Merge 'origin/master' into hipblas

.gitignore

@@ -32,6 +32,7 @@ models/*
 /vdot
 /Pipfile
 
+build-info.h
 arm_neon.h
 compile_commands.json
 

CMakeLists.txt

@@ -73,6 +73,41 @@ option(LLAMA_HIPBLAS                "llama: use hipBLAS"
 option(LLAMA_BUILD_TESTS            "llama: build tests"    ${LLAMA_STANDALONE})
 option(LLAMA_BUILD_EXAMPLES         "llama: build examples" ${LLAMA_STANDALONE})
 
+#
+# Build info header
+#
+
+# Write header template to binary dir to keep source directory clean
+file(WRITE "${CMAKE_BINARY_DIR}/BUILD_INFO.h.in" "\
+#ifndef BUILD_INFO_H\n\
+#define BUILD_INFO_H\n\
+\n\
+#define BUILD_NUMBER @BUILD_NUMBER@\n\
+#define BUILD_COMMIT \"@BUILD_COMMIT@\"\n\
+\n\
+#endif // BUILD_INFO_H\n\
+")
+
+# Generate initial build-info.h
+include(${CMAKE_CURRENT_SOURCE_DIR}/scripts/build-info.cmake)
+
+if(EXISTS "${CMAKE_CURRENT_SOURCE_DIR}/.git")
+    # Add a custom target for build-info.h
+    add_custom_target(BUILD_INFO ALL DEPENDS "${CMAKE_CURRENT_SOURCE_DIR}/build-info.h")
+
+    # Add a custom command to rebuild build-info.h when .git/index changes
+    add_custom_command(
+        OUTPUT "${CMAKE_