n_threads;
 
     // Set up a second graph computation to make sure we override the CPU cache lines
     // printf("Creating new tensor q12 & Running quantize\n");
@@ -199,8 +210,7 @@ int main(int argc, char ** argv)  {
 
     //printf("Creating compute graph\n");
     struct ggml_cgraph gf32 = ggml_build_forward(q32);
-    gf32.n_threads=benchmark_params.n_threads;
-    printf("cgraph->n_threads=%i\n",gf31.n_threads);
+    printf("n_threads=%i\n", benchmark_params.n_threads);
 
     const int dimx = sizex;
     const int dimy = sizey;
@@ -221,14 +231,15 @@ int main(int argc, char ** argv)  {
 
         long long int start = ggml_time_us();
         //printf("Running ggml_graph_compute\n");
-        ggml_graph_compute(ctx, &gf31);
+        ggml_graph_compute_helper(work_buffer, &gf31, benchmark_params.n_threads);
+
         long long int stop = ggml_time_us();
         long long int usec = stop-start;
         double gflops = (double)(flops_per_matrix)/usec/1000.0;
         gflops_sum += gflops;
         printf("%9i;%8i;%6i;%6i;%6i;%15lli;%18lli;%10.2f\n",
             i,
-            gf31.n_threads,
+            benchmark_params.n_threads,
             sizex, sizey, sizez, flops_per_matrix,
             usec,gflops);
 
@@ -253,7 +264,7 @@ int main(int argc, char ** argv)  {
         }
 
         // Running a different graph computation to make sure we override the CPU cache lines
-        ggml_graph_compute(ctx, &gf32);
+        ggml_graph_compute_helper(work_buffer, &gf32, benchmark_params.n_threads);
     }
     printf("\n");
     printf("Average%78.2f\n",gflops_sum/((double)benchmark_params.n_iterations));
diff --git a/examples/common.cpp b/examples/common.cpp
index 3278a064346b4..93159c6dfc6de 100644
--- a/examples/common.cpp
+++ b/examples/common.cpp
@@ -418,6 +418,8 @@ bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
 
     if (escape_prompt) {
         process_escapes(params.prompt);
+        process_escapes(params.input_prefix);
+        process_escapes(params.input_suffix);
     }
 
     return true;
diff --git a/examples/embd-input/embd-input-lib.cpp b/examples/embd-input/embd-input-lib.cpp
index 570e273fcfffc..5fa4942be7aaf 100644
--- a/examples/embd-input/embd-input-lib.cpp
+++ b/examples/embd-input/embd-input-lib.cpp
@@ -29,7 +29,7 @@ struct MyModel* create_mymodel(int argc, char ** argv) {
 
     fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT);
 
-    if (params.seed < 0) {
+    if (params.seed == LLAMA_DEFAULT_SEED) {
         params.seed = time(NULL);
     }
     fprintf(stderr, "%s: seed  = %d\n", __func__, params.seed);
diff --git a/examples/embedding/embedding.cpp b/examples/embedding/embedding.cpp
index 2b7eb39c51ff5..03e801c2a6d4b 100644
--- a/examples/embedding/embedding.cpp
+++ b/examples/embedding/embedding.cpp
@@ -18,7 +18,7 @@ int main(int argc, char ** argv) {
     params.embedding = true;
 
     if (params.n_ctx > 2048) {
-        fprintf(stderr, "%s: warning: model does not support context sizes greater than 2048 tokens (%d specified);"
+        fprintf(stderr, "%s: warning: model might not support context sizes greater than 2048 tokens (%d specified);"
                 "expect poor results\n", __func__, params.n_ctx);
     }
 
diff --git a/examples/main/main.cpp b/examples/main/main.cpp
index 3a171925ba510..0f6391acba45d 100644
--- a/examples/main/main.cpp
+++ b/examples/main/main.cpp
@@ -85,7 +85,7 @@ int main(int argc, char ** argv) {
     }
 
     if (params.n_ctx > 2048) {
-        fprintf(stderr, "%s: warning: model does not support context sizes greater than 2048 tokens (%d specified);"
+        fprintf(stderr, "%s: warning: model might not support context sizes greater than 2048 tokens (%d specified);"
                 "expect poor results\n", __func__, params.n_ctx);
     } else if (params.n_ctx < 8) {
         fprintf(stderr, "%s: warning: minimum context size is 8, using minimum size.\n", __func__);
diff --git a/examples/metal/metal.cpp b/examples/metal/metal.cpp
index cdfe4bfe97865..7438defdefcdf 100644
--- a/examples/metal/metal.cpp
+++ b/examples/metal/metal.cpp
@@ -35,10 +35,9 @@ int main(int argc, char ** argv) {
     struct ggml_context * ctx_eval = NULL;
 
     struct ggml_cgraph gf = ggml_graph_import(fname_cgraph, &ctx_data, &ctx_eval);
-    gf.n_threads = 1;
 
     // this allocates all Metal resources and memory buffers
-    auto * ctx_metal = ggml_metal_init();
+    auto * ctx_metal = ggml_metal_init(1);
 
     const size_t max_size_data = ggml_get_max_tensor_size(ctx_data);
     const size_t max_size_eval = ggml_get_max_tensor_size(ctx_eval);
diff --git a/examples/perplexity/perplexity.cpp b/examples/perplexity/perplexity.cpp
index dd54ed3c4bd6c..fd4b03cb261f6 100644
--- a/examples/perplexity/perplexity.cpp
+++ b/examples/perplexity/perplexity.cpp
@@ -130,7 +130,7 @@ int main(int argc, char ** argv) {
     params.n_batch = std::min(params.n_batch, params.n_ctx);
 
     if (params.n_ctx > 2048) {
-        fprintf(stderr, "%s: warning: model does not support context sizes greater than 2048 tokens (%d specified);"
+        fprintf(stderr, "%s: warning: model might not support context sizes greater than 2048 tokens (%d specified);"
                 "expect poor results\n", __func__, params.n_ctx);
     }
 
diff --git a/examples/quantize-stats/quantize-stats.cpp b/examples/quantize-stats/quantize-stats.cpp
index 9cea472dedb82..6aa06ec8fa115 100644
--- a/examples/quantize-stats/quantize-stats.cpp
+++ b/examples/quantize-stats/quantize-stats.cpp
@@ -147,7 +147,7 @@ void test_roundtrip_on_chunk(
         const ggml_tensor * layer,
         int64_t offset,
         int64_t chunk_size,
-        const quantize_fns_t & qfns,
+        const ggml_type_traits_t & qfns,
         bool use_reference,
         float * input_scratch,
         char * quantized_scratch,
@@ -163,11 +163,11 @@ void test_roundtrip_on_chunk(
     }
 
     if (use_reference) {
-        qfns.quantize_row_q_reference(input_scratch, quantized_scratch, chunk_size);
+        qfns.from_float_reference(input_scratch, quantized_scratch, chunk_size);
     } else {
-        qfns.quantize_row_q(input_scratch, quantized_scratch, chunk_size);
+        qfns.from_float(input_scratch, quantized_scratch, chunk_size);
     }
-    qfns.dequantize_row_q(quantized_scratch, output_scratch, chunk_size);
+    qfns.to_float(quantized_scratch, output_scratch, chunk_size);
 
     update_error_stats(chunk_size, input_scratch, output_scratch, stats);
 }
@@ -177,7 +177,7 @@ void test_roundtrip_on_chunk(
 void test_roundtrip_on_layer(
         std::string & name,
         bool print_layer_stats,
-        const quantize_fns_t & qfns,
+        const ggml_type_traits_t & qfns,
         bool use_reference,
         const ggml_tensor * layer,
         std::vector<float> & input_scratch,
@@ -388,8 +388,8 @@ int main(int argc, char ** argv) {
         if (!params.include_types.empty() && std::find(params.include_types.begin(), params.include_types.end(), i) == params.include_types.end()) {
             continue;
         }
-        quantize_fns_t qfns = ggml_internal_get_quantize_fn(i);
-        if (qfns.quantize_row_q && qfns.dequantize_row_q) {
+        ggml_type_traits_t qfns = ggml_internal_get_type_traits(type);
+        if (qfns.from_float && qfns.to_float) {
             if (params.verbose) {
                 printf("testing %s ...\n",  ggml_type_name(type));
             }
diff --git a/examples/server/README.md b/examples/server/README.md
index ba4b2fec9d1df..ad9b6bb081845 100644
--- a/examples/server/README.md
+++ b/examples/server/README.md
@@ -1,13 +1,13 @@
 # llama.cpp/example/server
 
-This example demonstrates a simple HTTP API server to interact with llama.cpp.
+This example demonstrates a simple HTTP API server and a simple web front end to interact with llama.cpp.
 
 Command line options:
 
 -   `--threads N`, `-t N`: Set the number of threads to use during computation.
 -   `-m FNAME`, `--model FNAME`: Specify the path to the LLaMA model file (e.g., `models/7B/ggml-model.bin`).
 -   `-m ALIAS`, `--alias ALIAS`: Set an alias for the model. The alias will be returned in API responses.
--   `-c N`, `--ctx-size N`: Set the size of the prompt context. The default is 512, but LLaMA models were built with a context of 2048, which will provide better results for longer input/inference.
+-   `-c N`, `--ctx-size N`: Set the size of the prompt context. The default is 512, but LLaMA models were built with a context of 2048, which will provide better results for longer input/inference. The size may differ in other models, for example, baichuan models were build with a context of 4096.
 -   `-ngl N`, `--n-gpu-layers N`: When compiled with appropriate support (currently CLBlast or cuBLAS), this option allows offloading some layers to the GPU for computation. Generally results in increased performance.
 -   `-mg i, --main-gpu i`: When using multiple GPUs this option controls which GPU is used for small tensors for which the overhead of splitting the computation across all GPUs is not worthwhile. The GPU in question will use slightly more VRAM to store a scratch buffer for temporary results. By default GPU 0 is used. Requires cuBLAS.
 -   `-ts SPLIT, --tensor-split SPLIT`: When using multiple GPUs this option controls how large tensors should be split across all GPUs. `SPLIT` is a comma-separated list of non-negative values that assigns the proportion of data that each GPU should get in order. For example, "3,2" will assign 60% of the data to GPU 0 and 40% to GPU 1. By default the data is split in proportion to VRAM but this may not be optimal for performance. Requires cuBLAS.
@@ -21,24 +21,22 @@ Command line options:
 -   `-to N`, `--timeout N`: Server read/write timeout in seconds. Default `600`.
 -   `--host`: Set the hostname or ip address to listen. Default `127.0.0.1`.
 -   `--port`: Set the port to listen. Default: `8080`.
+-   `--path`: path from which to serve static files (default examples/server/public)
 -   `--embedding`: Enable embedding extraction, Default: disabled.
 
 ## Build
 
-Build llama.cpp with server from repository root with either make or CMake.
+server is build alongside everything else from the root of the project
 
 - Using `make`:
 
   ```bash
-  LLAMA_BUILD_SERVER=1 make
+  make
   ```
 
 - Using `CMake`:
 
   ```bash
-  mkdir build-server
-  cd build-server
-  cmake -DLLAMA_BUILD_SERVER=ON ..
   cmake --build . --config Release
   ```
 
@@ -59,7 +57,7 @@ server.exe -m models\7B\ggml-model.bin -c 2048
 ```
 
 The above command will start a server that by default listens on `127.0.0.1:8080`.
-You can consume the endpoints with Postman or NodeJS with axios library.
+You can consume the endpoints with Postman or NodeJS with axios library. You can visit the web front end at the same url.
 
 ## Testing with CURL
 
@@ -190,3 +188,49 @@ Run with bash:
 ```sh
 bash chat.sh
 ```
+
+### API like OAI
+
+API example using Python Flask: [api_like_OAI.py](api_like_OAI.py)
+This example must be used with server.cpp
+
+```sh
+python api_like_OAI.py
+```
+
+After running the API server, you can use it in Python by setting the API base URL.
+```python
+openai.api_base = "http://<Your api-server IP>:port"
+```
+
+Then you can utilize llama.cpp as an OpenAI's **chat.completion** or **text_completion** API
+
+### Extending or building alternative Web Front End
+
+The default location for the static files is `examples/server/public`. You can extend the front end by running the server binary with `--path` set to `./your-directory` and importing `/completion.js` to get access to the llamaComplete() method.
+
+Read the documentation in `/completion.js` to see convenient ways to access llama.
+
+A simple example is below:
+
+```html
+<html>
+  <body>
+    <pre>
+      <script type="module">
+        import { llama } from '/completion.js'
+
+        const prompt = `### Instruction:
+Write dad jokes, each one paragraph.
+You can use html formatting if needed.
+
+### Response:`
+
+        for await (const chunk of llama(prompt)) {
+          document.write(chunk.data.content)
+        }
+      </script>
+    </pre>
+  </body>
+</html>
+```
diff --git a/examples/server/api_like_OAI.py b/examples/server/api_like_OAI.py
new file mode 100755
index 0000000000000..aa325a03ee444
--- /dev/null
+++ b/examples/server/api_like_OAI.py
@@ -0,0 +1,219 @@
+import argparse
+from flask import Flask, jsonify, request, Response
+import urllib.parse
+import requests
+import time
+import json
+
+
+app = Flask(__name__)
+
+parser = argparse.ArgumentParser(description="An example of using server.cpp with a similar API to OAI. It must be used together with server.cpp.")
+parser.add_argument("--chat-prompt", type=str, help="the top prompt in chat completions(default: 'A chat between a curious user and an artificial intelligence assistant. The assistant follows the given rules no matter what.\\n')", default='A chat between a curious user and an artificial intelligence assistant. The assistant follows the given rules no matter what.\\n')
+parser.add_argument("--user-name", type=str, help="USER name in chat completions(default: '\\nUSER: ')", default="\\nUSER: ")
+parser.add_argument("--ai-name", type=str, help="ASSISTANT name in chat completions(default: '\\nASSISTANT: ')", default="\\nASSISTANT: ")
+parser.add_argument("--system-name", type=str, help="SYSTEM name in chat completions(default: '\\nASSISTANT's RULE: ')", default="\\nASSISTANT's RULE: ")
+parser.add_argument("--stop", type=str, help="the end of response in chat completions(default: '</s>')", default="</s>")
+parser.add_argument("--llama-api", type=str, help="Set the address of server.cpp in llama.cpp(default: http://127.0.0.1:8080)", default='http://127.0.0.1:8080')
+parser.add_argument("--api-key", type=str, help="Set the api key to allow only few user(default: NULL)", default="")
+parser.add_argument("--host", type=str, help="Set the ip address to listen.(default: 127.0.0.1)", default='127.0.0.1')
+parser.add_argument("--port", type=int, help="Set the port to listen.(default: 8081)", default=8081)
+
+args = parser.parse_args()
+
+def is_present(json, key):
+    try:
+        buf = json[key]
+    except KeyError:
+        return False
+    return True
+
+
+
+#convert chat to prompt
+def convert_chat(messages):
+    prompt = "" + args.chat_prompt.replace("\\n", "\n")
+
+    system_n = args.system_name.replace("\\n", "\n")
+    user_n = args.user_name.replace("\\n", "\n")
+    ai_n = args.ai_name.replace("\\n", "\n")
+    stop = args.stop.replace("\\n", "\n")
+
+
+    for line in messages:
+        if (line["role"] == "system"):
+            prompt += f"{system_n}{line['content']}"
+        if (line["role"] == "user"):
+            prompt += f"{user_n}{line['content']}"
+        if (line["role"] == "assistant"):
+            prompt += f"{ai_n}{line['content']}{stop}"
+    prompt += ai_n.rstrip()
+
+    return prompt
+
+def make_postData(body, chat=False, stream=False):
+    postData = {}
+    if (chat):
+        postData["prompt"] = convert_chat(body["messages"])
+    else:
+        postData["prompt"] = body["prompt"]
+    if(is_present(body, "temperature")): postData["temperature"] = body["temperature"]
+    if(is_present(body, "top_k")): postData["top_k"] = body["top_k"]
+    if(is_present(body, "top_p")): postData["top_p"] = body["top_p"]
+    if(is_present(body, "max_tokens")): postData["n_predict"] = body["max_tokens"]
+    if(is_present(body, "presence_penalty")): postData["presence_penalty"] = body["presence_penalty"]
+    if(is_present(body, "frequency_penalty")): postData["frequency_penalty"] = body["frequency_penalty"]
+    if(is_present(body, "repeat_penalty")): postData["repeat_penalty"] = body["repeat_penalty"]
+    if(is_present(body, "mirostat")): postData["mirostat"] = body["mirostat"]
+    if(is_present(body, "mirostat_tau")): postData["mirostat_tau"] = body["mirostat_tau"]
+    if(is_present(body, "mirostat_eta")): postData["mirostat_eta"] = body["mirostat_eta"]
+    if(is_present(body, "seed")): postData["seed"] = body["seed"]
+    if(is_present(body, "logit_bias")): postData["logit_bias"] = [[int(token), body["logit_bias"][token]] for token in body["logit_bias"].keys()]
+    if (args.stop != ""):
+        postData["stop"] = [args.stop]
+    else:
+        postData["stop"] = []
+    if(is_present(body, "stop")): postData["stop"] += body["stop"]
+    postData["n_keep"] = -1
+    postData["stream"] = stream
+
+    return postData
+
+def make_resData(data, chat=False, promptToken=[]):
+    resData = {
+        "id": "chatcmpl" if (chat) else "cmpl",
+        "object": "chat.completion" if (chat) else "text_completion",
+        "created": int(time.time()),
+        "truncated": data["truncated"],
+        "model": "LLaMA_CPP",
+        "usage": {
+            "prompt_tokens": data["tokens_evaluated"],
+            "completion_tokens": data["tokens_predicted"],
+            "total_tokens": data["tokens_evaluated"] + data["tokens_predicted"]
+        }
+    }
+    if (len(promptToken) != 0):
+        resData["promptToken"] = promptToken
+    if (chat):
+        #only one choice is supported
+        resData["choices"] = [{
+            "index": 0,
+            "message": {
+                "role": "assistant",
+                "content": data["content"],
+            },
+            "finish_reason": "stop" if (data["stopped_eos"] or data["stopped_word"]) else "length"
+        }]
+    else:
+        #only one choice is supported
+        resData["choices"] = [{
+            "text": data["content"],
+            "index": 0,
+            "logprobs": None,
+            "finish_reason": "stop" if (data["stopped_eos"] or data["stopped_word"]) else "length"
+        }]
+    return resData
+
+def make_resData_stream(data, chat=False, time_now = 0, start=False):
+    resData = {
+        "id": "chatcmpl" if (chat) else "cmpl",
+        "object": "chat.completion.chunk" if (chat) else "text_completion.chunk",
+        "created": time_now,
+        "model": "LLaMA_CPP",
+        "choices": [
+            {
+                "finish_reason": None,
+                "index": 0
+            }
+        ]
+    }
+    if (chat):
+        if (start):
+            resData["choices"][0]["delta"] =  {
+                "role": "assistant"
+            }
+        else:
+            resData["choices"][0]["delta"] =  {
+                "content": data["content"]
+            }
+            if (data["stop"]):
+                resData["choices"][0]["finish_reason"] = "stop" if (data["stopped_eos"] or data["stopped_word"]) else "length"
+    else:
+        resData["choices"][0]["text"] = data["content"]
+        if (data["stop"]):
+            resData["choices"][0]["finish_reason"] = "stop" if (data["stopped_eos"] or data["stopped_word"]) else "length"
+
+    return resData
+
+
+@app.route('/chat/completions', methods=['POST'])
+@app.route('/v1/chat/completions', methods=['POST'])
+def chat_completions():
+    if (args.api_key != "" and request.headers["Authorization"].split()[1] != args.api_key):
+        return Response(status=403)
+    body = request.get_json()
+    stream = False
+    tokenize = False
+    if(is_present(body, "stream")): stream = body["stream"]
+    if(is_present(body, "tokenize")): tokenize = body["tokenize"]
+    postData = make_postData(body, chat=True, stream=stream)
+
+    promptToken = []
+    if (tokenize):
+        tokenData = requests.request("POST", urllib.parse.urljoin(args.llama_api, "/tokenize"), data=json.dumps({"content": postData["prompt"]})).json()
+        promptToken = tokenData["tokens"]
+
+    if (not stream):
+        data = requests.request("POST", urllib.parse.urljoin(args.llama_api, "/completion"), data=json.dumps(postData))
+        print(data.json())
+        resData = make_resData(data.json(), chat=True, promptToken=promptToken)
+        return jsonify(resData)
+    else:
+        def generate():
+            data = requests.request("POST", urllib.parse.urljoin(args.llama_api, "/completion"), data=json.dumps(postData), stream=True)
+            time_now = int(time.time())
+            resData = make_resData_stream({}, chat=True, time_now=time_now, start=True)
+            yield 'data: {}\n'.format(json.dumps(resData))
+            for line in data.iter_lines():
+                if line:
+                    decoded_line = line.decode('utf-8')
+                    resData = make_resData_stream(json.loads(decoded_line[6:]), chat=True, time_now=time_now)
+                    yield 'data: {}\n'.format(json.dumps(resData))
+        return Response(generate(), mimetype='text/event-stream')
+
+
+@app.route('/completions', methods=['POST'])
+@app.route('/v1/completions', methods=['POST'])
+def completion():
+    if (args.api_key != "" and request.headers["Authorization"].split()[1] != args.api_key):
+        return Response(status=403)
+    body = request.get_json()
+    stream = False
+    tokenize = False
+    if(is_present(body, "stream")): stream = body["stream"]
+    if(is_present(body, "tokenize")): tokenize = body["tokenize"]
+    postData = make_postData(body, chat=False, stream=stream)
+
+    promptToken = []
+    if (tokenize):
+        tokenData = requests.request("POST", urllib.parse.urljoin(args.llama_api, "/tokenize"), data=json.dumps({"content": postData["prompt"]})).json()
+        promptToken = tokenData["tokens"]
+
+    if (not stream):
+        data = requests.request("POST", urllib.parse.urljoin(args.llama_api, "/completion"), data=json.dumps(postData))
+        print(data.json())
+        resData = make_resData(data.json(), chat=False, promptToken=promptToken)
+        return jsonify(resData)
+    else:
+        def generate():
+            data = requests.request("POST", urllib.parse.urljoin(args.llama_api, "/completion"), data=json.dumps(postData), stream=True)
+            time_now = int(time.time())
+            for line in data.iter_lines():
+                if line:
+                    decoded_line = line.decode('utf-8')
+                    resData = make_resData_stream(json.loads(decoded_line[6:]), chat=False, time_now=time_now)
+                    yield 'data: {}\n'.format(json.dumps(resData))
+        return Response(generate(), mimetype='text/event-stream')
+
+if __name__ == '__main__':
+    app.run(args.host, port=args.port)
diff --git a/examples/server/completion.js.hpp b/examples/server/completion.js.hpp
index 002830cad7962..f399fb19a6bc8 100644
--- a/examples/server/completion.js.hpp
+++ b/examples/server/completion.js.hpp
@@ -7,187 +7,369 @@ unsigned char completion_js[] = {
   0x74, 0x65, 0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x3a,
   0x20, 0x30, 0x2e, 0x32, 0x2c, 0x0a, 0x20, 0x20, 0x73, 0x74, 0x6f, 0x70,
   0x3a, 0x20, 0x5b, 0x22, 0x3c, 0x2f, 0x73, 0x3e, 0x22, 0x5d, 0x0a, 0x7d,
-  0x3b, 0x0a, 0x0a, 0x2f, 0x2a, 0x2a, 0x0a, 0x20, 0x2a, 0x20, 0x54, 0x68,
-  0x69, 0x73, 0x20, 0x66, 0x75, 0x6e, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x20,
-  0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x65, 0x73, 0x20, 0x74, 0x68,
-  0x65, 0x20, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x20, 0x74, 0x65, 0x78, 0x74,
-  0x20, 0x75, 0x73, 0x69, 0x6e, 0x67, 0x20, 0x61, 0x20, 0x6c, 0x6c, 0x61,
-  0x6d, 0x61, 0x20, 0x64, 0x69, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x61, 0x72,
-  0x79, 0x2e, 0x0a, 0x20, 0x2a, 0x20, 0x40, 0x70, 0x61, 0x72, 0x61, 0x6d,
-  0x20, 0x7b, 0x6f, 0x62, 0x6a, 0x65, 0x63, 0x74, 0x7d, 0x20, 0x70, 0x61,
-  0x72, 0x61, 0x6d, 0x73, 0x20, 0x2d, 0x20, 0x54, 0x68, 0x65, 0x20, 0x70,
-  0x61, 0x72, 0x61, 0x6d, 0x65, 0x74, 0x65, 0x72, 0x73, 0x20, 0x66, 0x6f,
-  0x72, 0x20, 0x74, 0x68, 0x65, 0x20, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65,
-  0x74, 0x69, 0x6f, 0x6e, 0x20, 0x72, 0x65, 0x71, 0x75, 0x65, 0x73, 0x74,
-  0x2e, 0x0a, 0x20, 0x2a, 0x20, 0x40, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x20,
-  0x7b, 0x6f, 0x62, 0x6a, 0x65, 0x63, 0x74, 0x7d, 0x20, 0x63, 0x6f, 0x6e,
-  0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x20, 0x2d, 0x20, 0x61, 0x6e,
-  0x20, 0x69, 0x6e, 0x73, 0x74, 0x61, 0x6e, 0x63, 0x65, 0x20, 0x6f, 0x66,
-  0x20, 0x41, 0x62, 0x6f, 0x72, 0x74, 0x43, 0x6f, 0x6e, 0x74, 0x72, 0x6f,
-  0x6c, 0x6c, 0x65, 0x72, 0x20, 0x69, 0x66, 0x20, 0x79, 0x6f, 0x75, 0x20,
-  0x6e, 0x65, 0x65, 0x64, 0x20, 0x6f, 0x6e, 0x65, 0x2c, 0x20, 0x6f, 0x72,
-  0x20, 0x6e, 0x75, 0x6c, 0x6c, 0x2e, 0x0a, 0x20, 0x2a, 0x20, 0x40, 0x70,
-  0x61, 0x72, 0x61, 0x6d, 0x20, 0x7b, 0x66, 0x75, 0x6e, 0x63, 0x74, 0x69,
-  0x6f, 0x6e, 0x7d, 0x20, 0x63, 0x61, 0x6c, 0x6c, 0x62, 0x61, 0x63, 0x6b,
-  0x20, 0x2d, 0x20, 0x54, 0x68, 0x65, 0x20, 0x63, 0x61, 0x6c, 0x6c, 0x62,
-  0x61, 0x63, 0x6b, 0x20, 0x66, 0x75, 0x6e, 0x63, 0x74, 0x69, 0x6f, 0x6e,
-  0x20, 0x74, 0x6f, 0x20, 0x63, 0x61, 0x6c, 0x6c, 0x20, 0x77, 0x68, 0x65,
-  0x6e, 0x20, 0x74, 0x68, 0x65, 0x20, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65,
-  0x74, 0x69, 0x6f, 0x6e, 0x20, 0x69, 0x73, 0x20, 0x64, 0x6f, 0x6e, 0x65,
-  0x2e, 0x0a, 0x20, 0x2a, 0x20, 0x40, 0x72, 0x65, 0x74, 0x75, 0x72, 0x6e,
-  0x73, 0x20, 0x7b, 0x73, 0x74, 0x72, 0x69, 0x6e, 0x67, 0x7d, 0x20, 0x74,
-  0x68, 0x65, 0x20, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x65, 0x64,
-  0x20, 0x74, 0x65, 0x78, 0x74, 0x20, 0x61, 0x73, 0x20, 0x61, 0x20, 0x73,
-  0x74, 0x72, 0x69, 0x6e, 0x67, 0x2e, 0x20, 0x49, 0x64, 0x65, 0x61, 0x6c,
-  0x6c, 0x79, 0x20, 0x69, 0x67, 0x6e, 0x6f, 0x72, 0x65, 0x64, 0x2c, 0x20,
-  0x61, 0x6e, 0x64, 0x20, 0x79, 0x6f, 0x75, 0x20, 0x67, 0x65, 0x74, 0x20,
-  0x61, 0x74, 0x20, 0x69, 0x74, 0x20, 0x76, 0x69, 0x61, 0x20, 0x74, 0x68,
-  0x65, 0x20, 0x63, 0x61, 0x6c, 0x6c, 0x62, 0x61, 0x63, 0x6b, 0x2e, 0x0a,
-  0x20, 0x2a, 0x2f, 0x0a, 0x65, 0x78, 0x70, 0x6f, 0x72, 0x74, 0x20, 0x63,
-  0x6f, 0x6e, 0x73, 0x74, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x43, 0x6f,
-  0x6d, 0x70, 0x6c, 0x65, 0x74, 0x65, 0x20, 0x3d, 0x20, 0x61, 0x73, 0x79,
-  0x6e, 0x63, 0x20, 0x28, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2c, 0x20,
-  0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2c, 0x20,
-  0x63, 0x61, 0x6c, 0x6c, 0x62, 0x61, 0x63, 0x6b, 0x29, 0x20, 0x3d, 0x3e,
-  0x20, 0x7b, 0x0a, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x21, 0x63, 0x6f,
-  0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x29, 0x20, 0x7b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c,
-  0x65, 0x72, 0x20, 0x3d, 0x20, 0x6e, 0x65, 0x77, 0x20, 0x41, 0x62, 0x6f,
-  0x72, 0x74, 0x43, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72,
-  0x28, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x63, 0x6f,
-  0x6e, 0x73, 0x74, 0x20, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x69,
-  0x6f, 0x6e, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x20, 0x3d, 0x20, 0x7b,
-  0x20, 0x2e, 0x2e, 0x2e, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x44, 0x65, 0x66,
-  0x61, 0x75, 0x6c, 0x74, 0x73, 0x2c, 0x20, 0x2e, 0x2e, 0x2e, 0x70, 0x61,
-  0x72, 0x61, 0x6d, 0x73, 0x20, 0x7d, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x2f,
-  0x2f, 0x20, 0x77, 0x65, 0x20, 0x75, 0x73, 0x65, 0x20, 0x66, 0x65, 0x74,
-  0x63, 0x68, 0x20, 0x64, 0x69, 0x72, 0x65, 0x63, 0x74, 0x6c, 0x79, 0x20,
-  0x68, 0x65, 0x72, 0x65, 0x20, 0x62, 0x65, 0x63, 0x61, 0x73, 0x75, 0x65,
-  0x20, 0x74, 0x68, 0x65, 0x20, 0x62, 0x75, 0x69, 0x6c, 0x74, 0x20, 0x69,
-  0x6e, 0x20, 0x66, 0x65, 0x74, 0x63, 0x68, 0x45, 0x76, 0x65, 0x6e, 0x74,
-  0x53, 0x6f, 0x75, 0x72, 0x63, 0x65, 0x20, 0x64, 0x6f, 0x65, 0x73, 0x20,
-  0x6e, 0x6f, 0x74, 0x20, 0x73, 0x75, 0x70, 0x70, 0x6f, 0x72, 0x74, 0x20,
-  0x50, 0x4f, 0x53, 0x54, 0x0a, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74,
-  0x20, 0x72, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x20, 0x3d, 0x20,
-  0x61, 0x77, 0x61, 0x69, 0x74, 0x20, 0x66, 0x65, 0x74, 0x63, 0x68, 0x28,
-  0x22, 0x2f, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x69, 0x6f, 0x6e,
-  0x22, 0x2c, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x65, 0x74,
-  0x68, 0x6f, 0x64, 0x3a, 0x20, 0x27, 0x50, 0x4f, 0x53, 0x54, 0x27, 0x2c,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x62, 0x6f, 0x64, 0x79, 0x3a, 0x20, 0x4a,
-  0x53, 0x4f, 0x4e, 0x2e, 0x73, 0x74, 0x72, 0x69, 0x6e, 0x67, 0x69, 0x66,
-  0x79, 0x28, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x69, 0x6f, 0x6e,
-  0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x29, 0x2c, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x68, 0x65, 0x61, 0x64, 0x65, 0x72, 0x73, 0x3a, 0x20, 0x7b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x27, 0x43, 0x6f, 0x6e, 0x6e, 0x65,
-  0x63, 0x74, 0x69, 0x6f, 0x6e, 0x27, 0x3a, 0x20, 0x27, 0x6b, 0x65, 0x65,
-  0x70, 0x2d, 0x61, 0x6c, 0x69, 0x76, 0x65, 0x27, 0x2c, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x27, 0x43, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74,
-  0x2d, 0x54, 0x79, 0x70, 0x65, 0x27, 0x3a, 0x20, 0x27, 0x61, 0x70, 0x70,
-  0x6c, 0x69, 0x63, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x2f, 0x6a, 0x73, 0x6f,
-  0x6e, 0x27, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x27, 0x41,
-  0x63, 0x63, 0x65, 0x70, 0x74, 0x27, 0x3a, 0x20, 0x27, 0x74, 0x65, 0x78,
-  0x74, 0x2f, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x2d, 0x73, 0x74, 0x72, 0x65,
-  0x61, 0x6d, 0x27, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x2c, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x73, 0x69, 0x67, 0x6e, 0x61, 0x6c, 0x3a, 0x20, 0x63,
-  0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x73, 0x69,
-  0x67, 0x6e, 0x61, 0x6c, 0x2c, 0x0a, 0x20, 0x20, 0x7d, 0x29, 0x3b, 0x0a,
-  0x0a, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x72, 0x65, 0x61,
-  0x64, 0x65, 0x72, 0x20, 0x3d, 0x20, 0x72, 0x65, 0x73, 0x70, 0x6f, 0x6e,
-  0x73, 0x65, 0x2e, 0x62, 0x6f, 0x64, 0x79, 0x2e, 0x67, 0x65, 0x74, 0x52,
-  0x65, 0x61, 0x64, 0x65, 0x72, 0x28, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x63,
-  0x6f, 0x6e, 0x73, 0x74, 0x20, 0x64, 0x65, 0x63, 0x6f, 0x64, 0x65, 0x72,
-  0x20, 0x3d, 0x20, 0x6e, 0x65, 0x77, 0x20, 0x54, 0x65, 0x78, 0x74, 0x44,
-  0x65, 0x63, 0x6f, 0x64, 0x65, 0x72, 0x28, 0x29, 0x3b, 0x0a, 0x0a, 0x20,
-  0x20, 0x6c, 0x65, 0x74, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74,
-  0x20, 0x3d, 0x20, 0x22, 0x22, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x74, 0x72,
-  0x79, 0x20, 0x7b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x6c, 0x65, 0x74,
-  0x20, 0x63, 0x6f, 0x6e, 0x74, 0x20, 0x3d, 0x20, 0x74, 0x72, 0x75, 0x65,
-  0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x77, 0x68, 0x69, 0x6c, 0x65,
-  0x20, 0x28, 0x63, 0x6f, 0x6e, 0x74, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x72, 0x65,
-  0x73, 0x75, 0x6c, 0x74, 0x20, 0x3d, 0x20, 0x61, 0x77, 0x61, 0x69, 0x74,
-  0x20, 0x72, 0x65, 0x61, 0x64, 0x65, 0x72, 0x2e, 0x72, 0x65, 0x61, 0x64,
-  0x28, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x69, 0x66,
-  0x20, 0x28, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x2e, 0x64, 0x6f, 0x6e,
-  0x65, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x62, 0x72, 0x65, 0x61, 0x6b, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2f,
-  0x2f, 0x20, 0x73, 0x73, 0x65, 0x20, 0x61, 0x6e, 0x73, 0x77, 0x65, 0x72,
-  0x73, 0x20, 0x69, 0x6e, 0x20, 0x74, 0x68, 0x65, 0x20, 0x66, 0x6f, 0x72,
-  0x6d, 0x20, 0x6d, 0x75, 0x6c, 0x74, 0x69, 0x70, 0x6c, 0x65, 0x20, 0x6c,
-  0x69, 0x6e, 0x65, 0x73, 0x20, 0x6f, 0x66, 0x3a, 0x20, 0x76, 0x61, 0x6c,
-  0x75, 0x65, 0x5c, 0x6e, 0x20, 0x77, 0x69, 0x74, 0x68, 0x20, 0x64, 0x61,
-  0x74, 0x61, 0x20, 0x61, 0x6c, 0x77, 0x61, 0x79, 0x73, 0x20, 0x70, 0x72,
-  0x65, 0x73, 0x65, 0x6e, 0x74, 0x20, 0x61, 0x73, 0x20, 0x61, 0x20, 0x6b,
-  0x65, 0x79, 0x2e, 0x20, 0x69, 0x6e, 0x20, 0x6f, 0x75, 0x72, 0x20, 0x63,
-  0x61, 0x73, 0x65, 0x20, 0x77, 0x65, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x2f, 0x2f, 0x20, 0x6d, 0x61, 0x69, 0x6e, 0x6c, 0x79, 0x20, 0x63,
-  0x61, 0x72, 0x65, 0x20, 0x61, 0x62, 0x6f, 0x75, 0x74, 0x20, 0x74, 0x68,
-  0x65, 0x20, 0x64, 0x61, 0x74, 0x61, 0x3a, 0x20, 0x6b, 0x65, 0x79, 0x20,
-  0x68, 0x65, 0x72, 0x65, 0x2c, 0x20, 0x77, 0x68, 0x69, 0x63, 0x68, 0x20,
-  0x77, 0x65, 0x20, 0x65, 0x78, 0x70, 0x65, 0x63, 0x74, 0x20, 0x61, 0x73,
-  0x20, 0x6a, 0x73, 0x6f, 0x6e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x74, 0x65, 0x78, 0x74, 0x20, 0x3d,
-  0x20, 0x64, 0x65, 0x63, 0x6f, 0x64, 0x65, 0x72, 0x2e, 0x64, 0x65, 0x63,
-  0x6f, 0x64, 0x65, 0x28, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x2e, 0x76,
-  0x61, 0x6c, 0x75, 0x65, 0x29, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x2f, 0x2f, 0x20, 0x70, 0x61, 0x72, 0x73, 0x65, 0x20, 0x61,
-  0x6c, 0x6c, 0x20, 0x73, 0x73, 0x65, 0x20, 0x65, 0x76, 0x65, 0x6e, 0x74,
-  0x73, 0x20, 0x61, 0x6e, 0x64, 0x20, 0x61, 0x64, 0x64, 0x20, 0x74, 0x68,
-  0x65, 0x6d, 0x20, 0x74, 0x6f, 0x20, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74,
-  0x20, 0x72, 0x65, 0x67, 0x65, 0x78, 0x20, 0x3d, 0x20, 0x2f, 0x5e, 0x28,
-  0x5c, 0x53, 0x2b, 0x29, 0x3a, 0x5c, 0x73, 0x28, 0x2e, 0x2a, 0x29, 0x24,
-  0x2f, 0x67, 0x6d, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x66,
-  0x6f, 0x72, 0x20, 0x28, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x6d, 0x61,
-  0x74, 0x63, 0x68, 0x20, 0x6f, 0x66, 0x20, 0x74, 0x65, 0x78, 0x74, 0x2e,
-  0x6d, 0x61, 0x74, 0x63, 0x68, 0x41, 0x6c, 0x6c, 0x28, 0x72, 0x65, 0x67,
-  0x65, 0x78, 0x29, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x5b, 0x6d, 0x61,
-  0x74, 0x63, 0x68, 0x5b, 0x31, 0x5d, 0x5d, 0x20, 0x3d, 0x20, 0x6d, 0x61,
-  0x74, 0x63, 0x68, 0x5b, 0x32, 0x5d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f,
-  0x20, 0x73, 0x69, 0x6e, 0x63, 0x65, 0x20, 0x77, 0x65, 0x20, 0x6b, 0x6e,
-  0x6f, 0x77, 0x20, 0x74, 0x68, 0x69, 0x73, 0x20, 0x69, 0x73, 0x20, 0x6c,
-  0x6c, 0x61, 0x6d, 0x61, 0x2e, 0x63, 0x70, 0x70, 0x2c, 0x20, 0x6c, 0x65,
-  0x74, 0x27, 0x73, 0x20, 0x6a, 0x75, 0x73, 0x74, 0x20, 0x64, 0x65, 0x63,
-  0x6f, 0x64, 0x65, 0x20, 0x74, 0x68, 0x65, 0x20, 0x6a, 0x73, 0x6f, 0x6e,
-  0x20, 0x69, 0x6e, 0x20, 0x64, 0x61, 0x74, 0x61, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x2e, 0x64, 0x61,
-  0x74, 0x61, 0x20, 0x3d, 0x20, 0x4a, 0x53, 0x4f, 0x4e, 0x2e, 0x70, 0x61,
-  0x72, 0x73, 0x65, 0x28, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x2e, 0x64,
-  0x61, 0x74, 0x61, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x20, 0x2b, 0x3d, 0x20, 0x72,
-  0x65, 0x73, 0x75, 0x6c, 0x74, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x63,
-  0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20, 0x63, 0x61, 0x6c, 0x6c, 0x61, 0x63,
-  0x6b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28,
-  0x63, 0x61, 0x6c, 0x6c, 0x62, 0x61, 0x63, 0x6b, 0x29, 0x20, 0x7b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x74,
-  0x20, 0x3d, 0x20, 0x63, 0x61, 0x6c, 0x6c, 0x62, 0x61, 0x63, 0x6b, 0x28,
-  0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x29, 0x20, 0x21, 0x3d, 0x20, 0x66,
-  0x61, 0x6c, 0x73, 0x65, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20,
-  0x69, 0x66, 0x20, 0x77, 0x65, 0x20, 0x67, 0x6f, 0x74, 0x20, 0x61, 0x20,
-  0x73, 0x74, 0x6f, 0x70, 0x20, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x20, 0x66,
-  0x72, 0x6f, 0x6d, 0x20, 0x73, 0x65, 0x72, 0x76, 0x65, 0x72, 0x2c, 0x20,
-  0x77, 0x65, 0x20, 0x77, 0x69, 0x6c, 0x6c, 0x20, 0x62, 0x72, 0x65, 0x61,
-  0x6b, 0x20, 0x68, 0x65, 0x72, 0x65, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x69, 0x66, 0x20, 0x28, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x2e,
-  0x64, 0x61, 0x74, 0x61, 0x2e, 0x73, 0x74, 0x6f, 0x70, 0x29, 0x20, 0x7b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x62, 0x72, 0x65,
-  0x61, 0x6b, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x7d, 0x20, 0x63, 0x61,
-  0x74, 0x63, 0x68, 0x20, 0x28, 0x65, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x6f, 0x6c, 0x65, 0x2e, 0x65, 0x72,
-  0x72, 0x6f, 0x72, 0x28, 0x22, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x20, 0x65,
-  0x72, 0x72, 0x6f, 0x72, 0x3a, 0x20, 0x22, 0x2c, 0x20, 0x65, 0x29, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x74, 0x68, 0x72, 0x6f, 0x77, 0x20, 0x65,
-  0x3b, 0x0a, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x66, 0x69, 0x6e, 0x61,
-  0x6c, 0x6c, 0x79, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f,
-  0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x61, 0x62, 0x6f,
-  0x72, 0x74, 0x28, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20,
-  0x20, 0x72, 0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x63, 0x6f, 0x6e, 0x74,
-  0x65, 0x6e, 0x74, 0x3b, 0x0a, 0x7d, 0x0a
+  0x3b, 0x0a, 0x0a, 0x6c, 0x65, 0x74, 0x20, 0x67, 0x65, 0x6e, 0x65, 0x72,
+  0x61, 0x74, 0x69, 0x6f, 0x6e, 0x5f, 0x73, 0x65, 0x74, 0x74, 0x69, 0x6e,
+  0x67, 0x73, 0x20, 0x3d, 0x20, 0x6e, 0x75, 0x6c, 0x6c, 0x3b, 0x0a, 0x0a,
+  0x0a, 0x2f, 0x2f, 0x20, 0x43, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x65,
+  0x73, 0x20, 0x74, 0x68, 0x65, 0x20, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74,
+  0x20, 0x61, 0x73, 0x20, 0x61, 0x20, 0x67, 0x65, 0x6e, 0x65, 0x72, 0x61,
+  0x74, 0x6f, 0x72, 0x2e, 0x20, 0x52, 0x65, 0x63, 0x6f, 0x6d, 0x6d, 0x65,
+  0x6e, 0x64, 0x65, 0x64, 0x20, 0x66, 0x6f, 0x72, 0x20, 0x6d, 0x6f, 0x73,
+  0x74, 0x20, 0x75, 0x73, 0x65, 0x20, 0x63, 0x61, 0x73, 0x65, 0x73, 0x2e,
+  0x0a, 0x2f, 0x2f, 0x0a, 0x2f, 0x2f, 0x20, 0x45, 0x78, 0x61, 0x6d, 0x70,
+  0x6c, 0x65, 0x3a, 0x0a, 0x2f, 0x2f, 0x0a, 0x2f, 0x2f, 0x20, 0x20, 0x20,
+  0x20, 0x69, 0x6d, 0x70, 0x6f, 0x72, 0x74, 0x20, 0x7b, 0x20, 0x6c, 0x6c,
+  0x61, 0x6d, 0x61, 0x20, 0x7d, 0x20, 0x66, 0x72, 0x6f, 0x6d, 0x20, 0x27,
+  0x2f, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x69, 0x6f, 0x6e, 0x2e,
+  0x6a, 0x73, 0x27, 0x0a, 0x2f, 0x2f, 0x0a, 0x2f, 0x2f, 0x20, 0x20, 0x20,
+  0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x72, 0x65, 0x71, 0x75, 0x65,
+  0x73, 0x74, 0x20, 0x3d, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x28, 0x22,
+  0x54, 0x65, 0x6c, 0x6c, 0x20, 0x6d, 0x65, 0x20, 0x61, 0x20, 0x6a, 0x6f,
+  0x6b, 0x65, 0x22, 0x2c, 0x20, 0x7b, 0x6e, 0x5f, 0x70, 0x72, 0x65, 0x64,
+  0x69, 0x63, 0x74, 0x3a, 0x20, 0x38, 0x30, 0x30, 0x7d, 0x29, 0x0a, 0x2f,
+  0x2f, 0x20, 0x20, 0x20, 0x20, 0x66, 0x6f, 0x72, 0x20, 0x61, 0x77, 0x61,
+  0x69, 0x74, 0x20, 0x28, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63, 0x68,
+  0x75, 0x6e, 0x6b, 0x20, 0x6f, 0x66, 0x20, 0x72, 0x65, 0x71, 0x75, 0x65,
+  0x73, 0x74, 0x29, 0x20, 0x7b, 0x0a, 0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x64, 0x6f, 0x63, 0x75, 0x6d, 0x65, 0x6e, 0x74, 0x2e, 0x77,
+  0x72, 0x69, 0x74, 0x65, 0x28, 0x63, 0x68, 0x75, 0x6e, 0x6b, 0x2e, 0x64,
+  0x61, 0x74, 0x61, 0x2e, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x29,
+  0x0a, 0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x2f, 0x2f, 0x0a,
+  0x65, 0x78, 0x70, 0x6f, 0x72, 0x74, 0x20, 0x61, 0x73, 0x79, 0x6e, 0x63,
+  0x20, 0x66, 0x75, 0x6e, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x2a, 0x20, 0x6c,
+  0x6c, 0x61, 0x6d, 0x61, 0x28, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x2c,
+  0x20, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x20, 0x3d, 0x20, 0x7b, 0x7d,
+  0x2c, 0x20, 0x63, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x20, 0x3d, 0x20, 0x7b,
+  0x7d, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x6c, 0x65, 0x74, 0x20, 0x63,
+  0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x20, 0x3d, 0x20,
+  0x63, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x2e, 0x63, 0x6f, 0x6e, 0x74, 0x72,
+  0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x69, 0x66,
+  0x20, 0x28, 0x21, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65,
+  0x72, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e,
+  0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x20, 0x3d, 0x20, 0x6e, 0x65,
+  0x77, 0x20, 0x41, 0x62, 0x6f, 0x72, 0x74, 0x43, 0x6f, 0x6e, 0x74, 0x72,
+  0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x28, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x7d,
+  0x0a, 0x0a, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63, 0x6f,
+  0x6d, 0x70, 0x6c, 0x65, 0x74, 0x69, 0x6f, 0x6e, 0x50, 0x61, 0x72, 0x61,
+  0x6d, 0x73, 0x20, 0x3d, 0x20, 0x7b, 0x20, 0x2e, 0x2e, 0x2e, 0x70, 0x61,
+  0x72, 0x61, 0x6d, 0x44, 0x65, 0x66, 0x61, 0x75, 0x6c, 0x74, 0x73, 0x2c,
+  0x20, 0x2e, 0x2e, 0x2e, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2c, 0x20,
+  0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x20, 0x7d, 0x3b, 0x0a, 0x0a, 0x20,
+  0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x72, 0x65, 0x73, 0x70, 0x6f,
+  0x6e, 0x73, 0x65, 0x20, 0x3d, 0x20, 0x61, 0x77, 0x61, 0x69, 0x74, 0x20,
+  0x66, 0x65, 0x74, 0x63, 0x68, 0x28, 0x22, 0x2f, 0x63, 0x6f, 0x6d, 0x70,
+  0x6c, 0x65, 0x74, 0x69, 0x6f, 0x6e, 0x22, 0x2c, 0x20, 0x7b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x6d, 0x65, 0x74, 0x68, 0x6f, 0x64, 0x3a, 0x20, 0x27,
+  0x50, 0x4f, 0x53, 0x54, 0x27, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x62,
+  0x6f, 0x64, 0x79, 0x3a, 0x20, 0x4a, 0x53, 0x4f, 0x4e, 0x2e, 0x73, 0x74,
+  0x72, 0x69, 0x6e, 0x67, 0x69, 0x66, 0x79, 0x28, 0x63, 0x6f, 0x6d, 0x70,
+  0x6c, 0x65, 0x74, 0x69, 0x6f, 0x6e, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73,
+  0x29, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x68, 0x65, 0x61, 0x64, 0x65,
+  0x72, 0x73, 0x3a, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x27, 0x43, 0x6f, 0x6e, 0x6e, 0x65, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x27,
+  0x3a, 0x20, 0x27, 0x6b, 0x65, 0x65, 0x70, 0x2d, 0x61, 0x6c, 0x69, 0x76,
+  0x65, 0x27, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x27, 0x43,
+  0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x2d, 0x54, 0x79, 0x70, 0x65, 0x27,
+  0x3a, 0x20, 0x27, 0x61, 0x70, 0x70, 0x6c, 0x69, 0x63, 0x61, 0x74, 0x69,
+  0x6f, 0x6e, 0x2f, 0x6a, 0x73, 0x6f, 0x6e, 0x27, 0x2c, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x27, 0x41, 0x63, 0x63, 0x65, 0x70, 0x74, 0x27,
+  0x3a, 0x20, 0x27, 0x74, 0x65, 0x78, 0x74, 0x2f, 0x65, 0x76, 0x65, 0x6e,
+  0x74, 0x2d, 0x73, 0x74, 0x72, 0x65, 0x61, 0x6d, 0x27, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x7d, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x73, 0x69, 0x67,
+  0x6e, 0x61, 0x6c, 0x3a, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c,
+  0x6c, 0x65, 0x72, 0x2e, 0x73, 0x69, 0x67, 0x6e, 0x61, 0x6c, 0x2c, 0x0a,
+  0x20, 0x20, 0x7d, 0x29, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x63, 0x6f, 0x6e,
+  0x73, 0x74, 0x20, 0x72, 0x65, 0x61, 0x64, 0x65, 0x72, 0x20, 0x3d, 0x20,
+  0x72, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x73, 0x65, 0x2e, 0x62, 0x6f, 0x64,
+  0x79, 0x2e, 0x67, 0x65, 0x74, 0x52, 0x65, 0x61, 0x64, 0x65, 0x72, 0x28,
+  0x29, 0x3b, 0x0a, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x64,
+  0x65, 0x63, 0x6f, 0x64, 0x65, 0x72, 0x20, 0x3d, 0x20, 0x6e, 0x65, 0x77,
+  0x20, 0x54, 0x65, 0x78, 0x74, 0x44, 0x65, 0x63, 0x6f, 0x64, 0x65, 0x72,
+  0x28, 0x29, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x6c, 0x65, 0x74, 0x20, 0x63,
+  0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x20, 0x3d, 0x20, 0x22, 0x22, 0x3b,
+  0x0a, 0x0a, 0x20, 0x20, 0x74, 0x72, 0x79, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x6c, 0x65, 0x74, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x20, 0x3d,
+  0x20, 0x74, 0x72, 0x75, 0x65, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x77, 0x68, 0x69, 0x6c, 0x65, 0x20, 0x28, 0x63, 0x6f, 0x6e, 0x74, 0x29,
+  0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e,
+  0x73, 0x74, 0x20, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x20, 0x3d, 0x20,
+  0x61, 0x77, 0x61, 0x69, 0x74, 0x20, 0x72, 0x65, 0x61, 0x64, 0x65, 0x72,
+  0x2e, 0x72, 0x65, 0x61, 0x64, 0x28, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x72, 0x65, 0x73, 0x75, 0x6c,
+  0x74, 0x2e, 0x64, 0x6f, 0x6e, 0x65, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x62, 0x72, 0x65, 0x61, 0x6b, 0x3b,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20, 0x73, 0x73, 0x65, 0x20, 0x61,
+  0x6e, 0x73, 0x77, 0x65, 0x72, 0x73, 0x20, 0x69, 0x6e, 0x20, 0x74, 0x68,
+  0x65, 0x20, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x6d, 0x75, 0x6c, 0x74, 0x69,
+  0x70, 0x6c, 0x65, 0x20, 0x6c, 0x69, 0x6e, 0x65, 0x73, 0x20, 0x6f, 0x66,
+  0x3a, 0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x5c, 0x6e, 0x20, 0x77, 0x69,
+  0x74, 0x68, 0x20, 0x64, 0x61, 0x74, 0x61, 0x20, 0x61, 0x6c, 0x77, 0x61,
+  0x79, 0x73, 0x20, 0x70, 0x72, 0x65, 0x73, 0x65, 0x6e, 0x74, 0x20, 0x61,
+  0x73, 0x20, 0x61, 0x20, 0x6b, 0x65, 0x79, 0x2e, 0x20, 0x69, 0x6e, 0x20,
+  0x6f, 0x75, 0x72, 0x20, 0x63, 0x61, 0x73, 0x65, 0x20, 0x77, 0x65, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20, 0x6d, 0x61, 0x69,
+  0x6e, 0x6c, 0x79, 0x20, 0x63, 0x61, 0x72, 0x65, 0x20, 0x61, 0x62, 0x6f,
+  0x75, 0x74, 0x20, 0x74, 0x68, 0x65, 0x20, 0x64, 0x61, 0x74, 0x61, 0x3a,
+  0x20, 0x6b, 0x65, 0x79, 0x20, 0x68, 0x65, 0x72, 0x65, 0x2c, 0x20, 0x77,
+  0x68, 0x69, 0x63, 0x68, 0x20, 0x77, 0x65, 0x20, 0x65, 0x78, 0x70, 0x65,
+  0x63, 0x74, 0x20, 0x61, 0x73, 0x20, 0x6a, 0x73, 0x6f, 0x6e, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x74,
+  0x65, 0x78, 0x74, 0x20, 0x3d, 0x20, 0x64, 0x65, 0x63, 0x6f, 0x64, 0x65,
+  0x72, 0x2e, 0x64, 0x65, 0x63, 0x6f, 0x64, 0x65, 0x28, 0x72, 0x65, 0x73,
+  0x75, 0x6c, 0x74, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x29, 0x3b, 0x0a,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20, 0x70, 0x61,
+  0x72, 0x73, 0x65, 0x20, 0x61, 0x6c, 0x6c, 0x20, 0x73, 0x73, 0x65, 0x20,
+  0x65, 0x76, 0x65, 0x6e, 0x74, 0x73, 0x20, 0x61, 0x6e, 0x64, 0x20, 0x61,
+  0x64, 0x64, 0x20, 0x74, 0x68, 0x65, 0x6d, 0x20, 0x74, 0x6f, 0x20, 0x72,
+  0x65, 0x73, 0x75, 0x6c, 0x74, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x72, 0x65, 0x67, 0x65, 0x78, 0x20,
+  0x3d, 0x20, 0x2f, 0x5e, 0x28, 0x5c, 0x53, 0x2b, 0x29, 0x3a, 0x5c, 0x73,
+  0x28, 0x2e, 0x2a, 0x29, 0x24, 0x2f, 0x67, 0x6d, 0x3b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x66, 0x6f, 0x72, 0x20, 0x28, 0x63, 0x6f, 0x6e,
+  0x73, 0x74, 0x20, 0x6d, 0x61, 0x74, 0x63, 0x68, 0x20, 0x6f, 0x66, 0x20,
+  0x74, 0x65, 0x78, 0x74, 0x2e, 0x6d, 0x61, 0x74, 0x63, 0x68, 0x41, 0x6c,
+  0x6c, 0x28, 0x72, 0x65, 0x67, 0x65, 0x78, 0x29, 0x29, 0x20, 0x7b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x73, 0x75,
+  0x6c, 0x74, 0x5b, 0x6d, 0x61, 0x74, 0x63, 0x68, 0x5b, 0x31, 0x5d, 0x5d,
+  0x20, 0x3d, 0x20, 0x6d, 0x61, 0x74, 0x63, 0x68, 0x5b, 0x32, 0x5d, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20, 0x73, 0x69, 0x6e, 0x63, 0x65, 0x20,
+  0x77, 0x65, 0x20, 0x6b, 0x6e, 0x6f, 0x77, 0x20, 0x74, 0x68, 0x69, 0x73,
+  0x20, 0x69, 0x73, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x2e, 0x63, 0x70,
+  0x70, 0x2c, 0x20, 0x6c, 0x65, 0x74, 0x27, 0x73, 0x20, 0x6a, 0x75, 0x73,
+  0x74, 0x20, 0x64, 0x65, 0x63, 0x6f, 0x64, 0x65, 0x20, 0x74, 0x68, 0x65,
+  0x20, 0x6a, 0x73, 0x6f, 0x6e, 0x20, 0x69, 0x6e, 0x20, 0x64, 0x61, 0x74,
+  0x61, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x73, 0x75,
+  0x6c, 0x74, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x20, 0x3d, 0x20, 0x4a, 0x53,
+  0x4f, 0x4e, 0x2e, 0x70, 0x61, 0x72, 0x73, 0x65, 0x28, 0x72, 0x65, 0x73,
+  0x75, 0x6c, 0x74, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x29, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74,
+  0x20, 0x2b, 0x3d, 0x20, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x2e, 0x64,
+  0x61, 0x74, 0x61, 0x2e, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x3b,
+  0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20, 0x79,
+  0x69, 0x65, 0x6c, 0x64, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x79,
+  0x69, 0x65, 0x6c, 0x64, 0x20, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x3b,
+  0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20, 0x69,
+  0x66, 0x20, 0x77, 0x65, 0x20, 0x67, 0x6f, 0x74, 0x20, 0x61, 0x20, 0x73,
+  0x74, 0x6f, 0x70, 0x20, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x20, 0x66, 0x72,
+  0x6f, 0x6d, 0x20, 0x73, 0x65, 0x72, 0x76, 0x65, 0x72, 0x2c, 0x20, 0x77,
+  0x65, 0x20, 0x77, 0x69, 0x6c, 0x6c, 0x20, 0x62, 0x72, 0x65, 0x61, 0x6b,
+  0x20, 0x68, 0x65, 0x72, 0x65, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x69, 0x66, 0x20, 0x28, 0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x2e, 0x64,
+  0x61, 0x74, 0x61, 0x2e, 0x73, 0x74, 0x6f, 0x70, 0x29, 0x20, 0x7b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28,
+  0x72, 0x65, 0x73, 0x75, 0x6c, 0x74, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x2e,
+  0x67, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x5f, 0x73,
+  0x65, 0x74, 0x74, 0x69, 0x6e, 0x67, 0x73, 0x29, 0x20, 0x7b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x67, 0x65, 0x6e,
+  0x65, 0x72, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x5f, 0x73, 0x65, 0x74, 0x74,
+  0x69, 0x6e, 0x67, 0x73, 0x20, 0x3d, 0x20, 0x72, 0x65, 0x73, 0x75, 0x6c,
+  0x74, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x67, 0x65, 0x6e, 0x65, 0x72,
+  0x61, 0x74, 0x69, 0x6f, 0x6e, 0x5f, 0x73, 0x65, 0x74, 0x74, 0x69, 0x6e,
+  0x67, 0x73, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x62, 0x72,
+  0x65, 0x61, 0x6b, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x7d, 0x20, 0x63,
+  0x61, 0x74, 0x63, 0x68, 0x20, 0x28, 0x65, 0x29, 0x20, 0x7b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x65, 0x2e, 0x6e, 0x61, 0x6d,
+  0x65, 0x20, 0x21, 0x3d, 0x3d, 0x20, 0x27, 0x41, 0x62, 0x6f, 0x72, 0x74,
+  0x45, 0x72, 0x72, 0x6f, 0x72, 0x27, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x6f, 0x6c, 0x65, 0x2e,
+  0x65, 0x72, 0x72, 0x6f, 0x72, 0x28, 0x22, 0x6c, 0x6c, 0x61, 0x6d, 0x61,
+  0x20, 0x65, 0x72, 0x72, 0x6f, 0x72, 0x3a, 0x20, 0x22, 0x2c, 0x20, 0x65,
+  0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x74, 0x68, 0x72, 0x6f, 0x77, 0x20, 0x65, 0x3b, 0x0a, 0x20, 0x20,
+  0x7d, 0x0a, 0x20, 0x20, 0x66, 0x69, 0x6e, 0x61, 0x6c, 0x6c, 0x79, 0x20,
+  0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f,
+  0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x61, 0x62, 0x6f, 0x72, 0x74, 0x28, 0x29,
+  0x3b, 0x0a, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x72, 0x65, 0x74,
+  0x75, 0x72, 0x6e, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x3b,
+  0x0a, 0x7d, 0x0a, 0x0a, 0x2f, 0x2f, 0x20, 0x43, 0x61, 0x6c, 0x6c, 0x20,
+  0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x2c, 0x20, 0x72, 0x65, 0x74, 0x75, 0x72,
+  0x6e, 0x20, 0x61, 0x6e, 0x20, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x20, 0x74,
+  0x61, 0x72, 0x67, 0x65, 0x74, 0x20, 0x74, 0x68, 0x61, 0x74, 0x20, 0x79,
+  0x6f, 0x75, 0x20, 0x63, 0x61, 0x6e, 0x20, 0x73, 0x75, 0x62, 0x63, 0x72,
+  0x69, 0x62, 0x65, 0x20, 0x74, 0x6f, 0x0a, 0x2f, 0x2f, 0x0a, 0x2f, 0x2f,
+  0x20, 0x45, 0x78, 0x61, 0x6d, 0x70, 0x6c, 0x65, 0x3a, 0x0a, 0x2f, 0x2f,
+  0x0a, 0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20, 0x69, 0x6d, 0x70, 0x6f, 0x72,
+  0x74, 0x20, 0x7b, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x45, 0x76, 0x65,
+  0x6e, 0x74, 0x54, 0x61, 0x72, 0x67, 0x65, 0x74, 0x20, 0x7d, 0x20, 0x66,
+  0x72, 0x6f, 0x6d, 0x20, 0x27, 0x2f, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65,
+  0x74, 0x69, 0x6f, 0x6e, 0x2e, 0x6a, 0x73, 0x27, 0x0a, 0x2f, 0x2f, 0x0a,
+  0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20,
+  0x63, 0x6f, 0x6e, 0x6e, 0x20, 0x3d, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61,
+  0x45, 0x76, 0x65, 0x6e, 0x74, 0x54, 0x61, 0x72, 0x67, 0x65, 0x74, 0x28,
+  0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x29, 0x0a, 0x2f, 0x2f, 0x20, 0x20,
+  0x20, 0x20, 0x63, 0x6f, 0x6e, 0x6e, 0x2e, 0x61, 0x64, 0x64, 0x45, 0x76,
+  0x65, 0x6e, 0x74, 0x4c, 0x69, 0x73, 0x74, 0x65, 0x6e, 0x65, 0x72, 0x28,
+  0x22, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x22, 0x2c, 0x20, 0x28,
+  0x63, 0x68, 0x75, 0x6e, 0x6b, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a,
+  0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x64, 0x6f, 0x63, 0x75,
+  0x6d, 0x65, 0x6e, 0x74, 0x2e, 0x77, 0x72, 0x69, 0x74, 0x65, 0x28, 0x63,
+  0x68, 0x75, 0x6e, 0x6b, 0x2e, 0x64, 0x65, 0x74, 0x61, 0x69, 0x6c, 0x2e,
+  0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x29, 0x0a, 0x2f, 0x2f, 0x20,
+  0x20, 0x20, 0x20, 0x7d, 0x29, 0x0a, 0x2f, 0x2f, 0x0a, 0x65, 0x78, 0x70,
+  0x6f, 0x72, 0x74, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x6c, 0x6c,
+  0x61, 0x6d, 0x61, 0x45, 0x76, 0x65, 0x6e, 0x74, 0x54, 0x61, 0x72, 0x67,
+  0x65, 0x74, 0x20, 0x3d, 0x20, 0x28, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74,
+  0x2c, 0x20, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x20, 0x3d, 0x20, 0x7b,
+  0x7d, 0x2c, 0x20, 0x63, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x20, 0x3d, 0x20,
+  0x7b, 0x7d, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x63,
+  0x6f, 0x6e, 0x73, 0x74, 0x20, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x54, 0x61,
+  0x72, 0x67, 0x65, 0x74, 0x20, 0x3d, 0x20, 0x6e, 0x65, 0x77, 0x20, 0x45,
+  0x76, 0x65, 0x6e, 0x74, 0x54, 0x61, 0x72, 0x67, 0x65, 0x74, 0x28, 0x29,
+  0x3b, 0x0a, 0x20, 0x20, 0x28, 0x61, 0x73, 0x79, 0x6e, 0x63, 0x20, 0x28,
+  0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x6c,
+  0x65, 0x74, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x20, 0x3d,
+  0x20, 0x22, 0x22, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x66, 0x6f, 0x72,
+  0x20, 0x61, 0x77, 0x61, 0x69, 0x74, 0x20, 0x28, 0x63, 0x6f, 0x6e, 0x73,
+  0x74, 0x20, 0x63, 0x68, 0x75, 0x6e, 0x6b, 0x20, 0x6f, 0x66, 0x20, 0x6c,
+  0x6c, 0x61, 0x6d, 0x61, 0x28, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x2c,
+  0x20, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2c, 0x20, 0x63, 0x6f, 0x6e,
+  0x66, 0x69, 0x67, 0x29, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x63, 0x68, 0x75, 0x6e, 0x6b, 0x2e,
+  0x64, 0x61, 0x74, 0x61, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x20,
+  0x2b, 0x3d, 0x20, 0x63, 0x68, 0x75, 0x6e, 0x6b, 0x2e, 0x64, 0x61, 0x74,
+  0x61, 0x2e, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x65, 0x76, 0x65, 0x6e, 0x74,
+  0x54, 0x61, 0x72, 0x67, 0x65, 0x74, 0x2e, 0x64, 0x69, 0x73, 0x70, 0x61,
+  0x74, 0x63, 0x68, 0x45, 0x76, 0x65, 0x6e, 0x74, 0x28, 0x6e, 0x65, 0x77,
+  0x20, 0x43, 0x75, 0x73, 0x74, 0x6f, 0x6d, 0x45, 0x76, 0x65, 0x6e, 0x74,
+  0x28, 0x22, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x22, 0x2c, 0x20,
+  0x7b, 0x20, 0x64, 0x65, 0x74, 0x61, 0x69, 0x6c, 0x3a, 0x20, 0x63, 0x68,
+  0x75, 0x6e, 0x6b, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x20, 0x7d, 0x29, 0x29,
+  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x63, 0x68, 0x75, 0x6e,
+  0x6b, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x67, 0x65, 0x6e, 0x65, 0x72,
+  0x61, 0x74, 0x69, 0x6f, 0x6e, 0x5f, 0x73, 0x65, 0x74, 0x74, 0x69, 0x6e,
+  0x67, 0x73, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x54, 0x61, 0x72, 0x67, 0x65,
+  0x74, 0x2e, 0x64, 0x69, 0x73, 0x70, 0x61, 0x74, 0x63, 0x68, 0x45, 0x76,
+  0x65, 0x6e, 0x74, 0x28, 0x6e, 0x65, 0x77, 0x20, 0x43, 0x75, 0x73, 0x74,
+  0x6f, 0x6d, 0x45, 0x76, 0x65, 0x6e, 0x74, 0x28, 0x22, 0x67, 0x65, 0x6e,
+  0x65, 0x72, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x5f, 0x73, 0x65, 0x74, 0x74,
+  0x69, 0x6e, 0x67, 0x73, 0x22, 0x2c, 0x20, 0x7b, 0x20, 0x64, 0x65, 0x74,
+  0x61, 0x69, 0x6c, 0x3a, 0x20, 0x63, 0x68, 0x75, 0x6e, 0x6b, 0x2e, 0x64,
+  0x61, 0x74, 0x61, 0x2e, 0x67, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x69,
+  0x6f, 0x6e, 0x5f, 0x73, 0x65, 0x74, 0x74, 0x69, 0x6e, 0x67, 0x73, 0x20,
+  0x7d, 0x29, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x63,
+  0x68, 0x75, 0x6e, 0x6b, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x74, 0x69,
+  0x6d, 0x69, 0x6e, 0x67, 0x73, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x54, 0x61,
+  0x72, 0x67, 0x65, 0x74, 0x2e, 0x64, 0x69, 0x73, 0x70, 0x61, 0x74, 0x63,
+  0x68, 0x45, 0x76, 0x65, 0x6e, 0x74, 0x28, 0x6e, 0x65, 0x77, 0x20, 0x43,
+  0x75, 0x73, 0x74, 0x6f, 0x6d, 0x45, 0x76, 0x65, 0x6e, 0x74, 0x28, 0x22,
+  0x74, 0x69, 0x6d, 0x69, 0x6e, 0x67, 0x73, 0x22, 0x2c, 0x20, 0x7b, 0x20,
+  0x64, 0x65, 0x74, 0x61, 0x69, 0x6c, 0x3a, 0x20, 0x63, 0x68, 0x75, 0x6e,
+  0x6b, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x74, 0x69, 0x6d, 0x69, 0x6e,
+  0x67, 0x73, 0x20, 0x7d, 0x29, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x54, 0x61, 0x72, 0x67, 0x65,
+  0x74, 0x2e, 0x64, 0x69, 0x73, 0x70, 0x61, 0x74, 0x63, 0x68, 0x45, 0x76,
+  0x65, 0x6e, 0x74, 0x28, 0x6e, 0x65, 0x77, 0x20, 0x43, 0x75, 0x73, 0x74,
+  0x6f, 0x6d, 0x45, 0x76, 0x65, 0x6e, 0x74, 0x28, 0x22, 0x64, 0x6f, 0x6e,
+  0x65, 0x22, 0x2c, 0x20, 0x7b, 0x20, 0x64, 0x65, 0x74, 0x61, 0x69, 0x6c,
+  0x3a, 0x20, 0x7b, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x20,
+  0x7d, 0x20, 0x7d, 0x29, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x7d, 0x29, 0x28,
+  0x29, 0x3b, 0x0a, 0x20, 0x20, 0x72, 0x65, 0x74, 0x75, 0x72, 0x6e, 0x20,
+  0x65, 0x76, 0x65, 0x6e, 0x74, 0x54, 0x61, 0x72, 0x67, 0x65, 0x74, 0x3b,
+  0x0a, 0x7d, 0x0a, 0x0a, 0x2f, 0x2f, 0x20, 0x43, 0x61, 0x6c, 0x6c, 0x20,
+  0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x2c, 0x20, 0x72, 0x65, 0x74, 0x75, 0x72,
+  0x6e, 0x20, 0x61, 0x20, 0x70, 0x72, 0x6f, 0x6d, 0x69, 0x73, 0x65, 0x20,
+  0x74, 0x68, 0x61, 0x74, 0x20, 0x72, 0x65, 0x73, 0x6f, 0x6c, 0x76, 0x65,
+  0x73, 0x20, 0x74, 0x6f, 0x20, 0x74, 0x68, 0x65, 0x20, 0x63, 0x6f, 0x6d,
+  0x70, 0x6c, 0x65, 0x74, 0x65, 0x64, 0x20, 0x74, 0x65, 0x78, 0x74, 0x2e,
+  0x20, 0x54, 0x68, 0x69, 0x73, 0x20, 0x64, 0x6f, 0x65, 0x73, 0x20, 0x6e,
+  0x6f, 0x74, 0x20, 0x73, 0x75, 0x70, 0x70, 0x6f, 0x72, 0x74, 0x20, 0x73,
+  0x74, 0x72, 0x65, 0x61, 0x6d, 0x69, 0x6e, 0x67, 0x0a, 0x2f, 0x2f, 0x0a,
+  0x2f, 0x2f, 0x20, 0x45, 0x78, 0x61, 0x6d, 0x70, 0x6c, 0x65, 0x3a, 0x0a,
+  0x2f, 0x2f, 0x0a, 0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6c, 0x6c,
+  0x61, 0x6d, 0x61, 0x50, 0x72, 0x6f, 0x6d, 0x69, 0x73, 0x65, 0x28, 0x70,
+  0x72, 0x6f, 0x6d, 0x70, 0x74, 0x29, 0x2e, 0x74, 0x68, 0x65, 0x6e, 0x28,
+  0x28, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x29, 0x20, 0x3d, 0x3e,
+  0x20, 0x7b, 0x0a, 0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x64, 0x6f, 0x63, 0x75, 0x6d, 0x65, 0x6e, 0x74, 0x2e, 0x77, 0x72, 0x69,
+  0x74, 0x65, 0x28, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x29, 0x0a,
+  0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x29, 0x0a, 0x2f, 0x2f,
+  0x0a, 0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6f, 0x72, 0x0a, 0x2f,
+  0x2f, 0x0a, 0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e,
+  0x73, 0x74, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x20, 0x3d,
+  0x20, 0x61, 0x77, 0x61, 0x69, 0x74, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61,
+  0x50, 0x72, 0x6f, 0x6d, 0x69, 0x73, 0x65, 0x28, 0x70, 0x72, 0x6f, 0x6d,
+  0x70, 0x74, 0x29, 0x0a, 0x2f, 0x2f, 0x20, 0x20, 0x20, 0x20, 0x20, 0x64,
+  0x6f, 0x63, 0x75, 0x6d, 0x65, 0x6e, 0x74, 0x2e, 0x77, 0x72, 0x69, 0x74,
+  0x65, 0x28, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x29, 0x0a, 0x2f,
+  0x2f, 0x0a, 0x65, 0x78, 0x70, 0x6f, 0x72, 0x74, 0x20, 0x63, 0x6f, 0x6e,
+  0x73, 0x74, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x50, 0x72, 0x6f, 0x6d,
+  0x69, 0x73, 0x65, 0x20, 0x3d, 0x20, 0x28, 0x70, 0x72, 0x6f, 0x6d, 0x70,
+  0x74, 0x2c, 0x20, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x20, 0x3d, 0x20,
+  0x7b, 0x7d, 0x2c, 0x20, 0x63, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x20, 0x3d,
+  0x20, 0x7b, 0x7d, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x72, 0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x6e, 0x65, 0x77, 0x20, 0x50,
+  0x72, 0x6f, 0x6d, 0x69, 0x73, 0x65, 0x28, 0x61, 0x73, 0x79, 0x6e, 0x63,
+  0x20, 0x28, 0x72, 0x65, 0x73, 0x6f, 0x6c, 0x76, 0x65, 0x2c, 0x20, 0x72,
+  0x65, 0x6a, 0x65, 0x63, 0x74, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x6c, 0x65, 0x74, 0x20, 0x63, 0x6f, 0x6e, 0x74,
+  0x65, 0x6e, 0x74, 0x20, 0x3d, 0x20, 0x22, 0x22, 0x3b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x74, 0x72, 0x79, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x66, 0x6f, 0x72, 0x20, 0x61, 0x77, 0x61, 0x69, 0x74, 0x20,
+  0x28, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63, 0x68, 0x75, 0x6e, 0x6b,
+  0x20, 0x6f, 0x66, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x28, 0x70, 0x72,
+  0x6f, 0x6d, 0x70, 0x74, 0x2c, 0x20, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73,
+  0x2c, 0x20, 0x63, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x29, 0x29, 0x20, 0x7b,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e,
+  0x74, 0x65, 0x6e, 0x74, 0x20, 0x2b, 0x3d, 0x20, 0x63, 0x68, 0x75, 0x6e,
+  0x6b, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x63, 0x6f, 0x6e, 0x74, 0x65,
+  0x6e, 0x74, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x73, 0x6f, 0x6c, 0x76,
+  0x65, 0x28, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x29, 0x3b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x7d, 0x20, 0x63, 0x61, 0x74, 0x63, 0x68, 0x20,
+  0x28, 0x65, 0x72, 0x72, 0x6f, 0x72, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x6a, 0x65, 0x63, 0x74, 0x28, 0x65,
+  0x72, 0x72, 0x6f, 0x72, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d,
+  0x0a, 0x20, 0x20, 0x7d, 0x29, 0x3b, 0x0a, 0x7d, 0x3b, 0x0a, 0x0a, 0x2f,
+  0x2a, 0x2a, 0x0a, 0x20, 0x2a, 0x20, 0x28, 0x64, 0x65, 0x70, 0x72, 0x65,
+  0x63, 0x61, 0x74, 0x65, 0x64, 0x29, 0x0a, 0x20, 0x2a, 0x2f, 0x0a, 0x65,
+  0x78, 0x70, 0x6f, 0x72, 0x74, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20,
+  0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x43, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74,
+  0x65, 0x20, 0x3d, 0x20, 0x61, 0x73, 0x79, 0x6e, 0x63, 0x20, 0x28, 0x70,
+  0x61, 0x72, 0x61, 0x6d, 0x73, 0x2c, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x72,
+  0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2c, 0x20, 0x63, 0x61, 0x6c, 0x6c, 0x62,
+  0x61, 0x63, 0x6b, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x66, 0x6f, 0x72, 0x20, 0x61, 0x77, 0x61, 0x69, 0x74, 0x20, 0x28, 0x63,
+  0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63, 0x68, 0x75, 0x6e, 0x6b, 0x20, 0x6f,
+  0x66, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x28, 0x70, 0x61, 0x72, 0x61,
+  0x6d, 0x73, 0x2e, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x2c, 0x20, 0x70,
+  0x61, 0x72, 0x61, 0x6d, 0x73, 0x2c, 0x20, 0x7b, 0x20, 0x63, 0x6f, 0x6e,
+  0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x20, 0x7d, 0x29, 0x29, 0x20,
+  0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x61, 0x6c, 0x6c, 0x62, 0x61,
+  0x63, 0x6b, 0x28, 0x63, 0x68, 0x75, 0x6e, 0x6b, 0x29, 0x3b, 0x0a, 0x20,
+  0x20, 0x7d, 0x0a, 0x7d, 0x0a, 0x0a, 0x2f, 0x2f, 0x20, 0x47, 0x65, 0x74,
+  0x20, 0x74, 0x68, 0x65, 0x20, 0x6d, 0x6f, 0x64, 0x65, 0x6c, 0x20, 0x69,
+  0x6e, 0x66, 0x6f, 0x20, 0x66, 0x72, 0x6f, 0x6d, 0x20, 0x74, 0x68, 0x65,
+  0x20, 0x73, 0x65, 0x72, 0x76, 0x65, 0x72, 0x2e, 0x20, 0x54, 0x68, 0x69,
+  0x73, 0x20, 0x69, 0x73, 0x20, 0x75, 0x73, 0x65, 0x66, 0x75, 0x6c, 0x20,
+  0x66, 0x6f, 0x72, 0x20, 0x67, 0x65, 0x74, 0x74, 0x69, 0x6e, 0x67, 0x20,
+  0x74, 0x68, 0x65, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x78, 0x74, 0x20,
+  0x77, 0x69, 0x6e, 0x64, 0x6f, 0x77, 0x20, 0x61, 0x6e, 0x64, 0x20, 0x73,
+  0x6f, 0x20, 0x6f, 0x6e, 0x2e, 0x0a, 0x65, 0x78, 0x70, 0x6f, 0x72, 0x74,
+  0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61,
+  0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x49, 0x6e, 0x66, 0x6f, 0x20, 0x3d, 0x20,
+  0x61, 0x73, 0x79, 0x6e, 0x63, 0x20, 0x28, 0x29, 0x20, 0x3d, 0x3e, 0x20,
+  0x7b, 0x0a, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x21, 0x67, 0x65, 0x6e,
+  0x65, 0x72, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x5f, 0x73, 0x65, 0x74, 0x74,
+  0x69, 0x6e, 0x67, 0x73, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x67, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x5f, 0x73,
+  0x65, 0x74, 0x74, 0x69, 0x6e, 0x67, 0x73, 0x20, 0x3d, 0x20, 0x61, 0x77,
+  0x61, 0x69, 0x74, 0x20, 0x66, 0x65, 0x74, 0x63, 0x68, 0x28, 0x22, 0x2f,
+  0x6d, 0x6f, 0x64, 0x65, 0x6c, 0x2e, 0x6a, 0x73, 0x6f, 0x6e, 0x22, 0x29,
+  0x2e, 0x74, 0x68, 0x65, 0x6e, 0x28, 0x72, 0x20, 0x3d, 0x3e, 0x20, 0x72,
+  0x2e, 0x6a, 0x73, 0x6f, 0x6e, 0x28, 0x29, 0x29, 0x3b, 0x0a, 0x20, 0x20,
+  0x7d, 0x0a, 0x20, 0x20, 0x72, 0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x67,
+  0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x5f, 0x73, 0x65,
+  0x74, 0x74, 0x69, 0x6e, 0x67, 0x73, 0x3b, 0x0a, 0x7d, 0x0a
 };
-unsigned int completion_js_len = 2275;
+unsigned int completion_js_len = 4462;
diff --git a/examples/server/deps.sh b/examples/server/deps.sh
index cf995162a4de0..1e9fe964b961a 100755
--- a/examples/server/deps.sh
+++ b/examples/server/deps.sh
@@ -4,10 +4,6 @@
 # get the directory of this script file
 DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
 PUBLIC=$DIR/public
-OUTPUT=$DIR/templats.hpp
-
-echo "// Generated file, do not edit" > $OUTPUT
-echo "" > $OUTPUT
 
 echo "download js bundle files"
 curl https://npm.reversehttp.com/@preact/signals-core,@preact/signals,htm/preact,preact,preact/hooks > $PUBLIC/index.js
diff --git a/examples/server/index.html.hpp b/examples/server/index.html.hpp
index 832e9a3bb045f..42707fad967c2 100644
--- a/examples/server/index.html.hpp
+++ b/examples/server/index.html.hpp
@@ -13,138 +13,143 @@ unsigned char index_html[] = {
   0x3c, 0x74, 0x69, 0x74, 0x6c, 0x65, 0x3e, 0x6c, 0x6c, 0x61, 0x6d, 0x61,
   0x2e, 0x63, 0x70, 0x70, 0x20, 0x2d, 0x20, 0x63, 0x68, 0x61, 0x74, 0x3c,
   0x2f, 0x74, 0x69, 0x74, 0x6c, 0x65, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x3c,
-  0x73, 0x74, 0x79, 0x6c, 0x65, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x62, 0x6f, 0x64, 0x79, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x62, 0x61, 0x63, 0x6b, 0x67, 0x72, 0x6f, 0x75, 0x6e, 0x64, 0x2d,
-  0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x3a, 0x20, 0x23, 0x66, 0x66, 0x66, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6c, 0x6f, 0x72,
-  0x3a, 0x20, 0x23, 0x30, 0x30, 0x30, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x66, 0x6f, 0x6e, 0x74, 0x2d, 0x66, 0x61, 0x6d, 0x69, 0x6c,
-  0x79, 0x3a, 0x20, 0x73, 0x79, 0x73, 0x74, 0x65, 0x6d, 0x2d, 0x75, 0x69,
-  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x66, 0x6f, 0x6e, 0x74,
-  0x2d, 0x73, 0x69, 0x7a, 0x65, 0x3a, 0x20, 0x39, 0x30, 0x25, 0x3b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x23,
-  0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x20, 0x7b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x72, 0x67, 0x69, 0x6e,
-  0x3a, 0x20, 0x30, 0x65, 0x6d, 0x20, 0x61, 0x75, 0x74, 0x6f, 0x3b, 0x0a,
+  0x73, 0x74, 0x79, 0x6c, 0x65, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x62,
+  0x6f, 0x64, 0x79, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x62, 0x61, 0x63, 0x6b, 0x67, 0x72, 0x6f, 0x75, 0x6e, 0x64, 0x2d, 0x63,
+  0x6f, 0x6c, 0x6f, 0x72, 0x3a, 0x20, 0x23, 0x66, 0x66, 0x66, 0x3b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x3a,
+  0x20, 0x23, 0x30, 0x30, 0x30, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x66, 0x6f, 0x6e, 0x74, 0x2d, 0x66, 0x61, 0x6d, 0x69, 0x6c, 0x79,
+  0x3a, 0x20, 0x73, 0x79, 0x73, 0x74, 0x65, 0x6d, 0x2d, 0x75, 0x69, 0x3b,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x66, 0x6f, 0x6e, 0x74, 0x2d,
+  0x73, 0x69, 0x7a, 0x65, 0x3a, 0x20, 0x39, 0x30, 0x25, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x23, 0x63,
+  0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x20, 0x7b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x72, 0x67, 0x69, 0x6e, 0x3a,
+  0x20, 0x30, 0x65, 0x6d, 0x20, 0x61, 0x75, 0x74, 0x6f, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x64, 0x69, 0x73, 0x70, 0x6c, 0x61, 0x79,
+  0x3a, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x2d, 0x64, 0x69, 0x72, 0x65, 0x63,
+  0x74, 0x69, 0x6f, 0x6e, 0x3a, 0x20, 0x63, 0x6f, 0x6c, 0x75, 0x6d, 0x6e,
+  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6a, 0x75, 0x73, 0x74,
+  0x69, 0x66, 0x79, 0x2d, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x3a,
+  0x20, 0x73, 0x70, 0x61, 0x63, 0x65, 0x2d, 0x62, 0x65, 0x74, 0x77, 0x65,
+  0x65, 0x6e, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x68, 0x65,
+  0x69, 0x67, 0x68, 0x74, 0x3a, 0x20, 0x31, 0x30, 0x30, 0x25, 0x3b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x6d,
+  0x61, 0x69, 0x6e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x6d, 0x61, 0x72, 0x67, 0x69, 0x6e, 0x3a, 0x20, 0x33, 0x70, 0x78, 0x3b,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x64, 0x69, 0x73, 0x70, 0x6c,
+  0x61, 0x79, 0x3a, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x3b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x2d, 0x64, 0x69, 0x72,
+  0x65, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x3a, 0x20, 0x63, 0x6f, 0x6c, 0x75,
+  0x6d, 0x6e, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6a, 0x75,
+  0x73, 0x74, 0x69, 0x66, 0x79, 0x2d, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e,
+  0x74, 0x3a, 0x20, 0x73, 0x70, 0x61, 0x63, 0x65, 0x2d, 0x62, 0x65, 0x74,
+  0x77, 0x65, 0x65, 0x6e, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x67, 0x61, 0x70, 0x3a, 0x20, 0x31, 0x65, 0x6d, 0x3b, 0x0a, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x2d, 0x67, 0x72,
+  0x6f, 0x77, 0x3a, 0x20, 0x31, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x6f, 0x76, 0x65, 0x72, 0x66, 0x6c, 0x6f, 0x77, 0x2d, 0x79, 0x3a,
+  0x20, 0x61, 0x75, 0x74, 0x6f, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x62, 0x6f, 0x72, 0x64, 0x65, 0x72, 0x3a, 0x20, 0x31, 0x70,
+  0x78, 0x20, 0x73, 0x6f, 0x6c, 0x69, 0x64, 0x20, 0x23, 0x63, 0x63, 0x63,
+  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x62, 0x6f, 0x72, 0x64,
+  0x65, 0x72, 0x2d, 0x72, 0x61, 0x64, 0x69, 0x75, 0x73, 0x3a, 0x20, 0x35,
+  0x70, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x70, 0x61,
+  0x64, 0x64, 0x69, 0x6e, 0x67, 0x3a, 0x20, 0x30, 0x2e, 0x35, 0x65, 0x6d,
+  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x62, 0x6f, 0x64, 0x79, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x6d, 0x61, 0x78, 0x2d, 0x77, 0x69, 0x64, 0x74, 0x68, 0x3a,
+  0x20, 0x36, 0x30, 0x30, 0x70, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x6d, 0x69, 0x6e, 0x2d, 0x77, 0x69, 0x64, 0x74, 0x68, 0x3a,
+  0x20, 0x33, 0x30, 0x30, 0x70, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x6c, 0x69, 0x6e, 0x65, 0x2d, 0x68, 0x65, 0x69, 0x67, 0x68,
+  0x74, 0x3a, 0x20, 0x31, 0x2e, 0x32, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x6d, 0x61, 0x72, 0x67, 0x69, 0x6e, 0x3a, 0x20, 0x30, 0x20,
+  0x61, 0x75, 0x74, 0x6f, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x70, 0x61, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x3a, 0x20, 0x30, 0x20, 0x30,
+  0x2e, 0x35, 0x65, 0x6d, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x70, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x6f, 0x76, 0x65, 0x72, 0x66, 0x6c, 0x6f, 0x77, 0x2d,
+  0x77, 0x72, 0x61, 0x70, 0x3a, 0x20, 0x62, 0x72, 0x65, 0x61, 0x6b, 0x2d,
+  0x77, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x77, 0x6f, 0x72, 0x64, 0x2d, 0x77, 0x72, 0x61, 0x70, 0x3a, 0x20, 0x62,
+  0x72, 0x65, 0x61, 0x6b, 0x2d, 0x77, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x68, 0x79, 0x70, 0x68, 0x65, 0x6e, 0x73,
+  0x3a, 0x20, 0x61, 0x75, 0x74, 0x6f, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x6d, 0x61, 0x72, 0x67, 0x69, 0x6e, 0x2d, 0x74, 0x6f, 0x70,
+  0x3a, 0x20, 0x30, 0x2e, 0x35, 0x65, 0x6d, 0x3b, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x6d, 0x61, 0x72, 0x67, 0x69, 0x6e, 0x2d, 0x62, 0x6f,
+  0x74, 0x74, 0x6f, 0x6d, 0x3a, 0x20, 0x30, 0x2e, 0x35, 0x65, 0x6d, 0x3b,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x23, 0x77, 0x72, 0x69, 0x74, 0x65, 0x20, 0x66, 0x6f, 0x72, 0x6d, 0x20,
+  0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x72, 0x67,
+  0x69, 0x6e, 0x3a, 0x20, 0x31, 0x65, 0x6d, 0x20, 0x30, 0x20, 0x30, 0x20,
+  0x30, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x64, 0x69, 0x73,
+  0x70, 0x6c, 0x61, 0x79, 0x3a, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x3b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x2d, 0x64,
+  0x69, 0x72, 0x65, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x3a, 0x20, 0x63, 0x6f,
+  0x6c, 0x75, 0x6d, 0x6e, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x67, 0x61, 0x70, 0x3a, 0x20, 0x30, 0x2e, 0x35, 0x65, 0x6d, 0x3b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x61, 0x6c, 0x69, 0x67, 0x6e, 0x2d,
+  0x69, 0x74, 0x65, 0x6d, 0x73, 0x3a, 0x20, 0x73, 0x74, 0x72, 0x65, 0x74,
+  0x63, 0x68, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x2e, 0x72, 0x69, 0x67, 0x68, 0x74, 0x20, 0x7b, 0x0a,
   0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x64, 0x69, 0x73, 0x70, 0x6c, 0x61,
   0x79, 0x3a, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20,
   0x20, 0x20, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x2d, 0x64, 0x69, 0x72, 0x65,
-  0x63, 0x74, 0x69, 0x6f, 0x6e, 0x3a, 0x20, 0x63, 0x6f, 0x6c, 0x75, 0x6d,
-  0x6e, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6a, 0x75, 0x73,
-  0x74, 0x69, 0x66, 0x79, 0x2d, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74,
-  0x3a, 0x20, 0x73, 0x70, 0x61, 0x63, 0x65, 0x2d, 0x62, 0x65, 0x74, 0x77,
-  0x65, 0x65, 0x6e, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x68,
-  0x65, 0x69, 0x67, 0x68, 0x74, 0x3a, 0x20, 0x31, 0x30, 0x30, 0x25, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x68, 0x65, 0x61, 0x64, 0x65, 0x72, 0x2c, 0x20, 0x66, 0x6f, 0x6f, 0x74,
-  0x65, 0x72, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74,
-  0x65, 0x78, 0x74, 0x2d, 0x61, 0x6c, 0x69, 0x67, 0x6e, 0x3a, 0x20, 0x63,
-  0x65, 0x6e, 0x74, 0x65, 0x72, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d,
-  0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x69, 0x6e, 0x20, 0x7b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x72, 0x67, 0x69,
-  0x6e, 0x3a, 0x20, 0x33, 0x70, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x64, 0x69, 0x73, 0x70, 0x6c, 0x61, 0x79, 0x3a, 0x20, 0x66,
-  0x6c, 0x65, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x66,
-  0x6c, 0x65, 0x78, 0x2d, 0x64, 0x69, 0x72, 0x65, 0x63, 0x74, 0x69, 0x6f,
-  0x6e, 0x3a, 0x20, 0x63, 0x6f, 0x6c, 0x75, 0x6d, 0x6e, 0x3b, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x6a, 0x75, 0x73, 0x74, 0x69, 0x66, 0x79,
-  0x2d, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x3a, 0x20, 0x73, 0x70,
-  0x61, 0x63, 0x65, 0x2d, 0x62, 0x65, 0x74, 0x77, 0x65, 0x65, 0x6e, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x67, 0x61, 0x70, 0x3a, 0x20,
-  0x31, 0x65, 0x6d, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x66, 0x6c, 0x65, 0x78, 0x2d, 0x67, 0x72, 0x6f, 0x77, 0x3a, 0x20, 0x31,
-  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6f, 0x76, 0x65, 0x72,
-  0x66, 0x6c, 0x6f, 0x77, 0x2d, 0x79, 0x3a, 0x20, 0x61, 0x75, 0x74, 0x6f,
-  0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x62, 0x6f, 0x72,
-  0x64, 0x65, 0x72, 0x3a, 0x20, 0x31, 0x70, 0x78, 0x20, 0x73, 0x6f, 0x6c,
-  0x69, 0x64, 0x20, 0x23, 0x63, 0x63, 0x63, 0x3b, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x62, 0x6f, 0x72, 0x64, 0x65, 0x72, 0x2d, 0x72, 0x61,
-  0x64, 0x69, 0x75, 0x73, 0x3a, 0x20, 0x35, 0x70, 0x78, 0x3b, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x70, 0x61, 0x64, 0x64, 0x69, 0x6e, 0x67,
-  0x3a, 0x20, 0x30, 0x2e, 0x35, 0x65, 0x6d, 0x3b, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x62, 0x6f, 0x64, 0x79,
-  0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x78,
-  0x2d, 0x77, 0x69, 0x64, 0x74, 0x68, 0x3a, 0x20, 0x36, 0x30, 0x30, 0x70,
-  0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x69, 0x6e,
-  0x2d, 0x77, 0x69, 0x64, 0x74, 0x68, 0x3a, 0x20, 0x33, 0x30, 0x30, 0x70,
-  0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6c, 0x69, 0x6e,
-  0x65, 0x2d, 0x68, 0x65, 0x69, 0x67, 0x68, 0x74, 0x3a, 0x20, 0x31, 0x2e,
-  0x32, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x72,
-  0x67, 0x69, 0x6e, 0x3a, 0x20, 0x30, 0x20, 0x61, 0x75, 0x74, 0x6f, 0x3b,
+  0x63, 0x74, 0x69, 0x6f, 0x6e, 0x3a, 0x20, 0x72, 0x6f, 0x77, 0x3b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x67, 0x61, 0x70, 0x3a, 0x20, 0x30,
+  0x2e, 0x35, 0x65, 0x6d, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x6a, 0x75, 0x73, 0x74, 0x69, 0x66, 0x79, 0x2d, 0x63, 0x6f, 0x6e, 0x74,
+  0x65, 0x6e, 0x74, 0x3a, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x2d, 0x65, 0x6e,
+  0x64, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x66, 0x69, 0x65, 0x6c, 0x64, 0x73, 0x65, 0x74, 0x20, 0x7b,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x62, 0x6f, 0x72, 0x64, 0x65,
+  0x72, 0x3a, 0x20, 0x6e, 0x6f, 0x6e, 0x65, 0x3b, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x70, 0x61, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x3a, 0x20,
+  0x30, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x72,
+  0x67, 0x69, 0x6e, 0x3a, 0x20, 0x30, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x74, 0x65, 0x78, 0x74, 0x61,
+  0x72, 0x65, 0x61, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x70, 0x61, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x3a, 0x20, 0x35, 0x70, 0x78,
+  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x66, 0x6c, 0x65, 0x78,
+  0x2d, 0x67, 0x72, 0x6f, 0x77, 0x3a, 0x20, 0x31, 0x3b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x77, 0x69, 0x64, 0x74, 0x68, 0x3a, 0x20, 0x31,
+  0x30, 0x30, 0x25, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x70, 0x72, 0x65, 0x20, 0x63, 0x6f, 0x64, 0x65,
+  0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x64, 0x69, 0x73,
+  0x70, 0x6c, 0x61, 0x79, 0x3a, 0x20, 0x62, 0x6c, 0x6f, 0x63, 0x6b, 0x3b,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x62, 0x61, 0x63, 0x6b, 0x67,
+  0x72, 0x6f, 0x75, 0x6e, 0x64, 0x2d, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x3a,
+  0x20, 0x23, 0x32, 0x32, 0x32, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x3a, 0x20, 0x23, 0x64, 0x64, 0x64,
+  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x63, 0x6f, 0x64, 0x65, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x66, 0x6f, 0x6e, 0x74, 0x2d, 0x66, 0x61, 0x6d, 0x69, 0x6c, 0x79,
+  0x3a, 0x20, 0x6d, 0x6f, 0x6e, 0x6f, 0x73, 0x70, 0x61, 0x63, 0x65, 0x3b,
   0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x70, 0x61, 0x64, 0x64, 0x69,
-  0x6e, 0x67, 0x3a, 0x20, 0x30, 0x20, 0x30, 0x2e, 0x35, 0x65, 0x6d, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x70, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6f, 0x76,
-  0x65, 0x72, 0x66, 0x6c, 0x6f, 0x77, 0x2d, 0x77, 0x72, 0x61, 0x70, 0x3a,
-  0x20, 0x62, 0x72, 0x65, 0x61, 0x6b, 0x2d, 0x77, 0x6f, 0x72, 0x64, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x77, 0x6f, 0x72, 0x64, 0x2d,
-  0x77, 0x72, 0x61, 0x70, 0x3a, 0x20, 0x62, 0x72, 0x65, 0x61, 0x6b, 0x2d,
-  0x77, 0x6f, 0x72, 0x64, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x68, 0x79, 0x70, 0x68, 0x65, 0x6e, 0x73, 0x3a, 0x20, 0x61, 0x75, 0x74,
-  0x6f, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x72,
-  0x67, 0x69, 0x6e, 0x2d, 0x74, 0x6f, 0x70, 0x3a, 0x20, 0x30, 0x2e, 0x35,
-  0x65, 0x6d, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61,
-  0x72, 0x67, 0x69, 0x6e, 0x2d, 0x62, 0x6f, 0x74, 0x74, 0x6f, 0x6d, 0x3a,
-  0x20, 0x30, 0x2e, 0x35, 0x65, 0x6d, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x23, 0x77, 0x72, 0x69, 0x74,
-  0x65, 0x20, 0x66, 0x6f, 0x72, 0x6d, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x6d, 0x61, 0x72, 0x67, 0x69, 0x6e, 0x3a, 0x20, 0x31,
-  0x65, 0x6d, 0x20, 0x30, 0x20, 0x30, 0x20, 0x30, 0x3b, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x64, 0x69, 0x73, 0x70, 0x6c, 0x61, 0x79, 0x3a,
-  0x20, 0x66, 0x6c, 0x65, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x66, 0x6c, 0x65, 0x78, 0x2d, 0x64, 0x69, 0x72, 0x65, 0x63, 0x74,
-  0x69, 0x6f, 0x6e, 0x3a, 0x20, 0x63, 0x6f, 0x6c, 0x75, 0x6d, 0x6e, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x67, 0x61, 0x70, 0x3a, 0x20,
-  0x30, 0x2e, 0x35, 0x65, 0x6d, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x61, 0x6c, 0x69, 0x67, 0x6e, 0x2d, 0x69, 0x74, 0x65, 0x6d, 0x73,
-  0x3a, 0x20, 0x73, 0x74, 0x72, 0x65, 0x74, 0x63, 0x68, 0x3b, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x2e, 0x72,
-  0x69, 0x67, 0x68, 0x74, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x64, 0x69, 0x73, 0x70, 0x6c, 0x61, 0x79, 0x3a, 0x20, 0x66, 0x6c,
-  0x65, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x66, 0x6c,
-  0x65, 0x78, 0x2d, 0x64, 0x69, 0x72, 0x65, 0x63, 0x74, 0x69, 0x6f, 0x6e,
-  0x3a, 0x20, 0x72, 0x6f, 0x77, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x67, 0x61, 0x70, 0x3a, 0x20, 0x30, 0x2e, 0x35, 0x65, 0x6d, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6a, 0x75, 0x73, 0x74, 0x69,
-  0x66, 0x79, 0x2d, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x3a, 0x20,
-  0x66, 0x6c, 0x65, 0x78, 0x2d, 0x65, 0x6e, 0x64, 0x3b, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x66, 0x69, 0x65,
-  0x6c, 0x64, 0x73, 0x65, 0x74, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x62, 0x6f, 0x72, 0x64, 0x65, 0x72, 0x3a, 0x20, 0x6e, 0x6f,
-  0x6e, 0x65, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x70, 0x61,
-  0x64, 0x64, 0x69, 0x6e, 0x67, 0x3a, 0x20, 0x30, 0x3b, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x72, 0x67, 0x69, 0x6e, 0x3a, 0x20,
-  0x30, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x74, 0x65, 0x78, 0x74, 0x61, 0x72, 0x65, 0x61, 0x20, 0x7b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x70, 0x61, 0x64, 0x64, 0x69,
-  0x6e, 0x67, 0x3a, 0x20, 0x35, 0x70, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x66, 0x6c, 0x65, 0x78, 0x2d, 0x67, 0x72, 0x6f, 0x77,
-  0x3a, 0x20, 0x31, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x77,
-  0x69, 0x64, 0x74, 0x68, 0x3a, 0x20, 0x31, 0x30, 0x30, 0x25, 0x3b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x70,
-  0x72, 0x65, 0x20, 0x63, 0x6f, 0x64, 0x65, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x6e, 0x67, 0x3a, 0x20, 0x30, 0x2e, 0x31, 0x65, 0x6d, 0x20, 0x30, 0x2e,
+  0x33, 0x65, 0x6d, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x62,
+  0x6f, 0x72, 0x64, 0x65, 0x72, 0x2d, 0x72, 0x61, 0x64, 0x69, 0x75, 0x73,
+  0x3a, 0x20, 0x33, 0x70, 0x78, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d,
+  0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x66, 0x69, 0x65, 0x6c, 0x64, 0x73,
+  0x65, 0x74, 0x20, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x20, 0x7b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x61, 0x72, 0x67, 0x69, 0x6e, 0x3a,
+  0x20, 0x30, 0x2e, 0x35, 0x65, 0x6d, 0x20, 0x30, 0x3b, 0x0a, 0x20, 0x20,
   0x20, 0x20, 0x20, 0x20, 0x64, 0x69, 0x73, 0x70, 0x6c, 0x61, 0x79, 0x3a,
   0x20, 0x62, 0x6c, 0x6f, 0x63, 0x6b, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x62, 0x61, 0x63, 0x6b, 0x67, 0x72, 0x6f, 0x75, 0x6e, 0x64,
-  0x2d, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x3a, 0x20, 0x23, 0x32, 0x32, 0x32,
-  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6c, 0x6f,
-  0x72, 0x3a, 0x20, 0x23, 0x64, 0x64, 0x64, 0x3b, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x64, 0x65, 0x20,
-  0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x66, 0x6f, 0x6e, 0x74,
-  0x2d, 0x66, 0x61, 0x6d, 0x69, 0x6c, 0x79, 0x3a, 0x20, 0x6d, 0x6f, 0x6e,
-  0x6f, 0x73, 0x70, 0x61, 0x63, 0x65, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x70, 0x61, 0x64, 0x64, 0x69, 0x6e, 0x67, 0x3a, 0x20, 0x30,
-  0x2e, 0x31, 0x65, 0x6d, 0x20, 0x30, 0x2e, 0x33, 0x65, 0x6d, 0x3b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x62, 0x6f, 0x72, 0x64, 0x65, 0x72,
-  0x2d, 0x72, 0x61, 0x64, 0x69, 0x75, 0x73, 0x3a, 0x20, 0x33, 0x70, 0x78,
+  0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x68, 0x65, 0x61, 0x64, 0x65,
+  0x72, 0x2c, 0x20, 0x66, 0x6f, 0x6f, 0x74, 0x65, 0x72, 0x20, 0x7b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74, 0x65, 0x78, 0x74, 0x2d, 0x61,
+  0x6c, 0x69, 0x67, 0x6e, 0x3a, 0x20, 0x63, 0x65, 0x6e, 0x74, 0x65, 0x72,
   0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x66, 0x69, 0x65, 0x6c, 0x64, 0x73, 0x65, 0x74, 0x20, 0x6c, 0x61,
-  0x62, 0x65, 0x6c, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x6d, 0x61, 0x72, 0x67, 0x69, 0x6e, 0x3a, 0x20, 0x30, 0x2e, 0x35, 0x65,
-  0x6d, 0x20, 0x30, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x64,
-  0x69, 0x73, 0x70, 0x6c, 0x61, 0x79, 0x3a, 0x20, 0x62, 0x6c, 0x6f, 0x63,
-  0x6b, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x3c,
+  0x20, 0x66, 0x6f, 0x6f, 0x74, 0x65, 0x72, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x66, 0x6f, 0x6e, 0x74, 0x2d, 0x73, 0x69, 0x7a,
+  0x65, 0x3a, 0x20, 0x38, 0x30, 0x25, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x63, 0x6f, 0x6c, 0x6f, 0x72, 0x3a, 0x20, 0x23, 0x38, 0x38,
+  0x38, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x3c,
   0x2f, 0x73, 0x74, 0x79, 0x6c, 0x65, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x3c,
   0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x20, 0x74, 0x79, 0x70, 0x65, 0x3d,
   0x22, 0x6d, 0x6f, 0x64, 0x75, 0x6c, 0x65, 0x22, 0x3e, 0x0a, 0x20, 0x20,
@@ -159,440 +164,432 @@ unsigned char index_html[] = {
   0x72, 0x6f, 0x6d, 0x20, 0x27, 0x2f, 0x69, 0x6e, 0x64, 0x65, 0x78, 0x2e,
   0x6a, 0x73, 0x27, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x69, 0x6d,
   0x70, 0x6f, 0x72, 0x74, 0x20, 0x7b, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61,
-  0x43, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x65, 0x20, 0x7d, 0x20, 0x66,
-  0x72, 0x6f, 0x6d, 0x20, 0x27, 0x2f, 0x63, 0x6f, 0x6d, 0x70, 0x6c, 0x65,
-  0x74, 0x69, 0x6f, 0x6e, 0x2e, 0x6a, 0x73, 0x27, 0x3b, 0x0a, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x73, 0x65, 0x73,
-  0x73, 0x69, 0x6f, 0x6e, 0x20, 0x3d, 0x20, 0x73, 0x69, 0x67, 0x6e, 0x61,
-  0x6c, 0x28, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x70, 0x72,
-  0x6f, 0x6d, 0x70, 0x74, 0x3a, 0x20, 0x22, 0x54, 0x68, 0x69, 0x73, 0x20,
-  0x69, 0x73, 0x20, 0x61, 0x20, 0x63, 0x6f, 0x6e, 0x76, 0x65, 0x72, 0x73,
-  0x61, 0x74, 0x69, 0x6f, 0x6e, 0x20, 0x62, 0x65, 0x74, 0x77, 0x65, 0x65,
-  0x6e, 0x20, 0x75, 0x73, 0x65, 0x72, 0x20, 0x61, 0x6e, 0x64, 0x20, 0x6c,
-  0x6c, 0x61, 0x6d, 0x61, 0x2c, 0x20, 0x61, 0x20, 0x66, 0x72, 0x69, 0x65,
-  0x6e, 0x64, 0x6c, 0x79, 0x20, 0x63, 0x68, 0x61, 0x74, 0x62, 0x6f, 0x74,
-  0x2e, 0x20, 0x72, 0x65, 0x73, 0x70, 0x6f, 0x6e, 0x64, 0x20, 0x69, 0x6e,
-  0x20, 0x6d, 0x61, 0x72, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x2e, 0x22, 0x2c,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c,
-  0x61, 0x74, 0x65, 0x3a, 0x20, 0x22, 0x7b, 0x7b, 0x70, 0x72, 0x6f, 0x6d,
-  0x70, 0x74, 0x7d, 0x7d, 0x5c, 0x6e, 0x5c, 0x6e, 0x7b, 0x7b, 0x68, 0x69,
-  0x73, 0x74, 0x6f, 0x72, 0x79, 0x7d, 0x7d, 0x5c, 0x6e, 0x7b, 0x7b, 0x63,
-  0x68, 0x61, 0x72, 0x7d, 0x7d, 0x3a, 0x22, 0x2c, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x68, 0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x54, 0x65,
-  0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x3a, 0x20, 0x22, 0x7b, 0x7b, 0x6e,
-  0x61, 0x6d, 0x65, 0x7d, 0x7d, 0x3a, 0x20, 0x7b, 0x7b, 0x6d, 0x65, 0x73,
-  0x73, 0x61, 0x67, 0x65, 0x7d, 0x7d, 0x22, 0x2c, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70,
-  0x74, 0x3a, 0x20, 0x5b, 0x5d, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x74, 0x79, 0x70, 0x65, 0x3a, 0x20, 0x22, 0x63, 0x68, 0x61, 0x74,
-  0x22, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x68, 0x61,
-  0x72, 0x3a, 0x20, 0x22, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x22, 0x2c, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x75, 0x73, 0x65, 0x72, 0x3a, 0x20,
-  0x22, 0x55, 0x73, 0x65, 0x72, 0x22, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x7d, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73,
-  0x74, 0x20, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74,
-  0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x20, 0x3d, 0x20, 0x28, 0x74, 0x72,
-  0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x29, 0x20, 0x3d, 0x3e,
-  0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x73, 0x65, 0x73,
-  0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d,
-  0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2e,
-  0x2e, 0x2e, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61,
-  0x6c, 0x75, 0x65, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74,
-  0x20, 0x63, 0x68, 0x61, 0x74, 0x53, 0x74, 0x61, 0x72, 0x74, 0x65, 0x64,
-  0x20, 0x3d, 0x20, 0x63, 0x6f, 0x6d, 0x70, 0x75, 0x74, 0x65, 0x64, 0x28,
-  0x28, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f,
-  0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74, 0x72, 0x61, 0x6e,
-  0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x2e, 0x6c, 0x65, 0x6e, 0x67, 0x74,
-  0x68, 0x20, 0x3e, 0x20, 0x30, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73,
-  0x20, 0x3d, 0x20, 0x73, 0x69, 0x67, 0x6e, 0x61, 0x6c, 0x28, 0x7b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6e, 0x5f, 0x70, 0x72, 0x65, 0x64,
-  0x69, 0x63, 0x74, 0x3a, 0x20, 0x34, 0x30, 0x30, 0x2c, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x74, 0x65, 0x6d, 0x70, 0x65, 0x72, 0x61, 0x74,
-  0x75, 0x72, 0x65, 0x3a, 0x20, 0x30, 0x2e, 0x37, 0x2c, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x6c,
-  0x61, 0x73, 0x74, 0x5f, 0x6e, 0x3a, 0x20, 0x32, 0x35, 0x36, 0x2c, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74,
-  0x5f, 0x70, 0x65, 0x6e, 0x61, 0x6c, 0x74, 0x79, 0x3a, 0x20, 0x31, 0x2e,
-  0x31, 0x38, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74, 0x6f,
-  0x70, 0x5f, 0x6b, 0x3a, 0x20, 0x34, 0x30, 0x2c, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x74, 0x6f, 0x70, 0x5f, 0x70, 0x3a, 0x20, 0x30, 0x2e,
-  0x35, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x29, 0x0a, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63, 0x6f, 0x6e,
-  0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x20, 0x3d, 0x20, 0x73, 0x69,
-  0x67, 0x6e, 0x61, 0x6c, 0x28, 0x6e, 0x75, 0x6c, 0x6c, 0x29, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x67, 0x65, 0x6e,
-  0x65, 0x72, 0x61, 0x74, 0x69, 0x6e, 0x67, 0x20, 0x3d, 0x20, 0x63, 0x6f,
-  0x6d, 0x70, 0x75, 0x74, 0x65, 0x64, 0x28, 0x28, 0x29, 0x20, 0x3d, 0x3e,
-  0x20, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2e,
-  0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d, 0x3d, 0x20, 0x6e, 0x75, 0x6c,
-  0x6c, 0x20, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20,
-  0x73, 0x69, 0x6d, 0x70, 0x6c, 0x65, 0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c,
-  0x61, 0x74, 0x65, 0x20, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x74, 0x65,
-  0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x20, 0x3d, 0x20, 0x28, 0x73, 0x74,
-  0x72, 0x2c, 0x20, 0x65, 0x78, 0x74, 0x72, 0x61, 0x53, 0x65, 0x74, 0x74,
-  0x69, 0x6e, 0x67, 0x73, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x6c, 0x65, 0x74, 0x20, 0x73, 0x65, 0x74,
-  0x74, 0x69, 0x6e, 0x67, 0x73, 0x20, 0x3d, 0x20, 0x73, 0x65, 0x73, 0x73,
-  0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3b, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x65, 0x78, 0x74,
-  0x72, 0x61, 0x53, 0x65, 0x74, 0x74, 0x69, 0x6e, 0x67, 0x73, 0x29, 0x20,
-  0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x73, 0x65,
-  0x74, 0x74, 0x69, 0x6e, 0x67, 0x73, 0x20, 0x3d, 0x20, 0x7b, 0x20, 0x2e,
-  0x2e, 0x2e, 0x73, 0x65, 0x74, 0x74, 0x69, 0x6e, 0x67, 0x73, 0x2c, 0x20,
-  0x2e, 0x2e, 0x2e, 0x65, 0x78, 0x74, 0x72, 0x61, 0x53, 0x65, 0x74, 0x74,
-  0x69, 0x6e, 0x67, 0x73, 0x20, 0x7d, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65,
-  0x74, 0x75, 0x72, 0x6e, 0x20, 0x53, 0x74, 0x72, 0x69, 0x6e, 0x67, 0x28,
-  0x73, 0x74, 0x72, 0x29, 0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65,
-  0x41, 0x6c, 0x6c, 0x28, 0x2f, 0x5c, 0x7b, 0x5c, 0x7b, 0x28, 0x2e, 0x2a,
-  0x3f, 0x29, 0x5c, 0x7d, 0x5c, 0x7d, 0x2f, 0x67, 0x2c, 0x20, 0x28, 0x5f,
-  0x2c, 0x20, 0x6b, 0x65, 0x79, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x74, 0x65,
-  0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x28, 0x73, 0x65, 0x74, 0x74, 0x69,
-  0x6e, 0x67, 0x73, 0x5b, 0x6b, 0x65, 0x79, 0x5d, 0x29, 0x29, 0x3b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x2f,
-  0x2f, 0x20, 0x73, 0x65, 0x6e, 0x64, 0x20, 0x6d, 0x65, 0x73, 0x73, 0x61,
-  0x67, 0x65, 0x20, 0x74, 0x6f, 0x20, 0x73, 0x65, 0x72, 0x76, 0x65, 0x72,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63,
-  0x68, 0x61, 0x74, 0x20, 0x3d, 0x20, 0x61, 0x73, 0x79, 0x6e, 0x63, 0x20,
-  0x28, 0x6d, 0x73, 0x67, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x63, 0x6f, 0x6e,
+  0x20, 0x7d, 0x20, 0x66, 0x72, 0x6f, 0x6d, 0x20, 0x27, 0x2f, 0x63, 0x6f,
+  0x6d, 0x70, 0x6c, 0x65, 0x74, 0x69, 0x6f, 0x6e, 0x2e, 0x6a, 0x73, 0x27,
+  0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74,
+  0x20, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x20, 0x3d, 0x20, 0x73,
+  0x69, 0x67, 0x6e, 0x61, 0x6c, 0x28, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x3a, 0x20, 0x22, 0x54,
+  0x68, 0x69, 0x73, 0x20, 0x69, 0x73, 0x20, 0x61, 0x20, 0x63, 0x6f, 0x6e,
+  0x76, 0x65, 0x72, 0x73, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x20, 0x62, 0x65,
+  0x74, 0x77, 0x65, 0x65, 0x6e, 0x20, 0x75, 0x73, 0x65, 0x72, 0x20, 0x61,
+  0x6e, 0x64, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x2c, 0x20, 0x61, 0x20,
+  0x66, 0x72, 0x69, 0x65, 0x6e, 0x64, 0x6c, 0x79, 0x20, 0x63, 0x68, 0x61,
+  0x74, 0x62, 0x6f, 0x74, 0x2e, 0x20, 0x72, 0x65, 0x73, 0x70, 0x6f, 0x6e,
+  0x64, 0x20, 0x69, 0x6e, 0x20, 0x73, 0x69, 0x6d, 0x70, 0x6c, 0x65, 0x20,
+  0x6d, 0x61, 0x72, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x2e, 0x22, 0x2c, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61,
+  0x74, 0x65, 0x3a, 0x20, 0x22, 0x7b, 0x7b, 0x70, 0x72, 0x6f, 0x6d, 0x70,
+  0x74, 0x7d, 0x7d, 0x5c, 0x6e, 0x5c, 0x6e, 0x7b, 0x7b, 0x68, 0x69, 0x73,
+  0x74, 0x6f, 0x72, 0x79, 0x7d, 0x7d, 0x5c, 0x6e, 0x7b, 0x7b, 0x63, 0x68,
+  0x61, 0x72, 0x7d, 0x7d, 0x3a, 0x22, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x68, 0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x54, 0x65, 0x6d,
+  0x70, 0x6c, 0x61, 0x74, 0x65, 0x3a, 0x20, 0x22, 0x7b, 0x7b, 0x6e, 0x61,
+  0x6d, 0x65, 0x7d, 0x7d, 0x3a, 0x20, 0x7b, 0x7b, 0x6d, 0x65, 0x73, 0x73,
+  0x61, 0x67, 0x65, 0x7d, 0x7d, 0x22, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74,
+  0x3a, 0x20, 0x5b, 0x5d, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x74, 0x79, 0x70, 0x65, 0x3a, 0x20, 0x22, 0x63, 0x68, 0x61, 0x74, 0x22,
+  0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x68, 0x61, 0x72,
+  0x3a, 0x20, 0x22, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x22, 0x2c, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x75, 0x73, 0x65, 0x72, 0x3a, 0x20, 0x22,
+  0x55, 0x73, 0x65, 0x72, 0x22, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d,
+  0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74,
+  0x20, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x20, 0x3d, 0x20, 0x73, 0x69,
+  0x67, 0x6e, 0x61, 0x6c, 0x28, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x6e, 0x5f, 0x70, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x3a, 0x20,
+  0x34, 0x30, 0x30, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74,
+  0x65, 0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x3a, 0x20,
+  0x30, 0x2e, 0x37, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72,
+  0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x6c, 0x61, 0x73, 0x74, 0x5f, 0x6e,
+  0x3a, 0x20, 0x32, 0x35, 0x36, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x70, 0x65, 0x6e, 0x61,
+  0x6c, 0x74, 0x79, 0x3a, 0x20, 0x31, 0x2e, 0x31, 0x38, 0x2c, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x74, 0x6f, 0x70, 0x5f, 0x6b, 0x3a, 0x20,
+  0x34, 0x30, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74, 0x6f,
+  0x70, 0x5f, 0x70, 0x3a, 0x20, 0x30, 0x2e, 0x35, 0x2c, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x7d, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f,
+  0x6e, 0x73, 0x74, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x53, 0x74, 0x61,
+  0x74, 0x73, 0x20, 0x3d, 0x20, 0x73, 0x69, 0x67, 0x6e, 0x61, 0x6c, 0x28,
+  0x6e, 0x75, 0x6c, 0x6c, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f,
+  0x6e, 0x73, 0x74, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c,
+  0x65, 0x72, 0x20, 0x3d, 0x20, 0x73, 0x69, 0x67, 0x6e, 0x61, 0x6c, 0x28,
+  0x6e, 0x75, 0x6c, 0x6c, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63,
+  0x6f, 0x6e, 0x73, 0x74, 0x20, 0x67, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74,
+  0x69, 0x6e, 0x67, 0x20, 0x3d, 0x20, 0x63, 0x6f, 0x6d, 0x70, 0x75, 0x74,
+  0x65, 0x64, 0x28, 0x28, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x63, 0x6f, 0x6e,
   0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c, 0x75,
-  0x65, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x63, 0x6f, 0x6e, 0x73, 0x6f, 0x6c, 0x65, 0x2e, 0x6c, 0x6f, 0x67,
-  0x28, 0x27, 0x61, 0x6c, 0x72, 0x65, 0x61, 0x64, 0x79, 0x20, 0x72, 0x75,
-  0x6e, 0x6e, 0x69, 0x6e, 0x67, 0x2e, 0x2e, 0x2e, 0x27, 0x29, 0x3b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x74, 0x75,
-  0x72, 0x6e, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f,
-  0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d,
-  0x20, 0x6e, 0x65, 0x77, 0x20, 0x41, 0x62, 0x6f, 0x72, 0x74, 0x43, 0x6f,
-  0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x28, 0x29, 0x3b, 0x0a,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74, 0x72, 0x61, 0x6e, 0x73,
-  0x63, 0x72, 0x69, 0x70, 0x74, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x28,
-  0x5b, 0x2e, 0x2e, 0x2e, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e,
-  0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63,
-  0x72, 0x69, 0x70, 0x74, 0x2c, 0x20, 0x5b, 0x22, 0x7b, 0x7b, 0x75, 0x73,
-  0x65, 0x72, 0x7d, 0x7d, 0x22, 0x2c, 0x20, 0x6d, 0x73, 0x67, 0x5d, 0x5d,
-  0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e,
-  0x73, 0x74, 0x20, 0x70, 0x61, 0x79, 0x6c, 0x6f, 0x61, 0x64, 0x20, 0x3d,
-  0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x28, 0x73, 0x65,
-  0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e,
-  0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x2c, 0x20, 0x7b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x65, 0x73, 0x73,
-  0x61, 0x67, 0x65, 0x3a, 0x20, 0x6d, 0x73, 0x67, 0x2c, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x68, 0x69, 0x73, 0x74, 0x6f, 0x72,
-  0x79, 0x3a, 0x20, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76,
+  0x65, 0x20, 0x3d, 0x3d, 0x20, 0x6e, 0x75, 0x6c, 0x6c, 0x20, 0x29, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63, 0x68,
+  0x61, 0x74, 0x53, 0x74, 0x61, 0x72, 0x74, 0x65, 0x64, 0x20, 0x3d, 0x20,
+  0x63, 0x6f, 0x6d, 0x70, 0x75, 0x74, 0x65, 0x64, 0x28, 0x28, 0x29, 0x20,
+  0x3d, 0x3e, 0x20, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76,
   0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72,
-  0x69, 0x70, 0x74, 0x2e, 0x66, 0x6c, 0x61, 0x74, 0x4d, 0x61, 0x70, 0x28,
-  0x28, 0x5b, 0x6e, 0x61, 0x6d, 0x65, 0x2c, 0x20, 0x6d, 0x65, 0x73, 0x73,
-  0x61, 0x67, 0x65, 0x5d, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x74, 0x65, 0x6d,
-  0x70, 0x6c, 0x61, 0x74, 0x65, 0x28, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f,
-  0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x68, 0x69, 0x73, 0x74,
-  0x6f, 0x72, 0x79, 0x54, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x2c,
-  0x20, 0x7b, 0x6e, 0x61, 0x6d, 0x65, 0x2c, 0x20, 0x6d, 0x65, 0x73, 0x73,
-  0x61, 0x67, 0x65, 0x7d, 0x29, 0x29, 0x2e, 0x6a, 0x6f, 0x69, 0x6e, 0x28,
-  0x22, 0x5c, 0x6e, 0x22, 0x29, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x7d, 0x29, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x6c, 0x65, 0x74, 0x20, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x4d,
-  0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x20, 0x3d, 0x20, 0x27, 0x27, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74,
-  0x20, 0x68, 0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x20, 0x3d, 0x20, 0x73,
-  0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65,
-  0x2e, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x0a,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74,
-  0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73,
-  0x20, 0x3d, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x2e, 0x2e, 0x2e, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76,
+  0x69, 0x70, 0x74, 0x2e, 0x6c, 0x65, 0x6e, 0x67, 0x74, 0x68, 0x20, 0x3e,
+  0x20, 0x30, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e,
+  0x73, 0x74, 0x20, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70,
+  0x74, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x20, 0x3d, 0x20, 0x28, 0x74,
+  0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x29, 0x20, 0x3d,
+  0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x73, 0x65,
+  0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20,
+  0x3d, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x2e, 0x2e, 0x2e, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76,
   0x61, 0x6c, 0x75, 0x65, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x3a, 0x20, 0x70, 0x61,
-  0x79, 0x6c, 0x6f, 0x61, 0x64, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x73, 0x74, 0x6f, 0x70, 0x3a, 0x20, 0x5b, 0x22, 0x3c,
-  0x2f, 0x73, 0x3e, 0x22, 0x2c, 0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61,
-  0x74, 0x65, 0x28, 0x22, 0x7b, 0x7b, 0x63, 0x68, 0x61, 0x72, 0x7d, 0x7d,
-  0x3a, 0x22, 0x29, 0x2c, 0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74,
-  0x65, 0x28, 0x22, 0x7b, 0x7b, 0x75, 0x73, 0x65, 0x72, 0x7d, 0x7d, 0x3a,
-  0x22, 0x29, 0x5d, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d,
-  0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x61, 0x77, 0x61, 0x69,
-  0x74, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x43, 0x6f, 0x6d, 0x70, 0x6c,
-  0x65, 0x74, 0x65, 0x28, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x50, 0x61, 0x72,
-  0x61, 0x6d, 0x73, 0x2c, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c,
-  0x6c, 0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2c, 0x20, 0x28,
-  0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x29, 0x20, 0x3d, 0x3e, 0x20,
-  0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f,
-  0x6e, 0x73, 0x74, 0x20, 0x64, 0x61, 0x74, 0x61, 0x20, 0x3d, 0x20, 0x6d,
-  0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x75, 0x72,
-  0x72, 0x65, 0x6e, 0x74, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x20,
-  0x2b, 0x3d, 0x20, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x63, 0x6f, 0x6e, 0x74,
-  0x65, 0x6e, 0x74, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x2f, 0x2f, 0x20, 0x72, 0x65, 0x6d, 0x6f, 0x76, 0x65, 0x20, 0x6c,
-  0x65, 0x61, 0x64, 0x69, 0x6e, 0x67, 0x20, 0x77, 0x68, 0x69, 0x74, 0x65,
-  0x73, 0x70, 0x61, 0x63, 0x65, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x4d, 0x65, 0x73,
-  0x73, 0x61, 0x67, 0x65, 0x20, 0x3d, 0x20, 0x63, 0x75, 0x72, 0x72, 0x65,
-  0x6e, 0x74, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x2e, 0x72, 0x65,
-  0x70, 0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x5e, 0x5c, 0x73, 0x2b, 0x2f,
-  0x2c, 0x20, 0x22, 0x22, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70,
-  0x74, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x28, 0x5b, 0x2e, 0x2e, 0x2e,
-  0x68, 0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x2c, 0x20, 0x5b, 0x22, 0x7b,
-  0x7b, 0x63, 0x68, 0x61, 0x72, 0x7d, 0x7d, 0x22, 0x2c, 0x20, 0x63, 0x75,
-  0x72, 0x72, 0x65, 0x6e, 0x74, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65,
-  0x5d, 0x5d, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x69, 0x66, 0x20, 0x28, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x73, 0x74,
-  0x6f, 0x70, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x6f, 0x6c, 0x65, 0x2e,
-  0x6c, 0x6f, 0x67, 0x28, 0x22, 0x2d, 0x2d, 0x3e, 0x22, 0x2c, 0x20, 0x64,
-  0x61, 0x74, 0x61, 0x2c, 0x20, 0x27, 0x20, 0x72, 0x65, 0x73, 0x70, 0x6f,
-  0x6e, 0x73, 0x65, 0x20, 0x77, 0x61, 0x73, 0x3a, 0x27, 0x2c, 0x20, 0x63,
-  0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67,
-  0x65, 0x2c, 0x20, 0x27, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69,
-  0x70, 0x74, 0x20, 0x73, 0x74, 0x61, 0x74, 0x65, 0x3a, 0x27, 0x2c, 0x20,
-  0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75,
-  0x65, 0x2e, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74,
-  0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x29, 0x0a, 0x0a, 0x20,
+  0x20, 0x20, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20, 0x73,
+  0x69, 0x6d, 0x70, 0x6c, 0x65, 0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61,
+  0x74, 0x65, 0x20, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x74, 0x65, 0x6d,
+  0x70, 0x6c, 0x61, 0x74, 0x65, 0x20, 0x3d, 0x20, 0x28, 0x73, 0x74, 0x72,
+  0x2c, 0x20, 0x65, 0x78, 0x74, 0x72, 0x61, 0x53, 0x65, 0x74, 0x74, 0x69,
+  0x6e, 0x67, 0x73, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x6c, 0x65, 0x74, 0x20, 0x73, 0x65, 0x74, 0x74,
+  0x69, 0x6e, 0x67, 0x73, 0x20, 0x3d, 0x20, 0x73, 0x65, 0x73, 0x73, 0x69,
+  0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x65, 0x78, 0x74, 0x72,
+  0x61, 0x53, 0x65, 0x74, 0x74, 0x69, 0x6e, 0x67, 0x73, 0x29, 0x20, 0x7b,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x73, 0x65, 0x74,
+  0x74, 0x69, 0x6e, 0x67, 0x73, 0x20, 0x3d, 0x20, 0x7b, 0x20, 0x2e, 0x2e,
+  0x2e, 0x73, 0x65, 0x74, 0x74, 0x69, 0x6e, 0x67, 0x73, 0x2c, 0x20, 0x2e,
+  0x2e, 0x2e, 0x65, 0x78, 0x74, 0x72, 0x61, 0x53, 0x65, 0x74, 0x74, 0x69,
+  0x6e, 0x67, 0x73, 0x20, 0x7d, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x74,
+  0x75, 0x72, 0x6e, 0x20, 0x53, 0x74, 0x72, 0x69, 0x6e, 0x67, 0x28, 0x73,
+  0x74, 0x72, 0x29, 0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x41,
+  0x6c, 0x6c, 0x28, 0x2f, 0x5c, 0x7b, 0x5c, 0x7b, 0x28, 0x2e, 0x2a, 0x3f,
+  0x29, 0x5c, 0x7d, 0x5c, 0x7d, 0x2f, 0x67, 0x2c, 0x20, 0x28, 0x5f, 0x2c,
+  0x20, 0x6b, 0x65, 0x79, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x74, 0x65, 0x6d,
+  0x70, 0x6c, 0x61, 0x74, 0x65, 0x28, 0x73, 0x65, 0x74, 0x74, 0x69, 0x6e,
+  0x67, 0x73, 0x5b, 0x6b, 0x65, 0x79, 0x5d, 0x29, 0x29, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f,
+  0x20, 0x73, 0x65, 0x6e, 0x64, 0x20, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67,
+  0x65, 0x20, 0x74, 0x6f, 0x20, 0x73, 0x65, 0x72, 0x76, 0x65, 0x72, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63, 0x68,
+  0x61, 0x74, 0x20, 0x3d, 0x20, 0x61, 0x73, 0x79, 0x6e, 0x63, 0x20, 0x28,
+  0x6d, 0x73, 0x67, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x63, 0x6f, 0x6e, 0x74,
+  0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65,
+  0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x63, 0x6f, 0x6e, 0x73, 0x6f, 0x6c, 0x65, 0x2e, 0x6c, 0x6f, 0x67, 0x28,
+  0x27, 0x61, 0x6c, 0x72, 0x65, 0x61, 0x64, 0x79, 0x20, 0x72, 0x75, 0x6e,
+  0x6e, 0x69, 0x6e, 0x67, 0x2e, 0x2e, 0x2e, 0x27, 0x29, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x74, 0x75, 0x72,
+  0x6e, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20,
   0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c,
   0x6c, 0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d, 0x20,
-  0x6e, 0x75, 0x6c, 0x6c, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x66, 0x75, 0x6e, 0x63, 0x74, 0x69, 0x6f,
-  0x6e, 0x20, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x49, 0x6e, 0x70,
-  0x75, 0x74, 0x28, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x6d, 0x65, 0x73, 0x73, 0x61,
-  0x67, 0x65, 0x20, 0x3d, 0x20, 0x75, 0x73, 0x65, 0x53, 0x69, 0x67, 0x6e,
-  0x61, 0x6c, 0x28, 0x22, 0x22, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x73, 0x74, 0x6f, 0x70,
-  0x20, 0x3d, 0x20, 0x28, 0x65, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x65, 0x2e, 0x70, 0x72,
-  0x65, 0x76, 0x65, 0x6e, 0x74, 0x44, 0x65, 0x66, 0x61, 0x75, 0x6c, 0x74,
-  0x28, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x69, 0x66, 0x20, 0x28, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c,
-  0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x29, 0x20, 0x7b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f,
-  0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c,
-  0x75, 0x65, 0x2e, 0x61, 0x62, 0x6f, 0x72, 0x74, 0x28, 0x29, 0x3b, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f,
-  0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c,
-  0x75, 0x65, 0x20, 0x3d, 0x20, 0x6e, 0x75, 0x6c, 0x6c, 0x3b, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x72, 0x65, 0x73, 0x65, 0x74, 0x20,
+  0x6e, 0x65, 0x77, 0x20, 0x41, 0x62, 0x6f, 0x72, 0x74, 0x43, 0x6f, 0x6e,
+  0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x28, 0x29, 0x3b, 0x0a, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63,
+  0x72, 0x69, 0x70, 0x74, 0x55, 0x70, 0x64, 0x61, 0x74, 0x65, 0x28, 0x5b,
+  0x2e, 0x2e, 0x2e, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76,
+  0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72,
+  0x69, 0x70, 0x74, 0x2c, 0x20, 0x5b, 0x22, 0x7b, 0x7b, 0x75, 0x73, 0x65,
+  0x72, 0x7d, 0x7d, 0x22, 0x2c, 0x20, 0x6d, 0x73, 0x67, 0x5d, 0x5d, 0x29,
+  0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73,
+  0x74, 0x20, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x20, 0x3d, 0x20, 0x74,
+  0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x28, 0x73, 0x65, 0x73, 0x73,
+  0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74, 0x65,
+  0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x2c, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67,
+  0x65, 0x3a, 0x20, 0x6d, 0x73, 0x67, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x68, 0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x3a,
+  0x20, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c,
+  0x75, 0x65, 0x2e, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70,
+  0x74, 0x2e, 0x66, 0x6c, 0x61, 0x74, 0x4d, 0x61, 0x70, 0x28, 0x28, 0x5b,
+  0x6e, 0x61, 0x6d, 0x65, 0x2c, 0x20, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67,
+  0x65, 0x5d, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c,
+  0x61, 0x74, 0x65, 0x28, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e,
+  0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x68, 0x69, 0x73, 0x74, 0x6f, 0x72,
+  0x79, 0x54, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x2c, 0x20, 0x7b,
+  0x6e, 0x61, 0x6d, 0x65, 0x2c, 0x20, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67,
+  0x65, 0x7d, 0x29, 0x29, 0x2e, 0x6a, 0x6f, 0x69, 0x6e, 0x28, 0x22, 0x5c,
+  0x6e, 0x22, 0x29, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d,
+  0x29, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6c, 0x65,
+  0x74, 0x20, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x4d, 0x65, 0x73,
+  0x73, 0x61, 0x67, 0x65, 0x20, 0x3d, 0x20, 0x27, 0x27, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x68,
+  0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x20, 0x3d, 0x20, 0x73, 0x65, 0x73,
+  0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74,
+  0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x0a, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x6c,
+  0x6c, 0x61, 0x6d, 0x61, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x20, 0x3d,
+  0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2e,
+  0x2e, 0x2e, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c,
+  0x75, 0x65, 0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x73, 0x74, 0x6f, 0x70, 0x3a, 0x20, 0x5b, 0x22, 0x3c, 0x2f, 0x73, 0x3e,
+  0x22, 0x2c, 0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x28,
+  0x22, 0x7b, 0x7b, 0x63, 0x68, 0x61, 0x72, 0x7d, 0x7d, 0x3a, 0x22, 0x29,
+  0x2c, 0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x28, 0x22,
+  0x7b, 0x7b, 0x75, 0x73, 0x65, 0x72, 0x7d, 0x7d, 0x3a, 0x22, 0x29, 0x5d,
+  0x2c, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x66, 0x6f, 0x72, 0x20, 0x61, 0x77, 0x61,
+  0x69, 0x74, 0x20, 0x28, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63, 0x68,
+  0x75, 0x6e, 0x6b, 0x20, 0x6f, 0x66, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61,
+  0x28, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x2c, 0x20, 0x6c, 0x6c, 0x61,
+  0x6d, 0x61, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2c, 0x20, 0x7b, 0x20,
+  0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x3a, 0x20,
+  0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x76,
+  0x61, 0x6c, 0x75, 0x65, 0x20, 0x7d, 0x29, 0x29, 0x20, 0x7b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74,
+  0x20, 0x64, 0x61, 0x74, 0x61, 0x20, 0x3d, 0x20, 0x63, 0x68, 0x75, 0x6e,
+  0x6b, 0x2e, 0x64, 0x61, 0x74, 0x61, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x4d,
+  0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x20, 0x2b, 0x3d, 0x20, 0x64, 0x61,
+  0x74, 0x61, 0x2e, 0x63, 0x6f, 0x6e, 0x74, 0x65, 0x6e, 0x74, 0x3b, 0x0a,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20,
+  0x72, 0x65, 0x6d, 0x6f, 0x76, 0x65, 0x20, 0x6c, 0x65, 0x61, 0x64, 0x69,
+  0x6e, 0x67, 0x20, 0x77, 0x68, 0x69, 0x74, 0x65, 0x73, 0x70, 0x61, 0x63,
+  0x65, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x75,
+  0x72, 0x72, 0x65, 0x6e, 0x74, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65,
+  0x20, 0x3d, 0x20, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x4d, 0x65,
+  0x73, 0x73, 0x61, 0x67, 0x65, 0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63,
+  0x65, 0x28, 0x2f, 0x5e, 0x5c, 0x73, 0x2b, 0x2f, 0x2c, 0x20, 0x22, 0x22,
+  0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74,
+  0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x55, 0x70, 0x64,
+  0x61, 0x74, 0x65, 0x28, 0x5b, 0x2e, 0x2e, 0x2e, 0x68, 0x69, 0x73, 0x74,
+  0x6f, 0x72, 0x79, 0x2c, 0x20, 0x5b, 0x22, 0x7b, 0x7b, 0x63, 0x68, 0x61,
+  0x72, 0x7d, 0x7d, 0x22, 0x2c, 0x20, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e,
+  0x74, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x5d, 0x5d, 0x29, 0x0a,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20,
+  0x28, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x73, 0x74, 0x6f, 0x70, 0x29, 0x20,
+  0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x63, 0x6f, 0x6e, 0x73, 0x6f, 0x6c, 0x65, 0x2e, 0x6c, 0x6f, 0x67, 0x28,
+  0x22, 0x43, 0x6f, 0x6d, 0x70, 0x6c, 0x65, 0x74, 0x69, 0x6f, 0x6e, 0x20,
+  0x66, 0x69, 0x6e, 0x69, 0x73, 0x68, 0x65, 0x64, 0x3a, 0x20, 0x27, 0x22,
+  0x2c, 0x20, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x4d, 0x65, 0x73,
+  0x73, 0x61, 0x67, 0x65, 0x2c, 0x20, 0x22, 0x27, 0x2c, 0x20, 0x73, 0x75,
+  0x6d, 0x6d, 0x61, 0x72, 0x79, 0x3a, 0x20, 0x22, 0x2c, 0x20, 0x64, 0x61,
+  0x74, 0x61, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x69, 0x66, 0x20, 0x28, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x74, 0x69, 0x6d,
+  0x69, 0x6e, 0x67, 0x73, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x53,
+  0x74, 0x61, 0x74, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d,
+  0x20, 0x64, 0x61, 0x74, 0x61, 0x2e, 0x74, 0x69, 0x6d, 0x69, 0x6e, 0x67,
+  0x73, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c,
+  0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d, 0x20, 0x6e,
+  0x75, 0x6c, 0x6c, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x66, 0x75, 0x6e, 0x63, 0x74, 0x69, 0x6f, 0x6e,
+  0x20, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x49, 0x6e, 0x70, 0x75,
+  0x74, 0x28, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67,
+  0x65, 0x20, 0x3d, 0x20, 0x75, 0x73, 0x65, 0x53, 0x69, 0x67, 0x6e, 0x61,
+  0x6c, 0x28, 0x22, 0x22, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x73, 0x74, 0x6f, 0x70, 0x20,
   0x3d, 0x20, 0x28, 0x65, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x73, 0x74, 0x6f, 0x70, 0x28,
-  0x65, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x55, 0x70,
-  0x64, 0x61, 0x74, 0x65, 0x28, 0x5b, 0x5d, 0x29, 0x3b, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x73, 0x75, 0x62, 0x6d, 0x69,
-  0x74, 0x20, 0x3d, 0x20, 0x28, 0x65, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x73, 0x74, 0x6f,
-  0x70, 0x28, 0x65, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x63, 0x68, 0x61, 0x74, 0x28, 0x6d, 0x65, 0x73, 0x73, 0x61,
-  0x67, 0x65, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x29, 0x3b, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x65, 0x73, 0x73, 0x61,
-  0x67, 0x65, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d, 0x20, 0x22,
-  0x22, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20,
-  0x65, 0x6e, 0x74, 0x65, 0x72, 0x53, 0x75, 0x62, 0x6d, 0x69, 0x74, 0x73,
-  0x20, 0x3d, 0x20, 0x28, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x29, 0x20, 0x3d,
-  0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x69, 0x66, 0x20, 0x28, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x2e, 0x77, 0x68,
-  0x69, 0x63, 0x68, 0x20, 0x3d, 0x3d, 0x3d, 0x20, 0x31, 0x33, 0x20, 0x26,
-  0x26, 0x20, 0x21, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x2e, 0x73, 0x68, 0x69,
-  0x66, 0x74, 0x4b, 0x65, 0x79, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x73, 0x75, 0x62, 0x6d, 0x69,
-  0x74, 0x28, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x29, 0x3b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x65, 0x2e, 0x70, 0x72, 0x65,
+  0x76, 0x65, 0x6e, 0x74, 0x44, 0x65, 0x66, 0x61, 0x75, 0x6c, 0x74, 0x28,
+  0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x69,
+  0x66, 0x20, 0x28, 0x63, 0x6f, 0x6e, 0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65,
+  0x72, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x29, 0x20, 0x7b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e,
+  0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c, 0x75,
+  0x65, 0x2e, 0x61, 0x62, 0x6f, 0x72, 0x74, 0x28, 0x29, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e,
+  0x74, 0x72, 0x6f, 0x6c, 0x6c, 0x65, 0x72, 0x2e, 0x76, 0x61, 0x6c, 0x75,
+  0x65, 0x20, 0x3d, 0x20, 0x6e, 0x75, 0x6c, 0x6c, 0x3b, 0x0a, 0x20, 0x20,
   0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72,
-  0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x68, 0x74, 0x6d, 0x6c, 0x60, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x66, 0x6f, 0x72,
-  0x6d, 0x20, 0x6f, 0x6e, 0x73, 0x75, 0x62, 0x6d, 0x69, 0x74, 0x3d, 0x24,
-  0x7b, 0x73, 0x75, 0x62, 0x6d, 0x69, 0x74, 0x7d, 0x3e, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76,
+  0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63,
+  0x6f, 0x6e, 0x73, 0x74, 0x20, 0x72, 0x65, 0x73, 0x65, 0x74, 0x20, 0x3d,
+  0x20, 0x28, 0x65, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x73, 0x74, 0x6f, 0x70, 0x28, 0x65,
+  0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x74,
+  0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x55, 0x70, 0x64,
+  0x61, 0x74, 0x65, 0x28, 0x5b, 0x5d, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x73, 0x75, 0x62, 0x6d, 0x69, 0x74,
+  0x20, 0x3d, 0x20, 0x28, 0x65, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x73, 0x74, 0x6f, 0x70,
+  0x28, 0x65, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x63, 0x68, 0x61, 0x74, 0x28, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67,
+  0x65, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x29, 0x3b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67,
+  0x65, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d, 0x20, 0x22, 0x22,
+  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x65,
+  0x6e, 0x74, 0x65, 0x72, 0x53, 0x75, 0x62, 0x6d, 0x69, 0x74, 0x73, 0x20,
+  0x3d, 0x20, 0x28, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x29, 0x20, 0x3d, 0x3e,
+  0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x69,
+  0x66, 0x20, 0x28, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x2e, 0x77, 0x68, 0x69,
+  0x63, 0x68, 0x20, 0x3d, 0x3d, 0x3d, 0x20, 0x31, 0x33, 0x20, 0x26, 0x26,
+  0x20, 0x21, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x2e, 0x73, 0x68, 0x69, 0x66,
+  0x74, 0x4b, 0x65, 0x79, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x73, 0x75, 0x62, 0x6d, 0x69, 0x74,
+  0x28, 0x65, 0x76, 0x65, 0x6e, 0x74, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65,
+  0x74, 0x75, 0x72, 0x6e, 0x20, 0x68, 0x74, 0x6d, 0x6c, 0x60, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x66, 0x6f, 0x72, 0x6d,
+  0x20, 0x6f, 0x6e, 0x73, 0x75, 0x62, 0x6d, 0x69, 0x74, 0x3d, 0x24, 0x7b,
+  0x73, 0x75, 0x62, 0x6d, 0x69, 0x74, 0x7d, 0x3e, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x3c, 0x74, 0x65, 0x78, 0x74, 0x61, 0x72, 0x65, 0x61, 0x20, 0x74,
+  0x79, 0x70, 0x65, 0x3d, 0x22, 0x74, 0x65, 0x78, 0x74, 0x22, 0x20, 0x72,
+  0x6f, 0x77, 0x73, 0x3d, 0x32, 0x20, 0x6f, 0x6e, 0x6b, 0x65, 0x79, 0x70,
+  0x72, 0x65, 0x73, 0x73, 0x3d, 0x24, 0x7b, 0x65, 0x6e, 0x74, 0x65, 0x72,
+  0x53, 0x75, 0x62, 0x6d, 0x69, 0x74, 0x73, 0x7d, 0x20, 0x76, 0x61, 0x6c,
+  0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67,
+  0x65, 0x7d, 0x22, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d,
+  0x24, 0x7b, 0x28, 0x65, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x6d, 0x65, 0x73,
+  0x73, 0x61, 0x67, 0x65, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d,
+  0x20, 0x65, 0x2e, 0x74, 0x61, 0x72, 0x67, 0x65, 0x74, 0x2e, 0x76, 0x61,
+  0x6c, 0x75, 0x65, 0x7d, 0x20, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x68, 0x6f,
+  0x6c, 0x64, 0x65, 0x72, 0x3d, 0x22, 0x53, 0x61, 0x79, 0x20, 0x73, 0x6f,
+  0x6d, 0x65, 0x74, 0x68, 0x69, 0x6e, 0x67, 0x2e, 0x2e, 0x2e, 0x22, 0x2f,
   0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x3c, 0x74, 0x65, 0x78, 0x74, 0x61, 0x72, 0x65, 0x61, 0x20, 0x74, 0x79,
-  0x70, 0x65, 0x3d, 0x22, 0x74, 0x65, 0x78, 0x74, 0x22, 0x20, 0x72, 0x6f,
-  0x77, 0x73, 0x3d, 0x32, 0x20, 0x6f, 0x6e, 0x6b, 0x65, 0x79, 0x70, 0x72,
-  0x65, 0x73, 0x73, 0x3d, 0x24, 0x7b, 0x65, 0x6e, 0x74, 0x65, 0x72, 0x53,
-  0x75, 0x62, 0x6d, 0x69, 0x74, 0x73, 0x7d, 0x20, 0x76, 0x61, 0x6c, 0x75,
-  0x65, 0x3d, 0x22, 0x24, 0x7b, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65,
-  0x7d, 0x22, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24,
-  0x7b, 0x28, 0x65, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x6d, 0x65, 0x73, 0x73,
-  0x61, 0x67, 0x65, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d, 0x20,
-  0x65, 0x2e, 0x74, 0x61, 0x72, 0x67, 0x65, 0x74, 0x2e, 0x76, 0x61, 0x6c,
-  0x75, 0x65, 0x7d, 0x20, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x68, 0x6f, 0x6c,
-  0x64, 0x65, 0x72, 0x3d, 0x22, 0x53, 0x61, 0x79, 0x20, 0x73, 0x6f, 0x6d,
-  0x65, 0x74, 0x68, 0x69, 0x6e, 0x67, 0x2e, 0x2e, 0x2e, 0x22, 0x2f, 0x3e,
-  0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
   0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
   0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x20, 0x63, 0x6c,
   0x61, 0x73, 0x73, 0x3d, 0x22, 0x72, 0x69, 0x67, 0x68, 0x74, 0x22, 0x3e,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x3c, 0x62, 0x75, 0x74, 0x74, 0x6f, 0x6e, 0x20, 0x74, 0x79, 0x70,
+  0x65, 0x3d, 0x22, 0x73, 0x75, 0x62, 0x6d, 0x69, 0x74, 0x22, 0x20, 0x64,
+  0x69, 0x73, 0x61, 0x62, 0x6c, 0x65, 0x64, 0x3d, 0x24, 0x7b, 0x21, 0x67,
+  0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x69, 0x6e, 0x67, 0x2e, 0x76, 0x61,
+  0x6c, 0x75, 0x65, 0x7d, 0x20, 0x3e, 0x53, 0x65, 0x6e, 0x64, 0x3c, 0x2f,
+  0x62, 0x75, 0x74, 0x74, 0x6f, 0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x62, 0x75, 0x74,
+  0x74, 0x6f, 0x6e, 0x20, 0x6f, 0x6e, 0x63, 0x6c, 0x69, 0x63, 0x6b, 0x3d,
+  0x24, 0x7b, 0x73, 0x74, 0x6f, 0x70, 0x7d, 0x20, 0x64, 0x69, 0x73, 0x61,
+  0x62, 0x6c, 0x65, 0x64, 0x3d, 0x24, 0x7b, 0x67, 0x65, 0x6e, 0x65, 0x72,
+  0x61, 0x74, 0x69, 0x6e, 0x67, 0x7d, 0x3e, 0x53, 0x74, 0x6f, 0x70, 0x3c,
+  0x2f, 0x62, 0x75, 0x74, 0x74, 0x6f, 0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x62, 0x75,
+  0x74, 0x74, 0x6f, 0x6e, 0x20, 0x6f, 0x6e, 0x63, 0x6c, 0x69, 0x63, 0x6b,
+  0x3d, 0x24, 0x7b, 0x72, 0x65, 0x73, 0x65, 0x74, 0x7d, 0x3e, 0x52, 0x65,
+  0x73, 0x65, 0x74, 0x3c, 0x2f, 0x62, 0x75, 0x74, 0x74, 0x6f, 0x6e, 0x3e,
   0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
-  0x62, 0x75, 0x74, 0x74, 0x6f, 0x6e, 0x20, 0x74, 0x79, 0x70, 0x65, 0x3d,
-  0x22, 0x73, 0x75, 0x62, 0x6d, 0x69, 0x74, 0x22, 0x20, 0x64, 0x69, 0x73,
-  0x61, 0x62, 0x6c, 0x65, 0x64, 0x3d, 0x24, 0x7b, 0x21, 0x67, 0x65, 0x6e,
-  0x65, 0x72, 0x61, 0x74, 0x69, 0x6e, 0x67, 0x2e, 0x76, 0x61, 0x6c, 0x75,
-  0x65, 0x7d, 0x20, 0x3e, 0x53, 0x65, 0x6e, 0x64, 0x3c, 0x2f, 0x62, 0x75,
-  0x74, 0x74, 0x6f, 0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x3c, 0x62, 0x75, 0x74, 0x74, 0x6f, 0x6e, 0x20,
-  0x6f, 0x6e, 0x63, 0x6c, 0x69, 0x63, 0x6b, 0x3d, 0x24, 0x7b, 0x73, 0x74,
-  0x6f, 0x70, 0x7d, 0x20, 0x64, 0x69, 0x73, 0x61, 0x62, 0x6c, 0x65, 0x64,
-  0x3d, 0x24, 0x7b, 0x67, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x69, 0x6e,
-  0x67, 0x7d, 0x3e, 0x53, 0x74, 0x6f, 0x70, 0x3c, 0x2f, 0x62, 0x75, 0x74,
-  0x74, 0x6f, 0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x3c, 0x62, 0x75, 0x74, 0x74, 0x6f, 0x6e, 0x20, 0x6f,
-  0x6e, 0x63, 0x6c, 0x69, 0x63, 0x6b, 0x3d, 0x24, 0x7b, 0x72, 0x65, 0x73,
-  0x65, 0x74, 0x7d, 0x3e, 0x52, 0x65, 0x73, 0x65, 0x74, 0x3c, 0x2f, 0x62,
-  0x75, 0x74, 0x74, 0x6f, 0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x66, 0x6f,
-  0x72, 0x6d, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x60, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63,
-  0x6f, 0x6e, 0x73, 0x74, 0x20, 0x43, 0x68, 0x61, 0x74, 0x4c, 0x6f, 0x67,
-  0x20, 0x3d, 0x20, 0x28, 0x70, 0x72, 0x6f, 0x70, 0x73, 0x29, 0x20, 0x3d,
-  0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f,
-  0x6e, 0x73, 0x74, 0x20, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x73,
-  0x20, 0x3d, 0x20, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76,
-  0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74, 0x72, 0x61, 0x6e, 0x73, 0x63, 0x72,
-  0x69, 0x70, 0x74, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63,
-  0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e,
-  0x65, 0x72, 0x20, 0x3d, 0x20, 0x75, 0x73, 0x65, 0x52, 0x65, 0x66, 0x28,
-  0x6e, 0x75, 0x6c, 0x6c, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x75, 0x73, 0x65, 0x45, 0x66, 0x66, 0x65, 0x63, 0x74, 0x28, 0x28,
-  0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20, 0x73, 0x63, 0x72, 0x6f, 0x6c, 0x6c,
-  0x20, 0x74, 0x6f, 0x20, 0x62, 0x6f, 0x74, 0x74, 0x6f, 0x6d, 0x20, 0x28,
-  0x69, 0x66, 0x20, 0x6e, 0x65, 0x65, 0x64, 0x65, 0x64, 0x29, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x63,
-  0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x2e, 0x63, 0x75, 0x72,
-  0x72, 0x65, 0x6e, 0x74, 0x20, 0x26, 0x26, 0x20, 0x63, 0x6f, 0x6e, 0x74,
-  0x61, 0x69, 0x6e, 0x65, 0x72, 0x2e, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e,
-  0x74, 0x2e, 0x73, 0x63, 0x72, 0x6f, 0x6c, 0x6c, 0x48, 0x65, 0x69, 0x67,
-  0x68, 0x74, 0x20, 0x3c, 0x3d, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69,
-  0x6e, 0x65, 0x72, 0x2e, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x2e,
-  0x73, 0x63, 0x72, 0x6f, 0x6c, 0x6c, 0x54, 0x6f, 0x70, 0x20, 0x2b, 0x20,
-  0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x2e, 0x63, 0x75,
-  0x72, 0x72, 0x65, 0x6e, 0x74, 0x2e, 0x6f, 0x66, 0x66, 0x73, 0x65, 0x74,
-  0x48, 0x65, 0x69, 0x67, 0x68, 0x74, 0x20, 0x2b, 0x20, 0x33, 0x30, 0x30,
-  0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x2e,
+  0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x3c, 0x2f, 0x66, 0x6f, 0x72, 0x6d, 0x3e, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x60, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x43,
+  0x68, 0x61, 0x74, 0x4c, 0x6f, 0x67, 0x20, 0x3d, 0x20, 0x28, 0x70, 0x72,
+  0x6f, 0x70, 0x73, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x6d, 0x65,
+  0x73, 0x73, 0x61, 0x67, 0x65, 0x73, 0x20, 0x3d, 0x20, 0x73, 0x65, 0x73,
+  0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74,
+  0x72, 0x61, 0x6e, 0x73, 0x63, 0x72, 0x69, 0x70, 0x74, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63,
+  0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x20, 0x3d, 0x20, 0x75,
+  0x73, 0x65, 0x52, 0x65, 0x66, 0x28, 0x6e, 0x75, 0x6c, 0x6c, 0x29, 0x0a,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x75, 0x73, 0x65, 0x45, 0x66,
+  0x66, 0x65, 0x63, 0x74, 0x28, 0x28, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20,
+  0x73, 0x63, 0x72, 0x6f, 0x6c, 0x6c, 0x20, 0x74, 0x6f, 0x20, 0x62, 0x6f,
+  0x74, 0x74, 0x6f, 0x6d, 0x20, 0x28, 0x69, 0x66, 0x20, 0x6e, 0x65, 0x65,
+  0x64, 0x65, 0x64, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x69, 0x66, 0x20, 0x28, 0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e,
+  0x65, 0x72, 0x2e, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x20, 0x26,
+  0x26, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x2e,
   0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x2e, 0x73, 0x63, 0x72, 0x6f,
-  0x6c, 0x6c, 0x54, 0x6f, 0x28, 0x30, 0x2c, 0x20, 0x63, 0x6f, 0x6e, 0x74,
+  0x6c, 0x6c, 0x48, 0x65, 0x69, 0x67, 0x68, 0x74, 0x20, 0x3c, 0x3d, 0x20,
+  0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x2e, 0x63, 0x75,
+  0x72, 0x72, 0x65, 0x6e, 0x74, 0x2e, 0x73, 0x63, 0x72, 0x6f, 0x6c, 0x6c,
+  0x54, 0x6f, 0x70, 0x20, 0x2b, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69,
+  0x6e, 0x65, 0x72, 0x2e, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x2e,
+  0x6f, 0x66, 0x66, 0x73, 0x65, 0x74, 0x48, 0x65, 0x69, 0x67, 0x68, 0x74,
+  0x20, 0x2b, 0x20, 0x33, 0x30, 0x30, 0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x74,
   0x61, 0x69, 0x6e, 0x65, 0x72, 0x2e, 0x63, 0x75, 0x72, 0x72, 0x65, 0x6e,
-  0x74, 0x2e, 0x73, 0x63, 0x72, 0x6f, 0x6c, 0x6c, 0x48, 0x65, 0x69, 0x67,
-  0x68, 0x74, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x2c, 0x20, 0x5b,
-  0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x73, 0x5d, 0x29, 0x0a, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20,
-  0x63, 0x68, 0x61, 0x74, 0x4c, 0x69, 0x6e, 0x65, 0x20, 0x3d, 0x20, 0x28,
-  0x5b, 0x75, 0x73, 0x65, 0x72, 0x2c, 0x20, 0x6d, 0x73, 0x67, 0x5d, 0x29,
-  0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x72, 0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x68, 0x74, 0x6d,
-  0x6c, 0x60, 0x3c, 0x70, 0x20, 0x6b, 0x65, 0x79, 0x3d, 0x24, 0x7b, 0x6d,
-  0x73, 0x67, 0x7d, 0x3e, 0x3c, 0x73, 0x74, 0x72, 0x6f, 0x6e, 0x67, 0x3e,
-  0x24, 0x7b, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x28, 0x75,
-  0x73, 0x65, 0x72, 0x29, 0x7d, 0x3a, 0x3c, 0x2f, 0x73, 0x74, 0x72, 0x6f,
-  0x6e, 0x67, 0x3e, 0x20, 0x3c, 0x24, 0x7b, 0x4d, 0x61, 0x72, 0x6b, 0x64,
-  0x6f, 0x77, 0x6e, 0x7d, 0x20, 0x74, 0x65, 0x78, 0x74, 0x3d, 0x24, 0x7b,
-  0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x28, 0x6d, 0x73, 0x67,
-  0x29, 0x7d, 0x20, 0x2f, 0x3e, 0x3c, 0x2f, 0x70, 0x3e, 0x60, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x72, 0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x68, 0x74,
-  0x6d, 0x6c, 0x60, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x3c, 0x73, 0x65, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x20, 0x69, 0x64, 0x3d,
-  0x22, 0x63, 0x68, 0x61, 0x74, 0x22, 0x20, 0x72, 0x65, 0x66, 0x3d, 0x24,
-  0x7b, 0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x7d, 0x3e,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x24,
-  0x7b, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x73, 0x2e, 0x66, 0x6c,
-  0x61, 0x74, 0x4d, 0x61, 0x70, 0x28, 0x63, 0x68, 0x61, 0x74, 0x4c, 0x69,
-  0x6e, 0x65, 0x29, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x3c, 0x2f, 0x73, 0x65, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x3e, 0x60,
-  0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x3b, 0x0a, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x43, 0x6f, 0x6e, 0x66,
-  0x69, 0x67, 0x46, 0x6f, 0x72, 0x6d, 0x20, 0x3d, 0x20, 0x28, 0x70, 0x72,
-  0x6f, 0x70, 0x73, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x75, 0x70,
-  0x64, 0x61, 0x74, 0x65, 0x53, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x20,
-  0x3d, 0x20, 0x28, 0x65, 0x6c, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x73, 0x65,
-  0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20,
-  0x3d, 0x20, 0x7b, 0x20, 0x2e, 0x2e, 0x2e, 0x73, 0x65, 0x73, 0x73, 0x69,
-  0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2c, 0x20, 0x5b, 0x65,
-  0x6c, 0x2e, 0x74, 0x61, 0x72, 0x67, 0x65, 0x74, 0x2e, 0x6e, 0x61, 0x6d,
-  0x65, 0x5d, 0x3a, 0x20, 0x65, 0x6c, 0x2e, 0x74, 0x61, 0x72, 0x67, 0x65,
-  0x74, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x7d, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x75, 0x70,
-  0x64, 0x61, 0x74, 0x65, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x20, 0x3d,
-  0x20, 0x28, 0x65, 0x6c, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x70, 0x61, 0x72,
-  0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d, 0x20,
-  0x7b, 0x20, 0x2e, 0x2e, 0x2e, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e,
+  0x74, 0x2e, 0x73, 0x63, 0x72, 0x6f, 0x6c, 0x6c, 0x54, 0x6f, 0x28, 0x30,
+  0x2c, 0x20, 0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x2e,
+  0x63, 0x75, 0x72, 0x72, 0x65, 0x6e, 0x74, 0x2e, 0x73, 0x63, 0x72, 0x6f,
+  0x6c, 0x6c, 0x48, 0x65, 0x69, 0x67, 0x68, 0x74, 0x29, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x7d, 0x2c, 0x20, 0x5b, 0x6d, 0x65, 0x73, 0x73, 0x61, 0x67,
+  0x65, 0x73, 0x5d, 0x29, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x63, 0x68, 0x61, 0x74, 0x4c, 0x69,
+  0x6e, 0x65, 0x20, 0x3d, 0x20, 0x28, 0x5b, 0x75, 0x73, 0x65, 0x72, 0x2c,
+  0x20, 0x6d, 0x73, 0x67, 0x5d, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x74, 0x75,
+  0x72, 0x6e, 0x20, 0x68, 0x74, 0x6d, 0x6c, 0x60, 0x3c, 0x70, 0x20, 0x6b,
+  0x65, 0x79, 0x3d, 0x24, 0x7b, 0x6d, 0x73, 0x67, 0x7d, 0x3e, 0x3c, 0x73,
+  0x74, 0x72, 0x6f, 0x6e, 0x67, 0x3e, 0x24, 0x7b, 0x74, 0x65, 0x6d, 0x70,
+  0x6c, 0x61, 0x74, 0x65, 0x28, 0x75, 0x73, 0x65, 0x72, 0x29, 0x7d, 0x3a,
+  0x3c, 0x2f, 0x73, 0x74, 0x72, 0x6f, 0x6e, 0x67, 0x3e, 0x20, 0x3c, 0x24,
+  0x7b, 0x4d, 0x61, 0x72, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x69, 0x73, 0x68,
+  0x7d, 0x20, 0x74, 0x65, 0x78, 0x74, 0x3d, 0x24, 0x7b, 0x74, 0x65, 0x6d,
+  0x70, 0x6c, 0x61, 0x74, 0x65, 0x28, 0x6d, 0x73, 0x67, 0x29, 0x7d, 0x20,
+  0x2f, 0x3e, 0x3c, 0x2f, 0x70, 0x3e, 0x60, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x7d, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x72, 0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x68, 0x74, 0x6d, 0x6c, 0x60,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x73, 0x65,
+  0x63, 0x74, 0x69, 0x6f, 0x6e, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x63, 0x68,
+  0x61, 0x74, 0x22, 0x20, 0x72, 0x65, 0x66, 0x3d, 0x24, 0x7b, 0x63, 0x6f,
+  0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x7d, 0x3e, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x24, 0x7b, 0x6d, 0x65,
+  0x73, 0x73, 0x61, 0x67, 0x65, 0x73, 0x2e, 0x66, 0x6c, 0x61, 0x74, 0x4d,
+  0x61, 0x70, 0x28, 0x63, 0x68, 0x61, 0x74, 0x4c, 0x69, 0x6e, 0x65, 0x29,
+  0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f,
+  0x73, 0x65, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x3e, 0x60, 0x3b, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x7d, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63,
+  0x6f, 0x6e, 0x73, 0x74, 0x20, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x46,
+  0x6f, 0x72, 0x6d, 0x20, 0x3d, 0x20, 0x28, 0x70, 0x72, 0x6f, 0x70, 0x73,
+  0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x75, 0x70, 0x64, 0x61, 0x74,
+  0x65, 0x53, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x20, 0x3d, 0x20, 0x28,
+  0x65, 0x6c, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x73, 0x65, 0x73, 0x73, 0x69,
+  0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d, 0x20, 0x7b,
+  0x20, 0x2e, 0x2e, 0x2e, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e,
   0x76, 0x61, 0x6c, 0x75, 0x65, 0x2c, 0x20, 0x5b, 0x65, 0x6c, 0x2e, 0x74,
   0x61, 0x72, 0x67, 0x65, 0x74, 0x2e, 0x6e, 0x61, 0x6d, 0x65, 0x5d, 0x3a,
   0x20, 0x65, 0x6c, 0x2e, 0x74, 0x61, 0x72, 0x67, 0x65, 0x74, 0x2e, 0x76,
   0x61, 0x6c, 0x75, 0x65, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
   0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x75, 0x70, 0x64, 0x61, 0x74,
-  0x65, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x46, 0x6c, 0x6f, 0x61, 0x74,
-  0x20, 0x3d, 0x20, 0x28, 0x65, 0x6c, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x70,
-  0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20,
-  0x3d, 0x20, 0x7b, 0x20, 0x2e, 0x2e, 0x2e, 0x70, 0x61, 0x72, 0x61, 0x6d,
-  0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2c, 0x20, 0x5b, 0x65, 0x6c,
-  0x2e, 0x74, 0x61, 0x72, 0x67, 0x65, 0x74, 0x2e, 0x6e, 0x61, 0x6d, 0x65,
-  0x5d, 0x3a, 0x20, 0x70, 0x61, 0x72, 0x73, 0x65, 0x46, 0x6c, 0x6f, 0x61,
-  0x74, 0x28, 0x65, 0x6c, 0x2e, 0x74, 0x61, 0x72, 0x67, 0x65, 0x74, 0x2e,
-  0x76, 0x61, 0x6c, 0x75, 0x65, 0x29, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x68,
-  0x74, 0x6d, 0x6c, 0x60, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x3c, 0x66, 0x6f, 0x72, 0x6d, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x66, 0x69, 0x65, 0x6c, 0x64,
-  0x73, 0x65, 0x74, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x3c, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d,
-  0x22, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x22, 0x3e, 0x50, 0x72, 0x6f,
-  0x6d, 0x70, 0x74, 0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x3e, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x3c, 0x74, 0x65, 0x78, 0x74, 0x61, 0x72, 0x65, 0x61, 0x20,
-  0x74, 0x79, 0x70, 0x65, 0x3d, 0x22, 0x74, 0x65, 0x78, 0x74, 0x22, 0x20,
-  0x6e, 0x61, 0x6d, 0x65, 0x3d, 0x22, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74,
-  0x22, 0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x73,
-  0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65,
-  0x2e, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x7d, 0x22, 0x20, 0x72, 0x6f,
-  0x77, 0x73, 0x3d, 0x34, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74,
-  0x3d, 0x24, 0x7b, 0x75, 0x70, 0x64, 0x61, 0x74, 0x65, 0x53, 0x65, 0x73,
-  0x73, 0x69, 0x6f, 0x6e, 0x7d, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69,
-  0x76, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x3c, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22,
-  0x75, 0x73, 0x65, 0x72, 0x22, 0x3e, 0x55, 0x73, 0x65, 0x72, 0x20, 0x6e,
-  0x61, 0x6d, 0x65, 0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x3e, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x3c, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x20, 0x74, 0x79, 0x70,
-  0x65, 0x3d, 0x22, 0x74, 0x65, 0x78, 0x74, 0x22, 0x20, 0x6e, 0x61, 0x6d,
-  0x65, 0x3d, 0x22, 0x75, 0x73, 0x65, 0x72, 0x22, 0x20, 0x76, 0x61, 0x6c,
-  0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f,
-  0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x75, 0x73, 0x65, 0x72,
-  0x7d, 0x22, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24,
-  0x7b, 0x75, 0x70, 0x64, 0x61, 0x74, 0x65, 0x53, 0x65, 0x73, 0x73, 0x69,
-  0x6f, 0x6e, 0x7d, 0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76,
-  0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20,
+  0x65, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x20, 0x3d, 0x20, 0x28, 0x65,
+  0x6c, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73,
+  0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d, 0x20, 0x7b, 0x20, 0x2e,
+  0x2e, 0x2e, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c,
+  0x75, 0x65, 0x2c, 0x20, 0x5b, 0x65, 0x6c, 0x2e, 0x74, 0x61, 0x72, 0x67,
+  0x65, 0x74, 0x2e, 0x6e, 0x61, 0x6d, 0x65, 0x5d, 0x3a, 0x20, 0x65, 0x6c,
+  0x2e, 0x74, 0x61, 0x72, 0x67, 0x65, 0x74, 0x2e, 0x76, 0x61, 0x6c, 0x75,
+  0x65, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f,
+  0x6e, 0x73, 0x74, 0x20, 0x75, 0x70, 0x64, 0x61, 0x74, 0x65, 0x50, 0x61,
+  0x72, 0x61, 0x6d, 0x73, 0x46, 0x6c, 0x6f, 0x61, 0x74, 0x20, 0x3d, 0x20,
+  0x28, 0x65, 0x6c, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x70, 0x61, 0x72, 0x61,
+  0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x20, 0x3d, 0x20, 0x7b,
+  0x20, 0x2e, 0x2e, 0x2e, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76,
+  0x61, 0x6c, 0x75, 0x65, 0x2c, 0x20, 0x5b, 0x65, 0x6c, 0x2e, 0x74, 0x61,
+  0x72, 0x67, 0x65, 0x74, 0x2e, 0x6e, 0x61, 0x6d, 0x65, 0x5d, 0x3a, 0x20,
+  0x70, 0x61, 0x72, 0x73, 0x65, 0x46, 0x6c, 0x6f, 0x61, 0x74, 0x28, 0x65,
+  0x6c, 0x2e, 0x74, 0x61, 0x72, 0x67, 0x65, 0x74, 0x2e, 0x76, 0x61, 0x6c,
+  0x75, 0x65, 0x29, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x72, 0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x68, 0x74, 0x6d, 0x6c,
+  0x60, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x66,
+  0x6f, 0x72, 0x6d, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x3c, 0x66, 0x69, 0x65, 0x6c, 0x64, 0x73, 0x65, 0x74,
+  0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x6c,
+  0x61, 0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22, 0x70, 0x72,
+  0x6f, 0x6d, 0x70, 0x74, 0x22, 0x3e, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74,
+  0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x3e, 0x0a, 0x20, 0x20, 0x20,
   0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
-  0x6c, 0x61, 0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22, 0x62,
-  0x6f, 0x74, 0x22, 0x3e, 0x42, 0x6f, 0x74, 0x20, 0x6e, 0x61, 0x6d, 0x65,
+  0x74, 0x65, 0x78, 0x74, 0x61, 0x72, 0x65, 0x61, 0x20, 0x74, 0x79, 0x70,
+  0x65, 0x3d, 0x22, 0x74, 0x65, 0x78, 0x74, 0x22, 0x20, 0x6e, 0x61, 0x6d,
+  0x65, 0x3d, 0x22, 0x70, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x22, 0x20, 0x76,
+  0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x73, 0x65, 0x73, 0x73,
+  0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x70, 0x72,
+  0x6f, 0x6d, 0x70, 0x74, 0x7d, 0x22, 0x20, 0x72, 0x6f, 0x77, 0x73, 0x3d,
+  0x34, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24, 0x7b,
+  0x75, 0x70, 0x64, 0x61, 0x74, 0x65, 0x53, 0x65, 0x73, 0x73, 0x69, 0x6f,
+  0x6e, 0x7d, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x6c, 0x61,
+  0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22, 0x75, 0x73, 0x65,
+  0x72, 0x22, 0x3e, 0x55, 0x73, 0x65, 0x72, 0x20, 0x6e, 0x61, 0x6d, 0x65,
   0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x3e, 0x0a, 0x20, 0x20, 0x20,
   0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
   0x69, 0x6e, 0x70, 0x75, 0x74, 0x20, 0x74, 0x79, 0x70, 0x65, 0x3d, 0x22,
   0x74, 0x65, 0x78, 0x74, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3d, 0x22,
-  0x63, 0x68, 0x61, 0x72, 0x22, 0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d,
+  0x75, 0x73, 0x65, 0x72, 0x22, 0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d,
   0x22, 0x24, 0x7b, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76,
-  0x61, 0x6c, 0x75, 0x65, 0x2e, 0x63, 0x68, 0x61, 0x72, 0x7d, 0x22, 0x20,
+  0x61, 0x6c, 0x75, 0x65, 0x2e, 0x75, 0x73, 0x65, 0x72, 0x7d, 0x22, 0x20,
   0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24, 0x7b, 0x75, 0x70,
   0x64, 0x61, 0x74, 0x65, 0x53, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x7d,
   0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
@@ -600,247 +597,303 @@ unsigned char index_html[] = {
   0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
   0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
   0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x6c, 0x61, 0x62,
-  0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22, 0x74, 0x65, 0x6d, 0x70,
-  0x6c, 0x61, 0x74, 0x65, 0x22, 0x3e, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74,
-  0x20, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x3c, 0x2f, 0x6c,
+  0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22, 0x62, 0x6f, 0x74, 0x22,
+  0x3e, 0x42, 0x6f, 0x74, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3c, 0x2f, 0x6c,
   0x61, 0x62, 0x65, 0x6c, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x74, 0x65, 0x78,
-  0x74, 0x61, 0x72, 0x65, 0x61, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x74, 0x65,
-  0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65,
-  0x3d, 0x22, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x22, 0x20,
-  0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x73, 0x65, 0x73,
-  0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74,
-  0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x7d, 0x22, 0x20, 0x72, 0x6f,
-  0x77, 0x73, 0x3d, 0x34, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74,
-  0x3d, 0x24, 0x7b, 0x75, 0x70, 0x64, 0x61, 0x74, 0x65, 0x53, 0x65, 0x73,
-  0x73, 0x69, 0x6f, 0x6e, 0x7d, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69,
-  0x76, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x3c, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22,
-  0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x22, 0x3e, 0x43, 0x68,
-  0x61, 0x74, 0x20, 0x68, 0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x20, 0x74,
-  0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x3c, 0x2f, 0x6c, 0x61, 0x62,
-  0x65, 0x6c, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x74, 0x65, 0x78, 0x74, 0x61,
-  0x72, 0x65, 0x61, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x74, 0x65, 0x6d, 0x70,
-  0x6c, 0x61, 0x74, 0x65, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3d, 0x22,
-  0x68, 0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x54, 0x65, 0x6d, 0x70, 0x6c,
-  0x61, 0x74, 0x65, 0x22, 0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22,
-  0x24, 0x7b, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61,
-  0x6c, 0x75, 0x65, 0x2e, 0x68, 0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x54,
-  0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x7d, 0x22, 0x20, 0x72, 0x6f,
-  0x77, 0x73, 0x3d, 0x31, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74,
-  0x3d, 0x24, 0x7b, 0x75, 0x70, 0x64, 0x61, 0x74, 0x65, 0x53, 0x65, 0x73,
-  0x73, 0x69, 0x6f, 0x6e, 0x7d, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69,
-  0x76, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x3c, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22,
-  0x74, 0x65, 0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x22,
-  0x3e, 0x54, 0x65, 0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65,
-  0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x3e, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
-  0x69, 0x6e, 0x70, 0x75, 0x74, 0x20, 0x74, 0x79, 0x70, 0x65, 0x3d, 0x22,
-  0x72, 0x61, 0x6e, 0x67, 0x65, 0x22, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x74,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x69, 0x6e, 0x70,
+  0x75, 0x74, 0x20, 0x74, 0x79, 0x70, 0x65, 0x3d, 0x22, 0x74, 0x65, 0x78,
+  0x74, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3d, 0x22, 0x63, 0x68, 0x61,
+  0x72, 0x22, 0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b,
+  0x73, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75,
+  0x65, 0x2e, 0x63, 0x68, 0x61, 0x72, 0x7d, 0x22, 0x20, 0x6f, 0x6e, 0x69,
+  0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24, 0x7b, 0x75, 0x70, 0x64, 0x61, 0x74,
+  0x65, 0x53, 0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x7d, 0x20, 0x2f, 0x3e,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69,
+  0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x20,
+  0x66, 0x6f, 0x72, 0x3d, 0x22, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74,
+  0x65, 0x22, 0x3e, 0x50, 0x72, 0x6f, 0x6d, 0x70, 0x74, 0x20, 0x74, 0x65,
+  0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65,
+  0x6c, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x74, 0x65, 0x78, 0x74, 0x61, 0x72,
+  0x65, 0x61, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x74, 0x65, 0x6d, 0x70, 0x6c,
+  0x61, 0x74, 0x65, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3d, 0x22, 0x74,
+  0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65, 0x22, 0x20, 0x76, 0x61, 0x6c,
+  0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x73, 0x65, 0x73, 0x73, 0x69, 0x6f,
+  0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74, 0x65, 0x6d, 0x70,
+  0x6c, 0x61, 0x74, 0x65, 0x7d, 0x22, 0x20, 0x72, 0x6f, 0x77, 0x73, 0x3d,
+  0x34, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24, 0x7b,
+  0x75, 0x70, 0x64, 0x61, 0x74, 0x65, 0x53, 0x65, 0x73, 0x73, 0x69, 0x6f,
+  0x6e, 0x7d, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x6c, 0x61,
+  0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22, 0x74, 0x65, 0x6d,
+  0x70, 0x6c, 0x61, 0x74, 0x65, 0x22, 0x3e, 0x43, 0x68, 0x61, 0x74, 0x20,
+  0x68, 0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x20, 0x74, 0x65, 0x6d, 0x70,
+  0x6c, 0x61, 0x74, 0x65, 0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x3e,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x3c, 0x74, 0x65, 0x78, 0x74, 0x61, 0x72, 0x65, 0x61,
+  0x20, 0x69, 0x64, 0x3d, 0x22, 0x74, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74,
+  0x65, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3d, 0x22, 0x68, 0x69, 0x73,
+  0x74, 0x6f, 0x72, 0x79, 0x54, 0x65, 0x6d, 0x70, 0x6c, 0x61, 0x74, 0x65,
+  0x22, 0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x73,
+  0x65, 0x73, 0x73, 0x69, 0x6f, 0x6e, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65,
+  0x2e, 0x68, 0x69, 0x73, 0x74, 0x6f, 0x72, 0x79, 0x54, 0x65, 0x6d, 0x70,
+  0x6c, 0x61, 0x74, 0x65, 0x7d, 0x22, 0x20, 0x72, 0x6f, 0x77, 0x73, 0x3d,
+  0x31, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24, 0x7b,
+  0x75, 0x70, 0x64, 0x61, 0x74, 0x65, 0x53, 0x65, 0x73, 0x73, 0x69, 0x6f,
+  0x6e, 0x7d, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x6c, 0x61,
+  0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22, 0x74, 0x65, 0x6d,
+  0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x22, 0x3e, 0x54, 0x65,
+  0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x3c, 0x2f, 0x6c,
+  0x61, 0x62, 0x65, 0x6c, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x69, 0x6e, 0x70,
+  0x75, 0x74, 0x20, 0x74, 0x79, 0x70, 0x65, 0x3d, 0x22, 0x72, 0x61, 0x6e,
+  0x67, 0x65, 0x22, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x74, 0x65, 0x6d, 0x70,
+  0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x22, 0x20, 0x6d, 0x69, 0x6e,
+  0x3d, 0x22, 0x30, 0x2e, 0x30, 0x22, 0x20, 0x6d, 0x61, 0x78, 0x3d, 0x22,
+  0x31, 0x2e, 0x30, 0x22, 0x20, 0x73, 0x74, 0x65, 0x70, 0x3d, 0x22, 0x30,
+  0x2e, 0x30, 0x31, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3d, 0x22, 0x74,
   0x65, 0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x22, 0x20,
-  0x6d, 0x69, 0x6e, 0x3d, 0x22, 0x30, 0x2e, 0x30, 0x22, 0x20, 0x6d, 0x61,
-  0x78, 0x3d, 0x22, 0x31, 0x2e, 0x30, 0x22, 0x20, 0x73, 0x74, 0x65, 0x70,
-  0x3d, 0x22, 0x30, 0x2e, 0x30, 0x31, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65,
-  0x3d, 0x22, 0x74, 0x65, 0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72,
-  0x65, 0x22, 0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b,
-  0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65,
-  0x2e, 0x74, 0x65, 0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65,
-  0x7d, 0x22, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24,
-  0x7b, 0x75, 0x70, 0x64, 0x61, 0x74, 0x65, 0x50, 0x61, 0x72, 0x61, 0x6d,
-  0x73, 0x46, 0x6c, 0x6f, 0x61, 0x74, 0x7d, 0x20, 0x2f, 0x3e, 0x0a, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x3c, 0x73, 0x70, 0x61, 0x6e, 0x3e, 0x24, 0x7b, 0x70, 0x61, 0x72,
+  0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x70, 0x61, 0x72,
   0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74, 0x65,
-  0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x7d, 0x3c, 0x2f,
-  0x73, 0x70, 0x61, 0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e,
-  0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x6c,
-  0x61, 0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22, 0x6e, 0x50,
-  0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x22, 0x3e, 0x50, 0x72, 0x65, 0x64,
-  0x69, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x73, 0x3c, 0x2f, 0x6c, 0x61, 0x62,
-  0x65, 0x6c, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x69, 0x6e, 0x70, 0x75, 0x74,
-  0x20, 0x74, 0x79, 0x70, 0x65, 0x3d, 0x22, 0x72, 0x61, 0x6e, 0x67, 0x65,
-  0x22, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x6e, 0x50, 0x72, 0x65, 0x64, 0x69,
-  0x63, 0x74, 0x22, 0x20, 0x6d, 0x69, 0x6e, 0x3d, 0x22, 0x31, 0x22, 0x20,
-  0x6d, 0x61, 0x78, 0x3d, 0x22, 0x32, 0x30, 0x34, 0x38, 0x22, 0x20, 0x73,
-  0x74, 0x65, 0x70, 0x3d, 0x22, 0x31, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65,
-  0x3d, 0x22, 0x6e, 0x5f, 0x70, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x22,
-  0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x70, 0x61,
-  0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x6e,
-  0x5f, 0x70, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x7d, 0x22, 0x20, 0x6f,
-  0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24, 0x7b, 0x75, 0x70, 0x64,
-  0x61, 0x74, 0x65, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x46, 0x6c, 0x6f,
-  0x61, 0x74, 0x7d, 0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x73, 0x70,
-  0x61, 0x6e, 0x3e, 0x24, 0x7b, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e,
-  0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x6e, 0x5f, 0x70, 0x72, 0x65, 0x64,
-  0x69, 0x63, 0x74, 0x7d, 0x3c, 0x2f, 0x73, 0x70, 0x61, 0x6e, 0x3e, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76,
-  0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x3c, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x20, 0x66,
-  0x6f, 0x72, 0x3d, 0x22, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x70,
-  0x65, 0x6e, 0x61, 0x6c, 0x74, 0x79, 0x22, 0x3e, 0x50, 0x65, 0x6e, 0x61,
-  0x6c, 0x69, 0x7a, 0x65, 0x20, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x20,
-  0x73, 0x65, 0x71, 0x75, 0x65, 0x6e, 0x63, 0x65, 0x3c, 0x2f, 0x6c, 0x61,
-  0x62, 0x65, 0x6c, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x69, 0x6e, 0x70, 0x75,
-  0x74, 0x20, 0x74, 0x79, 0x70, 0x65, 0x3d, 0x22, 0x72, 0x61, 0x6e, 0x67,
-  0x65, 0x22, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x72, 0x65, 0x70, 0x65, 0x61,
-  0x74, 0x5f, 0x70, 0x65, 0x6e, 0x61, 0x6c, 0x74, 0x79, 0x22, 0x20, 0x6d,
-  0x69, 0x6e, 0x3d, 0x22, 0x30, 0x2e, 0x30, 0x22, 0x20, 0x6d, 0x61, 0x78,
-  0x3d, 0x22, 0x32, 0x2e, 0x30, 0x22, 0x20, 0x73, 0x74, 0x65, 0x70, 0x3d,
-  0x22, 0x30, 0x2e, 0x30, 0x31, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3d,
-  0x22, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x70, 0x65, 0x6e, 0x61,
-  0x6c, 0x74, 0x79, 0x22, 0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22,
-  0x24, 0x7b, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c,
-  0x75, 0x65, 0x2e, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x70, 0x65,
-  0x6e, 0x61, 0x6c, 0x74, 0x79, 0x7d, 0x22, 0x20, 0x6f, 0x6e, 0x69, 0x6e,
+  0x6d, 0x70, 0x65, 0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x7d, 0x22, 0x20,
+  0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24, 0x7b, 0x75, 0x70,
+  0x64, 0x61, 0x74, 0x65, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x46, 0x6c,
+  0x6f, 0x61, 0x74, 0x7d, 0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x73,
+  0x70, 0x61, 0x6e, 0x3e, 0x24, 0x7b, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73,
+  0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x74, 0x65, 0x6d, 0x70, 0x65,
+  0x72, 0x61, 0x74, 0x75, 0x72, 0x65, 0x7d, 0x3c, 0x2f, 0x73, 0x70, 0x61,
+  0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
+  0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x6c, 0x61, 0x62, 0x65,
+  0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22, 0x6e, 0x50, 0x72, 0x65, 0x64,
+  0x69, 0x63, 0x74, 0x22, 0x3e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74,
+  0x69, 0x6f, 0x6e, 0x73, 0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x3e,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x3c, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x20, 0x74, 0x79,
+  0x70, 0x65, 0x3d, 0x22, 0x72, 0x61, 0x6e, 0x67, 0x65, 0x22, 0x20, 0x69,
+  0x64, 0x3d, 0x22, 0x6e, 0x50, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x22,
+  0x20, 0x6d, 0x69, 0x6e, 0x3d, 0x22, 0x31, 0x22, 0x20, 0x6d, 0x61, 0x78,
+  0x3d, 0x22, 0x32, 0x30, 0x34, 0x38, 0x22, 0x20, 0x73, 0x74, 0x65, 0x70,
+  0x3d, 0x22, 0x31, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3d, 0x22, 0x6e,
+  0x5f, 0x70, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x22, 0x20, 0x76, 0x61,
+  0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x70, 0x61, 0x72, 0x61, 0x6d,
+  0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x6e, 0x5f, 0x70, 0x72,
+  0x65, 0x64, 0x69, 0x63, 0x74, 0x7d, 0x22, 0x20, 0x6f, 0x6e, 0x69, 0x6e,
   0x70, 0x75, 0x74, 0x3d, 0x24, 0x7b, 0x75, 0x70, 0x64, 0x61, 0x74, 0x65,
   0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x46, 0x6c, 0x6f, 0x61, 0x74, 0x7d,
   0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
   0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x73, 0x70, 0x61, 0x6e, 0x3e,
   0x24, 0x7b, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c,
-  0x75, 0x65, 0x2e, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x70, 0x65,
-  0x6e, 0x61, 0x6c, 0x74, 0x79, 0x7d, 0x3c, 0x2f, 0x73, 0x70, 0x61, 0x6e,
-  0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x64,
-  0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x6c, 0x61, 0x62, 0x65, 0x6c,
-  0x20, 0x66, 0x6f, 0x72, 0x3d, 0x22, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74,
-  0x5f, 0x6c, 0x61, 0x73, 0x74, 0x5f, 0x6e, 0x22, 0x3e, 0x43, 0x6f, 0x6e,
-  0x73, 0x69, 0x64, 0x65, 0x72, 0x20, 0x4e, 0x20, 0x74, 0x6f, 0x6b, 0x65,
-  0x6e, 0x73, 0x20, 0x66, 0x6f, 0x72, 0x20, 0x70, 0x65, 0x6e, 0x61, 0x6c,
-  0x69, 0x7a, 0x65, 0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x3e, 0x0a,
+  0x75, 0x65, 0x2e, 0x6e, 0x5f, 0x70, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74,
+  0x7d, 0x3c, 0x2f, 0x73, 0x70, 0x61, 0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64,
+  0x69, 0x76, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x20,
   0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x3c, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x20, 0x74, 0x79, 0x70,
-  0x65, 0x3d, 0x22, 0x72, 0x61, 0x6e, 0x67, 0x65, 0x22, 0x20, 0x69, 0x64,
-  0x3d, 0x22, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x6c, 0x61, 0x73,
-  0x74, 0x5f, 0x6e, 0x22, 0x20, 0x6d, 0x69, 0x6e, 0x3d, 0x22, 0x30, 0x2e,
-  0x30, 0x22, 0x20, 0x6d, 0x61, 0x78, 0x3d, 0x22, 0x32, 0x30, 0x34, 0x38,
-  0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3d, 0x22, 0x72, 0x65, 0x70, 0x65,
-  0x61, 0x74, 0x5f, 0x6c, 0x61, 0x73, 0x74, 0x5f, 0x6e, 0x22, 0x20, 0x76,
-  0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x70, 0x61, 0x72, 0x61,
-  0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x72, 0x65, 0x70,
-  0x65, 0x61, 0x74, 0x5f, 0x6c, 0x61, 0x73, 0x74, 0x5f, 0x6e, 0x7d, 0x22,
-  0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24, 0x7b, 0x75,
-  0x70, 0x64, 0x61, 0x74, 0x65, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x46,
-  0x6c, 0x6f, 0x61, 0x74, 0x7d, 0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x3c, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f, 0x72, 0x3d,
+  0x22, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x70, 0x65, 0x6e, 0x61,
+  0x6c, 0x74, 0x79, 0x22, 0x3e, 0x50, 0x65, 0x6e, 0x61, 0x6c, 0x69, 0x7a,
+  0x65, 0x20, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x20, 0x73, 0x65, 0x71,
+  0x75, 0x65, 0x6e, 0x63, 0x65, 0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65, 0x6c,
+  0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x3c, 0x69, 0x6e, 0x70, 0x75, 0x74, 0x20, 0x74,
+  0x79, 0x70, 0x65, 0x3d, 0x22, 0x72, 0x61, 0x6e, 0x67, 0x65, 0x22, 0x20,
+  0x69, 0x64, 0x3d, 0x22, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x70,
+  0x65, 0x6e, 0x61, 0x6c, 0x74, 0x79, 0x22, 0x20, 0x6d, 0x69, 0x6e, 0x3d,
+  0x22, 0x30, 0x2e, 0x30, 0x22, 0x20, 0x6d, 0x61, 0x78, 0x3d, 0x22, 0x32,
+  0x2e, 0x30, 0x22, 0x20, 0x73, 0x74, 0x65, 0x70, 0x3d, 0x22, 0x30, 0x2e,
+  0x30, 0x31, 0x22, 0x20, 0x6e, 0x61, 0x6d, 0x65, 0x3d, 0x22, 0x72, 0x65,
+  0x70, 0x65, 0x61, 0x74, 0x5f, 0x70, 0x65, 0x6e, 0x61, 0x6c, 0x74, 0x79,
+  0x22, 0x20, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x3d, 0x22, 0x24, 0x7b, 0x70,
+  0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e,
+  0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x70, 0x65, 0x6e, 0x61, 0x6c,
+  0x74, 0x79, 0x7d, 0x22, 0x20, 0x6f, 0x6e, 0x69, 0x6e, 0x70, 0x75, 0x74,
+  0x3d, 0x24, 0x7b, 0x75, 0x70, 0x64, 0x61, 0x74, 0x65, 0x50, 0x61, 0x72,
+  0x61, 0x6d, 0x73, 0x46, 0x6c, 0x6f, 0x61, 0x74, 0x7d, 0x20, 0x2f, 0x3e,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x3c, 0x73, 0x70, 0x61, 0x6e, 0x3e, 0x24, 0x7b, 0x70,
+  0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e,
+  0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x70, 0x65, 0x6e, 0x61, 0x6c,
+  0x74, 0x79, 0x7d, 0x3c, 0x2f, 0x73, 0x70, 0x61, 0x6e, 0x3e, 0x0a, 0x20,
   0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
-  0x73, 0x70, 0x61, 0x6e, 0x3e, 0x24, 0x7b, 0x70, 0x61, 0x72, 0x61, 0x6d,
-  0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x72, 0x65, 0x70, 0x65,
-  0x61, 0x74, 0x5f, 0x6c, 0x61, 0x73, 0x74, 0x5f, 0x6e, 0x7d, 0x3c, 0x2f,
-  0x73, 0x70, 0x61, 0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e,
-  0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x3c, 0x2f, 0x66, 0x69, 0x65, 0x6c, 0x64, 0x73, 0x65, 0x74, 0x3e, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x66, 0x6f,
-  0x72, 0x6d, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x60, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20,
-  0x4d, 0x61, 0x72, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x20, 0x3d, 0x20, 0x28,
-  0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b,
-  0x0a, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x6d, 0x64, 0x20,
-  0x3d, 0x20, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x74, 0x65, 0x78,
-  0x74, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61,
-  0x63, 0x65, 0x28, 0x2f, 0x5e, 0x23, 0x7b, 0x31, 0x2c, 0x36, 0x7d, 0x20,
-  0x28, 0x2e, 0x2a, 0x29, 0x24, 0x2f, 0x67, 0x69, 0x6d, 0x2c, 0x20, 0x27,
-  0x3c, 0x68, 0x33, 0x3e, 0x24, 0x31, 0x3c, 0x2f, 0x68, 0x33, 0x3e, 0x27,
-  0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61,
-  0x63, 0x65, 0x28, 0x2f, 0x5c, 0x2a, 0x5c, 0x2a, 0x28, 0x2e, 0x2a, 0x3f,
-  0x29, 0x5c, 0x2a, 0x5c, 0x2a, 0x2f, 0x67, 0x2c, 0x20, 0x27, 0x3c, 0x73,
+  0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x3e,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x3c, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x20, 0x66, 0x6f,
+  0x72, 0x3d, 0x22, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x6c, 0x61,
+  0x73, 0x74, 0x5f, 0x6e, 0x22, 0x3e, 0x43, 0x6f, 0x6e, 0x73, 0x69, 0x64,
+  0x65, 0x72, 0x20, 0x4e, 0x20, 0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x73, 0x20,
+  0x66, 0x6f, 0x72, 0x20, 0x70, 0x65, 0x6e, 0x61, 0x6c, 0x69, 0x7a, 0x65,
+  0x3c, 0x2f, 0x6c, 0x61, 0x62, 0x65, 0x6c, 0x3e, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
+  0x69, 0x6e, 0x70, 0x75, 0x74, 0x20, 0x74, 0x79, 0x70, 0x65, 0x3d, 0x22,
+  0x72, 0x61, 0x6e, 0x67, 0x65, 0x22, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x72,
+  0x65, 0x70, 0x65, 0x61, 0x74, 0x5f, 0x6c, 0x61, 0x73, 0x74, 0x5f, 0x6e,
+  0x22, 0x20, 0x6d, 0x69, 0x6e, 0x3d, 0x22, 0x30, 0x2e, 0x30, 0x22, 0x20,
+  0x6d, 0x61, 0x78, 0x3d, 0x22, 0x32, 0x30, 0x34, 0x38, 0x22, 0x20, 0x6e,
+  0x61, 0x6d, 0x65, 0x3d, 0x22, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f,
+  0x6c, 0x61, 0x73, 0x74, 0x5f, 0x6e, 0x22, 0x20, 0x76, 0x61, 0x6c, 0x75,
+  0x65, 0x3d, 0x22, 0x24, 0x7b, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e,
+  0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74,
+  0x5f, 0x6c, 0x61, 0x73, 0x74, 0x5f, 0x6e, 0x7d, 0x22, 0x20, 0x6f, 0x6e,
+  0x69, 0x6e, 0x70, 0x75, 0x74, 0x3d, 0x24, 0x7b, 0x75, 0x70, 0x64, 0x61,
+  0x74, 0x65, 0x50, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x46, 0x6c, 0x6f, 0x61,
+  0x74, 0x7d, 0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x73, 0x70, 0x61,
+  0x6e, 0x3e, 0x24, 0x7b, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x76,
+  0x61, 0x6c, 0x75, 0x65, 0x2e, 0x72, 0x65, 0x70, 0x65, 0x61, 0x74, 0x5f,
+  0x6c, 0x61, 0x73, 0x74, 0x5f, 0x6e, 0x7d, 0x3c, 0x2f, 0x73, 0x70, 0x61,
+  0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x66,
+  0x69, 0x65, 0x6c, 0x64, 0x73, 0x65, 0x74, 0x3e, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x66, 0x6f, 0x72, 0x6d, 0x3e,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x60, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x2f, 0x2f, 0x20, 0x70, 0x6f,
+  0x6f, 0x72, 0x20, 0x6d, 0x61, 0x6e, 0x73, 0x20, 0x6d, 0x61, 0x72, 0x6b,
+  0x64, 0x6f, 0x77, 0x6e, 0x20, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65,
+  0x6d, 0x65, 0x6e, 0x74, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e,
+  0x73, 0x74, 0x20, 0x4d, 0x61, 0x72, 0x6b, 0x64, 0x6f, 0x77, 0x6e, 0x69,
+  0x73, 0x68, 0x20, 0x3d, 0x20, 0x28, 0x70, 0x61, 0x72, 0x61, 0x6d, 0x73,
+  0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x6d, 0x64, 0x20, 0x3d, 0x20,
+  0x70, 0x61, 0x72, 0x61, 0x6d, 0x73, 0x2e, 0x74, 0x65, 0x78, 0x74, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2e, 0x72, 0x65, 0x70,
+  0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x5e, 0x23, 0x7b, 0x31, 0x2c, 0x36,
+  0x7d, 0x20, 0x28, 0x2e, 0x2a, 0x29, 0x24, 0x2f, 0x67, 0x69, 0x6d, 0x2c,
+  0x20, 0x27, 0x3c, 0x68, 0x33, 0x3e, 0x24, 0x31, 0x3c, 0x2f, 0x68, 0x33,
+  0x3e, 0x27, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x5c, 0x2a,
+  0x5c, 0x2a, 0x28, 0x2e, 0x2a, 0x3f, 0x29, 0x5c, 0x2a, 0x5c, 0x2a, 0x2f,
+  0x67, 0x2c, 0x20, 0x27, 0x3c, 0x73, 0x74, 0x72, 0x6f, 0x6e, 0x67, 0x3e,
+  0x24, 0x31, 0x3c, 0x2f, 0x73, 0x74, 0x72, 0x6f, 0x6e, 0x67, 0x3e, 0x27,
+  0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2e, 0x72,
+  0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x5f, 0x5f, 0x28, 0x2e,
+  0x2a, 0x3f, 0x29, 0x5f, 0x5f, 0x2f, 0x67, 0x2c, 0x20, 0x27, 0x3c, 0x73,
   0x74, 0x72, 0x6f, 0x6e, 0x67, 0x3e, 0x24, 0x31, 0x3c, 0x2f, 0x73, 0x74,
   0x72, 0x6f, 0x6e, 0x67, 0x3e, 0x27, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x5f, 0x5f,
-  0x28, 0x2e, 0x2a, 0x3f, 0x29, 0x5f, 0x5f, 0x2f, 0x67, 0x2c, 0x20, 0x27,
-  0x3c, 0x73, 0x74, 0x72, 0x6f, 0x6e, 0x67, 0x3e, 0x24, 0x31, 0x3c, 0x2f,
-  0x73, 0x74, 0x72, 0x6f, 0x6e, 0x67, 0x3e, 0x27, 0x29, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65,
+  0x28, 0x2f, 0x5c, 0x2a, 0x28, 0x2e, 0x2a, 0x3f, 0x29, 0x5c, 0x2a, 0x2f,
+  0x67, 0x2c, 0x20, 0x27, 0x3c, 0x65, 0x6d, 0x3e, 0x24, 0x31, 0x3c, 0x2f,
+  0x65, 0x6d, 0x3e, 0x27, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
   0x20, 0x20, 0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f,
-  0x5c, 0x2a, 0x28, 0x2e, 0x2a, 0x3f, 0x29, 0x5c, 0x2a, 0x2f, 0x67, 0x2c,
-  0x20, 0x27, 0x3c, 0x65, 0x6d, 0x3e, 0x24, 0x31, 0x3c, 0x2f, 0x65, 0x6d,
-  0x3e, 0x27, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x2e, 0x72, 0x65, 0x70,
-  0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x5f, 0x28, 0x2e, 0x2a, 0x3f, 0x29,
-  0x5f, 0x2f, 0x67, 0x2c, 0x20, 0x27, 0x3c, 0x65, 0x6d, 0x3e, 0x24, 0x31,
-  0x3c, 0x2f, 0x65, 0x6d, 0x3e, 0x27, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x60, 0x60,
-  0x60, 0x2e, 0x2a, 0x3f, 0x5c, 0x6e, 0x28, 0x5b, 0x5c, 0x73, 0x5c, 0x53,
-  0x5d, 0x2a, 0x3f, 0x29, 0x60, 0x60, 0x60, 0x2f, 0x67, 0x2c, 0x20, 0x27,
-  0x3c, 0x70, 0x72, 0x65, 0x3e, 0x3c, 0x63, 0x6f, 0x64, 0x65, 0x3e, 0x24,
-  0x31, 0x3c, 0x2f, 0x63, 0x6f, 0x64, 0x65, 0x3e, 0x3c, 0x2f, 0x70, 0x72,
-  0x65, 0x3e, 0x27, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x2e, 0x72, 0x65,
-  0x70, 0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x60, 0x28, 0x2e, 0x2a, 0x3f,
-  0x29, 0x60, 0x2f, 0x67, 0x2c, 0x20, 0x27, 0x3c, 0x63, 0x6f, 0x64, 0x65,
-  0x3e, 0x24, 0x31, 0x3c, 0x2f, 0x63, 0x6f, 0x64, 0x65, 0x3e, 0x27, 0x29,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x2e, 0x72, 0x65, 0x70, 0x6c, 0x61, 0x63,
-  0x65, 0x28, 0x2f, 0x5c, 0x6e, 0x2f, 0x67, 0x69, 0x6d, 0x2c, 0x20, 0x27,
-  0x3c, 0x62, 0x72, 0x20, 0x2f, 0x3e, 0x27, 0x29, 0x3b, 0x0a, 0x20, 0x20,
+  0x5f, 0x28, 0x2e, 0x2a, 0x3f, 0x29, 0x5f, 0x2f, 0x67, 0x2c, 0x20, 0x27,
+  0x3c, 0x65, 0x6d, 0x3e, 0x24, 0x31, 0x3c, 0x2f, 0x65, 0x6d, 0x3e, 0x27,
+  0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2e, 0x72,
+  0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x60, 0x60, 0x60, 0x2e,
+  0x2a, 0x3f, 0x5c, 0x6e, 0x28, 0x5b, 0x5c, 0x73, 0x5c, 0x53, 0x5d, 0x2a,
+  0x3f, 0x29, 0x60, 0x60, 0x60, 0x2f, 0x67, 0x2c, 0x20, 0x27, 0x3c, 0x70,
+  0x72, 0x65, 0x3e, 0x3c, 0x63, 0x6f, 0x64, 0x65, 0x3e, 0x24, 0x31, 0x3c,
+  0x2f, 0x63, 0x6f, 0x64, 0x65, 0x3e, 0x3c, 0x2f, 0x70, 0x72, 0x65, 0x3e,
+  0x27, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2e,
+  0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x60, 0x28, 0x2e,
+  0x2a, 0x3f, 0x29, 0x60, 0x2f, 0x67, 0x2c, 0x20, 0x27, 0x3c, 0x63, 0x6f,
+  0x64, 0x65, 0x3e, 0x24, 0x31, 0x3c, 0x2f, 0x63, 0x6f, 0x64, 0x65, 0x3e,
+  0x27, 0x29, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x2e,
+  0x72, 0x65, 0x70, 0x6c, 0x61, 0x63, 0x65, 0x28, 0x2f, 0x5c, 0x6e, 0x2f,
+  0x67, 0x69, 0x6d, 0x2c, 0x20, 0x27, 0x3c, 0x62, 0x72, 0x20, 0x2f, 0x3e,
+  0x27, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65,
+  0x74, 0x75, 0x72, 0x6e, 0x20, 0x68, 0x74, 0x6d, 0x6c, 0x60, 0x3c, 0x73,
+  0x70, 0x61, 0x6e, 0x20, 0x64, 0x61, 0x6e, 0x67, 0x65, 0x72, 0x6f, 0x75,
+  0x73, 0x6c, 0x79, 0x53, 0x65, 0x74, 0x49, 0x6e, 0x6e, 0x65, 0x72, 0x48,
+  0x54, 0x4d, 0x4c, 0x3d, 0x24, 0x7b, 0x7b, 0x20, 0x5f, 0x5f, 0x68, 0x74,
+  0x6d, 0x6c, 0x3a, 0x20, 0x6d, 0x64, 0x20, 0x7d, 0x7d, 0x20, 0x2f, 0x3e,
+  0x60, 0x3b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x3b, 0x0a, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x63, 0x6f, 0x6e, 0x73, 0x74, 0x20, 0x4d, 0x6f, 0x64,
+  0x65, 0x6c, 0x47, 0x65, 0x6e, 0x65, 0x72, 0x61, 0x74, 0x69, 0x6f, 0x6e,
+  0x49, 0x6e, 0x66, 0x6f, 0x20, 0x3d, 0x20, 0x28, 0x70, 0x61, 0x72, 0x61,
+  0x6d, 0x73, 0x29, 0x20, 0x3d, 0x3e, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x69, 0x66, 0x20, 0x28, 0x21, 0x6c, 0x6c, 0x61, 0x6d,
+  0x61, 0x53, 0x74, 0x61, 0x74, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65,
+  0x29, 0x20, 0x7b, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
   0x72, 0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x68, 0x74, 0x6d, 0x6c, 0x60,
-  0x3c, 0x73, 0x70, 0x61, 0x6e, 0x20, 0x64, 0x61, 0x6e, 0x67, 0x65, 0x72,
-  0x6f, 0x75, 0x73, 0x6c, 0x79, 0x53, 0x65, 0x74, 0x49, 0x6e, 0x6e, 0x65,
-  0x72, 0x48, 0x54, 0x4d, 0x4c, 0x3d, 0x24, 0x7b, 0x7b, 0x20, 0x5f, 0x5f,
-  0x68, 0x74, 0x6d, 0x6c, 0x3a, 0x20, 0x6d, 0x64, 0x20, 0x7d, 0x7d, 0x20,
-  0x2f, 0x3e, 0x60, 0x3b, 0x0a, 0x7d, 0x3b, 0x0a, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x66, 0x75, 0x6e, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x20, 0x41, 0x70,
-  0x70, 0x28, 0x70, 0x72, 0x6f, 0x70, 0x73, 0x29, 0x20, 0x7b, 0x0a, 0x0a,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x74, 0x75, 0x72, 0x6e,
-  0x20, 0x68, 0x74, 0x6d, 0x6c, 0x60, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x3c, 0x64, 0x69, 0x76, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x63, 0x6f,
-  0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72, 0x22, 0x3e, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x68, 0x65, 0x61, 0x64, 0x65,
-  0x72, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x3c, 0x68, 0x31, 0x3e, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x2e, 0x63,
-  0x70, 0x70, 0x3c, 0x2f, 0x68, 0x31, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x68, 0x65, 0x61, 0x64, 0x65, 0x72,
-  0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
-  0x6d, 0x61, 0x69, 0x6e, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x63, 0x6f, 0x6e,
-  0x74, 0x65, 0x6e, 0x74, 0x22, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x24, 0x7b, 0x63, 0x68, 0x61, 0x74,
-  0x53, 0x74, 0x61, 0x72, 0x74, 0x65, 0x64, 0x2e, 0x76, 0x61, 0x6c, 0x75,
-  0x65, 0x20, 0x3f, 0x20, 0x43, 0x68, 0x61, 0x74, 0x4c, 0x6f, 0x67, 0x20,
-  0x3a, 0x20, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x46, 0x6f, 0x72, 0x6d,
-  0x7d, 0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x3c, 0x2f, 0x6d, 0x61, 0x69, 0x6e, 0x3e, 0x0a, 0x0a, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x66, 0x6f, 0x6f, 0x74, 0x65,
-  0x72, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x77, 0x72, 0x69, 0x74, 0x65, 0x22,
-  0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
-  0x3c, 0x24, 0x7b, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x49, 0x6e,
-  0x70, 0x75, 0x74, 0x7d, 0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x66, 0x6f, 0x6f, 0x74, 0x65, 0x72,
-  0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
-  0x66, 0x6f, 0x6f, 0x74, 0x65, 0x72, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x70, 0x3e, 0x50, 0x6f, 0x77,
-  0x65, 0x72, 0x65, 0x64, 0x20, 0x62, 0x79, 0x20, 0x3c, 0x61, 0x20, 0x68,
-  0x72, 0x65, 0x66, 0x3d, 0x22, 0x68, 0x74, 0x74, 0x70, 0x73, 0x3a, 0x2f,
-  0x2f, 0x67, 0x69, 0x74, 0x68, 0x75, 0x62, 0x2e, 0x63, 0x6f, 0x6d, 0x2f,
-  0x67, 0x67, 0x65, 0x72, 0x67, 0x61, 0x6e, 0x6f, 0x76, 0x2f, 0x6c, 0x6c,
-  0x61, 0x6d, 0x61, 0x2e, 0x63, 0x70, 0x70, 0x22, 0x3e, 0x6c, 0x6c, 0x61,
-  0x6d, 0x61, 0x2e, 0x63, 0x70, 0x70, 0x3c, 0x2f, 0x61, 0x3e, 0x20, 0x61,
-  0x6e, 0x64, 0x20, 0x3c, 0x61, 0x20, 0x68, 0x72, 0x65, 0x66, 0x3d, 0x22,
-  0x68, 0x74, 0x74, 0x70, 0x73, 0x3a, 0x2f, 0x2f, 0x67, 0x67, 0x6d, 0x6c,
-  0x2e, 0x61, 0x69, 0x22, 0x3e, 0x67, 0x67, 0x6d, 0x6c, 0x2e, 0x61, 0x69,
-  0x3c, 0x2f, 0x61, 0x3e, 0x3c, 0x2f, 0x70, 0x3e, 0x0a, 0x20, 0x20, 0x20,
-  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x66, 0x6f, 0x6f, 0x74, 0x65,
-  0x72, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64,
-  0x69, 0x76, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x60, 0x3b,
-  0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20,
-  0x72, 0x65, 0x6e, 0x64, 0x65, 0x72, 0x28, 0x68, 0x28, 0x41, 0x70, 0x70,
-  0x29, 0x2c, 0x20, 0x64, 0x6f, 0x63, 0x75, 0x6d, 0x65, 0x6e, 0x74, 0x2e,
-  0x62, 0x6f, 0x64, 0x79, 0x29, 0x3b, 0x0a, 0x20, 0x20, 0x3c, 0x2f, 0x73,
-  0x63, 0x72, 0x69, 0x70, 0x74, 0x3e, 0x0a, 0x3c, 0x2f, 0x68, 0x65, 0x61,
-  0x64, 0x3e, 0x0a, 0x0a, 0x3c, 0x62, 0x6f, 0x64, 0x79, 0x3e, 0x0a, 0x3c,
-  0x2f, 0x62, 0x6f, 0x64, 0x79, 0x3e, 0x0a, 0x0a, 0x3c, 0x2f, 0x68, 0x74,
-  0x6d, 0x6c, 0x3e, 0x0a
+  0x3c, 0x73, 0x70, 0x61, 0x6e, 0x2f, 0x3e, 0x60, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x7d, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72,
+  0x65, 0x74, 0x75, 0x72, 0x6e, 0x20, 0x68, 0x74, 0x6d, 0x6c, 0x60, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x73, 0x70, 0x61,
+  0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x24, 0x7b, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x53, 0x74, 0x61, 0x74,
+  0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75, 0x65, 0x2e, 0x70, 0x72, 0x65, 0x64,
+  0x69, 0x63, 0x74, 0x65, 0x64, 0x5f, 0x70, 0x65, 0x72, 0x5f, 0x74, 0x6f,
+  0x6b, 0x65, 0x6e, 0x5f, 0x6d, 0x73, 0x2e, 0x74, 0x6f, 0x46, 0x69, 0x78,
+  0x65, 0x64, 0x28, 0x29, 0x7d, 0x6d, 0x73, 0x20, 0x70, 0x65, 0x72, 0x20,
+  0x74, 0x6f, 0x6b, 0x65, 0x6e, 0x2c, 0x20, 0x24, 0x7b, 0x6c, 0x6c, 0x61,
+  0x6d, 0x61, 0x53, 0x74, 0x61, 0x74, 0x73, 0x2e, 0x76, 0x61, 0x6c, 0x75,
+  0x65, 0x2e, 0x70, 0x72, 0x65, 0x64, 0x69, 0x63, 0x74, 0x65, 0x64, 0x5f,
+  0x70, 0x65, 0x72, 0x5f, 0x73, 0x65, 0x63, 0x6f, 0x6e, 0x64, 0x2e, 0x74,
+  0x6f, 0x46, 0x69, 0x78, 0x65, 0x64, 0x28, 0x32, 0x29, 0x7d, 0x20, 0x74,
+  0x6f, 0x6b, 0x65, 0x6e, 0x73, 0x20, 0x70, 0x65, 0x72, 0x20, 0x73, 0x65,
+  0x63, 0x6f, 0x6e, 0x64, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x3c, 0x2f, 0x73, 0x70, 0x61, 0x6e, 0x3e, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x60, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x7d, 0x0a, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x66, 0x75, 0x6e, 0x63, 0x74, 0x69, 0x6f, 0x6e,
+  0x20, 0x41, 0x70, 0x70, 0x28, 0x70, 0x72, 0x6f, 0x70, 0x73, 0x29, 0x20,
+  0x7b, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x74,
+  0x75, 0x72, 0x6e, 0x20, 0x68, 0x74, 0x6d, 0x6c, 0x60, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x64, 0x69, 0x76, 0x20, 0x69,
+  0x64, 0x3d, 0x22, 0x63, 0x6f, 0x6e, 0x74, 0x61, 0x69, 0x6e, 0x65, 0x72,
+  0x22, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x3c, 0x68, 0x65, 0x61, 0x64, 0x65, 0x72, 0x3e, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x68,
+  0x31, 0x3e, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x2e, 0x63, 0x70, 0x70, 0x3c,
+  0x2f, 0x68, 0x31, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x3c, 0x2f, 0x68, 0x65, 0x61, 0x64, 0x65, 0x72, 0x3e,
+  0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x3c, 0x6d, 0x61, 0x69, 0x6e, 0x20, 0x69, 0x64, 0x3d, 0x22, 0x63, 0x6f,
+  0x6e, 0x74, 0x65, 0x6e, 0x74, 0x22, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x24, 0x7b, 0x63,
+  0x68, 0x61, 0x74, 0x53, 0x74, 0x61, 0x72, 0x74, 0x65, 0x64, 0x2e, 0x76,
+  0x61, 0x6c, 0x75, 0x65, 0x20, 0x3f, 0x20, 0x43, 0x68, 0x61, 0x74, 0x4c,
+  0x6f, 0x67, 0x20, 0x3a, 0x20, 0x43, 0x6f, 0x6e, 0x66, 0x69, 0x67, 0x46,
+  0x6f, 0x72, 0x6d, 0x7d, 0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x6d, 0x61, 0x69, 0x6e,
+  0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x3c, 0x73, 0x65, 0x63, 0x74, 0x69, 0x6f, 0x6e, 0x20, 0x69, 0x64,
+  0x3d, 0x22, 0x77, 0x72, 0x69, 0x74, 0x65, 0x22, 0x3e, 0x0a, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x24,
+  0x7b, 0x4d, 0x65, 0x73, 0x73, 0x61, 0x67, 0x65, 0x49, 0x6e, 0x70, 0x75,
+  0x74, 0x7d, 0x20, 0x2f, 0x3e, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x73, 0x65, 0x63, 0x74, 0x69, 0x6f,
+  0x6e, 0x3e, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x3c, 0x66, 0x6f, 0x6f, 0x74, 0x65, 0x72, 0x3e, 0x0a, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
+  0x70, 0x3e, 0x3c, 0x24, 0x7b, 0x4d, 0x6f, 0x64, 0x65, 0x6c, 0x47, 0x65,
+  0x6e, 0x65, 0x72, 0x61, 0x74, 0x69, 0x6f, 0x6e, 0x49, 0x6e, 0x66, 0x6f,
+  0x7d, 0x20, 0x2f, 0x3e, 0x3c, 0x2f, 0x70, 0x3e, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x70, 0x3e,
+  0x50, 0x6f, 0x77, 0x65, 0x72, 0x65, 0x64, 0x20, 0x62, 0x79, 0x20, 0x3c,
+  0x61, 0x20, 0x68, 0x72, 0x65, 0x66, 0x3d, 0x22, 0x68, 0x74, 0x74, 0x70,
+  0x73, 0x3a, 0x2f, 0x2f, 0x67, 0x69, 0x74, 0x68, 0x75, 0x62, 0x2e, 0x63,
+  0x6f, 0x6d, 0x2f, 0x67, 0x67, 0x65, 0x72, 0x67, 0x61, 0x6e, 0x6f, 0x76,
+  0x2f, 0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x2e, 0x63, 0x70, 0x70, 0x22, 0x3e,
+  0x6c, 0x6c, 0x61, 0x6d, 0x61, 0x2e, 0x63, 0x70, 0x70, 0x3c, 0x2f, 0x61,
+  0x3e, 0x20, 0x61, 0x6e, 0x64, 0x20, 0x3c, 0x61, 0x20, 0x68, 0x72, 0x65,
+  0x66, 0x3d, 0x22, 0x68, 0x74, 0x74, 0x70, 0x73, 0x3a, 0x2f, 0x2f, 0x67,
+  0x67, 0x6d, 0x6c, 0x2e, 0x61, 0x69, 0x22, 0x3e, 0x67, 0x67, 0x6d, 0x6c,
+  0x2e, 0x61, 0x69, 0x3c, 0x2f, 0x61, 0x3e, 0x2e, 0x3c, 0x2f, 0x70, 0x3e,
+  0x0a, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x3c,
+  0x2f, 0x66, 0x6f, 0x6f, 0x74, 0x65, 0x72, 0x3e, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x3c, 0x2f, 0x64, 0x69, 0x76, 0x3e, 0x0a,
+  0x20, 0x20, 0x20, 0x20, 0x20, 0x20, 0x60, 0x3b, 0x0a, 0x20, 0x20, 0x20,
+  0x20, 0x7d, 0x0a, 0x0a, 0x20, 0x20, 0x20, 0x20, 0x72, 0x65, 0x6e, 0x64,
+  0x65, 0x72, 0x28, 0x68, 0x28, 0x41, 0x70, 0x70, 0x29, 0x2c, 0x20, 0x64,
+  0x6f, 0x63, 0x75, 0x6d, 0x65, 0x6e, 0x74, 0x2e, 0x62, 0x6f, 0x64, 0x79,
+  0x29, 0x3b, 0x0a, 0x20, 0x20, 0x3c, 0x2f, 0x73, 0x63, 0x72, 0x69, 0x70,
+  0x74, 0x3e, 0x0a, 0x3c, 0x2f, 0x68, 0x65, 0x61, 0x64, 0x3e, 0x0a, 0x0a,
+  0x3c, 0x62, 0x6f, 0x64, 0x79, 0x3e, 0x0a, 0x3c, 0x2f, 0x62, 0x6f, 0x64,
+  0x79, 0x3e, 0x0a, 0x0a, 0x3c, 0x2f, 0x68, 0x74, 0x6d, 0x6c, 0x3e, 0x0a
 };
-unsigned int index_html_len = 10108;
+unsigned int index_html_len = 10752;
diff --git a/examples/server/public/completion.js b/examples/server/public/completion.js
index 4f5005cfb92e1..a43d5a7d56753 100644
--- a/examples/server/public/completion.js
+++ b/examples/server/public/completion.js
@@ -5,20 +5,29 @@ const paramDefaults = {
   stop: ["</s>"]
 };
 
-/**
- * This function completes the input text using a llama dictionary.
- * @param {object} params - The parameters for the completion request.
- * @param {object} controller - an instance of AbortController if you need one, or null.
- * @param {function} callback - The callback function to call when the completion is done.
- * @returns {string} the completed text as a string. Ideally ignored, and you get at it via the callback.
- */
-export const llamaComplete = async (params, controller, callback) => {
+let generation_settings = null;
+
+
+// Completes the prompt as a generator. Recommended for most use cases.
+//
+// Example:
+//
+//    import { llama } from '/completion.js'
+//
+//    const request = llama("Tell me a joke", {n_predict: 800})
+//    for await (const chunk of request) {
+//      document.write(chunk.data.content)
+//    }
+//
+export async function* llama(prompt, params = {}, config = {}) {
+  let controller = config.controller;
+
   if (!controller) {
     controller = new AbortController();
   }
-  const completionParams = { ...paramDefaults, ...params };
 
-  // we use fetch directly here becasue the built in fetchEventSource does not support POST
+  const completionParams = { ...paramDefaults, ...params, prompt };
+
   const response = await fetch("/completion", {
     method: 'POST',
     body: JSON.stringify(completionParams),
@@ -36,7 +45,6 @@ export const llamaComplete = async (params, controller, callback) => {
   let content = "";
 
   try {
-
     let cont = true;
 
     while (cont) {
@@ -59,18 +67,21 @@ export const llamaComplete = async (params, controller, callback) => {
       result.data = JSON.parse(result.data);
       content += result.data.content;
 
-      // callack
-      if (callback) {
-        cont = callback(result) != false;
-      }
+      // yield
+      yield result;
 
       // if we got a stop token from server, we will break here
       if (result.data.stop) {
+        if (result.data.generation_settings) {
+          generation_settings = result.data.generation_settings;
+        }
         break;
       }
     }
   } catch (e) {
-    console.error("llama error: ", e);
+    if (e.name !== 'AbortError') {
+      console.error("llama error: ", e);
+    }
     throw e;
   }
   finally {
@@ -79,3 +90,79 @@ export const llamaComplete = async (params, controller, callback) => {
 
   return content;
 }
+
+// Call llama, return an event target that you can subcribe to
+//
+// Example:
+//
+//    import { llamaEventTarget } from '/completion.js'
+//
+//    const conn = llamaEventTarget(prompt)
+//    conn.addEventListener("message", (chunk) => {
+//      document.write(chunk.detail.content)
+//    })
+//
+export const llamaEventTarget = (prompt, params = {}, config = {}) => {
+  const eventTarget = new EventTarget();
+  (async () => {
+    let content = "";
+    for await (const chunk of llama(prompt, params, config)) {
+      if (chunk.data) {
+        content += chunk.data.content;
+        eventTarget.dispatchEvent(new CustomEvent("message", { detail: chunk.data }));
+      }
+      if (chunk.data.generation_settings) {
+        eventTarget.dispatchEvent(new CustomEvent("generation_settings", { detail: chunk.data.generation_settings }));
+      }
+      if (chunk.data.timings) {
+        eventTarget.dispatchEvent(new CustomEvent("timings", { detail: chunk.data.timings }));
+      }
+    }
+    eventTarget.dispatchEvent(new CustomEvent("done", { detail: { content } }));
+  })();
+  return eventTarget;
+}
+
+// Call llama, return a promise that resolves to the completed text. This does not support streaming
+//
+// Example:
+//
+//     llamaPromise(prompt).then((content) => {
+//       document.write(content)
+//     })
+//
+//     or
+//
+//     const content = await llamaPromise(prompt)
+//     document.write(content)
+//
+export const llamaPromise = (prompt, params = {}, config = {}) => {
+  return new Promise(async (resolve, reject) => {
+    let content = "";
+    try {
+      for await (const chunk of llama(prompt, params, config)) {
+        content += chunk.data.content;
+      }
+      resolve(content);
+    } catch (error) {
+      reject(error);
+    }
+  });
+};
+
+/**
+ * (deprecated)
+ */
+export const llamaComplete = async (params, controller, callback) => {
+  for await (const chunk of llama(params.prompt, params, { controller })) {
+    callback(chunk);
+  }
+}
+
+// Get the model info from the server. This is useful for getting the context window and so on.
+export const llamaModelInfo = async () => {
+  if (!generation_settings) {
+    generation_settings = await fetch("/model.json").then(r => r.json());
+  }
+  return generation_settings;
+}
diff --git a/examples/server/public/index.html b/examples/server/public/index.html
index 6393e2e75cf19..8ace0b0affe58 100644
--- a/examples/server/public/index.html
+++ b/examples/server/public/index.html
@@ -6,7 +6,6 @@
   <title>llama.cpp - chat</title>
 
   <style>
-
     body {
       background-color: #fff;
       color: #000;
@@ -22,10 +21,6 @@
       height: 100%;
     }
 
-    header, footer {
-      text-align: center;
-    }
-
     main {
       margin: 3px;
       display: flex;
@@ -99,6 +94,15 @@
       margin: 0.5em 0;
       display: block;
     }
+
+    header, footer {
+      text-align: center;
+    }
+
+    footer {
+      font-size: 80%;
+      color: #888;
+    }
   </style>
 
   <script type="module">
@@ -106,10 +110,10 @@
       html, h, signal, effect, computed, render, useSignal, useEffect, useRef
     } from '/index.js';
 
-    import { llamaComplete } from '/completion.js';
+    import { llama } from '/completion.js';
 
     const session = signal({
-      prompt: "This is a conversation between user and llama, a friendly chatbot. respond in markdown.",
+      prompt: "This is a conversation between user and llama, a friendly chatbot. respond in simple markdown.",
       template: "{{prompt}}\n\n{{history}}\n{{char}}:",
       historyTemplate: "{{name}}: {{message}}",
       transcript: [],
@@ -118,15 +122,6 @@
       user: "User",
     })
 
-    const transcriptUpdate = (transcript) => {
-      session.value = {
-        ...session.value,
-        transcript
-      }
-    }
-
-    const chatStarted = computed(() => session.value.transcript.length > 0)
-
     const params = signal({
       n_predict: 400,
       temperature: 0.7,
@@ -136,8 +131,18 @@
       top_p: 0.5,
     })
 
+    const llamaStats = signal(null)
     const controller = signal(null)
+
     const generating = computed(() => controller.value == null )
+    const chatStarted = computed(() => session.value.transcript.length > 0)
+
+    const transcriptUpdate = (transcript) => {
+      session.value = {
+        ...session.value,
+        transcript
+      }
+    }
 
     // simple template replace
     const template = (str, extraSettings) => {
@@ -158,7 +163,7 @@
 
       transcriptUpdate([...session.value.transcript, ["{{user}}", msg]])
 
-      const payload = template(session.value.template, {
+      const prompt = template(session.value.template, {
         message: msg,
         history: session.value.transcript.flatMap(([name, message]) => template(session.value.historyTemplate, {name, message})).join("\n"),
       });
@@ -168,22 +173,26 @@
 
       const llamaParams = {
         ...params.value,
-        prompt: payload,
         stop: ["</s>", template("{{char}}:"), template("{{user}}:")],
       }
 
-      await llamaComplete(llamaParams, controller.value, (message) => {
-        const data = message.data;
+      for await (const chunk of llama(prompt, llamaParams, { controller: controller.value })) {
+        const data = chunk.data;
         currentMessage += data.content;
+
         // remove leading whitespace
         currentMessage = currentMessage.replace(/^\s+/, "")
 
         transcriptUpdate([...history, ["{{char}}", currentMessage]])
 
         if (data.stop) {
-          console.log("-->", data, ' response was:', currentMessage, 'transcript state:', session.value.transcript);
+          console.log("Completion finished: '", currentMessage, "', summary: ", data);
         }
-      })
+
+        if (data.timings) {
+          llamaStats.value = data.timings;
+        }
+      }
 
       controller.value = null;
     }
@@ -219,13 +228,12 @@
       return html`
         <form onsubmit=${submit}>
           <div>
-          <textarea type="text" rows=2 onkeypress=${enterSubmits} value="${message}" oninput=${(e) => message.value = e.target.value} placeholder="Say something..."/>
-
+            <textarea type="text" rows=2 onkeypress=${enterSubmits} value="${message}" oninput=${(e) => message.value = e.target.value} placeholder="Say something..."/>
           </div>
           <div class="right">
-          <button type="submit" disabled=${!generating.value} >Send</button>
-          <button onclick=${stop} disabled=${generating}>Stop</button>
-          <button onclick=${reset}>Reset</button>
+            <button type="submit" disabled=${!generating.value} >Send</button>
+            <button onclick=${stop} disabled=${generating}>Stop</button>
+            <button onclick=${reset}>Reset</button>
           </div>
         </form>
       `
@@ -243,7 +251,7 @@
       }, [messages])
 
       const chatLine = ([user, msg]) => {
-        return html`<p key=${msg}><strong>${template(user)}:</strong> <${Markdown} text=${template(msg)} /></p>`
+        return html`<p key=${msg}><strong>${template(user)}:</strong> <${Markdownish} text=${template(msg)} /></p>`
       };
 
       return html`
@@ -313,39 +321,52 @@
         </form>
       `
     }
-const Markdown = (params) => {
-  const md = params.text
-    .replace(/^#{1,6} (.*)$/gim, '<h3>$1</h3>')
-    .replace(/\*\*(.*?)\*\*/g, '<strong>$1</strong>')
-    .replace(/__(.*?)__/g, '<strong>$1</strong>')
-    .replace(/\*(.*?)\*/g, '<em>$1</em>')
-    .replace(/_(.*?)_/g, '<em>$1</em>')
-    .replace(/```.*?\n([\s\S]*?)```/g, '<pre><code>$1</code></pre>')
-    .replace(/`(.*?)`/g, '<code>$1</code>')
-    .replace(/\n/gim, '<br />');
-  return html`<span dangerouslySetInnerHTML=${{ __html: md }} />`;
-};
+    // poor mans markdown replacement
+    const Markdownish = (params) => {
+      const md = params.text
+        .replace(/^#{1,6} (.*)$/gim, '<h3>$1</h3>')
+        .replace(/\*\*(.*?)\*\*/g, '<strong>$1</strong>')
+        .replace(/__(.*?)__/g, '<strong>$1</strong>')
+        .replace(/\*(.*?)\*/g, '<em>$1</em>')
+        .replace(/_(.*?)_/g, '<em>$1</em>')
+        .replace(/```.*?\n([\s\S]*?)```/g, '<pre><code>$1</code></pre>')
+        .replace(/`(.*?)`/g, '<code>$1</code>')
+        .replace(/\n/gim, '<br />');
+      return html`<span dangerouslySetInnerHTML=${{ __html: md }} />`;
+    };
+
+    const ModelGenerationInfo = (params) => {
+      if (!llamaStats.value) {
+        return html`<span/>`
+      }
+      return html`
+        <span>
+          ${llamaStats.value.predicted_per_token_ms.toFixed()}ms per token, ${llamaStats.value.predicted_per_second.toFixed(2)} tokens per second
+        </span>
+      `
+    }
 
     function App(props) {
 
       return html`
-      <div id="container">
-        <header>
-          <h1>llama.cpp</h1>
-        </header>
-
-        <main id="content">
-          <${chatStarted.value ? ChatLog : ConfigForm} />
-        </main>
-
-        <footer id="write">
-          <${MessageInput} />
-        </footer>
-
-        <footer>
-          <p>Powered by <a href="/ggerganov/llama.cpp">llama.cpp</a> and <a href="https://ggml.ai">ggml.ai</a></p>
-        </footer>
-      </div>
+        <div id="container">
+          <header>
+            <h1>llama.cpp</h1>
+          </header>
+
+          <main id="content">
+            <${chatStarted.value ? ChatLog : ConfigForm} />
+          </main>
+
+          <section id="write">
+            <${MessageInput} />
+          </section>
+
+          <footer>
+            <p><${ModelGenerationInfo} /></p>
+            <p>Powered by <a href="/ggerganov/llama.cpp">llama.cpp</a> and <a href="https://ggml.ai">ggml.ai</a>.</p>
+          </footer>
+        </div>
       `;
     }
 
diff --git a/examples/server/server.cpp b/examples/server/server.cpp
index 043e49750707b..2cbfc0018de3a 100644
--- a/examples/server/server.cpp
+++ b/examples/server/server.cpp
@@ -22,7 +22,8 @@
 using namespace httplib;
 using json = nlohmann::json;
 
-struct server_params {
+struct server_params
+{
     std::string hostname = "127.0.0.1";
     std::string public_path = "examples/server/public";
     int32_t port = 8080;
@@ -31,8 +32,10 @@ struct server_params {
 };
 
 // completion token output with probabilities
-struct completion_token_output {
-    struct token_prob {
+struct completion_token_output
+{
+    struct token_prob
+    {
         llama_token tok;
         float prob;
     };
@@ -41,30 +44,40 @@ struct completion_token_output {
     llama_token tok;
 };
 
-static size_t common_part(const std::vector<llama_token> & a, const std::vector<llama_token> & b) {
+static size_t common_part(const std::vector<llama_token> &a, const std::vector<llama_token> &b)
+{
     size_t i;
-    for (i = 0; i < a.size() && i < b.size() && a[i] == b[i]; i++) {}
+    for (i = 0; i < a.size() && i < b.size() && a[i] == b[i]; i++)
+    {
+    }
     return i;
 }
 
-enum stop_type {
+enum stop_type
+{
     STOP_FULL,
     STOP_PARTIAL,
 };
 
-static bool ends_with(const std::string & str, const std::string & suffix) {
+static bool ends_with(const std::string &str, const std::string &suffix)
+{
     return str.size() >= suffix.size() &&
-        0 == str.compare(str.size() - suffix.size(), suffix.size(), suffix);
+           0 == str.compare(str.size() - suffix.size(), suffix.size(), suffix);
 }
 
-static size_t find_partial_stop_string(const std::string & stop,
-                                       const std::string & text) {
-    if (!text.empty() && !stop.empty()) {
+static size_t find_partial_stop_string(const std::string &stop,
+                                       const std::string &text)
+{
+    if (!text.empty() && !stop.empty())
+    {
         const char text_last_char = text.back();
-        for (int64_t char_index = stop.size() - 1; char_index >= 0; char_index--) {
-            if (stop[char_index] == text_last_char) {
+        for (int64_t char_index = stop.size() - 1; char_index >= 0; char_index--)
+        {
+            if (stop[char_index] == text_last_char)
+            {
                 const std::string current_partial = stop.substr(0, char_index + 1);
-                if (ends_with(text, current_partial)) {
+                if (ends_with(text, current_partial))
+                {
                     return text.size() - char_index - 1;
                 }
             }
@@ -73,26 +86,30 @@ static size_t find_partial_stop_string(const std::string & stop,
     return std::string::npos;
 }
 
-template<class Iter>
-static std::string tokens_to_str(llama_context * ctx, Iter begin, Iter end) {
+template <class Iter>
+static std::string tokens_to_str(llama_context *ctx, Iter begin, Iter end)
+{
     std::string ret;
-    for (; begin != end; ++begin) {
+    for (; begin != end; ++begin)
+    {
         ret += llama_token_to_str(ctx, *begin);
     }
     return ret;
 }
 
-static void server_log(const char * level, const char * function, int line,
-                       const char * message, const nlohmann::ordered_json & extra) {
-    nlohmann::ordered_json log {
-        { "timestamp", time(nullptr) },
-        { "level", level },
-        { "function", function },
-        { "line", line },
-        { "message", message },
+static void server_log(const char *level, const char *function, int line,
+                       const char *message, const nlohmann::ordered_json &extra)
+{
+    nlohmann::ordered_json log{
+        {"timestamp", time(nullptr)},
+        {"level", level},
+        {"function", function},
+        {"line", line},
+        {"message", message},
     };
 
-    if (!extra.empty()) {
+    if (!extra.empty())
+    {
         log.merge_patch(extra);
     }
 
@@ -102,32 +119,37 @@ static void server_log(const char * level, const char * function, int line,
 }
 
 // format incomplete utf-8 multibyte character for output
-static std::string tokens_to_output_formatted_string(const llama_context * ctx, const llama_token token) {
+static std::string tokens_to_output_formatted_string(const llama_context *ctx, const llama_token token)
+{
     std::string out = token == -1 ? "" : llama_token_to_str(ctx, token);
     // if first bit is 1, meaning it's a partial character
-    if (out.size() > 0 && (out[0] & 0x80) == 0x80) {
+    if (out.size() > 0 && (out[0] & 0x80) == 0x80)
+    {
         std::stringstream ss;
-        ss<< std::hex << (out[0] & 0xff);
-        std::string res ( ss.str() );
+        ss << std::hex << (out[0] & 0xff);
+        std::string res(ss.str());
         out = "byte: \\x" + res;
     }
     return out;
 }
 
 // convert a vector of completion_token_output to json
-static json probs_vector_to_json(const llama_context * ctx, const std::vector<completion_token_output> probs) {
+static json probs_vector_to_json(const llama_context *ctx, const std::vector<completion_token_output> probs)
+{
     json out = json::array();
-    for (const auto & prob : probs) {
+    for (const auto &prob : probs)
+    {
         json probs_for_token = json::array();
-        for (const auto & p : prob.probs) {
+        for (const auto &p : prob.probs)
+        {
             std::string tok_str = tokens_to_output_formatted_string(ctx, p.tok);
-            probs_for_token.push_back(json {
-                { "tok_str", tok_str },
-                { "prob", p.prob },
+            probs_for_token.push_back(json{
+                {"tok_str", tok_str},
+                {"prob", p.prob},
             });
         }
         std::string tok_str = tokens_to_output_formatted_string(ctx, prob.tok);
-        out.push_back(json {
+        out.push_back(json{
             {"content", tok_str},
             {"probs", probs_for_token},
         });
@@ -138,26 +160,30 @@ static json probs_vector_to_json(const llama_context * ctx, const std::vector<co
 static bool server_verbose = false;
 
 #if SERVER_VERBOSE != 1
-#  define LOG_VERBOSE(MSG, ...)
+#define LOG_VERBOSE(MSG, ...)
 #else
-#  define LOG_VERBOSE(MSG, ...)                                          \
-    do {                                                                 \
-        if (server_verbose) {                                            \
+#define LOG_VERBOSE(MSG, ...)                                            \
+    do                                                                   \
+    {                                                                    \
+        if (server_verbose)                                              \
+        {                                                                \
             server_log("VERBOSE", __func__, __LINE__, MSG, __VA_ARGS__); \
         }                                                                \
-    } while(0)
+    } while (0)
 #endif
 
 #define LOG_ERROR(MSG, ...) server_log("ERROR", __func__, __LINE__, MSG, __VA_ARGS__)
 #define LOG_WARNING(MSG, ...) server_log("WARNING", __func__, __LINE__, MSG, __VA_ARGS__)
 #define LOG_INFO(MSG, ...) server_log("INFO", __func__, __LINE__, MSG, __VA_ARGS__)
 
-struct llama_server_context {
+struct llama_server_context
+{
     bool stream = false;
     bool has_next_token = false;
     std::string generated_text;
     std::vector<completion_token_output> generated_token_probs;
 
+    size_t num_prompt_tokens = 0;
     size_t num_tokens_predicted = 0;
     size_t n_past = 0;
     size_t n_remain = 0;
@@ -165,8 +191,8 @@ struct llama_server_context {
     std::vector<llama_token> embd;
     std::vector<llama_token> last_n_tokens;
 
-    llama_model * model = nullptr;
-    llama_context * ctx = nullptr;
+    llama_model *model = nullptr;
+    llama_context *ctx = nullptr;
     gpt_params params;
 
     bool truncated = false;
@@ -178,23 +204,29 @@ struct llama_server_context {
 
     std::mutex mutex;
 
-    std::unique_lock<std::mutex> lock() {
+    std::unique_lock<std::mutex> lock()
+    {
         return std::unique_lock<std::mutex>(mutex);
     }
 
-    ~llama_server_context() {
-        if (ctx) {
+    ~llama_server_context()
+    {
+        if (ctx)
+        {
             llama_free(ctx);
             ctx = nullptr;
         }
-        if (model) {
+        if (model)
+        {
             llama_free_model(model);
             model = nullptr;
         }
     }
 
-    void rewind() {
+    void rewind()
+    {
         params.antiprompt.clear();
+        num_prompt_tokens = 0;
         num_tokens_predicted = 0;
         generated_text = "";
         generated_text.reserve(params.n_ctx);
@@ -210,11 +242,13 @@ struct llama_server_context {
         n_past = 0;
     }
 
-    bool loadModel(const gpt_params & params_) {
+    bool loadModel(const gpt_params &params_)
+    {
         params = params_;
         std::tie(model, ctx) = llama_init_from_gpt_params(params);
-        if (model == nullptr) {
-            LOG_ERROR("unable to load model", { { "model", params_.model } });
+        if (model == nullptr)
+        {
+            LOG_ERROR("unable to load model", {{"model", params_.model}});
             return false;
         }
 
@@ -223,34 +257,40 @@ struct llama_server_context {
         return true;
     }
 
-    void loadPrompt() {
+    void loadPrompt()
+    {
         params.prompt.insert(0, 1, ' '); // always add a first space
         std::vector<llama_token> prompt_tokens = ::llama_tokenize(ctx, params.prompt, true);
+        num_prompt_tokens = prompt_tokens.size();
 
-        if (params.n_keep < 0) {
-            params.n_keep = (int)prompt_tokens.size();
+        if (params.n_keep < 0)
+        {
+            params.n_keep = (int)num_prompt_tokens;
         }
         params.n_keep = std::min(params.n_ctx - 4, params.n_keep);
 
         // if input prompt is too big, truncate like normal
-        if (prompt_tokens.size() >= (size_t)params.n_ctx) {
+        if (num_prompt_tokens >= (size_t)params.n_ctx)
+        {
             const int n_left = (params.n_ctx - params.n_keep) / 2;
             std::vector<llama_token> new_tokens(prompt_tokens.begin(), prompt_tokens.begin() + params.n_keep);
-            const int erased_blocks = (prompt_tokens.size() - params.n_keep - n_left - 1) / n_left;
+            const int erased_blocks = (num_prompt_tokens - params.n_keep - n_left - 1) / n_left;
             new_tokens.insert(new_tokens.end(), prompt_tokens.begin() + params.n_keep + erased_blocks * n_left, prompt_tokens.end());
             std::copy(prompt_tokens.end() - params.n_ctx, prompt_tokens.end(), last_n_tokens.begin());
 
             LOG_VERBOSE("input truncated", {
-                { "n_ctx", params.n_ctx },
-                { "n_keep", params.n_keep },
-                { "n_left", n_left },
-                { "new_tokens", tokens_to_str(ctx, new_tokens.cbegin(), new_tokens.cend()) },
-            });
+                                               {"n_ctx", params.n_ctx},
+                                               {"n_keep", params.n_keep},
+                                               {"n_left", n_left},
+                                               {"new_tokens", tokens_to_str(ctx, new_tokens.cbegin(), new_tokens.cend())},
+                                           });
 
             truncated = true;
             prompt_tokens = new_tokens;
-        } else {
-            const size_t ps = prompt_tokens.size();
+        }
+        else
+        {
+            const size_t ps = num_prompt_tokens;
             std::fill(last_n_tokens.begin(), last_n_tokens.end() - ps, 0);
             std::copy(prompt_tokens.begin(), prompt_tokens.end(), last_n_tokens.end() - ps);
         }
@@ -258,31 +298,35 @@ struct llama_server_context {
         // compare the evaluated prompt with the new prompt
         n_past = common_part(embd, prompt_tokens);
         embd = prompt_tokens;
-        if (n_past == prompt_tokens.size()) {
+        if (n_past == num_prompt_tokens)
+        {
             // we have to evaluate at least 1 token to generate logits.
             n_past--;
         }
 
         LOG_VERBOSE("prompt ingested", {
-            { "n_past", n_past },
-            { "cached", tokens_to_str(ctx, embd.cbegin(), embd.cbegin() + n_past) },
-            { "to_eval", tokens_to_str(ctx, embd.cbegin() + n_past, embd.cend()) },
-        });
+                                           {"n_past", n_past},
+                                           {"cached", tokens_to_str(ctx, embd.cbegin(), embd.cbegin() + n_past)},
+                                           {"to_eval", tokens_to_str(ctx, embd.cbegin() + n_past, embd.cend())},
+                                       });
 
         has_next_token = true;
     }
 
-    void beginCompletion() {
+    void beginCompletion()
+    {
         // number of tokens to keep when resetting context
         n_remain = params.n_predict;
         llama_set_rng_seed(ctx, params.seed);
     }
 
-    completion_token_output nextToken() {
+    completion_token_output nextToken()
+    {
         completion_token_output result;
         result.tok = -1;
 
-        if (embd.size() >= (size_t)params.n_ctx) {
+        if (embd.size() >= (size_t)params.n_ctx)
+        {
             // Reset context
             const int n_left = (params.n_ctx - params.n_keep) / 2;
 
@@ -292,32 +336,36 @@ struct llama_server_context {
             n_past = params.n_keep;
             truncated = true;
             LOG_VERBOSE("input truncated", {
-                { "n_ctx", params.n_ctx },
-                { "n_keep", params.n_keep },
-                { "n_left", n_left },
-                { "new_tokens", tokens_to_str(ctx, new_tokens.cbegin(), new_tokens.cend()) },
-            });
+                                               {"n_ctx", params.n_ctx},
+                                               {"n_keep", params.n_keep},
+                                               {"n_left", n_left},
+                                               {"new_tokens", tokens_to_str(ctx, new_tokens.cbegin(), new_tokens.cend())},
+                                           });
         }
 
-        while (n_past < embd.size()) {
+        while (n_past < embd.size())
+        {
             int n_eval = (int)embd.size() - n_past;
-            if (n_eval > params.n_batch) {
+            if (n_eval > params.n_batch)
+            {
                 n_eval = params.n_batch;
             }
-            if (llama_eval(ctx, &embd[n_past], n_eval, n_past, params.n_threads)) {
+            if (llama_eval(ctx, &embd[n_past], n_eval, n_past, params.n_threads))
+            {
                 LOG_ERROR("failed to eval", {
-                    { "n_eval", n_eval },
-                    { "n_past", n_past },
-                    { "n_threads", params.n_threads },
-                    { "embd", tokens_to_str(ctx, embd.cbegin() + n_past, embd.cend()) },
-                });
+                                                {"n_eval", n_eval},
+                                                {"n_past", n_past},
+                                                {"n_threads", params.n_threads},
+                                                {"embd", tokens_to_str(ctx, embd.cbegin() + n_past, embd.cend())},
+                                            });
                 has_next_token = false;
                 return result;
             }
             n_past += n_eval;
         }
 
-        if (params.n_predict == 0) {
+        if (params.n_predict == 0)
+        {
             has_next_token = false;
             result.tok = llama_token_eos();
             return result;
@@ -340,52 +388,64 @@ struct llama_server_context {
         const int32_t n_probs = params.n_probs;
 
         {
-            auto * logits = llama_get_logits(ctx);
+            auto *logits = llama_get_logits(ctx);
             auto n_vocab = llama_n_vocab(ctx);
 
             // Apply params.logit_bias map
-            for (const auto & it : params.logit_bias) {
+            for (const auto &it : params.logit_bias)
+            {
                 logits[it.first] += it.second;
             }
 
             std::vector<llama_token_data> candidates;
             candidates.reserve(n_vocab);
-            for (llama_token token_id = 0; token_id < n_vocab; token_id++) {
-                candidates.emplace_back(llama_token_data{ token_id, logits[token_id], 0.0f });
+            for (llama_token token_id = 0; token_id < n_vocab; token_id++)
+            {
+                candidates.emplace_back(llama_token_data{token_id, logits[token_id], 0.0f});
             }
 
-            llama_token_data_array candidates_p = { candidates.data(), candidates.size(), false };
+            llama_token_data_array candidates_p = {candidates.data(), candidates.size(), false};
 
             // Apply penalties
             float nl_logit = logits[llama_token_nl()];
             auto last_n_repeat = std::min(std::min((int)last_n_tokens.size(), repeat_last_n), params.n_ctx);
             llama_sample_repetition_penalty(ctx, &candidates_p,
-                last_n_tokens.data() + last_n_tokens.size() - last_n_repeat,
-                last_n_repeat, repeat_penalty);
+                                            last_n_tokens.data() + last_n_tokens.size() - last_n_repeat,
+                                            last_n_repeat, repeat_penalty);
             llama_sample_frequency_and_presence_penalties(ctx, &candidates_p,
-                last_n_tokens.data() + last_n_tokens.size() - last_n_repeat,
-                last_n_repeat, alpha_frequency, alpha_presence);
-            if (!penalize_nl) {
+                                                          last_n_tokens.data() + last_n_tokens.size() - last_n_repeat,
+                                                          last_n_repeat, alpha_frequency, alpha_presence);
+            if (!penalize_nl)
+            {
                 logits[llama_token_nl()] = nl_logit;
             }
 
-            if (temp <= 0) {
+            if (temp <= 0)
+            {
                 // Greedy sampling
                 result.tok = llama_sample_token_greedy(ctx, &candidates_p);
-                if (n_probs > 0) {
+                if (n_probs > 0)
+                {
                     llama_sample_softmax(ctx, &candidates_p);
                 }
-            } else {
-                if (mirostat == 1) {
+            }
+            else
+            {
+                if (mirostat == 1)
+                {
                     static float mirostat_mu = 2.0f * mirostat_tau;
                     const int mirostat_m = 100;
                     llama_sample_temperature(ctx, &candidates_p, temp);
                     result.tok = llama_sample_token_mirostat(ctx, &candidates_p, mirostat_tau, mirostat_eta, mirostat_m, &mirostat_mu);
-                } else if (mirostat == 2) {
+                }
+                else if (mirostat == 2)
+                {
                     static float mirostat_mu = 2.0f * mirostat_tau;
                     llama_sample_temperature(ctx, &candidates_p, temp);
                     result.tok = llama_sample_token_mirostat_v2(ctx, &candidates_p, mirostat_tau, mirostat_eta, &mirostat_mu);
-                } else {
+                }
+                else
+                {
                     // Temperature sampling
                     size_t min_keep = std::max(1, n_probs);
                     llama_sample_top_k(ctx, &candidates_p, top_k, min_keep);
@@ -397,7 +457,8 @@ struct llama_server_context {
                 }
             }
 
-            for (size_t i = 0; i < std::min(candidates_p.size, (size_t) n_probs); ++i) {
+            for (size_t i = 0; i < std::min(candidates_p.size, (size_t)n_probs); ++i)
+            {
                 result.probs.push_back({candidates_p.data[i].id, candidates_p.data[i].p});
             }
             last_n_tokens.erase(last_n_tokens.begin());
@@ -410,8 +471,9 @@ struct llama_server_context {
         // decrement remaining sampling budget
         --n_remain;
 
-        if (!embd.empty() && embd.back() == llama_token_eos()) {
-            //stopping_word = llama_token_to_str(ctx, embd.back());
+        if (!embd.empty() && embd.back() == llama_token_eos())
+        {
+            // stopping_word = llama_token_to_str(ctx, embd.back());
             has_next_token = false;
             stopped_eos = true;
             LOG_VERBOSE("eos token found", {});
@@ -422,22 +484,28 @@ struct llama_server_context {
         return result;
     }
 
-    size_t findStoppingStrings(const std::string & text, const size_t last_token_size,
-                               const stop_type type) {
+    size_t findStoppingStrings(const std::string &text, const size_t last_token_size,
+                               const stop_type type)
+    {
         size_t stop_pos = std::string::npos;
-        for (const std::string & word : params.antiprompt) {
+        for (const std::string &word : params.antiprompt)
+        {
             size_t pos;
-            if (type == STOP_FULL) {
+            if (type == STOP_FULL)
+            {
                 const size_t tmp = word.size() + last_token_size;
                 const size_t from_pos = text.size() > tmp ? text.size() - tmp : 0;
                 pos = text.find(word, from_pos);
             }
-            else {
+            else
+            {
                 pos = find_partial_stop_string(word, text);
             }
             if (pos != std::string::npos &&
-                (stop_pos == std::string::npos || pos < stop_pos)) {
-                if (type == STOP_FULL) {
+                (stop_pos == std::string::npos || pos < stop_pos))
+            {
+                if (type == STOP_FULL)
+                {
                     stopping_word = word;
                     stopped_word = true;
                     has_next_token = false;
@@ -448,74 +516,91 @@ struct llama_server_context {
         return stop_pos;
     }
 
-    completion_token_output doCompletion() {
+    completion_token_output doCompletion()
+    {
         const completion_token_output token_with_probs = nextToken();
 
         const std::string token_text = token_with_probs.tok == -1 ? "" : llama_token_to_str(ctx, token_with_probs.tok);
         generated_text += token_text;
 
-        if (params.n_probs > 0) {
+        if (params.n_probs > 0)
+        {
             generated_token_probs.push_back(token_with_probs);
         }
 
-        if (multibyte_pending > 0) {
+        if (multibyte_pending > 0)
+        {
             multibyte_pending -= token_text.size();
-        } else if (token_text.size() == 1) {
+        }
+        else if (token_text.size() == 1)
+        {
             const char c = token_text[0];
             // 2-byte characters: 110xxxxx 10xxxxxx
-            if ((c & 0xE0) == 0xC0) {
+            if ((c & 0xE0) == 0xC0)
+            {
                 multibyte_pending = 1;
-            // 3-byte characters: 1110xxxx 10xxxxxx 10xxxxxx
-            } else if ((c & 0xF0) == 0xE0) {
+                // 3-byte characters: 1110xxxx 10xxxxxx 10xxxxxx
+            }
+            else if ((c & 0xF0) == 0xE0)
+            {
                 multibyte_pending = 2;
-            // 4-byte characters: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
-            } else if ((c & 0xF8) == 0xF0) {
+                // 4-byte characters: 11110xxx 10xxxxxx 10xxxxxx 10xxxxxx
+            }
+            else if ((c & 0xF8) == 0xF0)
+            {
                 multibyte_pending = 3;
-            } else {
+            }
+            else
+            {
                 multibyte_pending = 0;
             }
         }
 
-        if (multibyte_pending > 0 && !has_next_token) {
+        if (multibyte_pending > 0 && !has_next_token)
+        {
             has_next_token = true;
             n_remain++;
         }
 
-        if (!has_next_token && n_remain == 0) {
+        if (!has_next_token && n_remain == 0)
+        {
             stopped_limit = true;
         }
 
         LOG_VERBOSE("next token", {
-            { "token", token_with_probs.tok },
-            { "token_text", tokens_to_output_formatted_string(ctx, token_with_probs.tok) },
-            { "has_next_token", has_next_token },
-            { "n_remain", n_remain },
-            { "num_tokens_predicted", num_tokens_predicted },
-            { "stopped_eos", stopped_eos },
-            { "stopped_word", stopped_word },
-            { "stopped_limit", stopped_limit },
-            { "stopping_word", stopping_word },
-        });
+                                      {"token", token_with_probs.tok},
+                                      {"token_text", tokens_to_output_formatted_string(ctx, token_with_probs.tok)},
+                                      {"has_next_token", has_next_token},
+                                      {"n_remain", n_remain},
+                                      {"num_tokens_predicted", num_tokens_predicted},
+                                      {"stopped_eos", stopped_eos},
+                                      {"stopped_word", stopped_word},
+                                      {"stopped_limit", stopped_limit},
+                                      {"stopping_word", stopping_word},
+                                  });
 
         return token_with_probs;
     }
 
-    std::vector<float> getEmbedding() {
+    std::vector<float> getEmbedding()
+    {
         static const int n_embd = llama_n_embd(ctx);
-        if (!params.embedding) {
+        if (!params.embedding)
+        {
             LOG_WARNING("embedding disabled", {
-                { "params.embedding", params.embedding },
-            });
+                                                  {"params.embedding", params.embedding},
+                                              });
             return std::vector<float>(n_embd, 0.0f);
         }
-        const float * data = llama_get_embeddings(ctx);
+        const float *data = llama_get_embeddings(ctx);
         std::vector<float> embedding(data, data + n_embd);
         return embedding;
     }
 };
 
-static void server_print_usage(const char * argv0, const gpt_params & params,
-                               const server_params & sparams) {
+static void server_print_usage(const char *argv0, const gpt_params &params,
+                               const server_params &sparams)
+{
     fprintf(stderr, "usage: %s [options]\n", argv0);
     fprintf(stderr, "\n");
     fprintf(stderr, "options:\n");
@@ -526,10 +611,12 @@ static void server_print_usage(const char * argv0, const gpt_params & params,
     fprintf(stderr, "  -b N, --batch-size N  batch size for prompt processing (default: %d)\n", params.n_batch);
     fprintf(stderr, "  --memory-f32          use f32 instead of f16 for memory key+value (default: disabled)\n");
     fprintf(stderr, "                        not recommended: doubles context memory required and no measurable increase in quality\n");
-    if (llama_mlock_supported()) {
+    if (llama_mlock_supported())
+    {
         fprintf(stderr, "  --mlock               force system to keep model in RAM rather than swapping or compressing\n");
     }
-    if (llama_mmap_supported()) {
+    if (llama_mmap_supported())
+    {
         fprintf(stderr, "  --no-mmap             do not memory-map model (slower load but may reduce pageouts if not using mlock)\n");
     }
 #ifdef LLAMA_SUPPORTS_GPU_OFFLOAD
@@ -555,78 +642,113 @@ static void server_print_usage(const char * argv0, const gpt_params & params,
     fprintf(stderr, "\n");
 }
 
-static void server_params_parse(int argc, char ** argv, server_params & sparams,
-                                gpt_params & params) {
+static void server_params_parse(int argc, char **argv, server_params &sparams,
+                                gpt_params &params)
+{
     gpt_params default_params;
     server_params default_sparams;
     std::string arg;
     bool invalid_param = false;
 
-    for (int i = 1; i < argc; i++) {
+    for (int i = 1; i < argc; i++)
+    {
         arg = argv[i];
-        if (arg == "--port") {
-            if (++i >= argc) {
+        if (arg == "--port")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             sparams.port = std::stoi(argv[i]);
-        } else if (arg == "--host") {
-            if (++i >= argc) {
+        }
+        else if (arg == "--host")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             sparams.hostname = argv[i];
-        } else if (arg == "--path") {
-            if (++i >= argc) {
+        }
+        else if (arg == "--path")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             sparams.public_path = argv[i];
-        } else if (arg == "--timeout" || arg == "-to") {
-            if (++i >= argc) {
+        }
+        else if (arg == "--timeout" || arg == "-to")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             sparams.read_timeout = std::stoi(argv[i]);
             sparams.write_timeout = std::stoi(argv[i]);
-        } else if (arg == "-m" || arg == "--model") {
-            if (++i >= argc) {
+        }
+        else if (arg == "-m" || arg == "--model")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             params.model = argv[i];
-        } else if (arg == "-a" || arg == "--alias") {
-            if (++i >= argc) {
+        }
+        else if (arg == "-a" || arg == "--alias")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             params.model_alias = argv[i];
-        } else if (arg == "-h" || arg == "--help") {
+        }
+        else if (arg == "-h" || arg == "--help")
+        {
             server_print_usage(argv[0], default_params, default_sparams);
             exit(0);
-        } else if (arg == "-c" || arg == "--ctx-size" || arg == "--ctx_size") {
-            if (++i >= argc) {
+        }
+        else if (arg == "-c" || arg == "--ctx-size" || arg == "--ctx_size")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             params.n_ctx = std::stoi(argv[i]);
-        } else if (arg == "--memory-f32" || arg == "--memory_f32") {
+        }
+        else if (arg == "--memory-f32" || arg == "--memory_f32")
+        {
             params.memory_f16 = false;
-        } else if (arg == "--threads" || arg == "-t") {
-            if (++i >= argc) {
+        }
+        else if (arg == "--threads" || arg == "-t")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             params.n_threads = std::stoi(argv[i]);
-        } else if (arg == "-b" || arg == "--batch-size") {
-            if (++i >= argc) {
+        }
+        else if (arg == "-b" || arg == "--batch-size")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             params.n_batch = std::stoi(argv[i]);
             params.n_batch = std::min(512, params.n_batch);
-        } else if (arg == "--gpu-layers" || arg == "-ngl" || arg == "--n-gpu-layers") {
-            if (++i >= argc) {
+        }
+        else if (arg == "--gpu-layers" || arg == "-ngl" || arg == "--n-gpu-layers")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
@@ -634,11 +756,14 @@ static void server_params_parse(int argc, char ** argv, server_params & sparams,
             params.n_gpu_layers = std::stoi(argv[i]);
 #else
             LOG_WARNING("Not compiled with GPU offload support, --n-gpu-layers option will be ignored. "
-                        "See main README.md for information on enabling GPU BLAS support", { { "n_gpu_layers", params.n_gpu_layers } });
+                        "See main README.md for information on enabling GPU BLAS support",
+                        {{"n_gpu_layers", params.n_gpu_layers}});
 #endif
         }
-        else if (arg == "--tensor-split" || arg == "-ts") {
-            if (++i >= argc) {
+        else if (arg == "--tensor-split" || arg == "-ts")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
@@ -646,16 +771,19 @@ static void server_params_parse(int argc, char ** argv, server_params & sparams,
             std::string arg_next = argv[i];
 
             // split string by , and /
-            const std::regex regex{ R"([,/]+)" };
-            std::sregex_token_iterator it{ arg_next.begin(), arg_next.end(), regex, -1 };
-            std::vector<std::string> split_arg{ it, {} };
+            const std::regex regex{R"([,/]+)"};
+            std::sregex_token_iterator it{arg_next.begin(), arg_next.end(), regex, -1};
+            std::vector<std::string> split_arg{it, {}};
             GGML_ASSERT(split_arg.size() <= LLAMA_MAX_DEVICES);
 
-            for (size_t i_device = 0; i_device < LLAMA_MAX_DEVICES; ++i_device) {
-                if (i_device < split_arg.size()) {
+            for (size_t i_device = 0; i_device < LLAMA_MAX_DEVICES; ++i_device)
+            {
+                if (i_device < split_arg.size())
+                {
                     params.tensor_split[i_device] = std::stof(split_arg[i_device]);
                 }
-                else {
+                else
+                {
                     params.tensor_split[i_device] = 0.0f;
                 }
             }
@@ -671,8 +799,10 @@ static void server_params_parse(int argc, char ** argv, server_params & sparams,
             fprintf(stderr, "warning: llama.cpp was compiled without cuBLAS. It is not possible to set lower vram usage.\n");
 #endif // GGML_USE_CUBLAS
         }
-        else if (arg == "--main-gpu" || arg == "-mg") {
-            if (++i >= argc) {
+        else if (arg == "--main-gpu" || arg == "-mg")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
@@ -681,124 +811,173 @@ static void server_params_parse(int argc, char ** argv, server_params & sparams,
 #else
             LOG_WARNING("llama.cpp was compiled without cuBLAS. It is not possible to set a main GPU.", {});
 #endif
-        } else if (arg == "--lora") {
-            if (++i >= argc) {
+        }
+        else if (arg == "--lora")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             params.lora_adapter = argv[i];
             params.use_mmap = false;
-        } else if (arg == "--lora-base") {
-            if (++i >= argc) {
+        }
+        else if (arg == "--lora-base")
+        {
+            if (++i >= argc)
+            {
                 invalid_param = true;
                 break;
             }
             params.lora_base = argv[i];
-        } else if (arg == "-v" || arg == "--verbose") {
+        }
+        else if (arg == "-v" || arg == "--verbose")
+        {
 #if SERVER_VERBOSE != 1
             LOG_WARNING("server.cpp is not built with verbose logging.", {});
 #else
             server_verbose = true;
 #endif
-        } else if (arg == "--mlock") {
+        }
+        else if (arg == "--mlock")
+        {
             params.use_mlock = true;
-        } else if (arg == "--no-mmap") {
+        }
+        else if (arg == "--no-mmap")
+        {
             params.use_mmap = false;
-        } else if (arg == "--embedding") {
+        }
+        else if (arg == "--embedding")
+        {
             params.embedding = true;
-        } else {
+        }
+        else
+        {
             fprintf(stderr, "error: unknown argument: %s\n", arg.c_str());
             server_print_usage(argv[0], default_params, default_sparams);
             exit(1);
         }
     }
 
-    if (invalid_param) {
+    if (invalid_param)
+    {
         fprintf(stderr, "error: invalid parameter for argument: %s\n", arg.c_str());
         server_print_usage(argv[0], default_params, default_sparams);
         exit(1);
     }
 }
 
-static json format_generation_settings(llama_server_context & llama) {
+static json format_generation_settings(llama_server_context &llama)
+{
     const auto eos_bias = llama.params.logit_bias.find(llama_token_eos());
     const bool ignore_eos = eos_bias != llama.params.logit_bias.end() &&
-        eos_bias->second < 0.0f && std::isinf(eos_bias->second);
-
-    return json {
-        { "seed", llama.params.seed },
-        { "temp", llama.params.temp },
-        { "top_k", llama.params.top_k },
-        { "top_p", llama.params.top_p },
-        { "tfs_z", llama.params.tfs_z },
-        { "typical_p", llama.params.typical_p },
-        { "repeat_last_n", llama.params.repeat_last_n },
-        { "repeat_penalty", llama.params.repeat_penalty },
-        { "presence_penalty", llama.params.presence_penalty },
-        { "frequency_penalty", llama.params.frequency_penalty },
-        { "mirostat", llama.params.mirostat },
-        { "mirostat_tau", llama.params.mirostat_tau },
-        { "mirostat_eta", llama.params.mirostat_eta },
-        { "penalize_nl", llama.params.penalize_nl },
-        { "stop", llama.params.antiprompt },
-        { "n_predict", llama.params.n_predict },
-        { "n_keep", llama.params.n_keep },
-        { "ignore_eos", ignore_eos },
-        { "stream", llama.stream },
-        { "logit_bias", llama.params.logit_bias },
-        { "n_probs", llama.params.n_probs },
+                            eos_bias->second < 0.0f && std::isinf(eos_bias->second);
+
+    return json{
+        {"n_ctx", llama.params.n_ctx},
+        {"model", llama.params.model_alias},
+        {"seed", llama.params.seed},
+        {"temp", llama.params.temp},
+        {"top_k", llama.params.top_k},
+        {"top_p", llama.params.top_p},
+        {"tfs_z", llama.params.tfs_z},
+        {"typical_p", llama.params.typical_p},
+        {"repeat_last_n", llama.params.repeat_last_n},
+        {"repeat_penalty", llama.params.repeat_penalty},
+        {"presence_penalty", llama.params.presence_penalty},
+        {"frequency_penalty", llama.params.frequency_penalty},
+        {"mirostat", llama.params.mirostat},
+        {"mirostat_tau", llama.params.mirostat_tau},
+        {"mirostat_eta", llama.params.mirostat_eta},
+        {"penalize_nl", llama.params.penalize_nl},
+        {"stop", llama.params.antiprompt},
+        {"n_predict", llama.params.n_predict},
+        {"n_keep", llama.params.n_keep},
+        {"ignore_eos", ignore_eos},
+        {"stream", llama.stream},
+        {"logit_bias", llama.params.logit_bias},
+        {"n_probs", llama.params.n_probs},
+    };
+}
+
+static json format_embedding_response(llama_server_context &llama)
+{
+    return json{
+        {"embedding", llama.getEmbedding()},
     };
 }
 
-static json format_embedding_response(llama_server_context & llama) {
-    return json {
-        { "embedding", llama.getEmbedding() },
+static json format_timings(llama_server_context &llama)
+{
+    const auto timings = llama_get_timings(llama.ctx);
+
+    assert(timings.n_eval == llama.num_tokens_predicted);
+
+    return json{
+        {"prompt_n", timings.n_eval},
+        {"prompt_ms", timings.t_p_eval_ms},
+        {"prompt_per_token_ms", timings.t_p_eval_ms / timings.n_p_eval},
+        {"prompt_per_second", 1e3 / timings.t_p_eval_ms * timings.n_p_eval},
+
+        {"predicted_n", timings.n_eval},
+        {"predicted_ms", timings.t_eval_ms},
+        {"predicted_per_token_ms", timings.t_eval_ms / timings.n_eval},
+        {"predicted_per_second", 1e3 / timings.t_eval_ms * timings.n_eval},
     };
 }
 
-static json format_final_response(llama_server_context & llama, const std::string & content, const std::vector<completion_token_output> & probs) {
-
-    json res = json {
-        { "content", content },
-        { "stop", true },
-        { "model", llama.params.model_alias },
-        { "tokens_predicted", llama.num_tokens_predicted },
-        { "generation_settings", format_generation_settings(llama) },
-        { "prompt", llama.params.prompt },
-        { "truncated", llama.truncated },
-        { "stopped_eos", llama.stopped_eos },
-        { "stopped_word", llama.stopped_word },
-        { "stopped_limit", llama.stopped_limit },
-        { "stopping_word", llama.stopping_word },
+static json format_final_response(llama_server_context &llama, const std::string &content, const std::vector<completion_token_output> &probs)
+{
+
+    json res = json{
+        {"content", content},
+        {"stop", true},
+        {"model", llama.params.model_alias},
+        {"tokens_predicted", llama.num_tokens_predicted},
+        {"tokens_evaluated", llama.num_prompt_tokens},
+        {"generation_settings", format_generation_settings(llama)},
+        {"prompt", llama.params.prompt},
+        {"truncated", llama.truncated},
+        {"stopped_eos", llama.stopped_eos},
+        {"stopped_word", llama.stopped_word},
+        {"stopped_limit", llama.stopped_limit},
+        {"stopping_word", llama.stopping_word},
+        {"tokens_cached", llama.n_past},
+        {"tokens_predicted", llama.num_tokens_predicted},
+        {"timings", format_timings(llama)},
     };
 
-    if (llama.params.n_probs > 0) {
+    if (llama.params.n_probs > 0)
+    {
         res["completion_probabilities"] = probs_vector_to_json(llama.ctx, probs);
     }
 
     return res;
 }
 
-static json format_partial_response(llama_server_context & llama, const std::string & content, const std::vector<completion_token_output> & probs) {
-    json res = json {
-        { "content", content },
-        { "stop", false },
+static json format_partial_response(llama_server_context &llama, const std::string &content, const std::vector<completion_token_output> &probs)
+{
+    json res = json{
+        {"content", content},
+        {"stop", false},
     };
 
-    if (llama.params.n_probs > 0) {
+    if (llama.params.n_probs > 0)
+    {
         res["completion_probabilities"] = probs_vector_to_json(llama.ctx, probs);
     }
 
     return res;
 }
 
-static json format_tokenizer_response(const std::vector<llama_token> & tokens) {
-    return json {
-        { "tokens", tokens }
-    };
+static json format_tokenizer_response(const std::vector<llama_token> &tokens)
+{
+    return json{
+        {"tokens", tokens}};
 }
 
-static void parse_options_completion(const json & body, llama_server_context & llama) {
+static void parse_options_completion(const json &body, llama_server_context &llama)
+{
     gpt_params default_params;
 
     llama.stream = body.value("stream", false);
@@ -822,20 +1001,28 @@ static void parse_options_completion(const json & body, llama_server_context & l
     llama.params.n_probs = body.value("n_probs", default_params.n_probs);
 
     llama.params.logit_bias.clear();
-    if (body.value("ignore_eos", false)) {
+    if (body.value("ignore_eos", false))
+    {
         llama.params.logit_bias[llama_token_eos()] = -INFINITY;
     }
 
-    const auto & logit_bias = body.find("logit_bias");
-    if (logit_bias != body.end() && logit_bias->is_array()) {
+    const auto &logit_bias = body.find("logit_bias");
+    if (logit_bias != body.end() && logit_bias->is_array())
+    {
         const int n_vocab = llama_n_vocab(llama.ctx);
-        for (const auto & el : *logit_bias) {
-            if (el.is_array() && el.size() == 2 && el[0].is_number_integer()) {
+        for (const auto &el : *logit_bias)
+        {
+            if (el.is_array() && el.size() == 2 && el[0].is_number_integer())
+            {
                 llama_token tok = el[0].get<llama_token>();
-                if (tok >= 0 && tok < n_vocab) {
-                    if (el[1].is_number()) {
+                if (tok >= 0 && tok < n_vocab)
+                {
+                    if (el[1].is_number())
+                    {
                         llama.params.logit_bias[tok] = el[1].get<float>();
-                    } else if (el[1].is_boolean() && !el[1].get<bool>()) {
+                    }
+                    else if (el[1].is_boolean() && !el[1].get<bool>())
+                    {
                         llama.params.logit_bias[tok] = -INFINITY;
                     }
                 }
@@ -844,10 +1031,13 @@ static void parse_options_completion(const json & body, llama_server_context & l
     }
 
     llama.params.antiprompt.clear();
-    const auto & stop = body.find("stop");
-    if (stop != body.end() && stop->is_array()) {
-        for (const auto & word : *stop) {
-            if (!word.empty()) {
+    const auto &stop = body.find("stop");
+    if (stop != body.end() && stop->is_array())
+    {
+        for (const auto &word : *stop)
+        {
+            if (!word.empty())
+            {
                 llama.params.antiprompt.push_back(word);
             }
         }
@@ -856,25 +1046,25 @@ static void parse_options_completion(const json & body, llama_server_context & l
     LOG_VERBOSE("completion parameters parsed", format_generation_settings(llama));
 }
 
-
-static void log_server_request(const Request & req, const Response & res) {
+static void log_server_request(const Request &req, const Response &res)
+{
     LOG_INFO("request", {
-        { "remote_addr", req.remote_addr },
-        { "remote_port", req.remote_port },
-        { "status", res.status },
-        { "method", req.method },
-        { "path", req.path },
-        { "params", req.params },
-    });
+                            {"remote_addr", req.remote_addr},
+                            {"remote_port", req.remote_port},
+                            {"status", res.status},
+                            {"method", req.method},
+                            {"path", req.path},
+                            {"params", req.params},
+                        });
 
     LOG_VERBOSE("request", {
-        { "request", req.body },
-        { "response", res.body },
-    });
+                               {"request", req.body},
+                               {"response", res.body},
+                           });
 }
 
-
-int main(int argc, char ** argv) {
+int main(int argc, char **argv)
+{
     // own arguments required by this example
     gpt_params params;
     server_params sparams;
@@ -884,54 +1074,53 @@ int main(int argc, char ** argv) {
 
     server_params_parse(argc, argv, sparams, params);
 
-    if (params.model_alias == "unknown") {
+    if (params.model_alias == "unknown")
+    {
         params.model_alias = params.model;
     }
 
     llama_init_backend(params.numa);
 
-    LOG_INFO("build info", {
-        { "build", BUILD_NUMBER },
-        { "commit", BUILD_COMMIT }
-    });
+    LOG_INFO("build info", {{"build", BUILD_NUMBER},
+                            {"commit", BUILD_COMMIT}});
     LOG_INFO("system info", {
-        { "n_threads", params.n_threads },
-        { "total_threads", std::thread::hardware_concurrency() },
-        { "system_info", llama_print_system_info() },
-    });
+                                {"n_threads", params.n_threads},
+                                {"total_threads", std::thread::hardware_concurrency()},
+                                {"system_info", llama_print_system_info()},
+                            });
 
     // load the model
-    if (!llama.loadModel(params)) {
+    if (!llama.loadModel(params))
+    {
         return 1;
     }
 
     Server svr;
 
-    svr.set_default_headers({
-        { "Server", "llama.cpp" },
-        { "Access-Control-Allow-Origin", "*" },
-        { "Access-Control-Allow-Headers", "content-type" }
-    });
-
-    // this is only called if no index.js is found in the public --path
-    svr.Get("/index.js", [](const Request &, Response & res) {
-        res.set_content(reinterpret_cast<const char *>(&index_js), index_js_len, "text/javascript");
-        return false;
-    });
+    svr.set_default_headers({{"Server", "llama.cpp"},
+                             {"Access-Control-Allow-Origin", "*"},
+                             {"Access-Control-Allow-Headers", "content-type"}});
 
     // this is only called if no index.html is found in the public --path
-    svr.Get("/", [](const Request &, Response & res) {
+    svr.Get("/", [](const Request &, Response &res)
+            {
         res.set_content(reinterpret_cast<const char*>(&index_html), index_html_len, "text/html");
-        return false;
-    });
+        return false; });
+
+    // this is only called if no index.js is found in the public --path
+    svr.Get("/index.js", [](const Request &, Response &res)
+            {
+        res.set_content(reinterpret_cast<const char *>(&index_js), index_js_len, "text/javascript");
+        return false; });
 
     // this is only called if no index.html is found in the public --path
-    svr.Get("/completion.js", [](const Request &, Response & res) {
+    svr.Get("/completion.js", [](const Request &, Response &res)
+            {
         res.set_content(reinterpret_cast<const char*>(&completion_js), completion_js_len, "application/javascript");
-        return false;
-    });
+        return false; });
 
-    svr.Post("/completion", [&llama](const Request & req, Response & res) {
+    svr.Post("/completion", [&llama](const Request &req, Response &res)
+             {
         auto lock = llama.lock();
 
         llama.rewind();
@@ -1036,24 +1225,28 @@ int main(int argc, char ** argv) {
                 return true;
             };
             res.set_chunked_content_provider("text/event-stream", chunked_content_provider);
-        }
-    });
+        } });
+
+    svr.Get("/model.json", [&llama](const Request &, Response &res)
+            {
+        const json data = format_generation_settings(llama);
+        return res.set_content(data.dump(), "application/json"); });
 
-    svr.Options(R"(/.*)", [](const Request &, Response & res) {
-        return res.set_content("", "application/json");
-    });
+    svr.Options(R"(/.*)", [](const Request &, Response &res)
+                { return res.set_content("", "application/json"); });
 
-    svr.Post("/tokenize", [&llama](const Request & req, Response & res) {
+    svr.Post("/tokenize", [&llama](const Request &req, Response &res)
+             {
         auto lock = llama.lock();
 
         const json body = json::parse(req.body);
         const std::string content = body.value("content", "");
         const std::vector<llama_token> tokens = llama_tokenize(llama.ctx, content, false);
         const json data = format_tokenizer_response(tokens);
-        return res.set_content(data.dump(), "application/json");
-    });
+        return res.set_content(data.dump(), "application/json"); });
 
-    svr.Post("/embedding", [&llama](const Request & req, Response & res) {
+    svr.Post("/embedding", [&llama](const Request &req, Response &res)
+             {
         auto lock = llama.lock();
 
         const json body = json::parse(req.body);
@@ -1067,12 +1260,12 @@ int main(int argc, char ** argv) {
         llama.doCompletion();
 
         const json data = format_embedding_response(llama);
-        return res.set_content(data.dump(), "application/json");
-    });
+        return res.set_content(data.dump(), "application/json"); });
 
     svr.set_logger(log_server_request);
 
-    svr.set_exception_handler([](const Request &, Response & res, std::exception_ptr ep) {
+    svr.set_exception_handler([](const Request &, Response &res, std::exception_ptr ep)
+                              {
         const auto * fmt = "500 Internal Server Error\n%s";
         char buf[BUFSIZ];
         try {
@@ -1083,20 +1276,19 @@ int main(int argc, char ** argv) {
             snprintf(buf, sizeof(buf), fmt, "Unknown Exception");
         }
         res.set_content(buf, "text/plain");
-        res.status = 500;
-    });
+        res.status = 500; });
 
-    svr.set_error_handler([](const Request &, Response & res) {
+    svr.set_error_handler([](const Request &, Response &res)
+                          {
         res.set_content("File Not Found", "text/plain");
-        res.status = 404;
-    });
-
+        res.status = 404; });
 
     // set timeouts and change hostname and port
     svr.set_read_timeout(sparams.read_timeout);
     svr.set_write_timeout(sparams.write_timeout);
 
-    if (!svr.bind_to_port(sparams.hostname, sparams.port)) {
+    if (!svr.bind_to_port(sparams.hostname, sparams.port))
+    {
         fprintf(stderr, "\ncouldn't bind to server socket: hostname=%s port=%d\n\n", sparams.hostname.c_str(), sparams.port);
         return 1;
     }
@@ -1108,11 +1300,12 @@ int main(int argc, char ** argv) {
     fprintf(stdout, "\nllama server listening at http://%s:%d\n\n", sparams.hostname.c_str(), sparams.port);
 
     LOG_INFO("HTTP server listening", {
-        { "hostname", sparams.hostname },
-        { "port", sparams.port },
-    });
+                                          {"hostname", sparams.hostname},
+                                          {"port", sparams.port},
+                                      });
 
-    if (!svr.listen_after_bind()) {
+    if (!svr.listen_after_bind())
+    {
         return 1;
     }
 
diff --git a/examples/train-text-from-scratch/train-text-from-scratch.cpp b/examples/train-text-from-scratch/train-text-from-scratch.cpp
index c50eeb343bcef..b96fdcdc44b57 100644
--- a/examples/train-text-from-scratch/train-text-from-scratch.cpp
+++ b/examples/train-text-from-scratch/train-text-from-scratch.cpp
@@ -60,6 +60,17 @@ float frand_uniform(struct random_uniform_distribution * rnd) {
     return rnd->rd(rnd->gen);
 }
 
+void ggml_graph_compute_helper(std::vector<uint8_t> & buf, ggml_cgraph * graph, int n_threads) {
+    struct ggml_cplan plan = ggml_graph_plan(graph, n_threads);
+
+    if (plan.work_size > 0) {
+        buf.resize(plan.work_size);
+        plan.work_data = buf.data();
+    }
+
+    ggml_graph_compute(graph, &plan);
+}
+
 struct ggml_tensor * randomize_tensor_normal(struct ggml_tensor * tensor, struct random_normal_distribution * rnd) {
     float scale = 1.0f; // xavier
     switch (tensor->n_dims) {
@@ -1426,11 +1437,9 @@ struct ggml_tensor * forward_batch_wo_cache_flash_attn_train(
 
     gf->n_nodes = 0;
     gf->n_leafs = 0;
-    gf->work_size = 0;
     gf->perf_runs = 0;
     gf->perf_cycles = 0;
     gf->perf_time_us = 0;
-    gf->work = NULL;
 
     const auto & hparams = model->hparams;
     //const int n_ctx      = hparams.n_ctx;
@@ -3162,6 +3171,7 @@ int main(int argc, char ** argv) {
     printf("used_mem model+cache: %zu bytes\n", ggml_used_mem(model.ctx));
     // ggml_print_tensor_objects(model.ctx);
 
+    // TODO: use std::vector<uint8_t> intead of "new"
     size_t    compute_size = 1024ll*1024ll*1024ll*((size_t) params.mem_compute_gb);
     uint8_t * compute_addr = new uint8_t[compute_size];
 
@@ -3183,6 +3193,8 @@ int main(int argc, char ** argv) {
         GGML_ASSERT(train_samples[i]+n_tokens-1 < (int) train_tokens.size());
     }
 
+    std::vector<uint8_t> work_buffer;
+
     printf("%s: begin training\n", __func__);
 
     for (int ex = 0; ex < params.n_examples; ++ex) {
@@ -3217,9 +3229,6 @@ int main(int argc, char ** argv) {
         struct ggml_cgraph * gf = (struct ggml_cgraph *) gfbuf->data;
         struct ggml_cgraph * gb = (struct ggml_cgraph *) gbbuf->data;
 
-        // ggml_cgraph gf = {};
-        gf->n_threads = params.n_threads;
-        gb->n_threads = params.n_threads;
 
         get_example_targets_batch(lctx, train_samples.data(), train_samples.size(), train_tokens.data(), train_tokens.size(), ex,  tokens_input, target_logits, target_probs);
 
@@ -3248,7 +3257,7 @@ int main(int argc, char ** argv) {
             *gb = ggml_build_backward(ctx0, gf, true);
         }
 
-        ggml_graph_compute(ctx0, gf);
+        ggml_graph_compute_helper(work_buffer, gf, params.n_threads);
 
         size_t used_mem_before_opt = ggml_used_mem(ctx0);
 
@@ -3272,7 +3281,7 @@ int main(int argc, char ** argv) {
         model.train_samples += n_batch;
         model.train_tokens  += n_batch * n_tokens;
 
-        ggml_graph_compute(ctx0, gf);
+        ggml_graph_compute_helper(work_buffer, gf, params.n_threads);
 
         float error_after_opt = ggml_get_f32_1d(loss, 0);
 
@@ -3354,13 +3363,12 @@ int main(int argc, char ** argv) {
             struct ggml_context * ctx0 = ggml_init(cparams);
 
             ggml_cgraph gf = {};
-            gf.n_threads = params.n_threads;
 
             int n_past = 0;
             struct ggml_tensor * logits = forward(&model, &kv_self, ctx0, &gf, tokens_input, sample_ctx, n_past);
 
             ggml_build_forward_expand(&gf, logits);
-            ggml_graph_compute(ctx0, &gf);
+            ggml_graph_compute_helper(work_buffer, &gf, params.n_threads);
 
             //struct ggml_tensor * best_samples = ggml_new_tensor_1d(ctx0, GGML_TYPE_I32, sample_ctx);
             //struct ggml_tensor * probs        = ggml_new_tensor_2d(ctx0, GGML_TYPE_F32, n_vocab, sample_ctx);
@@ -3386,6 +3394,7 @@ int main(int argc, char ** argv) {
     delete[] compute_addr;
     delete[] compute_buf_0;
     delete[] compute_buf_1;
+
     llama_free(lctx);
     llama_free_model(lmodel);
     ggml_free(model.ctx);
diff --git a/ggml-cuda.cu b/ggml-cuda.cu
index a9e5685f7965b..d05ffd7233eb2 100644
--- a/ggml-cuda.cu
+++ b/ggml-cuda.cu
@@ -115,8 +115,8 @@ typedef float2 dfloat2;
 #endif //GGML_CUDA_DMMV_F16
 
 typedef void (*dequantize_kernel_t)(const void * vx, const int ib, const int iqs, dfloat2 & v);
-typedef void (*to_fp32_cuda_t)(const void * x, float * y, int k, cudaStream_t stream);
-typedef void (*dot_kernel_k_t)(const void * vx, const int ib, const int iqs, const float * y, float & v);
+typedef void (*to_fp32_cuda_t)(const void * __restrict__ x, float * __restrict__ y, int k, cudaStream_t stream);
+typedef void (*dot_kernel_k_t)(const void * __restrict__ vx, const int ib, const int iqs, const float * __restrict__ y, float & v);
 typedef void (*cpy_kernel_t)(const char * cx, char * cdst);
 typedef void (*ggml_cuda_func_t)(const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst);
 typedef void (*ggml_cuda_op_t)(
@@ -126,9 +126,11 @@ typedef void (*ggml_cuda_op_t)(
 
 // QK = number of values after dequantization
 // QR = QK / number of values before dequantization
+// QI = number of 32 bit integers before dequantization
 
 #define QK4_0 32
 #define QR4_0 2
+#define QI4_0 4
 typedef struct {
     half    d;              // delta
     uint8_t qs[QK4_0 / 2];  // nibbles / quants
@@ -137,6 +139,7 @@ static_assert(sizeof(block_q4_0) == sizeof(ggml_fp16_t) + QK4_0 / 2, "wrong q4_0
 
 #define QK4_1 32
 #define QR4_1 2
+#define QI4_1 4
 typedef struct {
     half    d;              // delta
     half    m;              // min
@@ -146,6 +149,7 @@ static_assert(sizeof(block_q4_1) == sizeof(ggml_fp16_t) * 2 + QK4_1 / 2, "wrong
 
 #define QK5_0 32
 #define QR5_0 2
+#define QI5_0 4
 typedef struct {
     half d;                 // delta
     uint8_t qh[4];          // 5-th bit of quants
@@ -155,6 +159,7 @@ static_assert(sizeof(block_q5_0) == sizeof(ggml_fp16_t) + sizeof(uint32_t) + QK5
 
 #define QK5_1 32
 #define QR5_1 2
+#define QI5_1 4
 typedef struct {
     half d;                 // delta
     half m;                 // min
@@ -165,12 +170,25 @@ static_assert(sizeof(block_q5_1) == 2 * sizeof(ggml_fp16_t) + sizeof(uint32_t) +
 
 #define QK8_0 32
 #define QR8_0 1
+#define QI8_0 8
 typedef struct {
     half    d;              // delta
     int8_t  qs[QK8_0];      // quants
 } block_q8_0;
 static_assert(sizeof(block_q8_0) == sizeof(ggml_fp16_t) + QK8_0, "wrong q8_0 block size/padding");
 
+#define QK8_1 32
+#define QR8_1 1
+#define QI8_1 8
+typedef struct {
+    half    d;              // delta
+    half    s;              // unquantized sum
+    int8_t  qs[QK8_0];      // quants
+} block_q8_1;
+static_assert(sizeof(block_q8_1) == 2*sizeof(ggml_fp16_t) + QK8_0, "wrong q8_1 block size/padding");
+
+typedef float (*vec_dot_q_cuda_t)(const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int iqs);
+
 //================================= k-quants
 
 #ifdef GGML_QKK_64
@@ -246,6 +264,7 @@ typedef struct {
 static_assert(sizeof(block_q6_K) == sizeof(ggml_fp16_t) + 13*QK_K/16, "wrong q6_K block size/padding");
 
 #define WARP_SIZE 32
+#define MATRIX_ROW_PADDING 256 // last row of quant. matrices is a multiple of this to avoid out-of-bounds memory accesses
 
 #define CUDA_ADD_BLOCK_SIZE 256
 #define CUDA_MUL_BLOCK_SIZE 256
@@ -254,14 +273,15 @@ static_assert(sizeof(block_q6_K) == sizeof(ggml_fp16_t) + 13*QK_K/16, "wrong q6_
 #define CUDA_SCALE_BLOCK_SIZE 256
 #define CUDA_ROPE_BLOCK_SIZE 256
 #define CUDA_DIAG_MASK_INF_BLOCK_SIZE 32
+#define CUDA_QUANTIZE_BLOCK_SIZE 256
 #define CUDA_DEQUANTIZE_BLOCK_SIZE 256
 
 // dmmv = dequantize_mul_mat_vec
 #ifndef GGML_CUDA_DMMV_X
 #define GGML_CUDA_DMMV_X 32
 #endif
-#ifndef GGML_CUDA_DMMV_Y
-#define GGML_CUDA_DMMV_Y 1
+#ifndef GGML_CUDA_MMV_Y
+#define GGML_CUDA_MMV_Y 1
 #endif
 
 #ifndef K_QUANTS_PER_ITERATION
@@ -326,7 +346,6 @@ static __global__ void rms_norm_f32(const float * x, float * dst, const int ncol
     }
 
     // sum up partial sums
-    __syncthreads();
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
@@ -445,7 +464,7 @@ static __device__ __forceinline__ void dequantize_q8_0(const void * vx, const in
 
 //================================== k-quants
 
-static __global__ void dequantize_block_q2_K(const void * vx, float * yy) {
+static __global__ void dequantize_block_q2_K(const void * __restrict__ vx, float * __restrict__ yy) {
 
     const int i   = blockIdx.x;
     const block_q2_K * x = (const block_q2_K *) vx;
@@ -478,7 +497,7 @@ static __global__ void dequantize_block_q2_K(const void * vx, float * yy) {
 
 }
 
-static __global__ void dequantize_block_q3_K(const void * vx, float * yy) {
+static __global__ void dequantize_block_q3_K(const void * __restrict__ vx, float * __restrict__ yy) {
 
     const int i = blockIdx.x;
     const block_q3_K * x = (const block_q3_K *) vx;
@@ -542,7 +561,7 @@ static inline __device__ void get_scale_min_k4(int j, const uint8_t * q, uint8_t
 }
 #endif
 
-static __global__ void dequantize_block_q4_K(const void * vx, float * yy) {
+static __global__ void dequantize_block_q4_K(const void * __restrict__ vx, float * __restrict__ yy) {
     const block_q4_K * x = (const block_q4_K *) vx;
 
     const int i = blockIdx.x;
@@ -582,7 +601,7 @@ static __global__ void dequantize_block_q4_K(const void * vx, float * yy) {
 #endif
 }
 
-static __global__ void dequantize_block_q5_K(const void * vx, float * yy) {
+static __global__ void dequantize_block_q5_K(const void * __restrict__ vx, float * __restrict__ yy) {
     const block_q5_K * x = (const block_q5_K *) vx;
 
     const int i = blockIdx.x;
@@ -628,7 +647,7 @@ static __global__ void dequantize_block_q5_K(const void * vx, float * yy) {
 #endif
 }
 
-static __global__ void dequantize_block_q6_K(const void * vx, float * yy) {
+static __global__ void dequantize_block_q6_K(const void * __restrict__ vx, float * __restrict__ yy) {
     const block_q6_K * x = (const block_q6_K *) vx;
 
     const int i = blockIdx.x;
@@ -672,7 +691,7 @@ static __global__ void dequantize_block_q6_K(const void * vx, float * yy) {
 #endif
 }
 
-static __global__ void dequantize_mul_mat_vec_q2_k(const void * vx, const float * yy, float * dst, const int ncols, int nrows) {
+static __global__ void dequantize_mul_mat_vec_q2_k(const void * __restrict__ vx, const float * __restrict__ yy, float * __restrict__ dst, const int ncols, int nrows) {
 
     static_assert(16%K_QUANTS_PER_ITERATION == 0, "16 must be divisible by K_QUANTS_PER_ITERATION");
 
@@ -770,7 +789,6 @@ static __global__ void dequantize_mul_mat_vec_q2_k(const void * vx, const float
 #endif
 
     // sum up partial sums and write back result
-    __syncthreads();
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
@@ -781,7 +799,7 @@ static __global__ void dequantize_mul_mat_vec_q2_k(const void * vx, const float
     }
 }
 
-static __global__ void dequantize_mul_mat_vec_q3_k(const void * vx, const float * yy, float * dst, const int ncols, int nrows) {
+static __global__ void dequantize_mul_mat_vec_q3_k(const void * __restrict__ vx, const float * __restrict__ yy, float * __restrict__ dst, const int ncols, int nrows) {
 
     const int row = blockIdx.y*blockDim.y + threadIdx.y;
     if (row > nrows) return;
@@ -875,7 +893,6 @@ static __global__ void dequantize_mul_mat_vec_q3_k(const void * vx, const float
 #endif
 
     // sum up partial sums and write back result
-    __syncthreads();
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
@@ -886,7 +903,7 @@ static __global__ void dequantize_mul_mat_vec_q3_k(const void * vx, const float
     }
 }
 
-static __global__ void dequantize_mul_mat_vec_q4_k(const void * vx, const float * yy, float * dst, const int ncols, int nrows) {
+static __global__ void dequantize_mul_mat_vec_q4_k(const void * __restrict__ vx, const float * __restrict__ yy, float * __restrict__ dst, const int ncols, int nrows) {
 
     const int row = blockIdx.y*blockDim.y + threadIdx.y;
     if (row > nrows) return;
@@ -979,7 +996,6 @@ static __global__ void dequantize_mul_mat_vec_q4_k(const void * vx, const float
 #endif
 
     // sum up partial sums and write back result
-    __syncthreads();
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
@@ -990,7 +1006,7 @@ static __global__ void dequantize_mul_mat_vec_q4_k(const void * vx, const float
     }
 }
 
-static __global__ void dequantize_mul_mat_vec_q5_k(const void * vx, const float * yy, float * dst, const int ncols) {
+static __global__ void dequantize_mul_mat_vec_q5_k(const void * __restrict__ vx, const float * __restrict__ yy, float * __restrict__ dst, const int ncols) {
 
     const int row = blockIdx.x;
     const int num_blocks_per_row = ncols / QK_K;
@@ -1084,7 +1100,6 @@ static __global__ void dequantize_mul_mat_vec_q5_k(const void * vx, const float
 #endif
 
     // sum up partial sums and write back result
-    __syncthreads();
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
@@ -1095,7 +1110,7 @@ static __global__ void dequantize_mul_mat_vec_q5_k(const void * vx, const float
     }
 }
 
-static __global__ void dequantize_mul_mat_vec_q6_k(const void * vx, const float * yy, float * dst, const int ncols, int nrows) {
+static __global__ void dequantize_mul_mat_vec_q6_k(const void * __restrict__ vx, const float * __restrict__ yy, float * __restrict__ dst, const int ncols, int nrows) {
 
     static_assert(16%K_QUANTS_PER_ITERATION == 0, "16 must be divisible by K_QUANTS_PER_ITERATION");
 
@@ -1195,7 +1210,6 @@ static __global__ void dequantize_mul_mat_vec_q6_k(const void * vx, const float
 #endif
 
     // sum up partial sums and write back result
-    __syncthreads();
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
@@ -1214,8 +1228,43 @@ static __device__ void convert_f16(const void * vx, const int ib, const int iqs,
     v.y = x[ib + iqs + 1];
 }
 
+static __global__ void quantize_q8_1(const float * __restrict__ x, void * __restrict__ vy, const int ndata, const int k) {
+    const int i = blockDim.x*blockIdx.x + threadIdx.x;
+
+    if (i >= k) {
+        return;
+    }
+
+    block_q8_1 * y = (block_q8_1 *) vy;
+
+    const int ib = i / QK8_1; // block index
+    const int iqs = i % QK8_1; // quant index
+
+    const float xi = i < ndata ? x[i] : 0.0f;
+    float amax = fabsf(xi);
+    float sum = xi;
+
+#pragma unroll
+    for (int mask = 16; mask > 0; mask >>= 1) {
+        amax = fmaxf(amax, __shfl_xor_sync(0xffffffff, amax, mask, 32));
+        sum += __shfl_xor_sync(0xffffffff, sum, mask, 32);
+    }
+
+    const float d = amax / 127;
+    const int8_t q = amax == 0.0f ? 0 : roundf(xi / d);
+
+    y[ib].qs[iqs] = q;
+
+    if (iqs > 0) {
+        return;
+    }
+
+    y[ib].d = d;
+    y[ib].s = sum;
+}
+
 template <int qk, int qr, dequantize_kernel_t dequantize_kernel>
-static __global__ void dequantize_block(const void * vx, float * y, const int k) {
+static __global__ void dequantize_block(const void * __restrict__ vx, float * __restrict__ y, const int k) {
     const int i = blockDim.x*blockIdx.x + 2*threadIdx.x;
 
     if (i >= k) {
@@ -1235,8 +1284,184 @@ static __global__ void dequantize_block(const void * vx, float * y, const int k)
     y[iybs + iqs + y_offset] = v.y;
 }
 
+static __device__ __forceinline__ float vec_dot_q4_0_q8_1(const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int iqs) {
+#if __CUDA_ARCH__ >= 600 // lowest compute capability for integer intrinsics
+    const block_q4_0 * bq4_0 = (const block_q4_0 *) vbq;
+
+    int vi;
+    memcpy(&vi,  &bq4_0->qs[sizeof(int) * (iqs + 0)], sizeof(int));
+    const int ui0 = *((int *) &bq8_1->qs[sizeof(int) * (iqs + 0)]);
+    const int ui1 = *((int *) &bq8_1->qs[sizeof(int) * (iqs + QI4_0)]);
+
+    const float d = __half2float(bq4_0->d) * __half2float(bq8_1->d);
+
+    // subtract 8 from each quantized value
+    const int vi0 = __vsub4((vi >> 0) & 0x0F0F0F0F, 0x08080808);
+    const int vi1 = __vsub4((vi >> 4) & 0x0F0F0F0F, 0x08080808);
+
+    // SIMD dot product of quantized values
+    int sumi = __dp4a(vi0, ui0, 0);
+    sumi     = __dp4a(vi1, ui1, sumi);
+
+    return sumi*d;
+#else
+    return 0.0f; // only to satisfy the compiler
+#endif // __CUDA_ARCH__ >= 600
+}
+
+static __device__ __forceinline__ float vec_dot_q4_1_q8_1(const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int iqs) {
+#if __CUDA_ARCH__ >= 600 // lowest compute capability for integer intrinsics
+    const block_q4_1 * bq4_1 = (const block_q4_1 *) vbq;
+
+    const int vi  = *((int *) &bq4_1->qs[sizeof(int) * (iqs + 0)]);
+    const int ui0 = *((int *) &bq8_1->qs[sizeof(int) * (iqs + 0)]);
+    const int ui1 = *((int *) &bq8_1->qs[sizeof(int) * (iqs + QI4_1)]);
+
+    const float d = __half2float(bq4_1->d) * __half2float(bq8_1->d);
+    const float m = bq4_1->m;
+    const float s = bq8_1->s;
+
+    const int vi0 = (vi >> 0) & 0x0F0F0F0F;
+    const int vi1 = (vi >> 4) & 0x0F0F0F0F;
+
+    // SIMD dot product of quantized values
+    int sumi = __dp4a(vi0, ui0, 0);
+    sumi     = __dp4a(vi1, ui1, sumi);
+
+    return sumi*d + m*s / QI4_1; // scale sum by QI4_1 because there are QI4_1 threads working on this block
+#else
+    return 0.0f; // only to satisfy the compiler
+#endif // __CUDA_ARCH__ >= 600
+}
+
+static __device__ __forceinline__ float vec_dot_q5_0_q8_1(const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int iqs) {
+#if __CUDA_ARCH__ >= 600 // lowest compute capability for integer intrinsics
+    const block_q5_0 * bq5_0 = (const block_q5_0 *) vbq;
+
+    int qs;
+    memcpy(&qs, &bq5_0->qs[sizeof(int) * (iqs + 0)], sizeof(int));
+    const int qh0 = bq5_0->qh[iqs/2 + 0] >> 4*(iqs%2);
+    const int qh1 = bq5_0->qh[iqs/2 + 2] >> 4*(iqs%2);
+    const int ui0 = *((int *) &bq8_1->qs[sizeof(int) * (iqs + 0)]);
+    const int ui1 = *((int *) &bq8_1->qs[sizeof(int) * (iqs + QI5_0)]);
+
+    const float d = __half2float(bq5_0->d) * __half2float(bq8_1->d);
+
+    int vi0 = (qs  >>  0) & 0x0F0F0F0F; // lower 4 qs bits, still need qh0 as 5th bits
+    vi0    |= (qh0 <<  4) & 0x00000010; // 1 ->  5
+    vi0    |= (qh0 << 11) & 0x00001000; // 2 -> 13
+    vi0    |= (qh0 << 18) & 0x00100000; // 3 -> 21
+    vi0    |= (qh0 << 25) & 0x10000000; // 4 -> 29
+    vi0     = __vsub4(vi0,  0x10101010); // subtract 16 from quantized values
+    int sumi = __dp4a(vi0, ui0, 0); // SIMD dot product of quantized values
+
+    int vi1 = (qs  >>  4) & 0x0F0F0F0F; // upper 4 qs bits, still need qh1 as 5th bits
+    vi1    |= (qh1 <<  4) & 0x00000010; // 1 ->  5
+    vi1    |= (qh1 << 11) & 0x00001000; // 2 -> 13
+    vi1    |= (qh1 << 18) & 0x00100000; // 3 -> 21
+    vi1    |= (qh1 << 25) & 0x10000000; // 4 -> 29
+    vi1     = __vsub4(vi1,  0x10101010); // subtract 16 from quantized values
+    sumi = __dp4a(vi1, ui1, sumi); // SIMD dot product of quantized values
+
+    return sumi*d;
+#else
+    return 0.0f; // only to satisfy the compiler
+#endif // __CUDA_ARCH__ >= 600
+}
+
+static __device__ __forceinline__ float vec_dot_q5_1_q8_1(const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int iqs) {
+#if __CUDA_ARCH__ >= 600 // lowest compute capability for integer intrinsics
+    const block_q5_1 * bq5_1 = (const block_q5_1 *) vbq;
+
+    const int qs  = *((int *) &bq5_1->qs[sizeof(int) * (iqs + 0)]);
+    const int qh0 = bq5_1->qh[iqs/2 + 0] >> 4*(iqs%2);
+    const int qh1 = bq5_1->qh[iqs/2 + 2] >> 4*(iqs%2);
+    const int ui0 = *((int *) &bq8_1->qs[sizeof(int) * (iqs + 0)]);
+    const int ui1 = *((int *) &bq8_1->qs[sizeof(int) * (iqs + QI5_1)]);
+
+    const float d = __half2float(bq5_1->d) * __half2float(bq8_1->d);
+    const float m = bq5_1->m;
+    const float s = bq8_1->s;
+
+    int vi0 = (qs  >>  0) & 0x0F0F0F0F; // lower 4 qs bits, still need qh0 as 5th bits
+    vi0    |= (qh0 <<  4) & 0x00000010; // 1 ->  5
+    vi0    |= (qh0 << 11) & 0x00001000; // 2 -> 13
+    vi0    |= (qh0 << 18) & 0x00100000; // 3 -> 21
+    vi0    |= (qh0 << 25) & 0x10000000; // 4 -> 29
+    int sumi = __dp4a(vi0, ui0, 0); // SIMD dot product of quantized values
+
+    int vi1 = (qs  >>  4) & 0x0F0F0F0F; // upper 4 qs bits, still need qh1 as 5th bits
+    vi1    |= (qh1 <<  4) & 0x00000010; // 1 ->  5
+    vi1    |= (qh1 << 11) & 0x00001000; // 2 -> 13
+    vi1    |= (qh1 << 18) & 0x00100000; // 3 -> 21
+    vi1    |= (qh1 << 25) & 0x10000000; // 4 -> 29
+    sumi = __dp4a(vi1, ui1, sumi); // SIMD dot product of quantized values
+
+    return sumi*d + m*s / QI5_1; // scale sum by QI5_1 because there are QI5_1 threads working on this block
+#else
+    return 0.0f; // only to satisfy the compiler
+#endif // __CUDA_ARCH__ >= 600
+}
+
+static __device__ __forceinline__ float vec_dot_q8_0_q8_1(const void * __restrict__ vbq, const block_q8_1 * __restrict__ bq8_1, const int iqs) {
+#if __CUDA_ARCH__ >= 600 // lowest compute capability for integer intrinsics
+    const block_q8_0 * bq8_0 = (const block_q8_0 *) vbq;
+
+    int vi;
+    memcpy(&vi,  &bq8_0->qs[sizeof(int) * (iqs + 0)], sizeof(int));
+    const int ui = *((int *) &bq8_1->qs[sizeof(int) * (iqs + 0)]);
+
+    const float d = __half2float(bq8_0->d) * __half2float(bq8_1->d);
+
+    // SIMD dot product of quantized values
+    int sumi = __dp4a(vi, ui, 0);
+
+    return sumi*d;
+#else
+    return 0.0f; // only to satisfy the compiler
+#endif // __CUDA_ARCH__ >= 600
+}
+
+template <int qk, int qi, typename block_q_t, vec_dot_q_cuda_t vec_dot_q_cuda>
+static __global__ void mul_mat_vec_q(const void * __restrict__ vx, const void * __restrict__ vy, float * __restrict__ dst, const int ncols, const int nrows) {
+    const int row = blockIdx.y*blockDim.y + threadIdx.y;
+
+    if (row >= nrows) {
+        return;
+    }
+
+    const int blocks_per_row = ncols / qk;
+    const int blocks_per_warp = WARP_SIZE / qi;
+
+// partial sum for each thread
+    float tmp = 0.0f;
+
+    const block_q_t  * x = (const block_q_t  *) vx;
+    const block_q8_1 * y = (const block_q8_1 *) vy;
+
+    for (int i = 0; i < blocks_per_row; i += blocks_per_warp) {
+        const int ibx = row*blocks_per_row + i + threadIdx.x / qi; // x block index
+
+        const int iby = i + threadIdx.x / qi; // y block index
+
+        const int iqs  = threadIdx.x % qi; // x block quant index when casting the quants to int
+
+        tmp += vec_dot_q_cuda(&x[ibx], &y[iby], iqs);
+    }
+
+    // sum up partial sums and write back result
+#pragma unroll
+    for (int mask = 16; mask > 0; mask >>= 1) {
+        tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
+    }
+
+    if (threadIdx.x == 0) {
+        dst[row] = tmp;
+    }
+}
+
 template <int qk, int qr, dequantize_kernel_t dequantize_kernel>
-static __global__ void dequantize_mul_mat_vec(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows) {
+static __global__ void dequantize_mul_mat_vec(const void * __restrict__ vx, const dfloat * __restrict__ y, float * __restrict__ dst, const int ncols, const int nrows) {
     // qk = quantized weights per x block
     // qr = number of quantized weights per data value in x block
     const int row = blockIdx.y*blockDim.y + threadIdx.y;
@@ -1289,7 +1514,6 @@ static __global__ void dequantize_mul_mat_vec(const void * vx, const dfloat * y,
     }
 
     // sum up partial sums and write back result
-    __syncthreads();
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
@@ -1304,7 +1528,7 @@ static __global__ void dequantize_mul_mat_vec(const void * vx, const dfloat * y,
     }
 }
 
-static __global__ void mul_mat_p021_f16_f32(const void * vx, const float * y, float * dst, const int ncols_x, const int nrows_x, const int nchannels_x) {
+static __global__ void mul_mat_p021_f16_f32(const void * __restrict__ vx, const float * __restrict__ y, float * __restrict__ dst, const int ncols_x, const int nrows_x, const int nchannels_x) {
     const half * x = (const half *) vx;
 
     const int row_x = blockDim.y*blockIdx.y + threadIdx.y;
@@ -1340,7 +1564,6 @@ static __global__ void mul_mat_p021_f16_f32(const void * vx, const float * y, fl
     const int idst = channel*nrows_dst + row_dst;
 
     // sum up partial sums and write back result
-    __syncthreads();
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
@@ -1352,7 +1575,7 @@ static __global__ void mul_mat_p021_f16_f32(const void * vx, const float * y, fl
 }
 
 static __global__ void mul_mat_vec_nc_f16_f32( // nc == non-contiguous
-    const void * vx, const float * y, float * dst, const int ncols_x, const int nrows_x,
+    const void * __restrict__ vx, const float * __restrict__ y, float * __restrict__ dst, const int ncols_x, const int nrows_x,
     const int row_stride_x, const int channel_stride_x) {
 
     const half * x = (const half *) vx;
@@ -1386,7 +1609,6 @@ static __global__ void mul_mat_vec_nc_f16_f32( // nc == non-contiguous
     }
 
     // sum up partial sums and write back result
-    __syncthreads();
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
@@ -1496,7 +1718,6 @@ static __global__ void soft_max_f32(const float * x, float * dst, const int ncol
     }
 
     // sum up partial sums
-    __syncthreads();
 #pragma unroll
     for (int mask = 16; mask > 0; mask >>= 1) {
         tmp += __shfl_xor_sync(0xffffffff, tmp, mask, 32);
@@ -1550,6 +1771,11 @@ static void rms_norm_f32_cuda(const float * x, float * dst, const int ncols, con
     rms_norm_f32<<<nrows, block_dims, 0, stream>>>(x, dst, ncols);
 }
 
+static void quantize_row_q8_1_cuda(const float * x, void * vy, const int ndata, const int k, cudaStream_t stream) {
+    const int num_blocks = (k + CUDA_QUANTIZE_BLOCK_SIZE - 1) / CUDA_QUANTIZE_BLOCK_SIZE;
+    quantize_q8_1<<<num_blocks, CUDA_QUANTIZE_BLOCK_SIZE, 0, stream>>>(x, vy, ndata, k);
+}
+
 static void dequantize_row_q4_0_cuda(const void * vx, float * y, const int k, cudaStream_t stream) {
     const int num_blocks = (k + CUDA_DEQUANTIZE_BLOCK_SIZE - 1) / CUDA_DEQUANTIZE_BLOCK_SIZE;
     dequantize_block<QK4_0, QR4_0, dequantize_q4_0><<<num_blocks, CUDA_DEQUANTIZE_BLOCK_SIZE, 0, stream>>>(vx, y, k);
@@ -1618,45 +1844,45 @@ static void dequantize_row_q6_K_cuda(const void * vx, float * y, const int k, cu
 
 static void dequantize_mul_mat_vec_q4_0_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
     GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
-    const int block_num_y = (nrows + GGML_CUDA_DMMV_Y - 1) / GGML_CUDA_DMMV_Y;
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
     const dim3 block_nums(1, block_num_y, 1);
-    const dim3 block_dims(WARP_SIZE, GGML_CUDA_DMMV_Y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
     dequantize_mul_mat_vec<QK4_0, QR4_0, dequantize_q4_0>
         <<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
 }
 
 static void dequantize_mul_mat_vec_q4_1_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
     GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
-    const int block_num_y = (nrows + GGML_CUDA_DMMV_Y - 1) / GGML_CUDA_DMMV_Y;
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
     const dim3 block_nums(1, block_num_y, 1);
-    const dim3 block_dims(WARP_SIZE, GGML_CUDA_DMMV_Y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
     dequantize_mul_mat_vec<QK4_1, QR4_1, dequantize_q4_1>
         <<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
 }
 
 static void dequantize_mul_mat_vec_q5_0_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
     GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
-    const int block_num_y = (nrows + GGML_CUDA_DMMV_Y - 1) / GGML_CUDA_DMMV_Y;
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
     const dim3 block_nums(1, block_num_y, 1);
-    const dim3 block_dims(WARP_SIZE, GGML_CUDA_DMMV_Y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
     dequantize_mul_mat_vec<QK5_0, QR5_0, dequantize_q5_0>
         <<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
 }
 
 static void dequantize_mul_mat_vec_q5_1_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
     GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
-    const int block_num_y = (nrows + GGML_CUDA_DMMV_Y - 1) / GGML_CUDA_DMMV_Y;
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
     const dim3 block_nums(1, block_num_y, 1);
-    const dim3 block_dims(WARP_SIZE, GGML_CUDA_DMMV_Y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
     dequantize_mul_mat_vec<QK5_1, QR5_1, dequantize_q5_1>
         <<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
 }
 
 static void dequantize_mul_mat_vec_q8_0_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
     GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
-    const int block_num_y = (nrows + GGML_CUDA_DMMV_Y - 1) / GGML_CUDA_DMMV_Y;
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
     const dim3 block_nums(1, block_num_y, 1);
-    const dim3 block_dims(WARP_SIZE, GGML_CUDA_DMMV_Y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
     dequantize_mul_mat_vec<QK8_0, QR8_0, dequantize_q8_0>
         <<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
 }
@@ -1703,6 +1929,51 @@ static void dequantize_mul_mat_vec_q6_K_cuda(const void * vx, const float * y, f
     dequantize_mul_mat_vec_q6_k<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
 }
 
+static void mul_mat_vec_q4_0_q8_1_cuda(const void * vx, const void * vy, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
+    GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
+    const dim3 block_nums(1, block_num_y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
+    mul_mat_vec_q<QK4_0, QI4_0, block_q4_0, vec_dot_q4_0_q8_1>
+        <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols, nrows);
+}
+
+static void mul_mat_vec_q4_1_q8_1_cuda(const void * vx, const void * vy, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
+    GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
+    const dim3 block_nums(1, block_num_y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
+    mul_mat_vec_q<QK4_0, QI4_1, block_q4_1, vec_dot_q4_1_q8_1>
+        <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols, nrows);
+}
+
+static void mul_mat_vec_q5_0_q8_1_cuda(const void * vx, const void * vy, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
+    GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
+    const dim3 block_nums(1, block_num_y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
+    mul_mat_vec_q<QK5_0, QI5_0, block_q5_0, vec_dot_q5_0_q8_1>
+        <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols, nrows);
+}
+
+static void mul_mat_vec_q5_1_q8_1_cuda(const void * vx, const void * vy, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
+    GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
+    const dim3 block_nums(1, block_num_y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
+    mul_mat_vec_q<QK5_1, QI5_1, block_q5_1, vec_dot_q5_1_q8_1>
+        <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols, nrows);
+}
+
+static void mul_mat_vec_q8_0_q8_1_cuda(const void * vx, const void * vy, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
+    GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
+    const dim3 block_nums(1, block_num_y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
+    mul_mat_vec_q<QK8_0, QI8_0, block_q8_0, vec_dot_q8_0_q8_1>
+        <<<block_nums, block_dims, 0, stream>>>(vx, vy, dst, ncols, nrows);
+}
+
 static void convert_fp16_to_fp32_cuda(const void * vx, float * y, const int k, cudaStream_t stream) {
     const int num_blocks = (k + CUDA_DEQUANTIZE_BLOCK_SIZE - 1) / CUDA_DEQUANTIZE_BLOCK_SIZE;
     dequantize_block<1, 1, convert_f16><<<num_blocks, CUDA_DEQUANTIZE_BLOCK_SIZE, 0, stream>>>(vx, y, k);
@@ -1710,9 +1981,9 @@ static void convert_fp16_to_fp32_cuda(const void * vx, float * y, const int k, c
 
 static void convert_mul_mat_vec_f16_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
     GGML_ASSERT(ncols % GGML_CUDA_DMMV_X == 0);
-    const int block_num_y = (nrows + GGML_CUDA_DMMV_Y - 1) / GGML_CUDA_DMMV_Y;
+    const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
     const dim3 block_nums(1, block_num_y, 1);
-    const dim3 block_dims(WARP_SIZE, GGML_CUDA_DMMV_Y, 1);
+    const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
     dequantize_mul_mat_vec<1, 1, convert_f16>
         <<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
 }
@@ -1878,6 +2149,7 @@ static size_t g_scratch_offset = 0;
 
 static int g_device_count = -1;
 static int g_main_device = 0;
+static int g_compute_capabilities[GGML_CUDA_MAX_DEVICES];
 static float g_tensor_split[GGML_CUDA_MAX_DEVICES] = {0};
 
 static cublasHandle_t g_cublas_handles[GGML_CUDA_MAX_DEVICES] = {nullptr};
@@ -1895,9 +2167,12 @@ void ggml_init_cublas() {
         for (int id = 0; id < g_device_count; ++id) {
             cudaDeviceProp prop;
             CUDA_CHECK(cudaGetDeviceProperties(&prop, id));
-            fprintf(stderr, "  Device %d: %s\n", id, prop.name);
+            fprintf(stderr, "  Device %d: %s, compute capability %d.%d\n", id, prop.name, prop.major, prop.minor);
+
             g_tensor_split[id] = total_vram;
             total_vram += prop.totalGlobalMem;
+
+            g_compute_capabilities[id] = 100*prop.major + 10*prop.minor;
         }
         for (int id = 0; id < g_device_count; ++id) {
             g_tensor_split[id] /= total_vram;
@@ -2113,7 +2388,7 @@ inline void ggml_cuda_op_rms_norm(
     (void) i1;
 }
 
-inline void ggml_cuda_op_dequantize_mul_mat_vec(
+inline void ggml_cuda_op_mul_mat_vec(
     const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst, char * src0_ddq_i,
     float * src0_ddf_i, float * src1_ddf_i, float * dst_ddf_i, int64_t i02, int64_t i01_low, int64_t i01_high, int i1,
     cudaStream_t & cudaStream_main){
@@ -2125,69 +2400,115 @@ inline void ggml_cuda_op_dequantize_mul_mat_vec(
     const int64_t ne00 = src0->ne[0];
     const int64_t nrows = i01_high - i01_low;
 
-// on some GPUs it is faster to convert src1 to half and to use half precision intrinsics
-#ifdef GGML_CUDA_DMMV_F16
-    size_t ash;
-    dfloat * src1_dfloat = nullptr; // dfloat == half
+#ifdef GGML_CUDA_FORCE_DMMV
+    const bool use_mul_mat_vec_q = false;
+#else
+    int id;
+    CUDA_CHECK(cudaGetDevice(&id));
 
-    bool src1_convert_f16 = src0->type == GGML_TYPE_Q4_0 || src0->type == GGML_TYPE_Q4_1 ||
-        src0->type == GGML_TYPE_Q5_0 || src0->type == GGML_TYPE_Q5_1 ||
-        src0->type == GGML_TYPE_Q8_0 || src0->type == GGML_TYPE_F16;
+    const bool mul_mat_vec_q_implemented = src0->type == GGML_TYPE_Q4_0 ||
+        src0->type == GGML_TYPE_Q4_1 ||
+        src0->type == GGML_TYPE_Q5_0 ||
+        src0->type == GGML_TYPE_Q5_1 ||
+        src0->type == GGML_TYPE_Q8_0;
 
-    if (src1_convert_f16) {
-        src1_dfloat = (half *) ggml_cuda_pool_malloc(ne00*sizeof(half), &ash);
-        ggml_cpy_f32_f16_cuda((char *) src1_ddf_i, (char *) src1_dfloat, ne00,
-                                ne00, 1, sizeof(float), 0, 0,
-                                ne00, 1, sizeof(half),  0, 0, cudaStream_main);
-    }
+    const bool use_mul_mat_vec_q = g_compute_capabilities[id] >= 600 && mul_mat_vec_q_implemented;
+#endif
+
+    if (use_mul_mat_vec_q) {
+        int64_t padded_row_size = ne00 + MATRIX_ROW_PADDING - 1;
+        padded_row_size -= padded_row_size % MATRIX_ROW_PADDING;
+        size_t as;
+        void * src1_q8_1 = ggml_cuda_pool_malloc(padded_row_size*sizeof(block_q8_1)/QK8_1, &as);
+        quantize_row_q8_1_cuda(src1_ddf_i, src1_q8_1, ne00, padded_row_size, cudaStream_main);
+
+        switch (src0->type) {
+            case GGML_TYPE_Q4_0:
+                mul_mat_vec_q4_0_q8_1_cuda(src0_ddq_i, src1_q8_1, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q4_1:
+                mul_mat_vec_q4_1_q8_1_cuda(src0_ddq_i, src1_q8_1, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q5_0:
+                mul_mat_vec_q5_0_q8_1_cuda(src0_ddq_i, src1_q8_1, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q5_1:
+                mul_mat_vec_q5_1_q8_1_cuda(src0_ddq_i, src1_q8_1, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q8_0:
+                mul_mat_vec_q8_0_q8_1_cuda(src0_ddq_i, src1_q8_1, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            default:
+                GGML_ASSERT(false);
+                break;
+        }
+
+        ggml_cuda_pool_free(src1_q8_1, as);
+    } else {
+        // on some GPUs it is faster to convert src1 to half and to use half precision intrinsics
+#ifdef GGML_CUDA_DMMV_F16
+        size_t ash;
+        dfloat * src1_dfloat = nullptr; // dfloat == half
+
+        bool src1_convert_f16 = src0->type == GGML_TYPE_Q4_0 || src0->type == GGML_TYPE_Q4_1 ||
+            src0->type == GGML_TYPE_Q5_0 || src0->type == GGML_TYPE_Q5_1 ||
+            src0->type == GGML_TYPE_Q8_0 || src0->type == GGML_TYPE_F16;
+
+        if (src1_convert_f16) {
+            src1_dfloat = (half *) ggml_cuda_pool_malloc(ne00*sizeof(half), &ash);
+            ggml_cpy_f32_f16_cuda((char *) src1_ddf_i, (char *) src1_dfloat, ne00,
+                                    ne00, 1, sizeof(float), 0, 0,
+                                    ne00, 1, sizeof(half),  0, 0, cudaStream_main);
+        }
 #else
-    dfloat * src1_dfloat = src1_ddf_i; // dfloat == float, no conversion
+        dfloat * src1_dfloat = src1_ddf_i; // dfloat == float, no conversion
 #endif // GGML_CUDA_DMMV_F16
 
-    switch (src0->type) {
-        case GGML_TYPE_Q4_0:
-            dequantize_mul_mat_vec_q4_0_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        case GGML_TYPE_Q4_1:
-            dequantize_mul_mat_vec_q4_1_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        case GGML_TYPE_Q5_0:
-            dequantize_mul_mat_vec_q5_0_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        case GGML_TYPE_Q5_1:
-            dequantize_mul_mat_vec_q5_1_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        case GGML_TYPE_Q8_0:
-            dequantize_mul_mat_vec_q8_0_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        case GGML_TYPE_Q2_K:
-            dequantize_mul_mat_vec_q2_K_cuda(src0_ddq_i, src1_ddf_i, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        case GGML_TYPE_Q3_K:
-            dequantize_mul_mat_vec_q3_K_cuda(src0_ddq_i, src1_ddf_i, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        case GGML_TYPE_Q4_K:
-            dequantize_mul_mat_vec_q4_K_cuda(src0_ddq_i, src1_ddf_i, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        case GGML_TYPE_Q5_K:
-            dequantize_mul_mat_vec_q5_K_cuda(src0_ddq_i, src1_ddf_i, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        case GGML_TYPE_Q6_K:
-            dequantize_mul_mat_vec_q6_K_cuda(src0_ddq_i, src1_ddf_i, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        case GGML_TYPE_F16:
-            convert_mul_mat_vec_f16_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
-            break;
-        default:
-            GGML_ASSERT(false);
-            break;
-    }
+        switch (src0->type) {
+            case GGML_TYPE_Q4_0:
+                dequantize_mul_mat_vec_q4_0_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q4_1:
+                dequantize_mul_mat_vec_q4_1_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q5_0:
+                dequantize_mul_mat_vec_q5_0_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q5_1:
+                dequantize_mul_mat_vec_q5_1_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q8_0:
+                dequantize_mul_mat_vec_q8_0_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q2_K:
+                dequantize_mul_mat_vec_q2_K_cuda(src0_ddq_i, src1_ddf_i, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q3_K:
+                dequantize_mul_mat_vec_q3_K_cuda(src0_ddq_i, src1_ddf_i, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q4_K:
+                dequantize_mul_mat_vec_q4_K_cuda(src0_ddq_i, src1_ddf_i, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q5_K:
+                dequantize_mul_mat_vec_q5_K_cuda(src0_ddq_i, src1_ddf_i, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_Q6_K:
+                dequantize_mul_mat_vec_q6_K_cuda(src0_ddq_i, src1_ddf_i, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            case GGML_TYPE_F16:
+                convert_mul_mat_vec_f16_cuda(src0_ddq_i, src1_dfloat, dst_ddf_i, ne00, nrows, cudaStream_main);
+                break;
+            default:
+                GGML_ASSERT(false);
+                break;
+        }
 
 #ifdef GGML_CUDA_DMMV_F16
-    if (src1_convert_f16) {
-        ggml_cuda_pool_free(src1_dfloat, ash);
-    }
+        if (src1_convert_f16) {
+            ggml_cuda_pool_free(src1_dfloat, ash);
+        }
 #endif // GGML_CUDA_DMMV_F16
+    }
 
     (void) src1;
     (void) dst;
@@ -2757,8 +3078,8 @@ void ggml_cuda_mul_mat(const ggml_tensor * src0, const ggml_tensor * src1, ggml_
     }else if (src0->type == GGML_TYPE_F32) {
         ggml_cuda_op(src0, src1, dst, ggml_cuda_op_mul_mat_cublas, true, false);
     } else if (ggml_is_quantized(src0->type) || src0->type == GGML_TYPE_F16) {
-        if (src1->ne[1] == 1 && src0->ne[0] % GGML_CUDA_DMMV_X == 0 && src0->ne[1] % GGML_CUDA_DMMV_Y == 0) {
-            ggml_cuda_op(src0, src1, dst, ggml_cuda_op_dequantize_mul_mat_vec, false, false);
+        if (src1->ne[1] == 1 && src0->ne[0] % GGML_CUDA_DMMV_X == 0) {
+            ggml_cuda_op(src0, src1, dst, ggml_cuda_op_mul_mat_vec, false, false);
         } else {
             ggml_cuda_op(src0, src1, dst, ggml_cuda_op_mul_mat_cublas, true, false);
         }
@@ -2843,7 +3164,11 @@ void ggml_cuda_nop(const ggml_tensor * src0, const ggml_tensor * src1, ggml_tens
 
 void ggml_cuda_transform_tensor(void * data, struct ggml_tensor * tensor) {
     int nrows = ggml_nrows(tensor);
+
+    const int64_t ne0 = tensor->ne[0];
+
     const size_t nb1 = tensor->nb[1];
+
     ggml_backend backend = tensor->backend;
     struct ggml_tensor_extra_gpu * extra = new struct ggml_tensor_extra_gpu;
     memset(extra, 0, sizeof(*extra));
@@ -2872,11 +3197,24 @@ void ggml_cuda_transform_tensor(void * data, struct ggml_tensor * tensor) {
         int64_t nrows_split = row_high - row_low;
 
         const size_t offset_split = row_low*nb1;
-        const size_t size = ggml_nbytes_split(tensor, nrows_split);
+        size_t size = ggml_nbytes_split(tensor, nrows_split);
+        const size_t original_size = size;
 
-        void * buf;
+        // pad last row to a multiple of 256 elements to avoid out-of-bounds memory accesses
+        if (ne0 % MATRIX_ROW_PADDING != 0) {
+            size += (MATRIX_ROW_PADDING - ne0 % MATRIX_ROW_PADDING)
+                * ggml_type_size(tensor->type)/ggml_blck_size(tensor->type);
+        }
+
+        char * buf;
         CUDA_CHECK(cudaMalloc(&buf, size));
-        void * buf_host = (char*)data + offset_split;
+        char * buf_host = (char*)data + offset_split;
+
+        // set padding to 0 to avoid possible NaN values
+        if (size > original_size) {
+            CUDA_CHECK(cudaMemset(buf + original_size, 0, size - original_size));
+        }
+
 
         cudaMemcpy(buf, buf_host, size, cudaMemcpyHostToDevice);
 
diff --git a/ggml-metal.h b/ggml-metal.h
index b9e50ac745eb0..928f1705c381c 100644
--- a/ggml-metal.h
+++ b/ggml-metal.h
@@ -34,9 +34,13 @@ extern "C" {
 
 struct ggml_metal_context;
 
-struct ggml_metal_context * ggml_metal_init(void);
+// number of command buffers to use
+struct ggml_metal_context * ggml_metal_init(int n_cb);
 void ggml_metal_free(struct ggml_metal_context * ctx);
 
+// set the number of command buffers to use
+void ggml_metal_set_n_cb(struct ggml_metal_context * ctx, int n_cb);
+
 // creates a mapping between a host memory buffer and a device memory buffer
 // - make sure to map all buffers used in the graph before calling ggml_metal_graph_compute
 // - the mapping is used during computation to determine the arguments of the compute kernels
diff --git a/ggml-metal.m b/ggml-metal.m
index fd69c41fe357d..3f15f791f9f65 100644
--- a/ggml-metal.m
+++ b/ggml-metal.m
@@ -25,6 +25,8 @@
 };
 
 struct ggml_metal_context {
+    int n_cb;
+
     float * logits;
 
     id<MTLDevice>       device;
@@ -86,11 +88,12 @@ @interface GGMLMetalClass : NSObject
 @implementation GGMLMetalClass
 @end
 
-struct ggml_metal_context * ggml_metal_init(void) {
+struct ggml_metal_context * ggml_metal_init(int n_cb) {
     fprintf(stderr, "%s: allocating\n", __func__);
 
     struct ggml_metal_context * ctx = malloc(sizeof(struct ggml_metal_context));
 
+    ctx->n_cb   = n_cb;
     ctx->device = MTLCreateSystemDefaultDevice();
     ctx->queue  = [ctx->device newCommandQueue];
     ctx->n_buffers = 0;
@@ -208,6 +211,10 @@ void ggml_metal_free(struct ggml_metal_context * ctx) {
     free(ctx);
 }
 
+void ggml_metal_set_n_cb(struct ggml_metal_context * ctx, int n_cb) {
+    ctx->n_cb = n_cb;
+}
+
 // finds the Metal buffer that contains the tensor data on the GPU device
 // the assumption is that there is 1-to-1 mapping between the host and device memory buffers, so we can find the
 // Metal buffer based on the host memory pointer
@@ -354,7 +361,7 @@ void ggml_metal_graph_compute(
     // create multiple command buffers and enqueue them
     // then, we encode the graph into the command buffers in parallel
 
-    const int n_cb = gf->n_threads;
+    const int n_cb = ctx->n_cb;
 
     NSMutableArray * command_buffers = [NSMutableArray arrayWithCapacity:n_cb];
 
diff --git a/ggml-opencl.cpp b/ggml-opencl.cpp
index fa0bdbefb1de4..eb214a836489b 100644
--- a/ggml-opencl.cpp
+++ b/ggml-opencl.cpp
@@ -653,13 +653,17 @@ __kernel void dequantize_mul_mat_vec_q6_K(__global const struct block_q6_K * xx,
     const int im = tid/step;                             // 0 or 1. 0 computes 0..., 1 computes 128...
     const int in = tid - step*im;                        // 0...15 or 0...7
 
-#if K_QUANTS_PER_ITERATION == 1
+\n#if K_QUANTS_PER_ITERATION == 1\n
     const int l0 = K_QUANTS_PER_ITERATION*in;            // 0...15
     const int is = 0;
-#else
+
+\n#else\n
+
     const int l0 = 4 * in;                               // 0, 4, 8, ..., 28
     const int is = in / 4;
-#endif
+
+\n#endif\n
+
     const int ql_offset = 64*im + l0;
     const int qh_offset = 32*im + l0;
     const int s_offset  =  8*im + is;
@@ -676,7 +680,7 @@ __kernel void dequantize_mul_mat_vec_q6_K(__global const struct block_q6_K * xx,
 
         const float d = vload_half(0, &x[i].d);
 
-#if K_QUANTS_PER_ITERATION == 1
+\n#if K_QUANTS_PER_ITERATION == 1\n
         float sum = y[ 0] * s[0] * d * ((int8_t)((ql[ 0] & 0xF) | ((qh[ 0] & 0x03) << 4)) - 32)
                   + y[16] * s[1] * d * ((int8_t)((ql[16] & 0xF) | ((qh[16] & 0x03) << 4)) - 32)
                   + y[32] * s[2] * d * ((int8_t)((ql[32] & 0xF) | ((qh[ 0] & 0x0c) << 2)) - 32)
@@ -686,7 +690,7 @@ __kernel void dequantize_mul_mat_vec_q6_K(__global const struct block_q6_K * xx,
                   + y[96] * s[6] * d * ((int8_t)((ql[32]  >> 4) | ((qh[ 0] & 0xc0) >> 2)) - 32)
                   +y[112] * s[7] * d * ((int8_t)((ql[48]  >> 4) | ((qh[16] & 0xc0) >> 2)) - 32);
         tmp[16 * ix + tid] += sum;
-#else
+\n#else\n
         float sum = 0;
         for (int l = 0; l < 4; ++l) {
             sum += y[l+ 0] * s[0] * d * ((int8_t)((ql[l+ 0] & 0xF) | (((qh[l] >> 0) & 3) << 4)) - 32)
@@ -695,7 +699,7 @@ __kernel void dequantize_mul_mat_vec_q6_K(__global const struct block_q6_K * xx,
                  + y[l+96] * s[6] * d * ((int8_t)((ql[l+32]  >> 4) | (((qh[l] >> 6) & 3) << 4)) - 32);
         }
         tmp[16 * ix + tid] += sum;
-#endif
+\n#endif\n
 
     }
 
diff --git a/ggml.c b/ggml.c
index a8ea0e9f3e537..c10877a761dd8 100644
--- a/ggml.c
+++ b/ggml.c
@@ -220,21 +220,41 @@ inline static void* ggml_aligned_malloc(size_t size) {
 #define GGML_ALIGNED_FREE(ptr)     free(ptr)
 #endif
 
-#define UNUSED(x) (void)(x)
+#define UNUSED GGML_UNUSED
 #define SWAP(x, y, T) do { T SWAP = x; x = y; y = SWAP; } while (0)
 
+//
+// tensor access macros
+//
+
+#define GGML_TENSOR_UNARY_OP_LOCALS \
+    GGML_TENSOR_LOCALS(int64_t, ne0, src0, ne); \
+    GGML_TENSOR_LOCALS(size_t,  nb0, src0, nb); \
+    GGML_TENSOR_LOCALS(int64_t, ne,  dst,  ne); \
+    GGML_TENSOR_LOCALS(size_t,  nb,  dst,  nb);
+
+#define GGML_TENSOR_BINARY_OP_LOCALS \
+    GGML_TENSOR_LOCALS(int64_t, ne0, src0, ne); \
+    GGML_TENSOR_LOCALS(size_t,  nb0, src0, nb); \
+    GGML_TENSOR_LOCALS(int64_t, ne1, src1, ne); \
+    GGML_TENSOR_LOCALS(size_t,  nb1, src1, nb); \
+    GGML_TENSOR_LOCALS(int64_t, ne,  dst,  ne); \
+    GGML_TENSOR_LOCALS(size_t,  nb,  dst,  nb);
+
 #if defined(GGML_USE_ACCELERATE)
 #include <Accelerate/Accelerate.h>
 #if defined(GGML_USE_CLBLAST) // allow usage of CLBlast alongside Accelerate functions
 #include "ggml-opencl.h"
 #endif
 #elif defined(GGML_USE_OPENBLAS)
+#if defined(GGML_BLAS_USE_MKL)
+#include <mkl.h>
+#else
 #include <cblas.h>
 #endif
-#if defined(GGML_USE_CUBLAS)
+#elif defined(GGML_USE_CUBLAS)
 #include "ggml-cuda.h"
-#endif
-#if defined(GGML_USE_CLBLAST)
+#elif defined(GGML_USE_CLBLAST)
 #include "ggml-opencl.h"
 #endif
 
@@ -465,14 +485,14 @@ ggml_fp16_t ggml_fp32_to_fp16(float x) {
     return GGML_FP32_TO_FP16(x);
 }
 
-void ggml_fp16_to_fp32_row(const ggml_fp16_t * x, float * y, size_t n) {
-    for (size_t i = 0; i < n; i++) {
+void ggml_fp16_to_fp32_row(const ggml_fp16_t * x, float * y, int n) {
+    for (int i = 0; i < n; i++) {
         y[i] = GGML_FP16_TO_FP32(x[i]);
     }
 }
 
-void ggml_fp32_to_fp16_row(const float * x, ggml_fp16_t * y, size_t n) {
-    size_t i = 0;
+void ggml_fp32_to_fp16_row(const float * x, ggml_fp16_t * y, int n) {
+    int i = 0;
 #if defined(__F16C__)
     for (; i + 7 < n; i += 8) {
         __m256 x_vec = _mm256_loadu_ps(x + i);
@@ -1611,109 +1631,112 @@ static void dequantize_row_q8_0(const void * restrict vx, float * restrict y, in
     }
 }
 
+static void ggml_vec_dot_f32(const int n, float * restrict s, const float * restrict x, const float * restrict y);
+static void ggml_vec_dot_f16(const int n, float * restrict s, ggml_fp16_t * restrict x, ggml_fp16_t * restrict y);
 static void ggml_vec_dot_q4_0_q8_0(const int n, float * restrict s, const void * restrict vx, const void * restrict vy);
 static void ggml_vec_dot_q4_1_q8_1(const int n, float * restrict s, const void * restrict vx, const void * restrict vy);
 static void ggml_vec_dot_q5_0_q8_0(const int n, float * restrict s, const void * restrict vx, const void * restrict vy);
 static void ggml_vec_dot_q5_1_q8_1(const int n, float * restrict s, const void * restrict vx, const void * restrict vy);
 static void ggml_vec_dot_q8_0_q8_0(const int n, float * restrict s, const void * restrict vx, const void * restrict vy);
 
-static const quantize_fns_t quantize_fns[GGML_TYPE_COUNT] = {
+static const ggml_type_traits_t type_traits[GGML_TYPE_COUNT] = {
+    [GGML_TYPE_F32] = {
+        .vec_dot                  = (ggml_vec_dot_t) ggml_vec_dot_f32,
+        .vec_dot_type             = GGML_TYPE_F32,
+    },
+    [GGML_TYPE_F16] = {
+        .to_float                 = (ggml_to_float_t) ggml_fp16_to_fp32_row,
+        .from_float               = (ggml_from_float_t) ggml_fp32_to_fp16_row,
+        .from_float_reference     = (ggml_from_float_t) ggml_fp32_to_fp16_row,
+        .vec_dot                  = (ggml_vec_dot_t) ggml_vec_dot_f16,
+        .vec_dot_type             = GGML_TYPE_F16,
+    },
     [GGML_TYPE_Q4_0] = {
-        .dequantize_row_q         = (dequantize_row_q_t) dequantize_row_q4_0,
-        .quantize_row_q           = quantize_row_q4_0,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q4_0_reference,
-        .quantize_row_q_dot       = quantize_row_q8_0,
-        .vec_dot_q                = ggml_vec_dot_q4_0_q8_0,
+        .to_float                 = (ggml_to_float_t) dequantize_row_q4_0,
+        .from_float               = quantize_row_q4_0,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q4_0_reference,
+        .vec_dot                  = ggml_vec_dot_q4_0_q8_0,
         .vec_dot_type             = GGML_TYPE_Q8_0,
     },
     [GGML_TYPE_Q4_1] = {
-        .dequantize_row_q         = (dequantize_row_q_t)dequantize_row_q4_1,
-        .quantize_row_q           = quantize_row_q4_1,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q4_1_reference,
-        .quantize_row_q_dot       = quantize_row_q8_1,
-        .vec_dot_q                = ggml_vec_dot_q4_1_q8_1,
+        .to_float                 = (ggml_to_float_t) dequantize_row_q4_1,
+        .from_float               = quantize_row_q4_1,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q4_1_reference,
+        .vec_dot                  = ggml_vec_dot_q4_1_q8_1,
         .vec_dot_type             = GGML_TYPE_Q8_1,
     },
     [GGML_TYPE_Q5_0] = {
-        .dequantize_row_q         = (dequantize_row_q_t) dequantize_row_q5_0,
-        .quantize_row_q           = quantize_row_q5_0,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q5_0_reference,
-        .quantize_row_q_dot       = quantize_row_q8_0,
-        .vec_dot_q                = ggml_vec_dot_q5_0_q8_0,
+        .to_float                 = (ggml_to_float_t) dequantize_row_q5_0,
+        .from_float               = quantize_row_q5_0,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q5_0_reference,
+        .vec_dot                  = ggml_vec_dot_q5_0_q8_0,
         .vec_dot_type             = GGML_TYPE_Q8_0,
     },
     [GGML_TYPE_Q5_1] = {
-        .dequantize_row_q         = (dequantize_row_q_t) dequantize_row_q5_1,
-        .quantize_row_q           = quantize_row_q5_1,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q5_1_reference,
-        .quantize_row_q_dot       = quantize_row_q8_1,
-        .vec_dot_q                = ggml_vec_dot_q5_1_q8_1,
+        .to_float                 = (ggml_to_float_t) dequantize_row_q5_1,
+        .from_float               = quantize_row_q5_1,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q5_1_reference,
+        .vec_dot                  = ggml_vec_dot_q5_1_q8_1,
         .vec_dot_type             = GGML_TYPE_Q8_1,
     },
     [GGML_TYPE_Q8_0] = {
-        .dequantize_row_q         = dequantize_row_q8_0,
-        .quantize_row_q           = quantize_row_q8_0,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q8_0_reference,
-        .quantize_row_q_dot       = quantize_row_q8_0,
-        .vec_dot_q                = ggml_vec_dot_q8_0_q8_0,
+        .to_float                 = dequantize_row_q8_0,
+        .from_float               = quantize_row_q8_0,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q8_0_reference,
+        .vec_dot                  = ggml_vec_dot_q8_0_q8_0,
         .vec_dot_type             = GGML_TYPE_Q8_0,
     },
     [GGML_TYPE_Q8_1] = {
-        .dequantize_row_q         = NULL,   // TODO
-        .quantize_row_q           = quantize_row_q8_1,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q8_1_reference,
-        .quantize_row_q_dot       = quantize_row_q8_1,
-        .vec_dot_q                = NULL,   // TODO
+        .from_float               = quantize_row_q8_1,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q8_1_reference,
         .vec_dot_type             = GGML_TYPE_Q8_1,
     },
 #ifdef GGML_USE_K_QUANTS
     [GGML_TYPE_Q2_K] = {
-        .dequantize_row_q         = (dequantize_row_q_t) dequantize_row_q2_K,
-        .quantize_row_q           = quantize_row_q2_K,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q2_K_reference,
-        .quantize_row_q_dot       = quantize_row_q8_K,
-        .vec_dot_q                = ggml_vec_dot_q2_K_q8_K,
+        .to_float                 = (ggml_to_float_t) dequantize_row_q2_K,
+        .from_float               = quantize_row_q2_K,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q2_K_reference,
+        .vec_dot                  = ggml_vec_dot_q2_K_q8_K,
         .vec_dot_type             = GGML_TYPE_Q8_K,
     },
     [GGML_TYPE_Q3_K] = {
-        .dequantize_row_q         = (dequantize_row_q_t) dequantize_row_q3_K,
-        .quantize_row_q           = quantize_row_q3_K,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q3_K_reference,
-        .quantize_row_q_dot       = quantize_row_q8_K,
-        .vec_dot_q                = ggml_vec_dot_q3_K_q8_K,
+        .to_float                 = (ggml_to_float_t) dequantize_row_q3_K,
+        .from_float               = quantize_row_q3_K,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q3_K_reference,
+        .vec_dot                  = ggml_vec_dot_q3_K_q8_K,
         .vec_dot_type             = GGML_TYPE_Q8_K,
     },
     [GGML_TYPE_Q4_K] = {
-        .dequantize_row_q         = (dequantize_row_q_t) dequantize_row_q4_K,
-        .quantize_row_q           = quantize_row_q4_K,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q4_K_reference,
-        .quantize_row_q_dot       = quantize_row_q8_K,
-        .vec_dot_q                = ggml_vec_dot_q4_K_q8_K,
+        .to_float                 = (ggml_to_float_t) dequantize_row_q4_K,
+        .from_float               = quantize_row_q4_K,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q4_K_reference,
+        .vec_dot                  = ggml_vec_dot_q4_K_q8_K,
         .vec_dot_type             = GGML_TYPE_Q8_K,
     },
     [GGML_TYPE_Q5_K] = {
-        .dequantize_row_q         = (dequantize_row_q_t) dequantize_row_q5_K,
-        .quantize_row_q           = quantize_row_q5_K,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q5_K_reference,
-        .quantize_row_q_dot       = quantize_row_q8_K,
-        .vec_dot_q                = ggml_vec_dot_q5_K_q8_K,
+        .to_float                 = (ggml_to_float_t) dequantize_row_q5_K,
+        .from_float               = quantize_row_q5_K,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q5_K_reference,
+        .vec_dot                  = ggml_vec_dot_q5_K_q8_K,
         .vec_dot_type             = GGML_TYPE_Q8_K,
     },
     [GGML_TYPE_Q6_K] = {
-        .dequantize_row_q         = (dequantize_row_q_t) dequantize_row_q6_K,
-        .quantize_row_q           = quantize_row_q6_K,
-        .quantize_row_q_reference = (quantize_row_q_t) quantize_row_q6_K_reference,
-        .quantize_row_q_dot       = quantize_row_q8_K,
-        .vec_dot_q                = ggml_vec_dot_q6_K_q8_K,
+        .to_float                 = (ggml_to_float_t) dequantize_row_q6_K,
+        .from_float               = quantize_row_q6_K,
+        .from_float_reference     = (ggml_from_float_t) quantize_row_q6_K_reference,
+        .vec_dot                  = ggml_vec_dot_q6_K_q8_K,
         .vec_dot_type             = GGML_TYPE_Q8_K,
     },
+    [GGML_TYPE_Q8_K] = {
+        .from_float               = quantize_row_q8_K,
+    }
 #endif
 };
 
 // For internal test use
-quantize_fns_t ggml_internal_get_quantize_fn(size_t i) {
+ggml_type_traits_t ggml_internal_get_type_traits(enum ggml_type i) {
     GGML_ASSERT(i < GGML_TYPE_COUNT);
-    return quantize_fns[i];
+    return type_traits[i];
 }
 
 
@@ -2259,7 +2282,7 @@ inline static void ggml_vec_neg_f32 (const int n, float * y, const float * x)
 inline static void ggml_vec_mul_f32 (const int n, float * z, const float * x, const float * y) { for (int i = 0; i < n; ++i) z[i]  = x[i]*y[i];   }
 inline static void ggml_vec_div_f32 (const int n, float * z, const float * x, const float * y) { for (int i = 0; i < n; ++i) z[i]  = x[i]/y[i];   }
 
-inline static void ggml_vec_dot_f32(const int n, float * restrict s, const float * restrict x, const float * restrict y) {
+static void ggml_vec_dot_f32(const int n, float * restrict s, const float * restrict x, const float * restrict y) {
 #ifdef GGML_SIMD
     float sumf = 0.0f;
     const int np = (n & ~(GGML_F32_STEP - 1));
@@ -2296,7 +2319,7 @@ inline static void ggml_vec_dot_f32(const int n, float * restrict s, const float
     *s = sumf;
 }
 
-inline static void ggml_vec_dot_f16(const int n, float * restrict s, ggml_fp16_t * restrict x, ggml_fp16_t * restrict y) {
+static void ggml_vec_dot_f16(const int n, float * restrict s, ggml_fp16_t * restrict x, ggml_fp16_t * restrict y) {
     ggml_float sumf = 0.0;
 
 #if defined(GGML_SIMD)
@@ -3449,6 +3472,8 @@ inline static void ggml_vec_log_f32  (const int n, float * y, const float * x) {
 inline static void ggml_vec_abs_f32  (const int n, float * y, const float * x) { for (int i = 0; i < n; ++i) y[i] = fabsf(x[i]); }
 inline static void ggml_vec_sgn_f32  (const int n, float * y, const float * x) { for (int i = 0; i < n; ++i) y[i] = (x[i] > 0.f) ? 1.f : ((x[i] < 0.f) ? -1.f : 0.f); }
 inline static void ggml_vec_step_f32 (const int n, float * y, const float * x) { for (int i = 0; i < n; ++i) y[i] = (x[i] > 0.f) ? 1.f : 0.f; }
+inline static void ggml_vec_tanh_f32 (const int n, float * y, const float * x) { for (int i = 0; i < n; ++i) y[i] = tanhf(x[i]);  }
+inline static void ggml_vec_elu_f32  (const int n, float * y, const float * x) { for (int i = 0; i < n; ++i) y[i] = (x[i] > 0.f) ? x[i] : expf(x[i])-1; }
 inline static void ggml_vec_relu_f32 (const int n, float * y, const float * x) { for (int i = 0; i < n; ++i) y[i] = (x[i] > 0.f) ? x[i] : 0.f; }
 
 static const float GELU_COEF_A    = 0.044715f;
@@ -3600,6 +3625,16 @@ inline static void ggml_vec_norm_inv_f32(const int n, float * s, const float * x
     *s = 1.f/(*s);
 }
 
+inline static void ggml_vec_argmax_f32(const int n, int * s, const float * x) {
+    float max = -INFINITY;
+    int idx = 0;
+    for (int i = 0; i < n; ++i) {
+        max = MAX(max, x[i]);
+        if (max == x[i]) { idx = i; }
+    }
+    *s = idx;
+}
+
 //
 // data types
 //
@@ -3709,12 +3744,15 @@ static const char * GGML_OP_NAME[GGML_OP_COUNT] = {
     "SUM",
     "SUM_ROWS",
     "MEAN",
+    "ARGMAX",
     "REPEAT",
     "REPEAT_BACK",
     "ABS",
     "SGN",
     "NEG",
     "STEP",
+    "TANH",
+    "ELU",
     "RELU",
     "GELU",
     "GELU_QUICK",
@@ -3746,9 +3784,8 @@ static const char * GGML_OP_NAME[GGML_OP_COUNT] = {
     "ROPE_BACK",
     "ALIBI",
     "CLAMP",
-    "CONV_1D_S1_PH",
-    "CONV_1D_S2_PH",
-    "CONV_2D_SK_P0",
+    "CONV_1D",
+    "CONV_2D",
 
     "FLASH_ATTN",
     "FLASH_FF",
@@ -3767,7 +3804,7 @@ static const char * GGML_OP_NAME[GGML_OP_COUNT] = {
     "CROSS_ENTROPY_LOSS_BACK",
 };
 
-static_assert(GGML_OP_COUNT == 64, "GGML_OP_COUNT != 64");
+static_assert(GGML_OP_COUNT == 66, "GGML_OP_COUNT != 66");
 
 static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "none",
@@ -3785,12 +3822,15 @@ static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "Σx",
     "Σx_k",
     "Σx/n",
+    "argmax(x)",
     "repeat(x)",
     "repeat_back(x)",
     "abs(x)",
     "sgn(x)",
     "-x",
     "step(x)",
+    "tanh(x)",
+    "elu(x)",
     "relu(x)",
     "gelu(x)",
     "gelu_quick(x)",
@@ -3822,9 +3862,8 @@ static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "rope_back(x)",
     "alibi(x)",
     "clamp(x)",
-    "conv_1d_s1_ph(x)",
-    "conv_1d_s2_ph(x)",
-    "conv_2d_sk_p0(x)",
+    "conv_1d(x)",
+    "conv_2d(x)",
 
     "flash_attn(x)",
     "flash_ff(x)",
@@ -3843,7 +3882,7 @@ static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "cross_entropy_loss_back(x,y)",
 };
 
-static_assert(GGML_OP_COUNT == 64, "GGML_OP_COUNT != 64");
+static_assert(GGML_OP_COUNT == 66, "GGML_OP_COUNT != 66");
 
 static_assert(sizeof(struct ggml_object)%GGML_MEM_ALIGN == 0, "ggml_object size must be a multiple of GGML_MEM_ALIGN");
 static_assert(sizeof(struct ggml_tensor)%GGML_MEM_ALIGN == 0, "ggml_tensor size must be a multiple of GGML_MEM_ALIGN");
@@ -3869,9 +3908,8 @@ static void ggml_setup_op_has_task_pass(void) {
         p[GGML_OP_GET_ROWS_BACK          ] = true;
         p[GGML_OP_DIAG_MASK_INF          ] = true;
         p[GGML_OP_DIAG_MASK_ZERO         ] = true;
-        p[GGML_OP_CONV_1D_S1_PH          ] = true;
-        p[GGML_OP_CONV_1D_S2_PH          ] = true;
-        p[GGML_OP_CONV_2D_SK_P0          ] = true;
+        p[GGML_OP_CONV_1D                ] = true;
+        p[GGML_OP_CONV_2D                ] = true;
         p[GGML_OP_FLASH_ATTN_BACK        ] = true;
         p[GGML_OP_CROSS_ENTROPY_LOSS     ] = true;
     }
@@ -4549,14 +4587,13 @@ struct ggml_tensor * ggml_new_tensor_impl(
         /*.src0         =*/ NULL,
         /*.src1         =*/ NULL,
         /*.opt          =*/ { NULL },
-        /*.n_tasks      =*/ 0,
         /*.perf_runs    =*/ 0,
         /*.perf_cycles  =*/ 0,
         /*.perf_time_us =*/ 0,
         /*.data         =*/ (data == NULL && !ctx->no_alloc) ? (void *)(result + 1) : data,
         /*.name         =*/ { 0 },
         /*.extra        =*/ NULL,
-        /*.pad          =*/ { 0 },
+        /*.padding      =*/ { 0 },
     };
 
     // TODO: this should not be needed as long as we don't rely on aligned SIMD loads
@@ -5442,6 +5479,30 @@ struct ggml_tensor * ggml_mean(
     return result;
 }
 
+// ggml_argmax
+
+struct ggml_tensor * ggml_argmax(
+        struct ggml_context * ctx,
+        struct ggml_tensor * a) {
+    GGML_ASSERT(ggml_is_matrix(a));
+    bool is_node = false;
+
+    if (a->grad) {
+        GGML_ASSERT(false);
+        is_node = true;
+    }
+
+    int64_t ne[GGML_MAX_DIMS] = { a->ne[1], 1, 1, 1 };
+    struct ggml_tensor * result = ggml_new_tensor(ctx, GGML_TYPE_I32, a->n_dims, ne);
+
+    result->op   = GGML_OP_ARGMAX;
+    result->grad = is_node ? ggml_dup_tensor(ctx, result) : NULL;
+    result->src0 = a;
+    result->src1 = NULL;
+
+    return result;
+}
+
 // ggml_repeat
 
 struct ggml_tensor * ggml_repeat(
@@ -5635,6 +5696,74 @@ struct ggml_tensor * ggml_step_inplace(
     return ggml_step_impl(ctx, a, true);
 }
 
+// ggml_tanh
+
+struct ggml_tensor * ggml_tanh_impl(
+        struct ggml_context * ctx,
+        struct ggml_tensor * a,
+        bool inplace) {
+    bool is_node = false;
+
+    if (!inplace && (a->grad)) {
+        is_node = true;
+    }
+
+    struct ggml_tensor * result = inplace ? ggml_view_tensor(ctx, a) : ggml_dup_tensor(ctx, a);
+
+    result->op   = GGML_OP_TANH;
+    result->grad = is_node ? ggml_dup_tensor(ctx, result) : NULL;
+    result->src0 = a;
+    result->src1 = NULL;
+
+    return result;
+}
+
+struct ggml_tensor * ggml_tanh(
+        struct ggml_context * ctx,
+        struct ggml_tensor  * a) {
+    return ggml_tanh_impl(ctx, a, false);
+}
+
+struct ggml_tensor * ggml_tanh_inplace(
+        struct ggml_context * ctx,
+        struct ggml_tensor  * a) {
+    return ggml_tanh_impl(ctx, a, true);
+}
+
+// ggml_elu
+
+struct ggml_tensor * ggml_elu_impl(
+        struct ggml_context * ctx,
+        struct ggml_tensor * a,
+        bool inplace) {
+    bool is_node = false;
+
+    if (!inplace && (a->grad)) {
+        is_node = true;
+    }
+
+    struct ggml_tensor * result = inplace ? ggml_view_tensor(ctx, a) : ggml_dup_tensor(ctx, a);
+
+    result->op   = GGML_OP_ELU;
+    result->grad = is_node ? ggml_dup_tensor(ctx, result) : NULL;
+    result->src0 = a;
+    result->src1 = NULL;
+
+    return result;
+}
+
+struct ggml_tensor * ggml_elu(
+    struct ggml_context * ctx,
+    struct ggml_tensor  * a) {
+    return ggml_elu_impl(ctx, a, false);
+}
+
+struct ggml_tensor * ggml_elu_inplace(
+    struct ggml_context * ctx,
+    struct ggml_tensor  * a) {
+    return ggml_elu_impl(ctx, a, true);
+}
+
 // ggml_relu
 
 struct ggml_tensor * ggml_relu_impl(
@@ -6876,6 +7005,8 @@ struct ggml_tensor * ggml_rope_back(
         int                   n_dims,
         int                   mode) {
     GGML_ASSERT(n_past >= 0);
+    GGML_ASSERT((mode & 4) == 0 && "ggml_rope_back() for ChatGLM not implemented yet");
+
     bool is_node = false;
 
     if (a->grad) {
@@ -6976,15 +7107,21 @@ struct ggml_tensor * ggml_clamp(
     return result;
 }
 
-// ggml_conv_1d_s1_ph
+// ggml_conv_1d
+
+static int64_t ggml_calc_conv_output_size(int64_t ins, int64_t ks, int s, int p, int d) {
+    return (ins + 2 * p - d * (ks - 1) - 1) / s + 1;
+}
 
-struct ggml_tensor * ggml_conv_1d_s1_ph(
+GGML_API struct ggml_tensor * ggml_conv_1d(
         struct ggml_context * ctx,
         struct ggml_tensor  * a,
-        struct ggml_tensor  * b) {
+        struct ggml_tensor  * b,
+        int                   s0,
+        int                   p0,
+        int                   d0) {
     GGML_ASSERT(ggml_is_matrix(b));
     GGML_ASSERT(a->ne[1] == b->ne[1]);
-    GGML_ASSERT(a->ne[3] == 1);
     bool is_node = false;
 
     if (a->grad || b->grad) {
@@ -6992,26 +7129,43 @@ struct ggml_tensor * ggml_conv_1d_s1_ph(
         is_node = true;
     }
 
-    const int64_t ne[4] = { b->ne[0], a->ne[2], 1, 1, };
-    struct ggml_tensor * result = ggml_new_tensor(ctx, GGML_TYPE_F32, 2, ne);
+    const int64_t ne[4] = {
+        ggml_calc_conv_output_size(b->ne[0], a->ne[0], s0, p0, d0),
+        a->ne[2], 1, 1,
+    };
+    struct ggml_tensor* result = ggml_new_tensor(ctx, GGML_TYPE_F32, 2, ne);
+
+    ggml_scratch_save(ctx);
+    struct ggml_tensor* c = ggml_new_tensor_1d(ctx, GGML_TYPE_I32, 3);
+    ((int32_t*)c->data)[0] = s0;
+    ((int32_t*)c->data)[1] = p0;
+    ((int32_t*)c->data)[2] = d0;
+    ggml_scratch_load(ctx);
 
-    result->op   = GGML_OP_CONV_1D_S1_PH;
+    result->op = GGML_OP_CONV_1D;
     result->grad = is_node ? ggml_dup_tensor(ctx, result) : NULL;
     result->src0 = a;
     result->src1 = b;
+    result->opt[0] = c;
 
     return result;
 }
 
-// ggml_conv_1d_s2_ph
+// ggml_conv_2d
 
-struct ggml_tensor * ggml_conv_1d_s2_ph(
-        struct ggml_context * ctx,
-        struct ggml_tensor  * a,
-        struct ggml_tensor  * b) {
-    GGML_ASSERT(ggml_is_matrix(b));
-    GGML_ASSERT(a->ne[1] == b->ne[1]);
-    GGML_ASSERT(a->ne[3] == 1);
+struct ggml_tensor* ggml_conv_2d(
+    struct ggml_context* ctx,
+    struct ggml_tensor * a,
+    struct ggml_tensor * b,
+    int                  s0,
+    int                  s1,
+    int                  p0,
+    int                  p1,
+    int                  d0,
+    int                  d1) {
+
+    GGML_ASSERT(b->ne[3] == 1);
+    GGML_ASSERT(a->ne[2] == b->ne[2]);
     bool is_node = false;
 
     if (a->grad || b->grad) {
@@ -7019,43 +7173,42 @@ struct ggml_tensor * ggml_conv_1d_s2_ph(
         is_node = true;
     }
 
-    const int64_t ne[4] = { b->ne[0]/2, a->ne[2], 1, 1, };
-    struct ggml_tensor * result = ggml_new_tensor(ctx, GGML_TYPE_F32, 2, ne);
+    const int64_t ne[4] = {
+        ggml_calc_conv_output_size(b->ne[0], a->ne[0], s0, p0, d0),
+        ggml_calc_conv_output_size(b->ne[1], a->ne[1], s1, p1, d1),
+        a->ne[3], 1,
+    };
+    struct ggml_tensor* result = ggml_new_tensor(ctx, GGML_TYPE_F32, 4, ne);
+
+    ggml_scratch_save(ctx);
+    struct ggml_tensor* c = ggml_new_tensor_1d(ctx, GGML_TYPE_I32, 6);
+    ((int32_t*)c->data)[0] = s0;
+    ((int32_t*)c->data)[1] = s1;
+    ((int32_t*)c->data)[2] = p0;
+    ((int32_t*)c->data)[3] = p1;
+    ((int32_t*)c->data)[4] = d0;
+    ((int32_t*)c->data)[5] = d1;
+    ggml_scratch_load(ctx);
 
-    result->op   = GGML_OP_CONV_1D_S2_PH;
+    result->op = GGML_OP_CONV_2D;
     result->grad = is_node ? ggml_dup_tensor(ctx, result) : NULL;
     result->src0 = a;
     result->src1 = b;
+    result->opt[0] = c;
 
     return result;
+
 }
 
-// ggml_conv_2d_sk_p0
+// ggml_conv_1d_ph
 
-struct ggml_tensor * ggml_conv_2d_sk_p0(
+struct ggml_tensor* ggml_conv_1d_ph(
         struct ggml_context * ctx,
         struct ggml_tensor  * a,
-        struct ggml_tensor  * b) {
-    GGML_ASSERT(b->ne[3] == 1);
-    GGML_ASSERT(a->ne[2] == b->ne[2]);
-    GGML_ASSERT(b->ne[0] %  a->ne[0] == 0);
-    GGML_ASSERT(b->ne[1] %  a->ne[1] == 0);
-    bool is_node = false;
-
-    if (a->grad || b->grad) {
-        GGML_ASSERT(false); // TODO: implement backward
-        is_node = true;
-    }
-
-    const int64_t ne[4] = { b->ne[0]/a->ne[0], b->ne[1]/a->ne[1], a->ne[3], 1, };
-    struct ggml_tensor * result = ggml_new_tensor(ctx, GGML_TYPE_F32, 4, ne);
-
-    result->op   = GGML_OP_CONV_2D_SK_P0;
-    result->grad = is_node ? ggml_dup_tensor(ctx, result) : NULL;
-    result->src0 = a;
-    result->src1 = b;
-
-    return result;
+        struct ggml_tensor  * b,
+        int                   s,
+        int                   d) {
+    return ggml_conv_1d(ctx, a, b, s, a->ne[0] / 2, d);
 }
 
 // ggml_flash_attn
@@ -7605,25 +7758,7 @@ static void ggml_compute_forward_dup_f16(
         return;
     }
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
-    const int64_t ne3 = dst->ne[3];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     const int ith = params->ith; // thread index
     const int nth = params->nth; // number of threads
@@ -7696,8 +7831,8 @@ static void ggml_compute_forward_dup_f16(
                         id += ne00 * (ne01 - ir1);
                     }
                 }
-            } else if (ggml_is_quantized(dst->type)) {
-                quantize_row_q_t const quantize_row_q = quantize_fns[dst->type].quantize_row_q;
+            } else if (type_traits[dst->type].from_float) {
+                ggml_from_float_t const quantize_row_q = type_traits[dst->type].from_float;
                 float * src0_f32 = (float *) params->wdata + (ne00 + CACHE_LINE_SIZE_F32) * ith;
 
                 size_t id = 0;
@@ -7894,25 +8029,7 @@ static void ggml_compute_forward_dup_f32(
         return;
     }
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
-    const int64_t ne3 = dst->ne[3];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     const int ith = params->ith; // thread index
     const int nth = params->nth; // number of threads
@@ -7967,26 +8084,8 @@ static void ggml_compute_forward_dup_f32(
                         id += rs * (ne01 - ir1);
                     }
                 }
-            } else if (dst->type == GGML_TYPE_F16) {
-                size_t id = 0;
-                ggml_fp16_t * dst_ptr = (ggml_fp16_t *) dst->data;
-
-                for (int i03 = 0; i03 < ne03; i03++) {
-                    for (int i02 = 0; i02 < ne02; i02++) {
-                        id += ne00 * ir0;
-                        for (int i01 = ir0; i01 < ir1; i01++) {
-                            for (int i00 = 0; i00 < ne00; i00++) {
-                                const float * src0_ptr = (float *) ((char *) src0->data + i00*nb00 + i01*nb01 + i02*nb02 + i03*nb03);
-
-                                dst_ptr[id] = GGML_FP32_TO_FP16(*src0_ptr);
-                                id++;
-                            }
-                        }
-                        id += ne00 * (ne01 - ir1);
-                    }
-                }
-            } else if (ggml_is_quantized(dst->type)) {
-                quantize_row_q_t const quantize_row_q = quantize_fns[dst->type].quantize_row_q;
+            } else if (type_traits[dst->type].from_float) {
+                ggml_from_float_t const quantize_row_q = type_traits[dst->type].from_float;
 
                 size_t id = 0;
                 size_t rs = nb0 * (ne00 / GGML_BLCK_SIZE[dst->type]);
@@ -8210,24 +8309,8 @@ static void ggml_compute_forward_add_f32(
     const int nth = params->nth;
 
     const int nr  = ggml_nrows(src0);
-    const int64_t ne0 = src0->ne[0];
-    const int64_t ne1 = src0->ne[1];
-    const int64_t ne2 = src0->ne[2];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
 
-    const size_t nb10 = src1->nb[0];
-    const size_t nb11 = src1->nb[1];
-    const size_t nb12 = src1->nb[2];
-    const size_t nb13 = src1->nb[3];
-
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     GGML_ASSERT( nb0 == sizeof(float));
     GGML_ASSERT(nb00 == sizeof(float));
@@ -8296,28 +8379,12 @@ static void ggml_compute_forward_add_f16_f32(
     const int nth = params->nth;
 
     const int nr  = ggml_nrows(src0);
-    const int64_t ne0 = src0->ne[0];
-    const int64_t ne1 = src0->ne[1];
-    const int64_t ne2 = src0->ne[2];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
 
-    const size_t nb10 = src1->nb[0];
-    const size_t nb11 = src1->nb[1];
-    const size_t nb12 = src1->nb[2];
-    const size_t nb13 = src1->nb[3];
-
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     GGML_ASSERT(src0->type == GGML_TYPE_F16);
     GGML_ASSERT(src1->type == GGML_TYPE_F32);
-    GGML_ASSERT(dst->type == GGML_TYPE_F16);
+    GGML_ASSERT(dst->type  == GGML_TYPE_F16);
 
     GGML_ASSERT( nb0 == sizeof(ggml_fp16_t));
     GGML_ASSERT(nb00 == sizeof(ggml_fp16_t));
@@ -8366,24 +8433,8 @@ static void ggml_compute_forward_add_f16_f16(
     const int nth = params->nth;
 
     const int nr  = ggml_nrows(src0);
-    const int64_t ne0 = src0->ne[0];
-    const int64_t ne1 = src0->ne[1];
-    const int64_t ne2 = src0->ne[2];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const size_t nb10 = src1->nb[0];
-    const size_t nb11 = src1->nb[1];
-    const size_t nb12 = src1->nb[2];
-    const size_t nb13 = src1->nb[3];
 
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     GGML_ASSERT(src0->type == GGML_TYPE_F16);
     GGML_ASSERT(src1->type == GGML_TYPE_F16);
@@ -8433,32 +8484,15 @@ static void ggml_compute_forward_add_q_f32(
     }
 
     const int nr  = ggml_nrows(src0);
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    //const int64_t ne03 = src0->ne[3];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const size_t nb10 = src1->nb[0];
-    const size_t nb11 = src1->nb[1];
-    const size_t nb12 = src1->nb[2];
-    const size_t nb13 = src1->nb[3];
 
-    const size_t nb0  = dst->nb[0];
-    const size_t nb1  = dst->nb[1];
-    const size_t nb2  = dst->nb[2];
-    const size_t nb3  = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     const int ith = params->ith;
     const int nth = params->nth;
 
     const enum ggml_type type = src0->type;
-    dequantize_row_q_t const dequantize_row_q = quantize_fns[type].dequantize_row_q;
-    quantize_row_q_t const quantize_row_q = quantize_fns[type].quantize_row_q;
+    ggml_to_float_t const dequantize_row_q = type_traits[type].to_float;
+    ggml_from_float_t const quantize_row_q = type_traits[type].from_float;
 
     // we don't support permuted src0 or src1
     GGML_ASSERT(nb00 == GGML_TYPE_SIZE[type]);
@@ -8572,19 +8606,8 @@ static void ggml_compute_forward_add1_f32(
     const int nth = params->nth;
 
     const int nr  = ggml_nrows(src0);
-    const int64_t ne0 = src0->ne[0];
-    const int64_t ne1 = src0->ne[1];
-    const int64_t ne2 = src0->ne[2];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
 
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     GGML_ASSERT( nb0 == sizeof(float));
     GGML_ASSERT(nb00 == sizeof(float));
@@ -8638,23 +8661,12 @@ static void ggml_compute_forward_add1_f16_f32(
     const int nth = params->nth;
 
     const int nr  = ggml_nrows(src0);
-    const int64_t ne0 = src0->ne[0];
-    const int64_t ne1 = src0->ne[1];
-    const int64_t ne2 = src0->ne[2];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
 
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     GGML_ASSERT(src0->type == GGML_TYPE_F16);
     GGML_ASSERT(src1->type == GGML_TYPE_F32);
-    GGML_ASSERT(dst->type == GGML_TYPE_F16);
+    GGML_ASSERT(dst->type  == GGML_TYPE_F16);
 
     GGML_ASSERT( nb0 == sizeof(ggml_fp16_t));
     GGML_ASSERT(nb00 == sizeof(ggml_fp16_t));
@@ -8699,23 +8711,12 @@ static void ggml_compute_forward_add1_f16_f16(
     const int nth = params->nth;
 
     const int nr  = ggml_nrows(src0);
-    const int64_t ne0 = src0->ne[0];
-    const int64_t ne1 = src0->ne[1];
-    const int64_t ne2 = src0->ne[2];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
 
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     GGML_ASSERT(src0->type == GGML_TYPE_F16);
     GGML_ASSERT(src1->type == GGML_TYPE_F16);
-    GGML_ASSERT(dst->type == GGML_TYPE_F16);
+    GGML_ASSERT(dst->type  == GGML_TYPE_F16);
 
     GGML_ASSERT( nb0 == sizeof(ggml_fp16_t));
     GGML_ASSERT(nb00 == sizeof(ggml_fp16_t));
@@ -8760,23 +8761,12 @@ static void ggml_compute_forward_add1_q_f32(
     const int nth = params->nth;
 
     const int nr  = ggml_nrows(src0);
-    const int64_t ne0 = src0->ne[0];
-    const int64_t ne1 = src0->ne[1];
-    const int64_t ne2 = src0->ne[2];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
 
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     const enum ggml_type type = src0->type;
-    dequantize_row_q_t const dequantize_row_q = quantize_fns[type].dequantize_row_q;
-    quantize_row_q_t const quantize_row_q = quantize_fns[type].quantize_row_q;
+    ggml_to_float_t const dequantize_row_q = type_traits[type].to_float;
+    ggml_from_float_t const quantize_row_q = type_traits[type].from_float;
 
     // we don't support permuted src0
     GGML_ASSERT(nb00 == GGML_TYPE_SIZE[type]);
@@ -8904,15 +8894,8 @@ static void ggml_compute_forward_acc_f32(
     const int nr = ggml_nrows(src1);
     const int nc = src1->ne[0];
 
-    const int64_t ne10 = src1->ne[0];
-    const int64_t ne11 = src1->ne[1];
-    const int64_t ne12 = src1->ne[2];
-    const int64_t ne13 = src1->ne[3];
-
-    const size_t nb10 = src1->nb[0];
-    const size_t nb11 = src1->nb[1];
-    const size_t nb12 = src1->nb[2];
-    const size_t nb13 = src1->nb[3];
+    GGML_TENSOR_LOCALS(int64_t, ne1, src1, ne);
+    GGML_TENSOR_LOCALS(size_t,  nb1, src1, nb);
 
     // src0 and dst as viewed during acc
     const size_t nb0 = ggml_element_size(src0);
@@ -9001,24 +8984,8 @@ static void ggml_compute_forward_sub_f32(
     }
 
     const int nr  = ggml_nrows(src0);
-    const int64_t ne0 = src0->ne[0];
-    const int64_t ne1 = src0->ne[1];
-    const int64_t ne2 = src0->ne[2];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const size_t nb10 = src1->nb[0];
-    const size_t nb11 = src1->nb[1];
-    const size_t nb12 = src1->nb[2];
-    const size_t nb13 = src1->nb[3];
 
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     GGML_ASSERT( nb0 == sizeof(float));
     GGML_ASSERT(nb00 == sizeof(float));
@@ -9108,29 +9075,7 @@ static void ggml_compute_forward_mul_f32(
 
     const int64_t nr = ggml_nrows(src0);
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-
-    const int64_t ne10 = src1->ne[0];
-    const int64_t ne11 = src1->ne[1];
-    const int64_t ne12 = src1->ne[2];
-    const int64_t ne13 = src1->ne[3];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const size_t nb10 = src1->nb[0];
-    const size_t nb11 = src1->nb[1];
-    const size_t nb12 = src1->nb[2];
-    const size_t nb13 = src1->nb[3];
-
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     GGML_ASSERT( nb0 == sizeof(float));
     GGML_ASSERT(nb00 == sizeof(float));
@@ -9218,24 +9163,8 @@ static void ggml_compute_forward_div_f32(
     }
 
     const int nr  = ggml_nrows(src0);
-    const int64_t ne0 = src0->ne[0];
-    const int64_t ne1 = src0->ne[1];
-    const int64_t ne2 = src0->ne[2];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
 
-    const size_t nb10 = src1->nb[0];
-    const size_t nb11 = src1->nb[1];
-    const size_t nb12 = src1->nb[2];
-    const size_t nb13 = src1->nb[3];
-
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     GGML_ASSERT( nb0 == sizeof(float));
     GGML_ASSERT(nb00 == sizeof(float));
@@ -9442,14 +9371,8 @@ static void ggml_compute_forward_sum_f32(
     assert(ggml_is_scalar(dst));
     assert(src0->nb[0] == sizeof(float));
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
+    GGML_TENSOR_LOCALS(int64_t, ne0, src0, ne);
+    GGML_TENSOR_LOCALS(size_t,  nb0, src0, nb);
 
     ggml_float sum     = 0;
     ggml_float row_sum = 0;
@@ -9498,29 +9421,13 @@ static void ggml_compute_forward_sum_rows_f32(
     GGML_ASSERT(src0->nb[0] == sizeof(float));
     GGML_ASSERT(dst->nb[0] == sizeof(float));
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
-    const int64_t ne3 = dst->ne[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     GGML_ASSERT(ne0 == 1);
     GGML_ASSERT(ne1 == ne01);
     GGML_ASSERT(ne2 == ne02);
     GGML_ASSERT(ne3 == ne03);
 
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
-
     for (int64_t i3 = 0; i3 < ne03; i3++) {
         for (int64_t i2 = 0; i2 < ne02; i2++) {
             for (int64_t i1 = 0; i1 < ne01; i1++) {
@@ -9564,19 +9471,7 @@ static void ggml_compute_forward_mean_f32(
 
     assert(src0->nb[0] == sizeof(float));
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
-    const int64_t ne3 = dst->ne[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     assert(ne0 == 1);
     assert(ne1 == ne01);
@@ -9588,10 +9483,6 @@ static void ggml_compute_forward_mean_f32(
     UNUSED(ne2);
     UNUSED(ne3);
 
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
-
     for (int64_t i03 = 0; i03 < ne03; i03++) {
         for (int64_t i02 = 0; i02 < ne02; i02++) {
             for (int64_t i01 = 0; i01 < ne01; i01++) {
@@ -9621,38 +9512,66 @@ static void ggml_compute_forward_mean(
     }
 }
 
-// ggml_compute_forward_repeat
+// ggml_compute_forward_argmax
 
-static void ggml_compute_forward_repeat_f32(
+static void ggml_compute_forward_argmax_f32(
         const struct ggml_compute_params * params,
         const struct ggml_tensor * src0,
         struct ggml_tensor * dst) {
-    GGML_ASSERT(params->ith == 0);
-    GGML_ASSERT(ggml_can_repeat(src0, dst));
+    assert(params->ith == 0);
 
     if (params->type == GGML_TASK_INIT || params->type == GGML_TASK_FINALIZE) {
         return;
     }
 
-    const int64_t ne0  = dst->ne[0];
-    const int64_t ne1  = dst->ne[1];
-    const int64_t ne2  = dst->ne[2];
-    const int64_t ne3  = dst->ne[3];
+    assert(src0->nb[0] == sizeof(float));
+    assert(dst->nb[0] == sizeof(float));
 
     const int64_t ne00 = src0->ne[0];
     const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
 
-    const size_t nb0  = dst->nb[0];
-    const size_t nb1  = dst->nb[1];
-    const size_t nb2  = dst->nb[2];
-    const size_t nb3  = dst->nb[3];
-
-    const size_t nb00 = src0->nb[0];
     const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
+    const size_t nb0 = dst->nb[0];
+
+    for (int64_t i1 = 0; i1 < ne01; i1++) {
+        float * src = (float *) ((char *) src0->data + i1*nb01);
+        int32_t * dst_ = (int32_t *) ((char *)  dst->data + i1*nb0);
+        int v = 0;
+        ggml_vec_argmax_f32(ne00, &v, src);
+        dst_[0] = v;
+    }
+}
+
+static void ggml_compute_forward_argmax(
+        const struct ggml_compute_params * params,
+        const struct ggml_tensor * src0,
+        struct ggml_tensor * dst) {
+    switch (src0->type) {
+        case GGML_TYPE_F32:
+            {
+                ggml_compute_forward_argmax_f32(params, src0, dst);
+            } break;
+        default:
+            {
+                GGML_ASSERT(false);
+            } break;
+    }
+}
+
+// ggml_compute_forward_repeat
+
+static void ggml_compute_forward_repeat_f32(
+        const struct ggml_compute_params * params,
+        const struct ggml_tensor * src0,
+        struct ggml_tensor * dst) {
+    GGML_ASSERT(params->ith == 0);
+    GGML_ASSERT(ggml_can_repeat(src0, dst));
+
+    if (params->type == GGML_TASK_INIT || params->type == GGML_TASK_FINALIZE) {
+        return;
+    }
+
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     // guaranteed to be an integer due to the check in ggml_can_repeat
     const int nr0 = (int)(ne0/ne00);
@@ -9713,25 +9632,7 @@ static void ggml_compute_forward_repeat_back_f32(
         return;
     }
 
-    const int64_t ne0  = dst->ne[0];
-    const int64_t ne1  = dst->ne[1];
-    const int64_t ne2  = dst->ne[2];
-    const int64_t ne3  = dst->ne[3];
-
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const size_t nb0  = dst->nb[0];
-    const size_t nb1  = dst->nb[1];
-    const size_t nb2  = dst->nb[2];
-    const size_t nb3  = dst->nb[3];
-
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     // guaranteed to be an integer due to the check in ggml_can_repeat
     const int nr0 = (int)(ne00/ne0);
@@ -9961,6 +9862,90 @@ static void ggml_compute_forward_step(
     }
 }
 
+// ggml_compute_forward_tanh
+
+static void ggml_compute_forward_tanh_f32(
+        const struct ggml_compute_params * params,
+        const struct ggml_tensor * src0,
+        struct ggml_tensor * dst) {
+    assert(params->ith == 0);
+    assert(ggml_are_same_shape(src0, dst));
+
+    if (params->type == GGML_TASK_INIT || params->type == GGML_TASK_FINALIZE) {
+        return;
+    }
+
+    const int n  = ggml_nrows(src0);
+    const int nc = src0->ne[0];
+
+    assert(dst->nb[0]  == sizeof(float));
+    assert(src0->nb[0] == sizeof(float));
+
+    for (int i = 0; i < n; i++) {
+        ggml_vec_tanh_f32(nc,
+                (float *) ((char *) dst->data  + i*( dst->nb[1])),
+                (float *) ((char *) src0->data + i*(src0->nb[1])));
+    }
+}
+
+static void ggml_compute_forward_tanh(
+        const struct ggml_compute_params * params,
+        const struct ggml_tensor * src0,
+        struct ggml_tensor * dst) {
+    switch (src0->type) {
+        case GGML_TYPE_F32:
+            {
+                ggml_compute_forward_tanh_f32(params, src0, dst);
+            } break;
+        default:
+            {
+                GGML_ASSERT(false);
+            } break;
+    }
+}
+
+// ggml_compute_forward_elu
+
+static void ggml_compute_forward_elu_f32(
+        const struct ggml_compute_params * params,
+        const struct ggml_tensor * src0,
+        struct ggml_tensor * dst) {
+    assert(params->ith == 0);
+    assert(ggml_are_same_shape(src0, dst));
+
+    if (params->type == GGML_TASK_INIT || params->type == GGML_TASK_FINALIZE) {
+        return;
+    }
+
+    const int n  = ggml_nrows(src0);
+    const int nc = src0->ne[0];
+
+    assert(dst->nb[0]  == sizeof(float));
+    assert(src0->nb[0] == sizeof(float));
+
+    for (int i = 0; i < n; i++) {
+        ggml_vec_elu_f32(nc,
+                (float *) ((char *) dst->data  + i*( dst->nb[1])),
+                (float *) ((char *) src0->data + i*(src0->nb[1])));
+    }
+}
+
+static void ggml_compute_forward_elu(
+        const struct ggml_compute_params * params,
+        const struct ggml_tensor * src0,
+        struct ggml_tensor * dst) {
+    switch (src0->type) {
+        case GGML_TYPE_F32:
+            {
+                ggml_compute_forward_elu_f32(params, src0, dst);
+            } break;
+        default:
+            {
+                GGML_ASSERT(false);
+            } break;
+    }
+}
+
 // ggml_compute_forward_relu
 
 static void ggml_compute_forward_relu_f32(
@@ -10262,18 +10247,7 @@ static void ggml_compute_forward_norm_f32(
     const int ith = params->ith;
     const int nth = params->nth;
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     const float eps = 1e-5f; // TODO: make this a parameter
 
@@ -10339,18 +10313,7 @@ static void ggml_compute_forward_rms_norm_f32(
     const int ith = params->ith;
     const int nth = params->nth;
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     const float eps = 1e-6f; // TODO: make this a parameter
 
@@ -10415,22 +10378,7 @@ static void ggml_compute_forward_rms_norm_back_f32(
     const int ith = params->ith;
     const int nth = params->nth;
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const size_t nb11 = src1->nb[1];
-    const size_t nb12 = src1->nb[2];
-    const size_t nb13 = src1->nb[3];
-
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     const float eps = 1e-6f; // TODO: make this a parameter
 
@@ -10618,378 +10566,7 @@ static bool ggml_compute_forward_mul_mat_use_blas(
 }
 #endif
 
-static void ggml_compute_forward_mul_mat_f32(
-        const struct ggml_compute_params * params,
-        const struct ggml_tensor * src0,
-        const struct ggml_tensor * src1,
-              struct ggml_tensor * dst) {
-    int64_t t0 = ggml_perf_time_us();
-    UNUSED(t0);
-
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-#if defined(GGML_USE_ACCELERATE) || defined(GGML_USE_OPENBLAS)
-    const int64_t ne10 = src1->ne[0];
-#endif
-    const int64_t ne11 = src1->ne[1];
-#ifndef NDEBUG
-    const int64_t ne12 = src1->ne[2];
-    const int64_t ne13 = src1->ne[3];
-
-    const int64_t ne0  = dst->ne[0];
-    const int64_t ne1  = dst->ne[1];
-    const int64_t ne2  = dst->ne[2];
-    const int64_t ne3  = dst->ne[3];
-
-    const int nb00 = src0->nb[0];
-#endif
-    const int nb01 = src0->nb[1];
-    const int nb02 = src0->nb[2];
-    const int nb03 = src0->nb[3];
-
-#ifndef NDEBUG
-    const int nb10 = src1->nb[0];
-#endif
-    const int nb11 = src1->nb[1];
-    const int nb12 = src1->nb[2];
-    const int nb13 = src1->nb[3];
-
-    const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    const int nb2  = dst->nb[2];
-    const int nb3  = dst->nb[3];
-
-    const int ith = params->ith;
-    const int nth = params->nth;
-
-    assert(ne02 == ne12);
-    assert(ne03 == ne13);
-    assert(ne2  == ne12);
-    assert(ne3  == ne13);
-
-    // we don't support permuted src0 or src1
-    assert(nb00 == sizeof(float));
-    assert(nb10 == sizeof(float));
-
-    // dst cannot be transposed or permuted
-    assert(nb0 == sizeof(float));
-    assert(nb0 <= nb1);
-    assert(nb1 <= nb2);
-    assert(nb2 <= nb3);
-
-    assert(ne0 == ne01);
-    assert(ne1 == ne11);
-    assert(ne2 == ne02);
-    assert(ne3 == ne03);
-
-    // nb01 >= nb00 - src0 is not transposed
-    //   compute by src0 rows
-
-#if defined(GGML_USE_CLBLAST)
-    if (ggml_cl_can_mul_mat(src0, src1, dst)) {
-        if (params->ith == 0 && params->type == GGML_TASK_COMPUTE) {
-            ggml_cl_mul_mat(src0, src1, dst, params->wdata, params->wsize);
-        }
-        return;
-    }
-#endif
-
-#if defined(GGML_USE_ACCELERATE) || defined(GGML_USE_OPENBLAS)
-    if (ggml_compute_forward_mul_mat_use_blas(src0, src1, dst)) {
-        if (params->ith != 0) {
-            return;
-        }
-
-        if (params->type == GGML_TASK_INIT) {
-            return;
-        }
-
-        if (params->type == GGML_TASK_FINALIZE) {
-            return;
-        }
-
-        for (int64_t i03 = 0; i03 < ne03; i03++) {
-            for (int64_t i02 = 0; i02 < ne02; i02++) {
-                const float * x = (float *) ((char *) src0->data + i02*nb02 + i03*nb03);
-                const float * y = (float *) ((char *) src1->data + i02*nb12 + i03*nb13);
-                float * d = (float *) ((char *) dst->data + i02*nb2 + i03*nb3);
-
-                cblas_sgemm(CblasRowMajor, CblasNoTrans, CblasTrans,
-                        ne11, ne01, ne10,
-                        1.0f,    y, ne10,
-                                 x, ne00,
-                        0.0f,    d, ne01);
-            }
-        }
-        //printf("CBLAS F32 = %f ms, %d x %d x %d x %d\n", (ggml_perf_time_us() - t0)/1000.0, ne0, ne1, ne2, ne3);
-
-        return;
-    }
-#endif
-
-    if (params->type == GGML_TASK_INIT) {
-        return;
-    }
-
-    if (params->type == GGML_TASK_FINALIZE) {
-        return;
-    }
-
-    // parallelize by src0 rows using ggml_vec_dot_f32
-
-    // total rows in src0
-    const int nr = ne01*ne02*ne03;
-
-    // rows per thread
-    const int dr = (nr + nth - 1)/nth;
-
-    // row range for this thread
-    const int ir0 = dr*ith;
-    const int ir1 = MIN(ir0 + dr, nr);
-
-    for (int ir = ir0; ir < ir1; ++ir) {
-        // src0 indices
-        const int i03 = ir/(ne02*ne01);
-        const int i02 = (ir - i03*ne02*ne01)/ne01;
-        const int i01 = (ir - i03*ne02*ne01 - i02*ne01);
-
-        for (int64_t ic = 0; ic < ne11; ++ic) {
-            // src1 indices
-            const int i13 = i03;
-            const int i12 = i02;
-            const int i11 = ic;
-
-            // dst indices
-            const int i0 = i01;
-            const int i1 = i11;
-            const int i2 = i02;
-            const int i3 = i03;
-
-            ggml_vec_dot_f32(ne00,
-                    (float *) ((char *)  dst->data + (i0*nb0 + i1*nb1 + i2*nb2 + i3*nb3)),
-                    (float *) ((char *) src0->data + (i01*nb01 + i02*nb02 + i03*nb03)),
-                    (float *) ((char *) src1->data + (i11*nb11 + i12*nb12 + i13*nb13)));
-        }
-    }
-
-    //int64_t t1 = ggml_perf_time_us();
-    //static int64_t acc = 0;
-    //acc += t1 - t0;
-    //if (t1 - t0 > 10) {
-    //    printf("\n");
-    //    printf("ne00 = %5d, ne01 = %5d, ne02 = %5d, ne03 = %5d\n", ne00, ne01, ne02, ne03);
-    //    printf("nb00 = %5d, nb01 = %5d, nb02 = %5d, nb03 = %5d\n", nb00, nb01, nb02, nb03);
-    //    printf("ne10 = %5d, ne11 = %5d, ne12 = %5d, ne13 = %5d\n", ne10, ne11, ne12, ne13);
-    //    printf("nb10 = %5d, nb11 = %5d, nb12 = %5d, nb13 = %5d\n", nb10, nb11, nb12, nb13);
-
-    //    printf("XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX task %d/%d: %d us, acc = %d\n", ith, nth, (int) (t1 - t0), (int) acc);
-    //}
-}
-
-static void ggml_compute_forward_mul_mat_f16_f32(
-        const struct ggml_compute_params * params,
-        const struct ggml_tensor * src0,
-        const struct ggml_tensor * src1,
-              struct ggml_tensor * dst) {
-    int64_t t0 = ggml_perf_time_us();
-    UNUSED(t0);
-
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const int64_t ne10 = src1->ne[0];
-    const int64_t ne11 = src1->ne[1];
-    const int64_t ne12 = src1->ne[2];
-    const int64_t ne13 = src1->ne[3];
-
-    const int64_t ne0  = dst->ne[0];
-    const int64_t ne1  = dst->ne[1];
-    const int64_t ne2  = dst->ne[2];
-    const int64_t ne3  = dst->ne[3];
-    //const int64_t ne   = ne0*ne1*ne2*ne3;
-
-    const int nb00 = src0->nb[0];
-    const int nb01 = src0->nb[1];
-    const int nb02 = src0->nb[2];
-    const int nb03 = src0->nb[3];
-
-    const int nb10 = src1->nb[0];
-    const int nb11 = src1->nb[1];
-    const int nb12 = src1->nb[2];
-    const int nb13 = src1->nb[3];
-
-    const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    const int nb2  = dst->nb[2];
-    const int nb3  = dst->nb[3];
-
-    const int ith = params->ith;
-    const int nth = params->nth;
-
-    GGML_ASSERT(ne02 == ne12);
-    GGML_ASSERT(ne03 == ne13);
-    GGML_ASSERT(ne2  == ne12);
-    GGML_ASSERT(ne3  == ne13);
-
-    // TODO: we don't support permuted src0
-    GGML_ASSERT(nb00 == sizeof(ggml_fp16_t));
-
-    // dst cannot be transposed or permuted
-    GGML_ASSERT(nb0 == sizeof(float));
-    GGML_ASSERT(nb0 <= nb1);
-    GGML_ASSERT(nb1 <= nb2);
-    GGML_ASSERT(nb2 <= nb3);
-
-    GGML_ASSERT(ne0 == ne01);
-    GGML_ASSERT(ne1 == ne11);
-    GGML_ASSERT(ne2 == ne02);
-    GGML_ASSERT(ne3 == ne03);
-
-    // nb01 >= nb00 - src0 is not transposed
-    //   compute by src0 rows
-
-#if defined(GGML_USE_CLBLAST)
-    if (ggml_cl_can_mul_mat(src0, src1, dst)) {
-        if (params->ith == 0 && params->type == GGML_TASK_COMPUTE) {
-            ggml_cl_mul_mat(src0, src1, dst, params->wdata, params->wsize);
-        }
-        return;
-    }
-#endif
-
-#if defined(GGML_USE_ACCELERATE) || defined(GGML_USE_OPENBLAS)
-    if (ggml_compute_forward_mul_mat_use_blas(src0, src1, dst)) {
-        GGML_ASSERT(nb10 == sizeof(float));
-
-        if (params->ith != 0) {
-            return;
-        }
-
-        if (params->type == GGML_TASK_INIT) {
-            return;
-        }
-
-        if (params->type == GGML_TASK_FINALIZE) {
-            return;
-        }
-
-        for (int64_t i03 = 0; i03 < ne03; i03++) {
-            for (int64_t i02 = 0; i02 < ne02; i02++) {
-                float * const wdata = params->wdata;
-                {
-                    size_t id = 0;
-                    for (int64_t i01 = 0; i01 < ne01; ++i01) {
-                        for (int64_t i00 = 0; i00 < ne00; ++i00) {
-                            wdata[id++] = GGML_FP16_TO_FP32(*(ggml_fp16_t *) ((char *) src0->data + i03*nb03 + i02*nb02 + i01*nb01 + i00*nb00));
-                        }
-                    }
-
-                    assert(id*sizeof(float) <= params->wsize);
-                }
-
-                const float * x = wdata;
-                const float * y = (float *) ((char *) src1->data + i02*nb12 + i03*nb13);
-
-                float * d = (float *) ((char *) dst->data + i02*nb2 + i03*nb3);
-
-                // zT = y * xT
-                cblas_sgemm(CblasRowMajor, CblasNoTrans, CblasTrans,
-                        ne11, ne01, ne10,
-                        1.0f,    y, ne10,
-                                 x, ne00,
-                        0.0f,    d, ne01);
-            }
-        }
-
-        /*printf("CBLAS F16 = %f ms, %d x %d x %d x %d\n", (ggml_perf_time_us() - t0)/1000.0, ne0, ne1, ne2, ne3);*/
-
-        return;
-    }
-#endif
-
-    if (params->type == GGML_TASK_INIT) {
-        ggml_fp16_t * const wdata = params->wdata;
-
-        size_t id = 0;
-        for (int64_t i13 = 0; i13 < ne13; ++i13) {
-            for (int64_t i12 = 0; i12 < ne12; ++i12) {
-                for (int64_t i11 = 0; i11 < ne11; ++i11) {
-                    for (int64_t i10 = 0; i10 < ne10; ++i10) {
-                        wdata[id++] = GGML_FP32_TO_FP16(*(float *)((char *) src1->data + i13*nb13 + i12*nb12 + i11*nb11 + i10*nb10));
-                    }
-                }
-            }
-        }
-
-        GGML_ASSERT(id*sizeof(ggml_fp16_t) <= params->wsize);
-
-        return;
-    }
-
-    if (params->type == GGML_TASK_FINALIZE) {
-        return;
-    }
-
-    // fp16 -> half the size, so divide by 2
-    // TODO: do not support transposed src1
-    assert(nb10/2 == sizeof(ggml_fp16_t));
-
-    // parallelize by src0 rows using ggml_vec_dot_f16
-
-    // total rows in src0
-    const int nr = ne01*ne02*ne03;
-
-    // rows per thread
-    const int dr = (nr + nth - 1)/nth;
-
-    // row range for this thread
-    const int ir0 = dr*ith;
-    const int ir1 = MIN(ir0 + dr, nr);
-
-    ggml_fp16_t * wdata = params->wdata;
-
-    for (int ir = ir0; ir < ir1; ++ir) {
-        // src0 indices
-        const int i03 = ir/(ne02*ne01);
-        const int i02 = (ir - i03*ne02*ne01)/ne01;
-        const int i01 = (ir - i03*ne02*ne01 - i02*ne01);
-
-        const int i13 = i03;
-        const int i12 = i02;
-
-        const int i0 = i01;
-        const int i2 = i02;
-        const int i3 = i03;
-
-        ggml_fp16_t * src0_row = (ggml_fp16_t *) ((char *) src0->data + (i01*nb01 + i02*nb02 + i03*nb03));
-        ggml_fp16_t * src1_col =                                wdata + (       0 + i12*ne11 + i13*ne12*ne11)*ne00;
-
-        float * dst_col = (float *) ((char *) dst->data + (i0*nb0 + 0*nb1 + i2*nb2 + i3*nb3));
-
-        for (int64_t ic = 0; ic < ne11; ++ic) {
-            ggml_vec_dot_f16(ne00, &dst_col[ic*ne0], src0_row, src1_col + ic*ne00);
-        }
-    }
-
-    //int64_t t1 = ggml_time_us();
-    //static int64_t acc = 0;
-    //acc += t1 - t0;
-    //if (t1 - t0 > 10) {
-    //    printf("\n");
-    //    printf("ne00 = %5d, ne01 = %5d, ne02 = %5d, ne03 = %5d\n", ne00, ne01, ne02, ne03);
-    //    printf("nb00 = %5d, nb01 = %5d, nb02 = %5d, nb03 = %5d\n", nb00, nb01, nb02, nb03);
-    //    printf("ne10 = %5d, ne11 = %5d, ne12 = %5d, ne13 = %5d\n", ne10, ne11, ne12, ne13);
-
-    //    printf("XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX task %d/%d: %d us, acc = %d\n", ith, nth, (int) (t1 - t0), (int) acc);
-    //}
-}
-
-static void ggml_compute_forward_mul_mat_q_f32(
+static void ggml_compute_forward_mul_mat(
         const struct ggml_compute_params * params,
         const struct ggml_tensor * src0,
         const struct ggml_tensor * src1,
@@ -10997,35 +10574,7 @@ static void ggml_compute_forward_mul_mat_q_f32(
     int64_t t0 = ggml_perf_time_us();
     UNUSED(t0);
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
-
-    const int64_t ne10 = src1->ne[0];
-    const int64_t ne11 = src1->ne[1];
-    const int64_t ne12 = src1->ne[2];
-    const int64_t ne13 = src1->ne[3];
-
-    const int64_t ne0  = dst->ne[0];
-    const int64_t ne1  = dst->ne[1];
-    const int64_t ne2  = dst->ne[2];
-    const int64_t ne3  = dst->ne[3];
-
-    const int nb00 = src0->nb[0];
-    const int nb01 = src0->nb[1];
-    const int nb02 = src0->nb[2];
-    const int nb03 = src0->nb[3];
-
-    const int nb10 = src1->nb[0];
-    const int nb11 = src1->nb[1];
-    const int nb12 = src1->nb[2];
-    const int nb13 = src1->nb[3];
-
-    const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    const int nb2  = dst->nb[2];
-    const int nb3  = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -11036,12 +10585,13 @@ static void ggml_compute_forward_mul_mat_q_f32(
     GGML_ASSERT(ne3  == ne13);
 
     const enum ggml_type type = src0->type;
-    quantize_row_q_t const quantize_row_q_dot = quantize_fns[type].quantize_row_q_dot;
-    vec_dot_q_t      const vec_dot_q          = quantize_fns[type].vec_dot_q;
-    enum ggml_type   const vec_dot_type       = quantize_fns[type].vec_dot_type;
+
+    ggml_vec_dot_t    const vec_dot               = type_traits[type].vec_dot;
+    enum ggml_type    const vec_dot_type          = type_traits[type].vec_dot_type;
+    ggml_from_float_t const from_float_to_vec_dot = type_traits[vec_dot_type].from_float;
 
     // we don't support permuted src0 or src1
-    GGML_ASSERT(nb00 == (int) GGML_TYPE_SIZE[type]);
+    GGML_ASSERT(nb00 == GGML_TYPE_SIZE[type]);
     GGML_ASSERT(nb10 == sizeof(float));
 
     // dst cannot be transposed or permuted
@@ -11081,27 +10631,27 @@ static void ggml_compute_forward_mul_mat_q_f32(
             return;
         }
 
-        float * const wdata = params->wdata;
-        dequantize_row_q_t const dequantize_row_q = quantize_fns[type].dequantize_row_q;
-
         for (int64_t i03 = 0; i03 < ne03; i03++) {
             for (int64_t i02 = 0; i02 < ne02; i02++) {
+                const void * x = (char *) src0->data + i03*nb03 + i02*nb02;
                 const float * y = (float *) ((char *) src1->data + i02*nb12 + i03*nb13);
 
                 float * d = (float *) ((char *) dst->data + i02*nb2 + i03*nb3);
 
-                {
+                if (type != GGML_TYPE_F32) {
+                    float * const wdata = params->wdata;
+                    ggml_to_float_t const to_float = type_traits[type].to_float;
+
                     size_t id = 0;
                     for (int64_t i01 = 0; i01 < ne01; ++i01) {
-                        dequantize_row_q((char *) src0->data + i03*nb03 + i02*nb02 + i01*nb01, wdata + id, ne00);
+                        to_float((char *) src0->data + i03*nb03 + i02*nb02 + i01*nb01, wdata + id, ne00);
                         id += ne00;
                     }
 
                     assert(id*sizeof(float) <= params->wsize);
+                    x = wdata;
                 }
 
-                const float * x = wdata;
-
                 cblas_sgemm(CblasRowMajor, CblasNoTrans, CblasTrans,
                         ne11, ne01, ne10,
                         1.0f,    y, ne10,
@@ -11117,14 +10667,16 @@ static void ggml_compute_forward_mul_mat_q_f32(
 #endif
 
     if (params->type == GGML_TASK_INIT) {
-        char * wdata = params->wdata;
-        const size_t row_size = ne10*GGML_TYPE_SIZE[vec_dot_type]/GGML_BLCK_SIZE[vec_dot_type];
-
-        for (int64_t i13 = 0; i13 < ne13; ++i13) {
-            for (int64_t i12 = 0; i12 < ne12; ++i12) {
-                for (int64_t i11 = 0; i11 < ne11; ++i11) {
-                    quantize_row_q_dot((float *)((char *) src1->data + i13*nb13 + i12*nb12 + i11*nb11), (void *) wdata, ne10);
-                    wdata += row_size;
+        if (src1->type != vec_dot_type) {
+            char * wdata = params->wdata;
+            const size_t row_size = ne10*GGML_TYPE_SIZE[vec_dot_type]/GGML_BLCK_SIZE[vec_dot_type];
+
+            for (int64_t i13 = 0; i13 < ne13; ++i13) {
+                for (int64_t i12 = 0; i12 < ne12; ++i12) {
+                    for (int64_t i11 = 0; i11 < ne11; ++i11) {
+                        from_float_to_vec_dot((float *)((char *) src1->data + i13*nb13 + i12*nb12 + i11*nb11), (void *) wdata, ne10);
+                        wdata += row_size;
+                    }
                 }
             }
         }
@@ -11148,7 +10700,7 @@ static void ggml_compute_forward_mul_mat_q_f32(
     const int ir0 = dr*ith;
     const int ir1 = MIN(ir0 + dr, nr);
 
-    void * wdata = params->wdata;
+    void * wdata = (src1->type == vec_dot_type) ? src1->data : params->wdata;
     const size_t row_size = ne00*GGML_TYPE_SIZE[vec_dot_type]/GGML_BLCK_SIZE[vec_dot_type];
 
     for (int ir = ir0; ir < ir1; ++ir) {
@@ -11169,10 +10721,8 @@ static void ggml_compute_forward_mul_mat_q_f32(
 
         float * dst_col = (float *) ((char *) dst->data + (i0*nb0 + 0*nb1 + i2*nb2 + i3*nb3));
 
-        assert(ne00 % 32 == 0);
-
         for (int64_t ic = 0; ic < ne11; ++ic) {
-            vec_dot_q(ne00, &dst_col[ic*ne0], src0_row, (void *) (src1_col + ic*row_size));
+            vec_dot(ne00, &dst_col[ic*ne0], src0_row, (void *) (src1_col + ic*row_size));
         }
     }
 
@@ -11189,40 +10739,6 @@ static void ggml_compute_forward_mul_mat_q_f32(
     //}
 }
 
-static void ggml_compute_forward_mul_mat(
-        const struct ggml_compute_params * params,
-        const struct ggml_tensor * src0,
-        const struct ggml_tensor * src1,
-        struct ggml_tensor * dst) {
-    switch (src0->type) {
-        case GGML_TYPE_Q4_0:
-        case GGML_TYPE_Q4_1:
-        case GGML_TYPE_Q5_0:
-        case GGML_TYPE_Q5_1:
-        case GGML_TYPE_Q8_0:
-        case GGML_TYPE_Q8_1:
-        case GGML_TYPE_Q2_K:
-        case GGML_TYPE_Q3_K:
-        case GGML_TYPE_Q4_K:
-        case GGML_TYPE_Q5_K:
-        case GGML_TYPE_Q6_K:
-            {
-                ggml_compute_forward_mul_mat_q_f32(params, src0, src1, dst);
-            } break;
-        case GGML_TYPE_F16:
-            {
-                ggml_compute_forward_mul_mat_f16_f32(params, src0, src1, dst);
-            } break;
-        case GGML_TYPE_F32:
-            {
-                ggml_compute_forward_mul_mat_f32(params, src0, src1, dst);
-            } break;
-        default:
-            {
-                GGML_ASSERT(false);
-            } break;
-    }
-}
 
 // ggml_compute_forward_out_prod
 
@@ -11232,38 +10748,10 @@ static void ggml_compute_forward_out_prod_f32(
         const struct ggml_tensor * src0,
         const struct ggml_tensor * src1,
               struct ggml_tensor * dst) {
-    int64_t t0 = ggml_perf_time_us();
-    UNUSED(t0);
-
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3];
+    int64_t t0 = ggml_perf_time_us();
+    UNUSED(t0);
 
-    const int64_t ne10 = src1->ne[0];
-    //const int64_t ne11 = src1->ne[1];
-    const int64_t ne12 = src1->ne[2];
-    const int64_t ne13 = src1->ne[3];
-
-    const int64_t ne0  = dst->ne[0];
-    const int64_t ne1  = dst->ne[1];
-    const int64_t ne2  = dst->ne[2];
-    const int64_t ne3  = dst->ne[3];
-
-    const int nb00 = src0->nb[0];
-    const int nb01 = src0->nb[1];
-    const int nb02 = src0->nb[2];
-    const int nb03 = src0->nb[3];
-
-    const int nb10 = src1->nb[0];
-    const int nb11 = src1->nb[1];
-    const int nb12 = src1->nb[2];
-    const int nb13 = src1->nb[3];
-
-    const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    const int nb2  = dst->nb[2];
-    const int nb3  = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -11498,15 +10986,8 @@ static void ggml_compute_forward_set_f32(
     const int nr = ggml_nrows(src1);
     const int nc = src1->ne[0];
 
-    const int64_t ne10 = src1->ne[0];
-    const int64_t ne11 = src1->ne[1];
-    const int64_t ne12 = src1->ne[2];
-    const int64_t ne13 = src1->ne[3];
-
-    const size_t nb10 = src1->nb[0];
-    const size_t nb11 = src1->nb[1];
-    const size_t nb12 = src1->nb[2];
-    const size_t nb13 = src1->nb[3];
+    GGML_TENSOR_LOCALS(int64_t, ne1, src1, ne);
+    GGML_TENSOR_LOCALS(size_t,  nb1, src1, nb);
 
     // src0 and dst as viewed during set
     const size_t nb0 = ggml_element_size(src0);
@@ -11647,7 +11128,7 @@ static void ggml_compute_forward_get_rows_q(
     const int nc = src0->ne[0];
     const int nr = ggml_nelements(src1);
     const enum ggml_type type = src0->type;
-    dequantize_row_q_t const dequantize_row_q = quantize_fns[type].dequantize_row_q;
+    ggml_to_float_t const dequantize_row_q = type_traits[type].to_float;
 
     assert( dst->ne[0] == nc);
     assert( dst->ne[1] == nr);
@@ -11897,29 +11378,14 @@ static void ggml_compute_forward_diag_f32(
 
     // TODO: handle transposed/permuted matrices
 
-    const int ne00 = src0->ne[0];
-    const int ne01 = src0->ne[1];
-    const int ne02 = src0->ne[2];
-    const int ne03 = src0->ne[3];
-    const int ne0 = dst->ne[0];
-    const int ne1 = dst->ne[1];
-    const int ne2 = dst->ne[2];
-    const int ne3 = dst->ne[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
+
     GGML_ASSERT(ne00 == ne0);
     GGML_ASSERT(ne00 == ne1);
     GGML_ASSERT(ne01 == 1);
     GGML_ASSERT(ne02 == ne2);
     GGML_ASSERT(ne03 == ne3);
 
-    const int nb00 = src0->nb[0];
-    //const int nb01 = src0->nb[1];
-    const int nb02 = src0->nb[2];
-    const int nb03 = src0->nb[3];
-    const int nb0 = dst->nb[0];
-    const int nb1 = dst->nb[1];
-    const int nb2 = dst->nb[2];
-    const int nb3 = dst->nb[3];
-
     GGML_ASSERT(nb00 == sizeof(float));
     GGML_ASSERT(nb0  == sizeof(float));
 
@@ -12496,20 +11962,7 @@ static void ggml_compute_forward_rope_f32(
 
     assert(n_past >= 0);
 
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
-    const int64_t ne3 = dst->ne[3];
-
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     //printf("ne0: %d, ne1: %d, ne2: %d, ne3: %d\n", ne0, ne1, ne2, ne3);
     //printf("n_past = %d, ne2 = %d\n", n_past, ne2);
@@ -12636,20 +12089,7 @@ static void ggml_compute_forward_rope_f16(
 
     assert(n_past >= 0);
 
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
-    const int64_t ne3 = dst->ne[3];
-
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     //printf("ne0: %d, ne1: %d, ne2: %d, ne3: %d\n", ne0, ne1, ne2, ne3);
     //printf("n_past = %d, ne2 = %d\n", n_past, ne2);
@@ -12802,21 +12242,7 @@ static void ggml_compute_forward_rope_back_f32(
 
     assert(n_past >= 0);
 
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
-    const int64_t ne3 = dst->ne[3];
-
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
-
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     //printf("ne0: %d, ne1: %d, ne2: %d, ne3: %d\n", ne0, ne1, ne2, ne3);
     //printf("n_past = %d, ne2 = %d\n", n_past, ne2);
@@ -12915,21 +12341,7 @@ static void ggml_compute_forward_rope_back_f16(
 
     assert(n_past >= 0);
 
-    const size_t nb00 = src0->nb[0];
-    const size_t nb01 = src0->nb[1];
-    const size_t nb02 = src0->nb[2];
-    const size_t nb03 = src0->nb[3];
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
-    const int64_t ne3 = dst->ne[3];
-
-    const size_t nb0 = dst->nb[0];
-    const size_t nb1 = dst->nb[1];
-    const size_t nb2 = dst->nb[2];
-    const size_t nb3 = dst->nb[3];
-
+    GGML_TENSOR_UNARY_OP_LOCALS;
 
     //printf("ne0: %d, ne1: %d, ne2: %d, ne3: %d\n", ne0, ne1, ne2, ne3);
     //printf("n_past = %d, ne2 = %d\n", n_past, ne2);
@@ -13027,7 +12439,7 @@ static void ggml_compute_forward_rope_back(
     }
 }
 
-// ggml_compute_forward_conv_1d_s1_ph
+// ggml_compute_forward_conv_1d
 
 static void ggml_compute_forward_conv_1d_s1_ph_f16_f32(
         const struct ggml_compute_params * params,
@@ -13041,36 +12453,7 @@ static void ggml_compute_forward_conv_1d_s1_ph_f16_f32(
     int64_t t0 = ggml_perf_time_us();
     UNUSED(t0);
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    //const int64_t ne03 = src0->ne[3];
-
-    const int64_t ne10 = src1->ne[0];
-    const int64_t ne11 = src1->ne[1];
-    //const int64_t ne12 = src1->ne[2];
-    //const int64_t ne13 = src1->ne[3];
-
-    //const int64_t ne0  = dst->ne[0];
-    //const int64_t ne1  = dst->ne[1];
-    //const int64_t ne2  = dst->ne[2];
-    //const int64_t ne3  = dst->ne[3];
-    //const int64_t ne   = ne0*ne1*ne2*ne3;
-
-    const int nb00 = src0->nb[0];
-    const int nb01 = src0->nb[1];
-    const int nb02 = src0->nb[2];
-    //const int nb03 = src0->nb[3];
-
-    const int nb10 = src1->nb[0];
-    const int nb11 = src1->nb[1];
-    //const int nb12 = src1->nb[2];
-    //const int nb13 = src1->nb[3];
-
-    //const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    //const int nb2  = dst->nb[2];
-    //const int nb3  = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -13161,36 +12544,7 @@ static void ggml_compute_forward_conv_1d_s1_ph_f32(
     int64_t t0 = ggml_perf_time_us();
     UNUSED(t0);
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    //const int64_t ne03 = src0->ne[3];
-
-    const int64_t ne10 = src1->ne[0];
-    const int64_t ne11 = src1->ne[1];
-    //const int64_t ne12 = src1->ne[2];
-    //const int64_t ne13 = src1->ne[3];
-
-    //const int64_t ne0  = dst->ne[0];
-    //const int64_t ne1  = dst->ne[1];
-    //const int64_t ne2  = dst->ne[2];
-    //const int64_t ne3  = dst->ne[3];
-    //const int64_t ne   = ne0*ne1*ne2*ne3;
-
-    const int nb00 = src0->nb[0];
-    const int nb01 = src0->nb[1];
-    const int nb02 = src0->nb[2];
-    //const int nb03 = src0->nb[3];
-
-    const int nb10 = src1->nb[0];
-    const int nb11 = src1->nb[1];
-    //const int nb12 = src1->nb[2];
-    //const int nb13 = src1->nb[3];
-
-    //const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    //const int nb2  = dst->nb[2];
-    //const int nb3  = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -13290,8 +12644,6 @@ static void ggml_compute_forward_conv_1d_s1_ph(
     }
 }
 
-// ggml_compute_forward_conv_1d_s2_ph
-
 static void ggml_compute_forward_conv_1d_s2_ph_f16_f32(
         const struct ggml_compute_params * params,
         const struct ggml_tensor * src0,
@@ -13304,36 +12656,7 @@ static void ggml_compute_forward_conv_1d_s2_ph_f16_f32(
     int64_t t0 = ggml_perf_time_us();
     UNUSED(t0);
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    //const int64_t ne03 = src0->ne[3];
-
-    const int64_t ne10 = src1->ne[0];
-    const int64_t ne11 = src1->ne[1];
-    //const int64_t ne12 = src1->ne[2];
-    //const int64_t ne13 = src1->ne[3];
-
-    //const int64_t ne0  = dst->ne[0];
-    //const int64_t ne1  = dst->ne[1];
-    //const int64_t ne2  = dst->ne[2];
-    //const int64_t ne3  = dst->ne[3];
-    //const int64_t ne   = ne0*ne1*ne2*ne3;
-
-    const int nb00 = src0->nb[0];
-    const int nb01 = src0->nb[1];
-    const int nb02 = src0->nb[2];
-    //const int nb03 = src0->nb[3];
-
-    const int nb10 = src1->nb[0];
-    const int nb11 = src1->nb[1];
-    //const int nb12 = src1->nb[2];
-    //const int nb13 = src1->nb[3];
-
-    //const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    //const int nb2  = dst->nb[2];
-    //const int nb3  = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -13424,36 +12747,7 @@ static void ggml_compute_forward_conv_1d_s2_ph_f32(
     int64_t t0 = ggml_perf_time_us();
     UNUSED(t0);
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    //const int64_t ne03 = src0->ne[3];
-
-    const int64_t ne10 = src1->ne[0];
-    const int64_t ne11 = src1->ne[1];
-    //const int64_t ne12 = src1->ne[2];
-    //const int64_t ne13 = src1->ne[3];
-
-    //const int64_t ne0  = dst->ne[0];
-    //const int64_t ne1  = dst->ne[1];
-    //const int64_t ne2  = dst->ne[2];
-    //const int64_t ne3  = dst->ne[3];
-    //const int64_t ne   = ne0*ne1*ne2*ne3;
-
-    const int nb00 = src0->nb[0];
-    const int nb01 = src0->nb[1];
-    const int nb02 = src0->nb[2];
-    //const int nb03 = src0->nb[3];
-
-    const int nb10 = src1->nb[0];
-    const int nb11 = src1->nb[1];
-    //const int nb12 = src1->nb[2];
-    //const int nb13 = src1->nb[3];
-
-    //const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    //const int nb2  = dst->nb[2];
-    //const int nb3  = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -13553,6 +12847,28 @@ static void ggml_compute_forward_conv_1d_s2_ph(
     }
 }
 
+// ggml_compute_forward_conv_1d
+
+static void ggml_compute_forward_conv_1d(
+    const struct ggml_compute_params * params,
+    const struct ggml_tensor * src0,
+    const struct ggml_tensor * src1,
+    const struct ggml_tensor * opt0,
+    struct ggml_tensor * dst) {
+    const int32_t s0 = ((const int32_t*)(opt0->data))[0];
+    const int32_t p0 = ((const int32_t*)(opt0->data))[1];
+    const int32_t d0 = ((const int32_t*)(opt0->data))[2];
+    GGML_ASSERT(d0 == 1); // dilation not supported
+    GGML_ASSERT(p0 == src0->ne[0]/2); // only half padding supported
+    if (s0 == 1) {
+        ggml_compute_forward_conv_1d_s1_ph(params, src0, src1, dst);
+    } else if (s0 == 2) {
+        ggml_compute_forward_conv_1d_s2_ph(params, src0, src1, dst);
+    } else {
+        GGML_ASSERT(false); // only stride 1 and 2 supported
+    };
+}
+
 // ggml_compute_forward_conv_2d_sk_p0
 
 static void ggml_compute_forward_conv_2d_sk_p0_f16_f32(
@@ -13567,36 +12883,7 @@ static void ggml_compute_forward_conv_2d_sk_p0_f16_f32(
     int64_t t0 = ggml_perf_time_us();
     UNUSED(t0);
 
-    const int ne00 = src0->ne[0];
-    const int ne01 = src0->ne[1];
-    const int ne02 = src0->ne[2];
-    //const int ne03 = src0->ne[3];
-
-    const int ne10 = src1->ne[0];
-    //const int ne11 = src1->ne[1];
-    const int ne12 = src1->ne[2];
-    //const int ne13 = src1->ne[3];
-
-    const int ne0  = dst->ne[0];
-    const int ne1  = dst->ne[1];
-    const int ne2  = dst->ne[2];
-    //const int ne3  = dst->ne[3];
-    //const int ne   = ne0*ne1*ne2*ne3;
-
-    const int nb00 = src0->nb[0];
-    //const int nb01 = src0->nb[1];
-    //const int nb02 = src0->nb[2];
-    const int nb03 = src0->nb[3];
-
-    const int nb10 = src1->nb[0];
-    //const int nb11 = src1->nb[1];
-    const int nb12 = src1->nb[2];
-    //const int nb13 = src1->nb[3];
-
-    //const int nb0  = dst->nb[0];
-    //const int nb1  = dst->nb[1];
-    const int nb2  = dst->nb[2];
-    //const int nb3  = dst->nb[3];
+    GGML_TENSOR_BINARY_OP_LOCALS;
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -13689,6 +12976,34 @@ static void ggml_compute_forward_conv_2d_sk_p0(
     }
 }
 
+// ggml_compute_forward_conv_2d
+
+static void ggml_compute_forward_conv_2d(
+    const struct ggml_compute_params* params,
+    const struct ggml_tensor* src0,
+    const struct ggml_tensor* src1,
+    const struct ggml_tensor* opt0,
+    struct ggml_tensor* dst) {
+    const int32_t s0 = ((const int32_t*)(opt0->data))[0];
+    const int32_t s1 = ((const int32_t*)(opt0->data))[1];
+    const int32_t p0 = ((const int32_t*)(opt0->data))[2];
+    const int32_t p1 = ((const int32_t*)(opt0->data))[3];
+    const int32_t d0 = ((const int32_t*)(opt0->data))[4];
+    const int32_t d1 = ((const int32_t*)(opt0->data))[5];
+    GGML_ASSERT(d0 == 1); // dilation not supported
+    GGML_ASSERT(d1 == 1);
+    GGML_ASSERT(p0 == 0); // padding not supported
+    GGML_ASSERT(p1 == 0);
+
+    if (s0 == src0->ne[0] && s1 == src0->ne[1]) {
+        ggml_compute_forward_conv_2d_sk_p0(params, src0, src1, dst);
+    }
+    else {
+        GGML_ASSERT(false); // only stride equal to kernel size is supported
+    };
+}
+
+
 // ggml_compute_forward_flash_attn
 
 static void ggml_compute_forward_flash_attn_f32(
@@ -13701,45 +13016,14 @@ static void ggml_compute_forward_flash_attn_f32(
     int64_t t0 = ggml_perf_time_us();
     UNUSED(t0);
 
-    const int64_t neq0 = q->ne[0];
-    const int64_t neq1 = q->ne[1];
-    const int64_t neq2 = q->ne[2];
-    const int64_t neq3 = q->ne[3];
-
-    const int64_t nek0 = k->ne[0];
-    const int64_t nek1 = k->ne[1];
-    //const int64_t nek2 = k->ne[2];
-    //const int64_t nek3 = k->ne[3];
-
-    //const int64_t nev0 = v->ne[0];
-    const int64_t nev1 = v->ne[1];
-    //const int64_t nev2 = v->ne[2];
-    //const int64_t nev3 = v->ne[3];
-
-    const int64_t ne0  = dst->ne[0];
-    const int64_t ne1  = dst->ne[1];
-    //const int64_t ne2  = dst->ne[2];
-    //const int64_t ne3  = dst->ne[3];
-
-    const int nbk0 = k->nb[0];
-    const int nbk1 = k->nb[1];
-    const int nbk2 = k->nb[2];
-    const int nbk3 = k->nb[3];
-
-    const int nbq0 = q->nb[0];
-    const int nbq1 = q->nb[1];
-    const int nbq2 = q->nb[2];
-    const int nbq3 = q->nb[3];
-
-    const int nbv0 = v->nb[0];
-    const int nbv1 = v->nb[1];
-    const int nbv2 = v->nb[2];
-    const int nbv3 = v->nb[3];
-
-    const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    const int nb2  = dst->nb[2];
-    const int nb3  = dst->nb[3];
+    GGML_TENSOR_LOCALS(int64_t, neq, q,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nbq, q,   nb);
+    GGML_TENSOR_LOCALS(int64_t, nek, k,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nbk, k,   nb);
+    GGML_TENSOR_LOCALS(int64_t, nev, v,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nbv, v,   nb);
+    GGML_TENSOR_LOCALS(int64_t, ne,  dst, ne);
+    GGML_TENSOR_LOCALS(size_t,  nb,  dst, nb);
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -13910,45 +13194,14 @@ static void ggml_compute_forward_flash_attn_f16(
     int64_t t0 = ggml_perf_time_us();
     UNUSED(t0);
 
-    const int64_t neq0 = q->ne[0];
-    const int64_t neq1 = q->ne[1];
-    const int64_t neq2 = q->ne[2];
-    const int64_t neq3 = q->ne[3];
-
-    const int64_t nek0 = k->ne[0];
-    const int64_t nek1 = k->ne[1];
-    //const int64_t nek2 = k->ne[2];
-    //const int64_t nek3 = k->ne[3];
-
-    //const int64_t nev0 = v->ne[0];
-    const int64_t nev1 = v->ne[1];
-    //const int64_t nev2 = v->ne[2];
-    //const int64_t nev3 = v->ne[3];
-
-    const int64_t ne0  = dst->ne[0];
-    const int64_t ne1  = dst->ne[1];
-    //const int64_t ne2  = dst->ne[2];
-    //const int64_t ne3  = dst->ne[3];
-
-    const int nbk0 = k->nb[0];
-    const int nbk1 = k->nb[1];
-    const int nbk2 = k->nb[2];
-    const int nbk3 = k->nb[3];
-
-    const int nbq0 = q->nb[0];
-    const int nbq1 = q->nb[1];
-    const int nbq2 = q->nb[2];
-    const int nbq3 = q->nb[3];
-
-    const int nbv0 = v->nb[0];
-    const int nbv1 = v->nb[1];
-    const int nbv2 = v->nb[2];
-    const int nbv3 = v->nb[3];
-
-    const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    const int nb2  = dst->nb[2];
-    const int nb3  = dst->nb[3];
+    GGML_TENSOR_LOCALS(int64_t, neq, q,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nbq, q,   nb);
+    GGML_TENSOR_LOCALS(int64_t, nek, k,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nbk, k,   nb);
+    GGML_TENSOR_LOCALS(int64_t, nev, v,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nbv, v,   nb);
+    GGML_TENSOR_LOCALS(int64_t, ne,  dst, ne);
+    GGML_TENSOR_LOCALS(size_t,  nb,  dst, nb);
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -14182,65 +13435,18 @@ static void ggml_compute_forward_flash_ff_f16(
     int64_t t0 = ggml_perf_time_us();
     UNUSED(t0);
 
-    const int64_t nea0 = a->ne[0];
-    const int64_t nea1 = a->ne[1];
-    const int64_t nea2 = a->ne[2];
-    const int64_t nea3 = a->ne[3];
-
-    const int64_t neb00 = b0->ne[0];
-    const int64_t neb01 = b0->ne[1];
-    //const int64_t neb02 = b0->ne[2];
-    //const int64_t neb03 = b0->ne[3];
-
-    const int64_t neb10 = b1->ne[0];
-    const int64_t neb11 = b1->ne[1];
-    //const int64_t neb12 = b1->ne[2];
-    //const int64_t neb13 = b1->ne[3];
-
-    const int64_t nec00 = c0->ne[0];
-    const int64_t nec01 = c0->ne[1];
-    //const int64_t nec02 = c0->ne[2];
-    //const int64_t nec03 = c0->ne[3];
-
-    const int64_t nec10 = c1->ne[0];
-    const int64_t nec11 = c1->ne[1];
-    //const int64_t nec12 = c1->ne[2];
-    //const int64_t nec13 = c1->ne[3];
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
-    //const int64_t ne3 = dst->ne[3];
-
-    const int nba0 = a->nb[0];
-    const int nba1 = a->nb[1];
-    const int nba2 = a->nb[2];
-    const int nba3 = a->nb[3];
-
-    const int nbb00 = b0->nb[0];
-    const int nbb01 = b0->nb[1];
-    const int nbb02 = b0->nb[2];
-    const int nbb03 = b0->nb[3];
-
-    const int nbb10 = b1->nb[0];
-    //const int nbb11 = b1->nb[1];
-    //const int nbb12 = b1->nb[2];
-    //const int nbb13 = b1->nb[3];
-
-    const int nbc00 = c0->nb[0];
-    const int nbc01 = c0->nb[1];
-    const int nbc02 = c0->nb[2];
-    const int nbc03 = c0->nb[3];
-
-    const int nbc10 = c1->nb[0];
-    //const int nbc11 = c1->nb[1];
-    //const int nbc12 = c1->nb[2];
-    //const int nbc13 = c1->nb[3];
-
-    const int nb0 = dst->nb[0];
-    const int nb1 = dst->nb[1];
-    const int nb2 = dst->nb[2];
-    const int nb3 = dst->nb[3];
+    GGML_TENSOR_LOCALS(int64_t, nea,  a,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nba,  a,   nb);
+    GGML_TENSOR_LOCALS(int64_t, neb0, b0,  ne);
+    GGML_TENSOR_LOCALS(size_t,  nbb0, b0,  nb);
+    GGML_TENSOR_LOCALS(int64_t, neb1, b1,  ne);
+    GGML_TENSOR_LOCALS(size_t,  nbb1, b1,  nb);
+    GGML_TENSOR_LOCALS(int64_t, nec0, c0,  ne);
+    GGML_TENSOR_LOCALS(size_t,  nbc0, c0,  nb);
+    GGML_TENSOR_LOCALS(int64_t, nec1, c1,  ne);
+    GGML_TENSOR_LOCALS(size_t,  nbc1, c1,  nb);
+    GGML_TENSOR_LOCALS(int64_t, ne,   dst, ne);
+    GGML_TENSOR_LOCALS(size_t,  nb,   dst, nb);
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -14388,55 +13594,16 @@ static void ggml_compute_forward_flash_attn_back_f32(
     int64_t t0 = ggml_perf_time_us();
     UNUSED(t0);
 
-    const int64_t neq0 = q->ne[0];
-    const int64_t neq1 = q->ne[1];
-    const int64_t neq2 = q->ne[2];
-    const int64_t neq3 = q->ne[3];
-
-    const int64_t nek0 = k->ne[0];
-    const int64_t nek1 = k->ne[1];
-    //const int64_t nek2 = k->ne[2];
-    //const int64_t nek3 = k->ne[3];
-
-    const int64_t nev0 = v->ne[0];
-    const int64_t nev1 = v->ne[1];
-    //const int64_t nev2 = v->ne[2];
-    //const int64_t nev3 = v->ne[3];
-
-    const int64_t ned0 = d->ne[0];
-    const int64_t ned1 = d->ne[1];
-    //const int64_t ned2 = d->ne[2];
-    //const int64_t ned3 = d->ne[3];
-
-    const int64_t ne0  = dst->ne[0];
-    const int64_t ne1  = dst->ne[1];
-    const int64_t ne2  = dst->ne[2];
-    const int64_t ne3  = dst->ne[3];
-
-    const int nbk0 = k->nb[0];
-    const int nbk1 = k->nb[1];
-    const int nbk2 = k->nb[2];
-    const int nbk3 = k->nb[3];
-
-    const int nbq0 = q->nb[0];
-    const int nbq1 = q->nb[1];
-    const int nbq2 = q->nb[2];
-    const int nbq3 = q->nb[3];
-
-    const int nbv0 = v->nb[0];
-    const int nbv1 = v->nb[1];
-    const int nbv2 = v->nb[2];
-    const int nbv3 = v->nb[3];
-
-    const int nbd0 = d->nb[0];
-    const int nbd1 = d->nb[1];
-    const int nbd2 = d->nb[2];
-    const int nbd3 = d->nb[3];
-
-    const int nb0  = dst->nb[0];
-    const int nb1  = dst->nb[1];
-    const int nb2  = dst->nb[2];
-    const int nb3  = dst->nb[3];
+    GGML_TENSOR_LOCALS(int64_t, neq, q,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nbq, q,   nb);
+    GGML_TENSOR_LOCALS(int64_t, nek, k,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nbk, k,   nb);
+    GGML_TENSOR_LOCALS(int64_t, nev, v,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nbv, v,   nb);
+    GGML_TENSOR_LOCALS(int64_t, ned, d,   ne);
+    GGML_TENSOR_LOCALS(size_t,  nbd, d,   nb);
+    GGML_TENSOR_LOCALS(int64_t, ne,  dst, ne);
+    GGML_TENSOR_LOCALS(size_t,  nb,  dst, nb);
 
     const int ith = params->ith;
     const int nth = params->nth;
@@ -14794,15 +13961,8 @@ static void ggml_compute_forward_win_part_f32(
         return;
     }
 
-    const int64_t ne00 = src0->ne[0]; UNUSED(ne00);
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    const int64_t ne03 = src0->ne[3]; UNUSED(ne03);
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
-    const int64_t ne3 = dst->ne[3]; UNUSED(ne3);
+    GGML_TENSOR_LOCALS(int64_t, ne0, src0, ne);
+    GGML_TENSOR_LOCALS(int64_t, ne,  dst,  ne);
 
     const int32_t nep0 = ((const int32_t *)(opt0->data))[0];
     const int32_t nep1 = ((const int32_t *)(opt0->data))[1];
@@ -14865,14 +14025,8 @@ static void ggml_compute_forward_win_unpart_f32(
         return;
     }
 
-    const int64_t ne00 = src0->ne[0];
-    const int64_t ne01 = src0->ne[1];
-    const int64_t ne02 = src0->ne[2];
-    //const int64_t ne03 = src0->ne[3];
-
-    const int64_t ne0 = dst->ne[0];
-    const int64_t ne1 = dst->ne[1];
-    const int64_t ne2 = dst->ne[2];
+    GGML_TENSOR_LOCALS(int64_t, ne0, src0, ne);
+    GGML_TENSOR_LOCALS(int64_t, ne,  dst,  ne);
 
     const int32_t w = ((const int32_t *)(opt0->data))[0];
 
@@ -15470,6 +14624,10 @@ static void ggml_compute_forward(struct ggml_compute_params * params, struct ggm
             {
                 ggml_compute_forward_mean(params, tensor->src0, tensor);
             } break;
+        case GGML_OP_ARGMAX:
+            {
+                ggml_compute_forward_argmax(params, tensor->src0, tensor);
+            } break;
         case GGML_OP_REPEAT:
             {
                 ggml_compute_forward_repeat(params, tensor->src0, tensor);
@@ -15494,6 +14652,14 @@ static void ggml_compute_forward(struct ggml_compute_params * params, struct ggm
             {
                 ggml_compute_forward_step(params, tensor->src0, tensor);
             } break;
+        case GGML_OP_TANH:
+            {
+                ggml_compute_forward_tanh(params, tensor->src0, tensor);
+            } break;
+        case GGML_OP_ELU:
+            {
+                ggml_compute_forward_elu(params, tensor->src0, tensor);
+            } break;
         case GGML_OP_RELU:
             {
                 ggml_compute_forward_relu(params, tensor->src0, tensor);
@@ -15610,17 +14776,13 @@ static void ggml_compute_forward(struct ggml_compute_params * params, struct ggm
             {
                 ggml_compute_forward_clamp(params, tensor->src0, tensor->src1, tensor);
             } break;
-        case GGML_OP_CONV_1D_S1_PH:
-            {
-                ggml_compute_forward_conv_1d_s1_ph(params, tensor->src0, tensor->src1, tensor);
-            } break;
-        case GGML_OP_CONV_1D_S2_PH:
+        case GGML_OP_CONV_1D:
             {
-                ggml_compute_forward_conv_1d_s2_ph(params, tensor->src0, tensor->src1, tensor);
+                ggml_compute_forward_conv_1d(params, tensor->src0, tensor->src1, tensor->opt[0], tensor);
             } break;
-        case GGML_OP_CONV_2D_SK_P0:
+        case GGML_OP_CONV_2D:
             {
-                ggml_compute_forward_conv_2d_sk_p0(params, tensor->src0, tensor->src1, tensor);
+                ggml_compute_forward_conv_2d(params, tensor->src0, tensor->src1, tensor->opt[0], tensor);
             } break;
         case GGML_OP_FLASH_ATTN:
             {
@@ -15869,6 +15031,7 @@ static void ggml_compute_backward(struct ggml_context * ctx, struct ggml_tensor
                 }
             } break;
         case GGML_OP_MEAN:
+        case GGML_OP_ARGMAX:
             {
                 GGML_ASSERT(false); // TODO: implement
             } break;
@@ -15922,6 +15085,14 @@ static void ggml_compute_backward(struct ggml_context * ctx, struct ggml_tensor
                     // noop
                 }
             } break;
+        case GGML_OP_TANH:
+            {
+                GGML_ASSERT(false); // TODO: not implemented
+            } break;
+        case GGML_OP_ELU:
+            {
+                GGML_ASSERT(false); // TODO: not implemented
+            } break;
         case GGML_OP_RELU:
             {
                 if (src0->grad) {
@@ -15941,14 +15112,6 @@ static void ggml_compute_backward(struct ggml_context * ctx, struct ggml_tensor
             {
                 GGML_ASSERT(false); // TODO: not implemented
             } break;
-        case GGML_OP_ALIBI:
-            {
-                GGML_ASSERT(false); // TODO: not implemented
-            } break;
-        case GGML_OP_CLAMP:
-            {
-                GGML_ASSERT(false); // TODO: not implemented
-            } break;
         case GGML_OP_SILU:
             {
                 // necessary for llama
@@ -16265,7 +15428,7 @@ static void ggml_compute_backward(struct ggml_context * ctx, struct ggml_tensor
                 // necessary for llama
                 if (src0->grad) {
                     assert(src1->type == GGML_TYPE_I32);
-                    assert(ggml_nelements(src1) == 3);
+                    assert(ggml_nelements(src1) == 4);
                     const int n_past = ((int32_t *) src1->data)[0];
                     const int n_dims = ((int32_t *) src1->data)[1];
                     const int mode   = ((int32_t *) src1->data)[2];
@@ -16305,15 +15468,19 @@ static void ggml_compute_backward(struct ggml_context * ctx, struct ggml_tensor
                     // noop
                 }
             } break;
-        case GGML_OP_CONV_1D_S1_PH:
+        case GGML_OP_ALIBI:
+            {
+                GGML_ASSERT(false); // TODO: not implemented
+            } break;
+        case GGML_OP_CLAMP:
             {
                 GGML_ASSERT(false); // TODO: not implemented
             } break;
-        case GGML_OP_CONV_1D_S2_PH:
+        case GGML_OP_CONV_1D:
             {
                 GGML_ASSERT(false); // TODO: not implemented
             } break;
-        case GGML_OP_CONV_2D_SK_P0:
+        case GGML_OP_CONV_2D:
             {
                 GGML_ASSERT(false); // TODO: not implemented
             } break;
@@ -16606,9 +15773,6 @@ struct ggml_cgraph ggml_build_forward(struct ggml_tensor * tensor) {
     struct ggml_cgraph result = {
         /*.n_nodes      =*/ 0,
         /*.n_leafs      =*/ 0,
-        /*.n_threads    =*/ GGML_DEFAULT_N_THREADS,
-        /*.work_size    =*/ 0,
-        /*.work         =*/ NULL,
         /*.nodes        =*/ { NULL },
         /*.grads        =*/ { NULL },
         /*.leafs        =*/ { NULL },
@@ -16779,12 +15943,13 @@ void clear_numa_thread_affinity(void) {}
 #endif
 
 struct ggml_compute_state_shared {
-    struct ggml_cgraph * cgraph;
+    const struct ggml_cgraph * cgraph;
+    const struct ggml_cplan  * cplan;
 
     int64_t perf_node_start_cycles;
     int64_t perf_node_start_time_us;
 
-    int n_threads;
+    const int n_threads;
 
     // synchronization primitives
     atomic_int n_active; // num active threads
@@ -16808,9 +15973,13 @@ static void ggml_graph_compute_perf_stats_node(struct ggml_tensor * node, const
 
 static thread_ret_t ggml_graph_compute_thread(void * data) {
     struct ggml_compute_state * state = (struct ggml_compute_state *) data;
-    struct ggml_cgraph * cgraph = state->shared->cgraph;
 
-    const int n_threads = state->shared->n_threads;
+    const struct ggml_cgraph * cgraph = state->shared->cgraph;
+    const struct ggml_cplan  * cplan  = state->shared->cplan;
+
+    const int * n_tasks_arr = cplan->n_tasks;
+    const int   n_threads   = state->shared->n_threads;
+
     set_numa_thread_affinity(state->ith, n_threads);
 
     int node_n = -1;
@@ -16823,15 +15992,15 @@ static thread_ret_t ggml_graph_compute_thread(void * data) {
                 /*.type  =*/ GGML_TASK_FINALIZE,
                 /*.ith   =*/ 0,
                 /*.nth   =*/ 0,
-                /*.wsize =*/ cgraph->work ? ggml_nbytes(cgraph->work) : 0,
-                /*.wdata =*/ cgraph->work ? cgraph->work->data : NULL,
+                /*.wsize =*/ cplan->work_size,
+                /*.wdata =*/ cplan->work_data,
             };
 
             if (node_n != -1) {
                 /* FINALIZE */
                 struct ggml_tensor * node = state->shared->cgraph->nodes[node_n];
                 if (GGML_OP_HAS_FINALIZE[node->op]) {
-                    params.nth = node->n_tasks;
+                    params.nth = n_tasks_arr[node_n];
                     ggml_compute_forward(&params, node);
                     ggml_graph_compute_perf_stats_node(node, state->shared);
                 }
@@ -16842,11 +16011,12 @@ static thread_ret_t ggml_graph_compute_thread(void * data) {
                 GGML_PRINT_DEBUG_5("%s: %d/%d\n", __func__, node_n, cgraph->n_nodes);
 
                 struct ggml_tensor * node = cgraph->nodes[node_n];
+                const int n_tasks = n_tasks_arr[node_n];
 
                 state->shared->perf_node_start_cycles  = ggml_perf_cycles();
                 state->shared->perf_node_start_time_us = ggml_perf_time_us();
 
-                params.nth = node->n_tasks;
+                params.nth = n_tasks;
 
                 /* INIT */
                 if (GGML_OP_HAS_INIT[node->op]) {
@@ -16854,7 +16024,7 @@ static thread_ret_t ggml_graph_compute_thread(void * data) {
                     ggml_compute_forward(&params, node);
                 }
 
-                if (node->n_tasks == 1) {
+                if (n_tasks == 1) {
                     // TODO: maybe push node_n to the atomic but if other threads see n_tasks is 1,
                     // they do something more efficient than spinning (?)
                     params.type = GGML_TASK_COMPUTE;
@@ -16876,7 +16046,7 @@ static thread_ret_t ggml_graph_compute_thread(void * data) {
             // wait for other threads to finish
             const int last = node_n;
             do {
-                sched_yield();
+                //sched_yield();
                 node_n = atomic_load(&state->shared->node_n);
             } while (node_n == last);
         }
@@ -16886,16 +16056,17 @@ static thread_ret_t ggml_graph_compute_thread(void * data) {
 
         /* COMPUTE */
         struct ggml_tensor * node = cgraph->nodes[node_n];
+        const int n_tasks = n_tasks_arr[node_n];
 
         struct ggml_compute_params params = {
             /*.type  =*/ GGML_TASK_COMPUTE,
             /*.ith   =*/ state->ith,
-            /*.nth   =*/ node->n_tasks,
-            /*.wsize =*/ cgraph->work ? ggml_nbytes(cgraph->work) : 0,
-            /*.wdata =*/ cgraph->work ? cgraph->work->data : NULL,
+            /*.nth   =*/ n_tasks,
+            /*.wsize =*/ cplan->work_size,
+            /*.wdata =*/ cplan->work_data,
         };
 
-        if (state->ith < node->n_tasks) {
+        if (state->ith < n_tasks) {
             ggml_compute_forward(&params, node);
         }
     }
@@ -16903,365 +16074,372 @@ static thread_ret_t ggml_graph_compute_thread(void * data) {
     return 0;
 }
 
-void ggml_graph_compute(struct ggml_context * ctx, struct ggml_cgraph * cgraph) {
-    const int n_threads = cgraph->n_threads;
+struct ggml_cplan ggml_graph_plan(struct ggml_cgraph * cgraph, int n_threads) {
+    if (n_threads <= 0) {
+        n_threads = GGML_DEFAULT_N_THREADS;
+    }
 
-    struct ggml_compute_state_shared state_shared = {
-        /*.cgraph                  =*/ cgraph,
-        /*.perf_node_start_cycles  =*/ 0,
-        /*.perf_node_start_time_us =*/ 0,
-        /*.n_threads               =*/ n_threads,
-        /*.n_active                =*/ n_threads,
-        /*.node_n                  =*/ -1,
-    };
-    struct ggml_compute_state * workers = alloca(sizeof(struct ggml_compute_state)*n_threads);
+    size_t work_size = 0;
 
-    // initialize tasks + work buffer
-    {
-        size_t work_size = 0;
+    struct ggml_cplan cplan;
+    memset(&cplan, 0, sizeof(struct ggml_cplan));
 
-        // thread scheduling for the different operations
-        for (int i = 0; i < cgraph->n_nodes; i++) {
-            struct ggml_tensor * node = cgraph->nodes[i];
+    // thread scheduling for the different operations + work buffer size estimation
+    for (int i = 0; i < cgraph->n_nodes; i++) {
+        int n_tasks = 1;
 
-            switch (node->op) {
-                case GGML_OP_CPY:
-                case GGML_OP_DUP:
-                    {
-                        node->n_tasks = n_threads;
+        struct ggml_tensor * node = cgraph->nodes[i];
 
-                        size_t cur = 0;
-                        if (ggml_is_quantized(node->type)) {
-                            cur = GGML_TYPE_SIZE[GGML_TYPE_F32] * node->ne[0] * n_threads;
-                        }
+        switch (node->op) {
+            case GGML_OP_CPY:
+            case GGML_OP_DUP:
+                {
+                    n_tasks = n_threads;
 
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_ADD:
-                case GGML_OP_ADD1:
-                    {
-                        node->n_tasks = n_threads;
+                    size_t cur = 0;
+                    if (ggml_is_quantized(node->type)) {
+                        cur = GGML_TYPE_SIZE[GGML_TYPE_F32] * node->ne[0] * n_tasks;
+                    }
+
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_ADD:
+            case GGML_OP_ADD1:
+                {
+                    n_tasks = n_threads;
 
-                        size_t cur = 0;
+                    size_t cur = 0;
 
-                        if (ggml_is_quantized(node->src0->type)) {
-                            cur = GGML_TYPE_SIZE[GGML_TYPE_F32] * node->src0->ne[0] * n_threads;
-                        }
+                    if (ggml_is_quantized(node->src0->type)) {
+                        cur = GGML_TYPE_SIZE[GGML_TYPE_F32] * node->src0->ne[0] * n_tasks;
+                    }
 
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_ACC:
-                    {
-                        node->n_tasks = n_threads;
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_ACC:
+                {
+                    n_tasks = n_threads;
 
-                        size_t cur = 0;
+                    size_t cur = 0;
 
-                        if (ggml_is_quantized(node->src0->type)) {
-                            cur = GGML_TYPE_SIZE[GGML_TYPE_F32] * node->src1->ne[0] * n_threads;
-                        }
+                    if (ggml_is_quantized(node->src0->type)) {
+                        cur = GGML_TYPE_SIZE[GGML_TYPE_F32] * node->src1->ne[0] * n_tasks;
+                    }
+
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_SUB:
+            case GGML_OP_DIV:
+            case GGML_OP_SQR:
+            case GGML_OP_SQRT:
+            case GGML_OP_LOG:
+            case GGML_OP_SUM:
+            case GGML_OP_SUM_ROWS:
+            case GGML_OP_MEAN:
+            case GGML_OP_ARGMAX:
+            case GGML_OP_REPEAT:
+            case GGML_OP_REPEAT_BACK:
+            case GGML_OP_ABS:
+            case GGML_OP_SGN:
+            case GGML_OP_NEG:
+            case GGML_OP_STEP:
+            case GGML_OP_TANH:
+            case GGML_OP_ELU:
+            case GGML_OP_RELU:
+                {
+                    n_tasks = 1;
+                } break;
+            case GGML_OP_MUL:
+            case GGML_OP_GELU:
+            case GGML_OP_GELU_QUICK:
+            case GGML_OP_SILU:
+            case GGML_OP_SILU_BACK:
+            case GGML_OP_NORM:
+            case GGML_OP_RMS_NORM:
+            case GGML_OP_RMS_NORM_BACK:
+                {
+                    n_tasks = n_threads;
+                } break;
+            case GGML_OP_MUL_MAT:
+            case GGML_OP_OUT_PROD:
+                {
+                    n_tasks = n_threads;
 
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_SUB:
-                case GGML_OP_DIV:
-                case GGML_OP_SQR:
-                case GGML_OP_SQRT:
-                case GGML_OP_LOG:
-                case GGML_OP_SUM:
-                case GGML_OP_SUM_ROWS:
-                case GGML_OP_MEAN:
-                case GGML_OP_REPEAT:
-                case GGML_OP_REPEAT_BACK:
-                case GGML_OP_ABS:
-                case GGML_OP_SGN:
-                case GGML_OP_NEG:
-                case GGML_OP_STEP:
-                case GGML_OP_RELU:
-                    {
-                        node->n_tasks = 1;
-                    } break;
-                case GGML_OP_MUL:
-                case GGML_OP_GELU:
-                case GGML_OP_GELU_QUICK:
-                case GGML_OP_SILU:
-                case GGML_OP_SILU_BACK:
-                case GGML_OP_NORM:
-                case GGML_OP_RMS_NORM:
-                case GGML_OP_RMS_NORM_BACK:
-                    {
-                        node->n_tasks = n_threads;
-                    } break;
-                case GGML_OP_MUL_MAT:
-                case GGML_OP_OUT_PROD:
-                    {
-                        node->n_tasks = n_threads;
-
-                        // TODO: use different scheduling for different matrix sizes
-                        //const int nr0 = ggml_nrows(node->src0);
-                        //const int nr1 = ggml_nrows(node->src1);
-
-                        //node->n_tasks = MIN(n_threads, MAX(1, nr0/128));
-                        //printf("nr0 = %8d, nr1 = %8d, nr0*nr1 = %8d, n_tasks = %d\n", nr0, nr1, nr0*nr1, node->n_tasks);
-
-                        size_t cur = 0;
+                    // TODO: use different scheduling for different matrix sizes
+                    //const int nr0 = ggml_nrows(node->src0);
+                    //const int nr1 = ggml_nrows(node->src1);
+
+                    //n_tasks = MIN(n_threads, MAX(1, nr0/128));
+                    //printf("nr0 = %8d, nr1 = %8d, nr0*nr1 = %8d, n_tasks%d\n", nr0, nr1, nr0*nr1, n_tasks);
+
+                    size_t cur = 0;
+                    const enum ggml_type vec_dot_type = type_traits[node->src0->type].vec_dot_type;
 
 #if defined(GGML_USE_CUBLAS)
-                        if (ggml_cuda_can_mul_mat(node->src0, node->src1, node)) {
-                            node->n_tasks = 1; // TODO: this actually is doing nothing
-                                                //       the threads are still spinning
-                        }
-                        else
+                    if (ggml_cuda_can_mul_mat(node->src0, node->src1, node)) {
+                        n_tasks = 1; // TODO: this actually is doing nothing
+                                     //       the threads are still spinning
+                    } else
 #elif defined(GGML_USE_CLBLAST)
-                        if (ggml_cl_can_mul_mat(node->src0, node->src1, node)) {
-                            node->n_tasks = 1; // TODO: this actually is doing nothing
-                                                //       the threads are still spinning
-                            cur = ggml_cl_mul_mat_get_wsize(node->src0, node->src1, node);
-                        }
-                        else
-#endif
-                        if (node->src0->type == GGML_TYPE_F16 && node->src1->type == GGML_TYPE_F32) {
-#if defined(GGML_USE_ACCELERATE) || defined(GGML_USE_OPENBLAS)
-                            if (ggml_compute_forward_mul_mat_use_blas(node->src0, node->src1, node)) {
-                                node->n_tasks = 1; // TODO: this actually is doing nothing
-                                                   //       the threads are still spinning
-                                // here we need memory just for single 2D matrix from src0
-                                cur = GGML_TYPE_SIZE[GGML_TYPE_F32]*(node->src0->ne[0]*node->src0->ne[1]);
-                            } else {
-                                cur = GGML_TYPE_SIZE[GGML_TYPE_F16]*ggml_nelements(node->src1);
-                            }
-#else
-                            cur = GGML_TYPE_SIZE[GGML_TYPE_F16]*ggml_nelements(node->src1);
-#endif
-                        } else if (node->src0->type == GGML_TYPE_F32 && node->src1->type == GGML_TYPE_F32) {
-                            cur = 0;
-#if defined(GGML_USE_ACCELERATE) || defined(GGML_USE_OPENBLAS)
-                            if (ggml_compute_forward_mul_mat_use_blas(node->src0, node->src1, node)) {
-                                node->n_tasks = 1;
-                            }
+                    if (ggml_cl_can_mul_mat(node->src0, node->src1, node)) {
+                        n_tasks = 1; // TODO: this actually is doing nothing
+                                     //       the threads are still spinning
+                        cur = ggml_cl_mul_mat_get_wsize(node->src0, node->src1, node);
+                    } else
 #endif
-                        } else if (ggml_is_quantized(node->src0->type) && node->src1->type == GGML_TYPE_F32) {
 #if defined(GGML_USE_ACCELERATE) || defined(GGML_USE_OPENBLAS)
-                            if (ggml_compute_forward_mul_mat_use_blas(node->src0, node->src1, node)) {
-                                node->n_tasks = 1;
-                                cur = GGML_TYPE_SIZE[GGML_TYPE_F32]*(node->src0->ne[0]*node->src0->ne[1]);
-                            } else
-#endif
-                            {
-                                const enum ggml_type type_q = quantize_fns[node->src0->type].vec_dot_type;
-                                cur = GGML_TYPE_SIZE[type_q]*ggml_nelements(node->src1)/GGML_BLCK_SIZE[type_q];
-                            }
-                        } else {
-                            GGML_ASSERT(false);
+                    if (ggml_compute_forward_mul_mat_use_blas(node->src0, node->src1, node)) {
+                        n_tasks = 1; // TODO: this actually is doing nothing
+                                     //       the threads are still spinning
+                        if (node->src0->type != GGML_TYPE_F32) {
+                            // here we need memory just for single 2D matrix from src0
+                            cur = GGML_TYPE_SIZE[GGML_TYPE_F32]*(node->src0->ne[0]*node->src0->ne[1]);
                         }
+                    } else
+#endif
+                    if (node->src1->type != vec_dot_type) {
+                        cur = GGML_TYPE_SIZE[vec_dot_type]*ggml_nelements(node->src1)/GGML_BLCK_SIZE[vec_dot_type];
+                    } else {
+                        cur = 0;
+                    }
+
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_SCALE:
+                {
+                    n_tasks = 1;
+                } break;
+            case GGML_OP_SET:
+            case GGML_OP_CONT:
+            case GGML_OP_RESHAPE:
+            case GGML_OP_VIEW:
+            case GGML_OP_PERMUTE:
+            case GGML_OP_TRANSPOSE:
+            case GGML_OP_GET_ROWS:
+            case GGML_OP_GET_ROWS_BACK:
+            case GGML_OP_DIAG:
+            case GGML_OP_DIAG_MASK_ZERO:
+                {
+                    n_tasks = 1;
+                } break;
+            case GGML_OP_DIAG_MASK_INF:
+            case GGML_OP_SOFT_MAX:
+            case GGML_OP_SOFT_MAX_BACK:
+            case GGML_OP_ROPE:
+            case GGML_OP_ROPE_BACK:
+                {
+                    n_tasks = n_threads;
+                } break;
+            case GGML_OP_ALIBI:
+                {
+                    n_tasks = 1; //TODO
+                } break;
+            case GGML_OP_CLAMP:
+                {
+                    n_tasks = 1; //TODO
+                } break;
+            case GGML_OP_CONV_1D:
+                {
+                    n_tasks = n_threads;
+
+                    GGML_ASSERT(node->src0->ne[3] == 1);
+                    GGML_ASSERT(node->src1->ne[2] == 1);
+                    GGML_ASSERT(node->src1->ne[3] == 1);
+
+                    size_t cur = 0;
+                    const int nk = node->src0->ne[0];
 
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_SCALE:
-                    {
-                        node->n_tasks = 1;
-                    } break;
-                case GGML_OP_SET:
-                case GGML_OP_CONT:
-                case GGML_OP_RESHAPE:
-                case GGML_OP_VIEW:
-                case GGML_OP_PERMUTE:
-                case GGML_OP_TRANSPOSE:
-                case GGML_OP_GET_ROWS:
-                case GGML_OP_GET_ROWS_BACK:
-                case GGML_OP_DIAG:
-                case GGML_OP_DIAG_MASK_ZERO:
-                    {
-                        node->n_tasks = 1;
-                    } break;
-                case GGML_OP_DIAG_MASK_INF:
-                case GGML_OP_SOFT_MAX:
-                case GGML_OP_SOFT_MAX_BACK:
-                case GGML_OP_ROPE:
-                case GGML_OP_ROPE_BACK:
-                    {
-                        node->n_tasks = n_threads;
-                    } break;
-                case GGML_OP_ALIBI:
-                    {
-                        node->n_tasks = 1; //TODO
-                    } break;
-                case GGML_OP_CLAMP:
-                    {
-                        node->n_tasks = 1; //TODO
-                    } break;
-                case GGML_OP_CONV_1D_S1_PH:
-                case GGML_OP_CONV_1D_S2_PH:
-                    {
-                        node->n_tasks = n_threads;
-
-                        GGML_ASSERT(node->src0->ne[3] == 1);
-                        GGML_ASSERT(node->src1->ne[2] == 1);
-                        GGML_ASSERT(node->src1->ne[3] == 1);
-
-                        size_t cur = 0;
-                        const int nk = node->src0->ne[0];
-
-                        if (node->src0->type == GGML_TYPE_F16 &&
+                    if (node->src0->type == GGML_TYPE_F16 &&
                             node->src1->type == GGML_TYPE_F32) {
-                            cur = sizeof(ggml_fp16_t)*(
-                                    nk*ggml_up32(node->src0->ne[1])*node->src0->ne[2] +
-                                    ( 2*(nk/2) + node->src1->ne[0])*node->src1->ne[1]
-                                    );
-                        } else if (node->src0->type == GGML_TYPE_F32 &&
-                                   node->src1->type == GGML_TYPE_F32) {
-                            cur = sizeof(float)*(
-                                    nk*ggml_up32(node->src0->ne[1])*node->src0->ne[2] +
-                                    ( 2*(nk/2) + node->src1->ne[0])*node->src1->ne[1]
-                                    );
-                        } else {
-                            GGML_ASSERT(false);
-                        }
+                        cur = sizeof(ggml_fp16_t)*(
+                                nk*ggml_up32(node->src0->ne[1])*node->src0->ne[2] +
+                                ( 2*(nk/2) + node->src1->ne[0])*node->src1->ne[1]
+                                );
+                    } else if (node->src0->type == GGML_TYPE_F32 &&
+                            node->src1->type == GGML_TYPE_F32) {
+                        cur = sizeof(float)*(
+                                nk*ggml_up32(node->src0->ne[1])*node->src0->ne[2] +
+                                ( 2*(nk/2) + node->src1->ne[0])*node->src1->ne[1]
+                                );
+                    } else {
+                        GGML_ASSERT(false);
+                    }
 
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_CONV_2D_SK_P0:
-                    {
-                        node->n_tasks = n_threads;
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_CONV_2D:
+                {
+                    n_tasks = n_threads;
 
-                        GGML_ASSERT(node->src1->ne[3] == 1);
+                    GGML_ASSERT(node->src1->ne[3] == 1);
 
-                        const int64_t ne00 = node->src0->ne[0]; // W
-                        const int64_t ne01 = node->src0->ne[1]; // H
-                        const int64_t ne02 = node->src0->ne[2]; // C
-                        const int64_t ne03 = node->src0->ne[3]; // N
+                    const int64_t ne00 = node->src0->ne[0]; // W
+                    const int64_t ne01 = node->src0->ne[1]; // H
+                    const int64_t ne02 = node->src0->ne[2]; // C
+                    const int64_t ne03 = node->src0->ne[3]; // N
 
-                        const int64_t ne10 = node->src1->ne[0]; // W
-                        const int64_t ne11 = node->src1->ne[1]; // H
-                        const int64_t ne12 = node->src1->ne[2]; // C
+                    const int64_t ne10 = node->src1->ne[0]; // W
+                    const int64_t ne11 = node->src1->ne[1]; // H
+                    const int64_t ne12 = node->src1->ne[2]; // C
 
-                        const int64_t nk = ne00*ne01;
+                    const int64_t nk = ne00*ne01;
 
-                        UNUSED(ne02);
-                        UNUSED(ne03);
-                        UNUSED(nk);
+                    UNUSED(ne02);
+                    UNUSED(ne03);
+                    UNUSED(nk);
 
-                        size_t cur = 0;
+                    size_t cur = 0;
 
-                        if (node->src0->type == GGML_TYPE_F16 &&
+                    if (node->src0->type == GGML_TYPE_F16 &&
                             node->src1->type == GGML_TYPE_F32) {
-                            cur = sizeof(ggml_fp16_t)*(ne10*ne11*ne12);
-                        } else if (node->src0->type == GGML_TYPE_F32 &&
-                                   node->src1->type == GGML_TYPE_F32) {
-                            cur = sizeof(float)*      (ne10*ne11*ne12);
-                        } else {
-                            GGML_ASSERT(false);
-                        }
+                        cur = sizeof(ggml_fp16_t)*(ne10*ne11*ne12);
+                    } else if (node->src0->type == GGML_TYPE_F32 &&
+                            node->src1->type == GGML_TYPE_F32) {
+                        cur = sizeof(float)*      (ne10*ne11*ne12);
+                    } else {
+                        GGML_ASSERT(false);
+                    }
 
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_FLASH_ATTN:
-                    {
-                        node->n_tasks = n_threads;
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_FLASH_ATTN:
+                {
+                    n_tasks = n_threads;
 
-                        size_t cur = 0;
+                    size_t cur = 0;
 
-                        const int64_t ne11 = ggml_up(node->src1->ne[1], GGML_SOFT_MAX_UNROLL);
+                    const int64_t ne11 = ggml_up(node->src1->ne[1], GGML_SOFT_MAX_UNROLL);
 
-                        if (node->src1->type == GGML_TYPE_F32) {
-                            cur  = sizeof(float)*ne11*node->n_tasks; // TODO: this can become (n_tasks-1)
-                            cur += sizeof(float)*ne11*node->n_tasks; // this is overestimated by x2
-                        }
+                    if (node->src1->type == GGML_TYPE_F32) {
+                        cur  = sizeof(float)*ne11*n_tasks; // TODO: this can become (n_tasks-1)
+                        cur += sizeof(float)*ne11*n_tasks; // this is overestimated by x2
+                    }
 
-                        if (node->src1->type == GGML_TYPE_F16) {
-                            cur  = sizeof(float)*ne11*node->n_tasks; // TODO: this can become (n_tasks-1)
-                            cur += sizeof(float)*ne11*node->n_tasks; // this is overestimated by x2
-                        }
+                    if (node->src1->type == GGML_TYPE_F16) {
+                        cur  = sizeof(float)*ne11*n_tasks; // TODO: this can become (n_tasks-1)
+                        cur += sizeof(float)*ne11*n_tasks; // this is overestimated by x2
+                    }
 
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_FLASH_FF:
-                    {
-                        node->n_tasks = n_threads;
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_FLASH_FF:
+                {
+                    n_tasks = n_threads;
 
-                        size_t cur = 0;
+                    size_t cur = 0;
 
-                        if (node->src1->type == GGML_TYPE_F32) {
-                            cur  = sizeof(float)*node->src1->ne[1]*node->n_tasks; // TODO: this can become (n_tasks-1)
-                            cur += sizeof(float)*node->src1->ne[1]*node->n_tasks; // this is overestimated by x2
-                        }
+                    if (node->src1->type == GGML_TYPE_F32) {
+                        cur  = sizeof(float)*node->src1->ne[1]*n_tasks; // TODO: this can become (n_tasks-1)
+                        cur += sizeof(float)*node->src1->ne[1]*n_tasks; // this is overestimated by x2
+                    }
 
-                        if (node->src1->type == GGML_TYPE_F16) {
-                            cur  = sizeof(float)*node->src1->ne[1]*node->n_tasks; // TODO: this can become (n_tasks-1)
-                            cur += sizeof(float)*node->src1->ne[1]*node->n_tasks; // this is overestimated by x2
-                        }
+                    if (node->src1->type == GGML_TYPE_F16) {
+                        cur  = sizeof(float)*node->src1->ne[1]*n_tasks; // TODO: this can become (n_tasks-1)
+                        cur += sizeof(float)*node->src1->ne[1]*n_tasks; // this is overestimated by x2
+                    }
 
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_FLASH_ATTN_BACK:
-                    {
-                        node->n_tasks = n_threads;
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_FLASH_ATTN_BACK:
+                {
+                    n_tasks = n_threads;
 
-                        size_t cur = 0;
+                    size_t cur = 0;
 
-                        const int64_t    D = node->src0->ne[0];
-                        const int64_t ne11 = ggml_up(node->src1->ne[1], GGML_SOFT_MAX_UNROLL);
-                        const int64_t mxDn = MAX(D, ne11) * 2; // *2 because of S and SM in ggml_compute_forward_flash_attn_back
-                        if (node->src1->type == GGML_TYPE_F32) {
-                            cur  = sizeof(float)*mxDn*node->n_tasks; // TODO: this can become (n_tasks-1)
-                            cur += sizeof(float)*mxDn*node->n_tasks; // this is overestimated by x2
-                        }
+                    const int64_t    D = node->src0->ne[0];
+                    const int64_t ne11 = ggml_up(node->src1->ne[1], GGML_SOFT_MAX_UNROLL);
+                    const int64_t mxDn = MAX(D, ne11) * 2; // *2 because of S and SM in ggml_compute_forward_flash_attn_back
+                    if (node->src1->type == GGML_TYPE_F32) {
+                        cur  = sizeof(float)*mxDn*n_tasks; // TODO: this can become (n_tasks-1)
+                        cur += sizeof(float)*mxDn*n_tasks; // this is overestimated by x2
+                    }
 
-                        if (node->src1->type == GGML_TYPE_F16) {
-                            cur  = sizeof(float)*mxDn*node->n_tasks; // TODO: this can become (n_tasks-1)
-                            cur += sizeof(float)*mxDn*node->n_tasks; // this is overestimated by x2
-                        }
+                    if (node->src1->type == GGML_TYPE_F16) {
+                        cur  = sizeof(float)*mxDn*n_tasks; // TODO: this can become (n_tasks-1)
+                        cur += sizeof(float)*mxDn*n_tasks; // this is overestimated by x2
+                    }
 
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_WIN_PART:
-                case GGML_OP_WIN_UNPART:
-                case GGML_OP_MAP_UNARY:
-                case GGML_OP_MAP_BINARY:
-                case GGML_OP_MAP_CUSTOM1:
-                case GGML_OP_MAP_CUSTOM2:
-                case GGML_OP_MAP_CUSTOM3:
-                    {
-                        node->n_tasks = 1;
-                    } break;
-                case GGML_OP_CROSS_ENTROPY_LOSS:
-                    {
-                        node->n_tasks = n_threads;
-
-                        size_t cur = ggml_type_size(node->type)*(node->n_tasks + node->src0->ne[0]*node->n_tasks);
-
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_CROSS_ENTROPY_LOSS_BACK:
-                    {
-                        node->n_tasks = n_threads;
-
-                        size_t cur = ggml_type_size(node->type)*node->src0->ne[0]*node->n_tasks;
-
-                        work_size = MAX(work_size, cur);
-                    } break;
-                case GGML_OP_NONE:
-                    {
-                        node->n_tasks = 1;
-                    } break;
-                case GGML_OP_COUNT:
-                    {
-                        GGML_ASSERT(false);
-                    } break;
-            }
-        }
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_WIN_PART:
+            case GGML_OP_WIN_UNPART:
+            case GGML_OP_MAP_UNARY:
+            case GGML_OP_MAP_BINARY:
+            case GGML_OP_MAP_CUSTOM1:
+            case GGML_OP_MAP_CUSTOM2:
+            case GGML_OP_MAP_CUSTOM3:
+                {
+                    n_tasks = 1;
+                } break;
+            case GGML_OP_CROSS_ENTROPY_LOSS:
+                {
+                    n_tasks = n_threads;
+
+                    size_t cur = ggml_type_size(node->type)*(n_tasks + node->src0->ne[0]*n_tasks);
+
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_CROSS_ENTROPY_LOSS_BACK:
+                {
+                    n_tasks = n_threads;
 
-        if (cgraph->work != NULL && work_size > cgraph->work_size) {
-            GGML_ASSERT(false); // TODO: better handling
+                    size_t cur = ggml_type_size(node->type)*node->src0->ne[0]*n_tasks;
+
+                    work_size = MAX(work_size, cur);
+                } break;
+            case GGML_OP_NONE:
+                {
+                    n_tasks = 1;
+                } break;
+            case GGML_OP_COUNT:
+                {
+                    GGML_ASSERT(false);
+                } break;
         }
 
-        if (work_size > 0 && cgraph->work == NULL) {
-            cgraph->work_size = work_size + CACHE_LINE_SIZE*(n_threads - 1);
+        cplan.n_tasks[i] = n_tasks;
+    }
+
+    if (work_size > 0) {
+        work_size += CACHE_LINE_SIZE*(n_threads - 1);
+    }
+
+    cplan.n_threads = n_threads;
+    cplan.work_size = work_size;
+    cplan.work_data = NULL;
 
-            GGML_PRINT_DEBUG("%s: allocating work buffer for graph (%zu bytes)\n", __func__, cgraph->work_size);
-            cgraph->work = ggml_new_tensor_1d(ctx, GGML_TYPE_I8, cgraph->work_size);
+    return cplan;
+}
+
+void ggml_graph_compute(struct ggml_cgraph * cgraph, struct ggml_cplan * cplan) {
+    {
+        GGML_ASSERT(cplan);
+        GGML_ASSERT(cplan->n_threads > 0);
+
+        if (cplan->work_size > 0) {
+            GGML_ASSERT(cplan->work_data);
+        }
+
+        for (int i = 0; i < cgraph->n_nodes; ++i) {
+            if (cgraph->nodes[i]->op != GGML_OP_NONE) {
+                GGML_ASSERT(cplan->n_tasks[i] > 0);
+            }
         }
     }
 
+    const int n_threads = cplan->n_threads;
+
+    struct ggml_compute_state_shared state_shared = {
+        /*.cgraph                  =*/ cgraph,
+        /*.cgraph_plan             =*/ cplan,
+        /*.perf_node_start_cycles  =*/ 0,
+        /*.perf_node_start_time_us =*/ 0,
+        /*.n_threads               =*/ n_threads,
+        /*.n_active                =*/ n_threads,
+        /*.node_n                  =*/ -1,
+    };
+    struct ggml_compute_state * workers = alloca(sizeof(struct ggml_compute_state)*n_threads);
+
     // create thread pool
     if (n_threads > 1) {
         for (int j = 1; j < n_threads; ++j) {
@@ -17323,6 +16501,17 @@ void ggml_graph_reset(struct ggml_cgraph * cgraph) {
     }
 }
 
+void ggml_graph_compute_with_ctx(struct ggml_context * ctx, struct ggml_cgraph * cgraph, int n_threads) {
+    struct ggml_cplan cplan = ggml_graph_plan(cgraph, n_threads);
+
+    struct ggml_tensor * buf = ggml_new_tensor_1d(ctx, GGML_TYPE_I8, cplan.work_size);
+    GGML_ASSERT(buf);
+
+    cplan.work_data = buf->data;
+
+    ggml_graph_compute(cgraph, &cplan);
+}
+
 struct ggml_tensor * ggml_graph_get_tensor(struct ggml_cgraph * cgraph, const char * name) {
     for (int i = 0; i < cgraph->n_leafs; i++) {
         struct ggml_tensor * leaf = cgraph->leafs[i];
@@ -17361,14 +16550,13 @@ static void ggml_graph_export_node(const struct ggml_tensor * tensor, const char
     const int64_t * ne = tensor->ne;
     const size_t  * nb = tensor->nb;
 
-    fprintf(fout, "%-6s %-6s %-12s %8d %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %16zu %16zu %16zu %16zu %8d %16p %32s\n",
+    fprintf(fout, "%-6s %-6s %-12s %8d %" PRId64 " %" PRId64 " %" PRId64 " %" PRId64 " %16zu %16zu %16zu %16zu %16p %32s\n",
             arg,
             ggml_type_name(tensor->type),
             ggml_op_name  (tensor->op),
             tensor->n_dims,
             ne[0], ne[1], ne[2], ne[3],
             nb[0], nb[1], nb[2], nb[3],
-            tensor->n_tasks,
             tensor->data,
             tensor->name);
 }
@@ -17481,13 +16669,6 @@ void ggml_graph_export(const struct ggml_cgraph * cgraph, const char * fname) {
                     fwrite(&nb, sizeof(uint64_t), 1, fout);
                 }
 
-                // store the pointer address
-                {
-                    const uint64_t ptr = (uint64_t) tensor->data;
-
-                    fwrite(&ptr, sizeof(uint64_t), 1, fout);
-                }
-
                 fwrite(tensor->name, sizeof(char), GGML_MAX_NAME, fout);
 
                 // dump the data
@@ -17521,13 +16702,6 @@ void ggml_graph_export(const struct ggml_cgraph * cgraph, const char * fname) {
                     fwrite(&nb, sizeof(uint64_t), 1, fout);
                 }
 
-                // store the pointer address
-                {
-                    const uint64_t ptr = (uint64_t) tensor->data;
-
-                    fwrite(&ptr, sizeof(uint64_t), 1, fout);
-                }
-
                 fwrite(tensor->name, sizeof(char), GGML_MAX_NAME, fout);
 
                 // output the op arguments
@@ -17712,8 +16886,6 @@ struct ggml_cgraph ggml_graph_import(const char * fname, struct ggml_context **
 
                 tensor->op = (enum ggml_op) op;
 
-                uint64_t ptr_cur = *(const uint64_t *) ptr; ptr += sizeof(ptr_cur);
-
                 memcpy(tensor->name, ptr, GGML_MAX_NAME); ptr += GGML_MAX_NAME;
 
                 tensor->data = (void *) ptr;
@@ -17759,8 +16931,6 @@ struct ggml_cgraph ggml_graph_import(const char * fname, struct ggml_context **
                     nb[j] = nb_cur;
                 }
 
-                uint64_t ptr_cur = *(const uint64_t *) ptr; ptr += sizeof(ptr_cur); // TODO: not yet used
-
                 const char * ptr_name = ptr; ptr += GGML_MAX_NAME;
 
                 const int32_t * ptr_arg_idx = (const int32_t *) ptr; ptr += (2 + GGML_MAX_OPT)*sizeof(int32_t);
@@ -18122,9 +17292,6 @@ static enum ggml_opt_result ggml_opt_adam(
         struct ggml_cgraph * gb) {
     GGML_ASSERT(ggml_is_scalar(f));
 
-    gf->n_threads = params.n_threads;
-    gb->n_threads = params.n_threads;
-
     // these will store the parameters we want to optimize
     struct ggml_tensor * ps[GGML_MAX_PARAMS];
 
@@ -18171,7 +17338,8 @@ static enum ggml_opt_result ggml_opt_adam(
     // compute the function value
     ggml_graph_reset  (gf);
     ggml_set_f32      (f->grad, 1.0f);
-    ggml_graph_compute(ctx, gb);
+
+    ggml_graph_compute_with_ctx(ctx, gb, params.n_threads);
 
     opt->adam.fx_prev = ggml_get_f32_1d(f, 0);
     opt->adam.fx_best = opt->adam.fx_prev;
@@ -18251,7 +17419,8 @@ static enum ggml_opt_result ggml_opt_adam(
 
         ggml_graph_reset  (gf);
         ggml_set_f32      (f->grad, 1.0f);
-        ggml_graph_compute(ctx, gb);
+
+        ggml_graph_compute_with_ctx(ctx, gb, params.n_threads);
 
         const float fx = ggml_get_f32_1d(f, 0);
 
@@ -18373,7 +17542,8 @@ static enum ggml_opt_result linesearch_backtracking(
 
             ggml_graph_reset  (gf);
             ggml_set_f32      (f->grad, 1.0f);
-            ggml_graph_compute(ctx, gb);
+
+            ggml_graph_compute_with_ctx(ctx, gb, params->n_threads);
 
             ggml_opt_get_grad(np, ps, g);
 
@@ -18441,9 +17611,6 @@ static enum ggml_opt_result ggml_opt_lbfgs(
         }
     }
 
-    gf->n_threads = params.n_threads;
-    gb->n_threads = params.n_threads;
-
     const int m = params.lbfgs.m;
 
     // these will store the parameters we want to optimize
@@ -18495,7 +17662,8 @@ static enum ggml_opt_result ggml_opt_lbfgs(
 
         ggml_graph_reset  (gf);
         ggml_set_f32      (f->grad, 1.0f);
-        ggml_graph_compute(ctx, gb);
+
+        ggml_graph_compute_with_ctx(ctx, gb, params.n_threads);
 
         ggml_opt_get_grad(np, ps, g);
 
diff --git a/ggml.h b/ggml.h
index 11b51f8bd656a..ab84bef68747e 100644
--- a/ggml.h
+++ b/ggml.h
@@ -65,7 +65,7 @@
 //       ggml_set_f32(a, 3.0f);
 //       ggml_set_f32(b, 4.0f);
 //
-//       ggml_graph_compute(ctx0, &gf);
+//       ggml_graph_compute_with_ctx(ctx, &gf, n_threads);
 //
 //       printf("f = %f\n", ggml_get_f32_1d(f, 0));
 //
@@ -201,6 +201,8 @@
 #define GGML_MAX_NAME          48
 #define GGML_DEFAULT_N_THREADS 4
 
+#define GGML_UNUSED(x) (void)(x)
+
 #define GGML_ASSERT(x) \
     do { \
         if (!(x)) { \
@@ -209,6 +211,30 @@
         } \
     } while (0)
 
+// used to copy the number of elements and stride in bytes of tensors into local variables.
+// main purpose is to reduce code duplication and improve readability.
+//
+// example:
+//
+//    GGML_TENSOR_LOCALS(int64_t, ne1, src1, ne);
+//    GGML_TENSOR_LOCALS(size_t,  nb1, src1, nb);
+//
+#define GGML_TENSOR_LOCALS_1(type, prefix, pointer, array) \
+    const type prefix##0 = (pointer)->array[0]; \
+    GGML_UNUSED(prefix##0);
+#define GGML_TENSOR_LOCALS_2(type, prefix, pointer, array) \
+    GGML_TENSOR_LOCALS_1    (type, prefix, pointer, array) \
+    const type prefix##1 = (pointer)->array[1]; \
+    GGML_UNUSED(prefix##1);
+#define GGML_TENSOR_LOCALS_3(type, prefix, pointer, array) \
+    GGML_TENSOR_LOCALS_2    (type, prefix, pointer, array) \
+    const type prefix##2 = (pointer)->array[2]; \
+    GGML_UNUSED(prefix##2);
+#define GGML_TENSOR_LOCALS(type, prefix, pointer, array) \
+    GGML_TENSOR_LOCALS_3  (type, prefix, pointer, array) \
+    const type prefix##3 = (pointer)->array[3]; \
+    GGML_UNUSED(prefix##3);
+
 #ifdef  __cplusplus
 extern "C" {
 #endif
@@ -224,8 +250,8 @@ extern "C" {
     GGML_API float       ggml_fp16_to_fp32(ggml_fp16_t x);
     GGML_API ggml_fp16_t ggml_fp32_to_fp16(float x);
 
-    GGML_API void ggml_fp16_to_fp32_row(const ggml_fp16_t * x, float * y, size_t n);
-    GGML_API void ggml_fp32_to_fp16_row(const float * x, ggml_fp16_t * y, size_t n);
+    GGML_API void ggml_fp16_to_fp32_row(const ggml_fp16_t * x, float * y, int n);
+    GGML_API void ggml_fp32_to_fp16_row(const float * x, ggml_fp16_t * y, int n);
 
     struct ggml_object;
     struct ggml_context;
@@ -295,12 +321,15 @@ extern "C" {
         GGML_OP_SUM,
         GGML_OP_SUM_ROWS,
         GGML_OP_MEAN,
+        GGML_OP_ARGMAX,
         GGML_OP_REPEAT,
         GGML_OP_REPEAT_BACK,
         GGML_OP_ABS,
         GGML_OP_SGN,
         GGML_OP_NEG,
         GGML_OP_STEP,
+        GGML_OP_TANH,
+        GGML_OP_ELU,
         GGML_OP_RELU,
         GGML_OP_GELU,
         GGML_OP_GELU_QUICK,
@@ -332,9 +361,8 @@ extern "C" {
         GGML_OP_ROPE_BACK,
         GGML_OP_ALIBI,
         GGML_OP_CLAMP,
-        GGML_OP_CONV_1D_S1_PH,
-        GGML_OP_CONV_1D_S2_PH,
-        GGML_OP_CONV_2D_SK_P0,
+        GGML_OP_CONV_1D,
+        GGML_OP_CONV_2D,
 
         GGML_OP_FLASH_ATTN,
         GGML_OP_FLASH_FF,
@@ -390,9 +418,6 @@ extern "C" {
         struct ggml_tensor * src1;
         struct ggml_tensor * opt[GGML_MAX_OPT];
 
-        // thread scheduling
-        int n_tasks;
-
         // performance
         int     perf_runs;
         int64_t perf_cycles;
@@ -404,19 +429,27 @@ extern "C" {
 
         void * extra; // extra things e.g. for ggml-cuda.cu
 
-        char padding[4];
+        char padding[8];
     };
 
     static const size_t GGML_TENSOR_SIZE = sizeof(struct ggml_tensor);
 
+    // the compute plan that needs to be prepared for ggml_graph_compute()
+    // since https://github.com/ggerganov/ggml/issues/287
+    struct ggml_cplan {
+        size_t    work_size; // size of work buffer, calculated by `ggml_graph_plan()`
+        uint8_t * work_data; // work buffer, to be allocated by caller before calling to `ggml_graph_compute()`
+
+        int n_threads;
+
+        // the `n_tasks` of nodes, 1:1 mapping to cgraph nodes
+        int n_tasks[GGML_MAX_NODES];
+    };
+
     // computation graph
     struct ggml_cgraph {
         int n_nodes;
         int n_leafs;
-        int n_threads;
-
-        size_t work_size;
-        struct ggml_tensor * work;
 
         struct ggml_tensor * nodes[GGML_MAX_NODES];
         struct ggml_tensor * grads[GGML_MAX_NODES];
@@ -690,6 +723,11 @@ extern "C" {
             struct ggml_context * ctx,
             struct ggml_tensor  * a);
 
+    // argmax along rows
+    GGML_API struct ggml_tensor * ggml_argmax(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a);
+
     // if a is the same shape as b, and a is not parameter, return a
     // otherwise, return a new tensor: repeat(a) to fit in b
     GGML_API struct ggml_tensor * ggml_repeat(
@@ -734,6 +772,22 @@ extern "C" {
             struct ggml_context * ctx,
             struct ggml_tensor  * a);
 
+    GGML_API struct ggml_tensor * ggml_tanh(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a);
+
+    GGML_API struct ggml_tensor * ggml_tanh_inplace(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a);
+
+    GGML_API struct ggml_tensor * ggml_elu(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a);
+
+    GGML_API struct ggml_tensor * ggml_elu_inplace(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a);
+
     GGML_API struct ggml_tensor * ggml_relu(
             struct ggml_context * ctx,
             struct ggml_tensor  * a);
@@ -1084,58 +1138,33 @@ extern "C" {
             float                 min,
             float                 max);
 
-    // TODO: implement general-purpose convolutions
-    // GGML_API struct ggml_tensor * ggml_conv_1d(
-    //        struct ggml_context * ctx,
-    //        struct ggml_tensor  * a,
-    //        struct ggml_tensor  * b,
-    //        int                   s0
-    //        int                   p0,
-    //        int                   d0);
-    //
-    // GGML_API struct ggml_tensor * ggml_conv_2d(
-    //        struct ggml_context * ctx,
-    //        struct ggml_tensor  * a,
-    //        struct ggml_tensor  * b,
-    //        int                   s0,
-    //        int                   s1,
-    //        int                   p0,
-    //        int                   p1,
-    //        int                   d0,
-    //        int                   d1);
-
-    // padding = half
-    // TODO: we don't support extra parameters for now
-    //       that's why we are hard-coding the stride, padding, and dilation
-    //       not great ..
-    // example:
-    // a:      3   80  768    1
-    // b:   3000   80    1    1
-    // res: 3000  768    1    1
-    // used in whisper
-    GGML_API struct ggml_tensor * ggml_conv_1d_s1_ph(
+    GGML_API struct ggml_tensor * ggml_conv_1d(
             struct ggml_context * ctx,
             struct ggml_tensor  * a,
-            struct ggml_tensor  * b);
+            struct ggml_tensor  * b,
+            int                   s0,  // stride
+            int                   p0,  // padding
+            int                   d0); // dilation
 
-    // used in whisper
-    GGML_API struct ggml_tensor * ggml_conv_1d_s2_ph(
+    GGML_API struct ggml_tensor * ggml_conv_2d(
             struct ggml_context * ctx,
             struct ggml_tensor  * a,
-            struct ggml_tensor  * b);
+            struct ggml_tensor  * b,
+            int                   s0,
+            int                   s1,
+            int                   p0,
+            int                   p1,
+            int                   d0,
+            int                   d1);
 
-    // kernel size is a->ne[0] x a->ne[1]
-    // stride is equal to kernel size
-    // padding is zero
-    // example:
-    // a:     16   16    3  768
-    // b:   1024 1024    3    1
-    // res:   64   64  768    1
-    // used in sam
-    GGML_API struct ggml_tensor * ggml_conv_2d_sk_p0(
+    // conv_1d with padding = half
+    // alias for ggml_conv_1d(a, b, s, a->ne[0]/2, d)
+    GGML_API struct ggml_tensor* ggml_conv_1d_ph(
             struct ggml_context * ctx,
             struct ggml_tensor  * a,
-            struct ggml_tensor  * b);
+            struct ggml_tensor  * b,
+            int                   s,
+            int                   d);
 
     GGML_API struct ggml_tensor * ggml_flash_attn(
             struct ggml_context * ctx,
@@ -1266,15 +1295,22 @@ extern "C" {
 
     GGML_API void ggml_set_param(
             struct ggml_context * ctx,
-            struct ggml_tensor * tensor);
+            struct ggml_tensor  * tensor);
 
     GGML_API void ggml_build_forward_expand(struct ggml_cgraph * cgraph, struct ggml_tensor * tensor);
 
     GGML_API struct ggml_cgraph ggml_build_forward (struct ggml_tensor * tensor);
     GGML_API struct ggml_cgraph ggml_build_backward(struct ggml_context * ctx, struct ggml_cgraph * gf, bool keep);
 
-    GGML_API void ggml_graph_compute(struct ggml_context * ctx, struct ggml_cgraph * cgraph);
-    GGML_API void ggml_graph_reset  (struct ggml_cgraph * cgraph);
+    // ggml_graph_plan() has to be called before ggml_graph_compute()
+    // when plan.work_size > 0, caller must allocate memory for plan.work_data
+    GGML_API struct ggml_cplan ggml_graph_plan   (struct ggml_cgraph * cgraph, int n_threads /*= GGML_DEFAULT_N_THREADS*/);
+    GGML_API              void ggml_graph_compute(struct ggml_cgraph * cgraph, struct ggml_cplan * cplan);
+    GGML_API              void ggml_graph_reset  (struct ggml_cgraph * cgraph);
+
+    // same as ggml_graph_compute() but the work data is allocated as a part of the context
+    // note: the drawback of this API is that you must have ensured that the context has enough memory for the work data
+    GGML_API void ggml_graph_compute_with_ctx(struct ggml_context * ctx, struct ggml_cgraph * cgraph, int n_threads);
 
     GGML_API struct ggml_tensor * ggml_graph_get_tensor(struct ggml_cgraph * cgraph, const char * name);
 
@@ -1491,25 +1527,24 @@ extern "C" {
     //
 
 #ifdef  __cplusplus
-    // restrict not standard in C++
+// restrict not standard in C++
 #define GGML_RESTRICT
 #else
 #define GGML_RESTRICT restrict
 #endif
-    typedef void (*dequantize_row_q_t)(const void * GGML_RESTRICT x, float * GGML_RESTRICT y, int k);
-    typedef void (*quantize_row_q_t)  (const float * GGML_RESTRICT x, void * GGML_RESTRICT y, int k);
-    typedef void (*vec_dot_q_t)       (const int n, float * GGML_RESTRICT s, const void * GGML_RESTRICT x, const void * GGML_RESTRICT y);
+    typedef void (*ggml_to_float_t)  (const void  * GGML_RESTRICT x, float * GGML_RESTRICT y, int k);
+    typedef void (*ggml_from_float_t)(const float * GGML_RESTRICT x, void  * GGML_RESTRICT y, int k);
+    typedef void (*ggml_vec_dot_t)   (const int n, float * GGML_RESTRICT s, const void * GGML_RESTRICT x, const void * GGML_RESTRICT y);
 
     typedef struct {
-        dequantize_row_q_t dequantize_row_q;
-        quantize_row_q_t   quantize_row_q;
-        quantize_row_q_t   quantize_row_q_reference;
-        quantize_row_q_t   quantize_row_q_dot;
-        vec_dot_q_t        vec_dot_q;
-        enum ggml_type     vec_dot_type;
-    } quantize_fns_t;
-
-    quantize_fns_t ggml_internal_get_quantize_fn(size_t i);
+        ggml_to_float_t   to_float;
+        ggml_from_float_t from_float;
+        ggml_from_float_t from_float_reference;
+        ggml_vec_dot_t    vec_dot;
+        enum ggml_type    vec_dot_type;
+    } ggml_type_traits_t;
+
+    ggml_type_traits_t ggml_internal_get_type_traits(enum ggml_type i);
 
 #ifdef  __cplusplus
 }
diff --git a/llama.cpp b/llama.cpp
index 7419b03b61dc3..a491f1c7e15c4 100644
--- a/llama.cpp
+++ b/llama.cpp
@@ -79,6 +79,25 @@ void llama_nop(struct ggml_tensor * tensor) { // don't offload by default
     (void) tensor;
 }
 
+//
+// ggml helpers
+//
+
+static void ggml_graph_compute_helper(std::vector<uint8_t> & buf, ggml_cgraph * graph, int n_threads) {
+    struct ggml_cplan plan = ggml_graph_plan(graph, n_threads);
+
+    if (plan.work_size > 0) {
+        buf.resize(plan.work_size);
+        plan.work_data = buf.data();
+    }
+
+    ggml_graph_compute(graph, &plan);
+}
+
+//
+// memory sizes
+//
+
 static const std::map<e_model, size_t> & MEM_REQ_SCRATCH0()
 {
     static std::map<e_model, size_t> k_sizes = {
@@ -321,6 +340,9 @@ struct llama_context {
     // input embedding (1-dimensional array: [n_embd])
     std::vector<float> embedding;
 
+    // reusable buffer for `struct ggml_graph_plan.work_data`
+    std::vector<uint8_t> work_buffer;
+
     // memory buffers used to evaluate the model
     // TODO: move in llama_state
     llama_ctx_buffer buf_compute;
@@ -758,7 +780,6 @@ struct llama_model_loader {
 
 };
 
-
 //
 // kv cache
 //
@@ -1156,6 +1177,7 @@ static void llama_model_load_internal(
             }
         }
 #endif // GGML_USE_CUBLAS
+
 #if defined(GGML_USE_CUBLAS) || defined(GGML_USE_CLBLAST)
         const int n_gpu = std::min(n_gpu_layers, int(hparams.n_layer));
 
@@ -1164,6 +1186,10 @@ static void llama_model_load_internal(
             fprintf(stderr, "%s: offloading non-repeating layers to GPU\n", __func__);
         }
         size_t vram_kv_cache = 0;
+
+#ifdef GGML_USE_CUBLAS
+        const int max_backend_supported_layers = hparams.n_layer + 3;
+        const int max_offloadable_layers = low_vram ? hparams.n_layer + 1 : hparams.n_layer + 3;
         if (n_gpu_layers > (int) hparams.n_layer + 1) {
             if (low_vram) {
                 fprintf(stderr, "%s: cannot offload v cache to GPU due to low VRAM option\n", __func__);
@@ -1180,14 +1206,18 @@ static void llama_model_load_internal(
                 vram_kv_cache += MEM_REQ_KV_SELF().at(model.type) / 2;
             }
         }
-        const int max_offloadable_layers = low_vram ? hparams.n_layer + 1 : hparams.n_layer + 3;
+#elif defined(GGML_USE_CLBLAST)
+        const int max_backend_supported_layers = hparams.n_layer + 1;
+        const int max_offloadable_layers = hparams.n_layer + 1;
+#endif // GGML_USE_CUBLAS
+
         fprintf(stderr, "%s: offloaded %d/%d layers to GPU\n",
-                __func__, std::min(n_gpu_layers, max_offloadable_layers), hparams.n_layer + 3);
+                __func__, std::min(n_gpu_layers, max_offloadable_layers), max_backend_supported_layers);
         fprintf(stderr, "%s: total VRAM used: %zu MB\n",
                 __func__, (vram_weights + vram_scratch + vram_kv_cache + MB - 1) / MB); // round up
 #else
         (void) n_gpu_layers;
-#endif
+#endif // defined(GGML_USE_CUBLAS) || defined(GGML_USE_CLBLAST)
     }
 
     // populate `tensors_by_name`
@@ -1256,17 +1286,11 @@ static bool llama_eval_internal(
            const float * embd,
              const int   n_tokens,
              const int   n_past,
-             const int   n_threads,
+                   int   n_threads,
             const char * cgraph_fname) {
 
     LLAMA_ASSERT((!tokens && embd) || (tokens && !embd));
 
-    // enforce that the first token is BOS
-    if (tokens && n_past == 0 && tokens[0] != llama_token_bos()) {
-        fprintf(stderr, "%s: first token must be BOS\n", __func__);
-        return false;
-    }
-
     const int64_t t_start_us = ggml_time_us();
 
     const int N = n_tokens;
@@ -1297,10 +1321,11 @@ static bool llama_eval_internal(
 
     struct ggml_context * ctx0 = ggml_init(params);
 
+    ggml_cgraph gf = {};
+
     // for big prompts, if BLAS is enabled, it is better to use only one thread
     // otherwise, the threads are spin-lock waiting for the BLAS calls and are degrading the performance
-    ggml_cgraph gf = {};
-    gf.n_threads = N >= 32 && ggml_cpu_has_blas() && !ggml_cpu_has_gpublas() ? 1 : n_threads;
+    n_threads = N >= 32 && ggml_cpu_has_blas() && !ggml_cpu_has_gpublas() ? 1 : n_threads;
 
     struct ggml_tensor * cur;
     struct ggml_tensor * inpL;
@@ -1584,6 +1609,7 @@ static bool llama_eval_internal(
 
 #ifdef GGML_USE_METAL
     if (lctx.ctx_metal && N == 1) {
+        ggml_metal_set_n_cb     (lctx.ctx_metal, n_threads);
         ggml_metal_graph_compute(lctx.ctx_metal, &gf);
         ggml_metal_get_tensor   (lctx.ctx_metal, cur);
     } else {
@@ -1603,10 +1629,10 @@ static bool llama_eval_internal(
             ggml_metal_get_tensor(lctx.ctx_metal, kv_self.v);
         }
 
-        ggml_graph_compute(ctx0, &gf);
+        ggml_graph_compute_helper(lctx.work_buffer, &gf, n_threads);
     }
 #else
-    ggml_graph_compute(ctx0, &gf);
+    ggml_graph_compute_helper(lctx.work_buffer, &gf, n_threads);
 #endif
 
     if (cgraph_fname) {
@@ -1896,10 +1922,10 @@ void llama_sample_top_p(struct llama_context * ctx, llama_token_data_array * can
         return;
     }
 
-    const int64_t t_start_sample_us = ggml_time_us();
-
     llama_sample_softmax(ctx, candidates);
 
+    const int64_t t_start_sample_us = ggml_time_us();
+
     // Compute the cumulative probabilities
     float cum_sum = 0.0f;
     size_t last_idx = candidates->size;
@@ -1928,9 +1954,8 @@ void llama_sample_tail_free(struct llama_context * ctx, llama_token_data_array *
         return;
     }
 
-    const int64_t t_start_sample_us = ggml_time_us();
-
     llama_sample_softmax(nullptr, candidates);
+    const int64_t t_start_sample_us = ggml_time_us();
 
     // Compute the first and second derivatives
     std::vector<float> first_derivatives(candidates->size - 1);
@@ -1982,11 +2007,11 @@ void llama_sample_typical(struct llama_context * ctx, llama_token_data_array * c
         return;
     }
 
-    const int64_t t_start_sample_us = ggml_time_us();
-
     // Compute the softmax of logits and calculate entropy
     llama_sample_softmax(nullptr, candidates);
 
+    const int64_t t_start_sample_us = ggml_time_us();
+
     float entropy = 0.0f;
     for (size_t i = 0; i < candidates->size; ++i) {
         entropy += -candidates->data[i].p * logf(candidates->data[i].p);
@@ -2155,13 +2180,11 @@ llama_token llama_sample_token_mirostat(struct llama_context * ctx, llama_token_
 
     if (ctx) {
         ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
-        ctx->n_sample++;
     }
     return X;
 }
 
 llama_token llama_sample_token_mirostat_v2(struct llama_context * ctx, llama_token_data_array * candidates, float tau, float eta, float * mu) {
-    assert(ctx);
     int64_t t_start_sample_us;
     t_start_sample_us = ggml_time_us();
 
@@ -2176,13 +2199,14 @@ llama_token llama_sample_token_mirostat_v2(struct llama_context * ctx, llama_tok
         candidates->size = 1;
     }
 
+    if (ctx) {
+        ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
+    }
+
     // Normalize the probabilities of the remaining words
     llama_sample_softmax(ctx, candidates);
 
     // Sample the next word X from the remaining words
-    if (ctx) {
-        ctx->t_sample_us += ggml_time_us() - t_start_sample_us;
-    }
     llama_token X = llama_sample_token(ctx, candidates);
     t_start_sample_us = ggml_time_us();
 
@@ -2250,10 +2274,10 @@ static void llama_convert_tensor_internal(const llama_load_tensor & tensor, llam
     }
     float * f32_output = (float *) output.addr;
 
-    quantize_fns_t qtype;
+    ggml_type_traits_t qtype;
     if (ggml_is_quantized(tensor.type)) {
-        qtype = ggml_internal_get_quantize_fn(tensor.type);
-        if (qtype.dequantize_row_q == NULL) {
+        qtype = ggml_internal_get_type_traits(tensor.type);
+        if (qtype.to_float == NULL) {
             throw std::runtime_error(format("type %s unsupported for integer quantization: no dequantization available", ggml_type_name(tensor.type)));
         }
     } else if (tensor.type != GGML_TYPE_F16) {
@@ -2264,7 +2288,7 @@ static void llama_convert_tensor_internal(const llama_load_tensor & tensor, llam
         if (tensor.type == GGML_TYPE_F16) {
             ggml_fp16_to_fp32_row((ggml_fp16_t *)tensor.data, f32_output, nelements);
         } else if (ggml_is_quantized(tensor.type)) {
-            qtype.dequantize_row_q(tensor.data, f32_output, nelements);
+            qtype.to_float(tensor.data, f32_output, nelements);
         } else {
             LLAMA_ASSERT(false); // unreachable
         }
@@ -2289,7 +2313,7 @@ static void llama_convert_tensor_internal(const llama_load_tensor & tensor, llam
             if (typ == GGML_TYPE_F16) {
                 ggml_fp16_to_fp32_row((ggml_fp16_t *)inbuf, outbuf, nels);
             } else {
-                qtype.dequantize_row_q(inbuf, outbuf, nels);
+                qtype.to_float(inbuf, outbuf, nels);
             }
         };
         workers.push_back(std::thread(compute, tensor.type, tensor.data + in_buff_offs, f32_output + out_buff_offs, thr_elems));
@@ -2568,8 +2592,8 @@ void llama_free_model(struct llama_model * model) {
 }
 
 struct llama_context * llama_new_context_with_model(
-                             struct llama_model * model,
-            struct llama_context_params   params) {
+                 struct llama_model * model,
+        struct llama_context_params   params) {
 
     if (!model) {
         return nullptr;
@@ -2638,7 +2662,7 @@ struct llama_context * llama_new_context_with_model(
 #ifdef GGML_USE_METAL
     if (params.n_gpu_layers > 0) {
         // this allocates all Metal resources and memory buffers
-        ctx->ctx_metal = ggml_metal_init();
+        ctx->ctx_metal = ggml_metal_init(1);
 
         void * data_ptr  = NULL;
         size_t data_size = 0;
@@ -2795,6 +2819,9 @@ int llama_apply_lora_from_file_internal(const struct llama_model & model, const
     // read tensors and apply
     bool warned = false;
     int n_tensors = 0;
+
+    std::vector<uint8_t> work_buffer;
+
     while (true) {
         int32_t n_dims;
         int32_t length;
@@ -2959,8 +2986,8 @@ int llama_apply_lora_from_file_internal(const struct llama_model & model, const
             }
 
             struct ggml_cgraph gf = ggml_build_forward(r);
-            gf.n_threads = n_threads;
-            ggml_graph_compute(lora_ctx, &gf);
+
+            ggml_graph_compute_helper(work_buffer, &gf, n_threads);
 
             // we won't need these tensors again, reset the context to save memory
             ggml_free(lora_ctx);
@@ -3113,7 +3140,6 @@ size_t llama_copy_state_data(struct llama_context * ctx, uint8_t * dst) {
 
             ggml_context * cpy_ctx = ggml_init({ 4096, NULL, /* no_alloc */ true });
             ggml_cgraph gf{};
-            gf.n_threads = 1;
 
             ggml_tensor * kout3d = ggml_new_tensor_3d(cpy_ctx, kv_self.k->type, n_embd, kv_ntok, n_layer);
             kout3d->data = out;
@@ -3133,7 +3159,7 @@ size_t llama_copy_state_data(struct llama_context * ctx, uint8_t * dst) {
 
             ggml_build_forward_expand(&gf, ggml_cpy(cpy_ctx, k3d, kout3d));
             ggml_build_forward_expand(&gf, ggml_cpy(cpy_ctx, v3d, vout3d));
-            ggml_graph_compute(cpy_ctx, &gf);
+            ggml_graph_compute_helper(ctx->work_buffer, &gf, /*n_threads*/ 1);
 
             ggml_free(cpy_ctx);
         }
@@ -3219,7 +3245,6 @@ size_t llama_set_state_data(struct llama_context * ctx, uint8_t * src) {
 
             ggml_context * cpy_ctx = ggml_init({ 4096, NULL, /* no_alloc */ true });
             ggml_cgraph gf{};
-            gf.n_threads = 1;
 
             ggml_tensor * kin3d = ggml_new_tensor_3d(cpy_ctx, kv_self.k->type, n_embd, kv_ntok, n_layer);
             kin3d->data = (void *) inp;
@@ -3239,7 +3264,7 @@ size_t llama_set_state_data(struct llama_context * ctx, uint8_t * src) {
 
             ggml_build_forward_expand(&gf, ggml_cpy(cpy_ctx, kin3d, k3d));
             ggml_build_forward_expand(&gf, ggml_cpy(cpy_ctx, vin3d, v3d));
-            ggml_graph_compute(cpy_ctx, &gf);
+            ggml_graph_compute_helper(ctx->work_buffer, &gf, /*n_threads*/ 1);
 
             ggml_free(cpy_ctx);
         }
@@ -3473,23 +3498,35 @@ llama_token llama_token_nl() {
     return 13;
 }
 
+struct llama_timings llama_get_timings(struct llama_context * ctx) {
+    struct llama_timings result = {
+        /*.t_start_ms  =*/ 1e-3 * ctx->t_start_us,
+        /*.t_end_ms    =*/ 1.00 * ggml_time_ms(),
+        /*.t_load_ms   =*/ 1e-3 * ctx->t_load_us,
+        /*.t_sample_ms =*/ 1e-3 * ctx->t_sample_us,
+        /*.t_p_eval_ms =*/ 1e-3 * ctx->t_p_eval_us,
+        /*.t_eval_ms   =*/ 1e-3 * ctx->t_eval_us,
 
-void llama_print_timings(struct llama_context * ctx) {
-    const int64_t t_end_us = ggml_time_us();
+        /*.n_sample =*/ std::max(1, ctx->n_sample),
+        /*.n_p_eval =*/ std::max(1, ctx->n_p_eval),
+        /*.n_eval   =*/ std::max(1, ctx->n_eval),
+    };
 
-    const int32_t n_sample = std::max(1, ctx->n_sample);
-    const int32_t n_eval   = std::max(1, ctx->n_eval);
-    const int32_t n_p_eval = std::max(1, ctx->n_p_eval);
+    return result;
+}
+
+void llama_print_timings(struct llama_context * ctx) {
+    const llama_timings timings = llama_get_timings(ctx);
 
     fprintf(stderr, "\n");
-    fprintf(stderr, "%s:        load time = %8.2f ms\n", __func__, ctx->t_load_us / 1000.0);
+    fprintf(stderr, "%s:        load time = %8.2f ms\n", __func__, timings.t_load_ms);
     fprintf(stderr, "%s:      sample time = %8.2f ms / %5d runs   (%8.2f ms per token, %8.2f tokens per second)\n",
-            __func__, 1e-3 * ctx->t_sample_us, n_sample, 1e-3 * ctx->t_sample_us / n_sample, 1e6 / ctx->t_sample_us * n_sample);
+            __func__, timings.t_sample_ms, timings.n_sample, timings.t_sample_ms / timings.n_sample, 1e3 / timings.t_sample_ms * timings.n_sample);
     fprintf(stderr, "%s: prompt eval time = %8.2f ms / %5d tokens (%8.2f ms per token, %8.2f tokens per second)\n",
-            __func__, 1e-3 * ctx->t_p_eval_us, n_p_eval, 1e-3 * ctx->t_p_eval_us / n_p_eval, 1e6 / ctx->t_p_eval_us * n_p_eval);
+            __func__, timings.t_p_eval_ms, timings.n_p_eval, timings.t_p_eval_ms / timings.n_p_eval, 1e3 / timings.t_p_eval_ms * timings.n_p_eval);
     fprintf(stderr, "%s:        eval time = %8.2f ms / %5d runs   (%8.2f ms per token, %8.2f tokens per second)\n",
-            __func__, 1e-3 * ctx->t_eval_us,   n_eval,   1e-3 * ctx->t_eval_us   / n_eval,   1e6 / ctx->t_eval_us   * n_eval);
-    fprintf(stderr, "%s:       total time = %8.2f ms\n", __func__, (t_end_us - ctx->t_start_us)/1000.0);
+            __func__, timings.t_eval_ms, timings.n_eval, timings.t_eval_ms / timings.n_eval, 1e3 / timings.t_eval_ms * timings.n_eval);
+    fprintf(stderr, "%s:       total time = %8.2f ms\n", __func__, (timings.t_end_ms - timings.t_start_ms));
 }
 
 void llama_reset_timings(struct llama_context * ctx) {
diff --git a/llama.h b/llama.h
index 5bb1964bd390d..c1e7dab9f5a9b 100644
--- a/llama.h
+++ b/llama.h
@@ -134,6 +134,20 @@ extern "C" {
         bool quantize_output_tensor; // quantize output.weight
     } llama_model_quantize_params;
 
+    // performance timing information
+    struct llama_timings {
+        double t_start_ms;
+        double t_end_ms;
+        double t_load_ms;
+        double t_sample_ms;
+        double t_p_eval_ms;
+        double t_eval_ms;
+
+        int32_t n_sample;
+        int32_t n_p_eval;
+        int32_t n_eval;
+    };
+
     LLAMA_API struct llama_context_params llama_context_default_params();
     LLAMA_API struct llama_model_quantize_params llama_model_quantize_default_params();
 
@@ -331,6 +345,7 @@ extern "C" {
     LLAMA_API llama_token llama_sample_token(struct llama_context * ctx, llama_token_data_array * candidates);
 
     // Performance information
+    LLAMA_API struct llama_timings llama_get_timings(struct llama_context * ctx);
     LLAMA_API void llama_print_timings(struct llama_context * ctx);
     LLAMA_API void llama_reset_timings(struct llama_context * ctx);
 
diff --git a/pocs/vdot/q8dot.cpp b/pocs/vdot/q8dot.cpp
index 5748c8ac22193..4e0e023575322 100644
--- a/pocs/vdot/q8dot.cpp
+++ b/pocs/vdot/q8dot.cpp
@@ -136,7 +136,7 @@ int main(int argc, char** argv) {
 
     auto ggml_type = type == 0 ? GGML_TYPE_Q4_0 : GGML_TYPE_Q4_1;
 
-    auto funcs = ggml_internal_get_quantize_fn(ggml_type);
+    auto funcs = ggml_internal_get_type_traits(ggml_type);
 
     Stat simple, ggml;
 
@@ -156,8 +156,8 @@ int main(int argc, char** argv) {
 
         t1 = std::chrono::high_resolution_clock::now();
         float fs;
-        if (type == 0) funcs.vec_dot_q(kVecSize * QK4_1, &fs, x40.data(), y.data());
-        else funcs.vec_dot_q(kVecSize * QK4_1, &fs, x41.data(), y.data());
+        if (type == 0) funcs.vec_dot(kVecSize * QK4_1, &fs, x40.data(), y.data());
+        else funcs.vec_dot(kVecSize * QK4_1, &fs, x41.data(), y.data());
         t2 = std::chrono::high_resolution_clock::now();
         t = 1e-3*std::chrono::duration_cast<std::chrono::nanoseconds>(t2-t1).count();
         if (iloop > 3) ggml.addResult(fs, t);
diff --git a/pocs/vdot/vdot.cpp b/pocs/vdot/vdot.cpp
index 7b18090d66b6b..48758cda81fdf 100644
--- a/pocs/vdot/vdot.cpp
+++ b/pocs/vdot/vdot.cpp
@@ -235,7 +235,7 @@ int main(int argc, char** argv) {
     int n4 = useQ4_1 ? kVecSize / QK4_1 : kVecSize / QK4_0; n4 = 64*((n4 + 63)/64);
     int n8 = kVecSize / QK8_0; n8 = 64*((n8 + 63)/64);
 
-    auto funcs = useQ4_1 ? ggml_internal_get_quantize_fn(GGML_TYPE_Q4_1) : ggml_internal_get_quantize_fn(GGML_TYPE_Q4_0);
+    auto funcs = useQ4_1 ? ggml_internal_get_type_traits(GGML_TYPE_Q4_1) : ggml_internal_get_type_traits(GGML_TYPE_Q4_0);
 
     std::vector<block_q4_0> q40;
     std::vector<block_q4_1> q41;
@@ -261,9 +261,9 @@ int main(int argc, char** argv) {
         // Note, we do not include this in the timing as in practical application
         // we already have the quantized model weights.
         if (useQ4_1) {
-            funcs.quantize_row_q(x1.data(), q41.data(), kVecSize);
+            funcs.from_float(x1.data(), q41.data(), kVecSize);
         } else {
-            funcs.quantize_row_q(x1.data(), q40.data(), kVecSize);
+            funcs.from_float(x1.data(), q40.data(), kVecSize);
         }
 
         // Now measure time the dot product needs using the "scalar" version above
@@ -282,9 +282,10 @@ int main(int argc, char** argv) {
             dot_q4_q8(kVecSize, &result, q40.data(), q8.data());
         }
         else {
-            funcs.quantize_row_q_dot(y1.data(), q8.data(), kVecSize);
-            if (useQ4_1) funcs.vec_dot_q(kVecSize, &result, q41.data(), q8.data());
-            else funcs.vec_dot_q(kVecSize, &result, q40.data(), q8.data());
+            auto vdot = ggml_internal_get_type_traits(funcs.vec_dot_type);
+            vdot.from_float(y1.data(), q8.data(), kVecSize);
+            if (useQ4_1) funcs.vec_dot(kVecSize, &result, q41.data(), q8.data());
+            else funcs.vec_dot(kVecSize, &result, q40.data(), q8.data());
         }
         sumq += result;
         t2 = std::chrono::high_resolution_clock::now();
diff --git a/scripts/sync-ggml.sh b/scripts/sync-ggml.sh
index e6e39ff8f1dd9..02ea6ec15241b 100755
--- a/scripts/sync-ggml.sh
+++ b/scripts/sync-ggml.sh
@@ -1,6 +1,14 @@
 #!/bin/bash
 
-cp -rpv ../ggml/src/ggml.c          ./ggml.c
-cp -rpv ../ggml/src/ggml-cuda.cu    ./ggml-cuda.cu
-cp -rpv ../ggml/src/ggml-cuda.h     ./ggml-cuda.h
-cp -rpv ../ggml/include/ggml/ggml.h ./ggml.h
+cp -rpv ../ggml/src/ggml.c           ./ggml.c
+cp -rpv ../ggml/src/ggml-cuda.h      ./ggml-cuda.h
+cp -rpv ../ggml/src/ggml-cuda.cu     ./ggml-cuda.cu
+cp -rpv ../ggml/src/ggml-opencl.h    ./ggml-opencl.h
+cp -rpv ../ggml/src/ggml-opencl.cpp  ./ggml-opencl.cpp
+cp -rpv ../ggml/src/ggml-metal.h     ./ggml-metal.h
+cp -rpv ../ggml/src/ggml-metal.m     ./ggml-metal.m
+cp -rpv ../ggml/src/ggml-metal.metal ./ggml-metal.metal
+cp -rpv ../ggml/include/ggml/ggml.h  ./ggml.h
+
+cp -rpv ../ggml/tests/test-opt.c    ./tests/test-opt.c
+cp -rpv ../ggml/tests/test-grad0.c  ./tests/test-grad0.c
diff --git a/tests/CMakeLists.txt b/tests/CMakeLists.txt
index 4171c126c7b7d..1acf050a743e4 100644
--- a/tests/CMakeLists.txt
+++ b/tests/CMakeLists.txt
@@ -10,5 +10,5 @@ llama_add_test(test-quantize-fns.cpp)
 llama_add_test(test-quantize-perf.cpp)
 llama_add_test(test-sampling.cpp)
 llama_add_test(test-tokenizer-0.cpp ${CMAKE_CURRENT_SOURCE_DIR}/../models/ggml-vocab.bin)
-# llama_add_test(test-grad0.c) # SLOW
+llama_add_test(test-grad0.c) # SLOW
 # llama_add_test(test-opt.c) # SLOW
diff --git a/tests/test-grad0.c b/tests/test-grad0.c
index b5a499c1db57e..da4001ce5269f 100644
--- a/tests/test-grad0.c
+++ b/tests/test-grad0.c
@@ -10,6 +10,8 @@
 #pragma warning(disable: 4244 4267) // possible loss of data
 #endif
 
+#pragma GCC diagnostic ignored "-Wdouble-promotion"
+
 #define MAX_NARGS 3
 
 #undef MIN
@@ -49,7 +51,7 @@ float frand(void) {
 
 int irand(int n) {
     if (n == 0) return 0;
-    else return rand()%n;
+    return rand()%n;
 }
 
 void get_random_dims(int64_t * dims, int ndims) {
@@ -159,12 +161,14 @@ struct ggml_tensor * get_random_tensor_int(
 float get_element(const struct ggml_tensor * t, int idx) {
     if (t->type == GGML_TYPE_F32) {
         return ((float *)t->data)[idx];
-    } else if (t->type == GGML_TYPE_I32) {
+    }
+
+    if (t->type == GGML_TYPE_I32) {
         return ((int32_t *)t->data)[idx];
-    } else {
-        assert(false);
-        return INFINITY;
     }
+
+    assert(false);
+    return INFINITY;
 }
 
 void set_element(struct ggml_tensor * t, int idx, float value) {
@@ -215,15 +219,14 @@ bool check_gradient(
     }
 
     struct ggml_cgraph gf = ggml_build_forward (f);
-    gf.n_threads = n_threads;
-
     struct ggml_cgraph gb = ggml_build_backward(ctx0, &gf, false);
-    gb.n_threads = n_threads;
 
-    ggml_graph_compute(ctx0, &gf);
+    ggml_graph_compute_with_ctx(ctx0, &gf, n_threads);
+
     ggml_graph_reset  (&gf);
     ggml_set_f32      (f->grad, 1.0f);
-    ggml_graph_compute(ctx0, &gb);
+
+    ggml_graph_compute_with_ctx(ctx0, &gb, n_threads);
 
     // ggml_graph_dump_dot(&gf, NULL, "test-grad0-forward.dot");
     // ggml_graph_dump_dot(&gb, &gf,  "test-grad0-backward.dot");
@@ -236,15 +239,16 @@ bool check_gradient(
             const float xm = x0 - eps;
             const float xp = x0 + eps;
             set_element(x[i], k, xp);
-            ggml_graph_compute(ctx0, &gf);
+
+            ggml_graph_compute_with_ctx(ctx0, &gf, n_threads);
 
             const float f0 = ggml_get_f32_1d(f, 0);
 
             set_element(x[i], k, xm);
-            ggml_graph_compute(ctx0, &gf);
 
-            const float f1 = ggml_get_f32_1d(f, 0);
+            ggml_graph_compute_with_ctx(ctx0, &gf, n_threads);
 
+            const float f1 = ggml_get_f32_1d(f, 0);
             const float g0 = (f0 - f1)/(2.0f*eps);
 
             set_element(x[i], k, x0);
@@ -252,12 +256,13 @@ bool check_gradient(
             // compute gradient using backward graph
             ggml_graph_reset  (&gf);
             ggml_set_f32      (f->grad, 1.0f);
-            ggml_graph_compute(ctx0, &gb);
+
+            ggml_graph_compute_with_ctx(ctx0, &gb, n_threads);
 
             const float g1 = get_element(x[i]->grad, k);
 
             const float error_abs = fabsf(g0 - g1);
-            const float error_rel = g0 != 0 ? fabsf(g0 - g1)/fabs(g0) : 0;
+            const float error_rel = g0 != 0 ? fabsf(g0 - g1)/fabsf(g0) : 0;
 
             if (error_abs > max_error_abs || error_rel > max_error_rel) {
                 printf("%s: ndims=%d, i=%d, k=%d, x0=%f, xm=%f, xp=%f, f0=%f, f1=%f, g0=%f, g1=%f, eps=%f, error_abs=%f, error_rel=%f\n",
@@ -1154,7 +1159,7 @@ int main(int argc, const char ** argv) {
                             continue;
                         }
 
-                        struct ggml_tensor * f = ggml_sum(ctx0, ggml_rope(ctx0, x[0], n_past, n_rot, mode));
+                        struct ggml_tensor * f = ggml_sum(ctx0, ggml_rope(ctx0, x[0], n_past, n_rot, mode, 0));
 
                         GGML_PRINT_DEBUG("rope: n_past: %d n_rot: %d mode: %d\n", n_past, n_rot, mode);
                         check_gradient("rope", ctx0, x, f, ndims, nargs, 1e-2f, 1e-3f, INFINITY);
diff --git a/tests/test-opt.c b/tests/test-opt.c
index d001615ee353b..e928a7df7ee68 100644
--- a/tests/test-opt.c
+++ b/tests/test-opt.c
@@ -7,6 +7,7 @@
 
 #define MAX_NARGS 2
 
+#pragma GCC diagnostic ignored "-Wdouble-promotion"
 
 //
 // logging
@@ -33,7 +34,7 @@
 #define GGML_PRINT(...) printf(__VA_ARGS__)
 
 
-float frand() {
+float frand(void) {
     return (float)rand()/(float)RAND_MAX;
 }
 
@@ -114,7 +115,7 @@ void set_element(struct ggml_tensor * t, int idx, float value) {
     ((float *)t->data)[idx] = value;
 }
 
-int main(int argc, const char ** argv) {
+int main(void) {
     struct ggml_init_params params = {
         .mem_size   = 1024*1024*1024,
         .mem_buffer = NULL,
@@ -137,10 +138,11 @@ int main(int argc, const char ** argv) {
     struct ggml_tensor * d  = ggml_sub(ctx, c, ab);
     struct ggml_tensor * e  = ggml_sum(ctx, ggml_sqr(ctx, d));
 
-
     struct ggml_cgraph ge = ggml_build_forward(e);
-    ggml_graph_reset  (&ge);
-    ggml_graph_compute(ctx, &ge);
+    ggml_graph_reset(&ge);
+
+    ggml_graph_compute_with_ctx(ctx, &ge, /*n_threads*/ 1);
+
     const float fe = ggml_get_f32_1d(e, 0);
     printf("%s: e = %.4f\n", __func__, fe);
 
@@ -148,8 +150,10 @@ int main(int argc, const char ** argv) {
 
     ggml_opt(ctx, opt_params, e);
 
-    ggml_graph_reset  (&ge);
-    ggml_graph_compute(ctx, &ge);
+    ggml_graph_reset(&ge);
+
+    ggml_graph_compute_with_ctx(ctx, &ge, /*n_threads*/ 1);
+
     const float fe_opt = ggml_get_f32_1d(e, 0);
     printf("%s: original  e = %.4f\n", __func__, fe);
     printf("%s: optimized e = %.4f\n", __func__, fe_opt);
diff --git a/tests/test-quantize-fns.cpp b/tests/test-quantize-fns.cpp
index c40f1b29c7c36..8d3c162d2bfa0 100644
--- a/tests/test-quantize-fns.cpp
+++ b/tests/test-quantize-fns.cpp
@@ -40,26 +40,26 @@ float array_rmse(const float * a1, const float * a2, size_t n) {
 }
 
 // Total quantization error on test data
-float total_quantization_error(quantize_fns_t & qfns, size_t test_size, const float * test_data) {
+float total_quantization_error(ggml_type_traits_t & qfns, size_t test_size, const float * test_data) {
     std::vector<uint8_t> tmp_q(2*test_size);
     std::vector<float> tmp_out(test_size);
 
-    qfns.quantize_row_q(test_data, tmp_q.data(), test_size);
-    qfns.dequantize_row_q(tmp_q.data(), tmp_out.data(), test_size);
+    qfns.from_float(test_data, tmp_q.data(), test_size);
+    qfns.to_float(tmp_q.data(), tmp_out.data(), test_size);
     return array_rmse(test_data, tmp_out.data(), test_size);
 }
 
 // Total quantization error on test data
-float reference_quantization_error(quantize_fns_t & qfns, size_t test_size, const float * test_data) {
+float reference_quantization_error(ggml_type_traits_t & qfns, size_t test_size, const float * test_data) {
     std::vector<uint8_t> tmp_q(2*test_size);
     std::vector<float> tmp_out(test_size);
     std::vector<float> tmp_out_ref(test_size);
 
-    qfns.quantize_row_q(test_data, tmp_q.data(), test_size);
-    qfns.dequantize_row_q(tmp_q.data(), tmp_out.data(), test_size);
+    qfns.from_float(test_data, tmp_q.data(), test_size);
+    qfns.to_float(tmp_q.data(), tmp_out.data(), test_size);
 
-    qfns.quantize_row_q_reference(test_data, tmp_q.data(), test_size);
-    qfns.dequantize_row_q(tmp_q.data(), tmp_out_ref.data(), test_size);
+    qfns.from_float_reference(test_data, tmp_q.data(), test_size);
+    qfns.to_float(tmp_q.data(), tmp_out_ref.data(), test_size);
 
     return array_rmse(tmp_out.data(), tmp_out_ref.data(), test_size);
 }
@@ -73,15 +73,17 @@ float dot_product(const float * a1, const float * a2, size_t test_size) {
 }
 
 // Total dot product error
-float dot_product_error(quantize_fns_t & qfns, size_t test_size, const float * test_data1, const float *test_data2) {
+float dot_product_error(ggml_type_traits_t & qfns, size_t test_size, const float * test_data1, const float *test_data2) {
     std::vector<uint8_t> tmp_q1(2*test_size);
     std::vector<uint8_t> tmp_q2(2*test_size);
 
-    qfns.quantize_row_q    (test_data1, tmp_q1.data(), test_size);
-    qfns.quantize_row_q_dot(test_data2, tmp_q2.data(), test_size);
+    auto vdot = ggml_internal_get_type_traits(qfns.vec_dot_type);
+
+    qfns.from_float(test_data1, tmp_q1.data(), test_size);
+    vdot.from_float(test_data2, tmp_q2.data(), test_size);
 
     float result = INFINITY;
-    qfns.vec_dot_q(test_size, &result, tmp_q1.data(), tmp_q2.data());
+    qfns.vec_dot(test_size, &result, tmp_q1.data(), tmp_q2.data());
 
     const float dot_ref = dot_product(test_data1, test_data2, test_size);
 
@@ -123,9 +125,9 @@ int main(int argc, char * argv[]) {
 
     for (int i = 0; i < GGML_TYPE_COUNT; i++) {
         ggml_type type = (ggml_type) i;
-        quantize_fns_t qfns = ggml_internal_get_quantize_fn(i);
+        ggml_type_traits_t qfns = ggml_internal_get_type_traits(type);
 
-        if (qfns.quantize_row_q && qfns.dequantize_row_q) {
+        if (qfns.from_float && qfns.to_float) {
             const float total_error = total_quantization_error(qfns, test_size, test_data.data());
             const float max_quantization_error =
                 type == GGML_TYPE_Q2_K ? MAX_QUANTIZATION_TOTAL_ERROR_2BITS :
diff --git a/tests/test-quantize-perf.cpp b/tests/test-quantize-perf.cpp
index c0e361e92313f..0bb9537f693ed 100644
--- a/tests/test-quantize-perf.cpp
+++ b/tests/test-quantize-perf.cpp
@@ -123,9 +123,9 @@ void usage(char * argv[]) {
     printf("  --type TYPE           set test type as");
     for (int i = 0; i < GGML_TYPE_COUNT; i++) {
         ggml_type type = (ggml_type) i;
-        quantize_fns_t qfns = ggml_internal_get_quantize_fn(type);
+        ggml_type_traits_t qfns = ggml_internal_get_type_traits(type);
         if (ggml_type_name(type) != NULL) {
-            if (qfns.quantize_row_q && qfns.dequantize_row_q) {
+            if (qfns.from_float && qfns.to_float) {
                 printf(" %s", ggml_type_name(type));
             }
         }
@@ -271,12 +271,12 @@ int main(int argc, char * argv[]) {
 
     for (int i = 0; i < GGML_TYPE_COUNT; i++) {
         ggml_type type = (ggml_type) i;
-        quantize_fns_t qfns = ggml_internal_get_quantize_fn(i);
+        ggml_type_traits_t qfns = ggml_internal_get_type_traits(type);
         if (!params.include_types.empty() && ggml_type_name(type) && std::find(params.include_types.begin(), params.include_types.end(), ggml_type_name(type)) == params.include_types.end()) {
             continue;
         }
 
-        if (qfns.quantize_row_q && qfns.dequantize_row_q) {
+        if (qfns.from_float && qfns.to_float) {
             printf("%s\n", ggml_type_name(type));
 
             if (params.op_quantize_row_q_reference) {
@@ -284,7 +284,7 @@ int main(int argc, char * argv[]) {
                 for (size_t size : params.test_sizes) {
                     printf("    %zu values (%.2f MB)\n", size, 4*size/(float)(1024*1024));
                     auto quantize_fn = [&](void ) {
-                        qfns.quantize_row_q_reference(test_data1, test_q1, size);
+                        qfns.from_float_reference(test_data1, test_q1, size);
                         return test_q1[0];
                     };
                     size_t quantized_size = size / ggml_blck_size(type) * ggml_type_size(type);
@@ -298,7 +298,7 @@ int main(int argc, char * argv[]) {
                 for (size_t size : params.test_sizes) {
                     printf("    %zu values (%.2f MB)\n", size, 4*size/(float)(1024*1024));
                     auto quantize_fn = [&](void ) {
-                        qfns.quantize_row_q(test_data1, test_q1, size);
+                        qfns.from_float(test_data1, test_q1, size);
                         return test_q1[0];
                     };
                     size_t quantized_size = size / ggml_blck_size(type) * ggml_type_size(type);
@@ -309,11 +309,11 @@ int main(int argc, char * argv[]) {
 
             if (params.op_dequantize_row_q) {
                 printf("  dequantize_row_q\n");
-                qfns.quantize_row_q(test_data1, test_q1, largest);
+                qfns.from_float(test_data1, test_q1, largest);
                 for (size_t size : params.test_sizes) {
                     printf("    %zu values (%.2f MB)\n", size, 4*size/(float)(1024*1024));
                     auto quantize_fn = [&](void ) {
-                        qfns.dequantize_row_q(test_q1, test_out, size);
+                        qfns.to_float(test_q1, test_out, size);
                         return test_out[0];
                     };
                     size_t quantized_size = size / ggml_blck_size(type) * ggml_type_size(type);
@@ -327,7 +327,8 @@ int main(int argc, char * argv[]) {
                 for (size_t size : params.test_sizes) {
                     printf("    %zu values (%.2f MB)\n", size, 4*size/(float)(1024*1024));
                     auto quantize_fn = [&](void ) {
-                        qfns.quantize_row_q_dot(test_data1, test_q1, size);
+                        auto vdot = ggml_internal_get_type_traits(qfns.vec_dot_type);
+                        vdot.from_float(test_data1, test_q1, size);
                         return test_q1[0];
                     };
                     size_t quantized_size = size / ggml_blck_size(type) * ggml_type_size(type);
@@ -338,13 +339,13 @@ int main(int argc, char * argv[]) {
 
             if (params.op_vec_dot_q) {
                 printf("  vec_dot_q\n");
-                qfns.quantize_row_q(test_data1, test_q1, largest);
-                qfns.quantize_row_q(test_data2, test_q2, largest);
+                qfns.from_float(test_data1, test_q1, largest);
+                qfns.from_float(test_data2, test_q2, largest);
                 for (size_t size : params.test_sizes) {
                     printf("    %zu values (%.2f MB)\n", size, 4*size/(float)(1024*1024));
                     auto quantize_fn = [&](void ) {
                         float result;
-                        qfns.vec_dot_q(size, &result, test_q1, test_q2);
+                        qfns.vec_dot(size, &result, test_q1, test_q2);
                         return result;
                     };
                     size_t quantized_size = size / ggml_blck_size(type) * ggml_type_size(type);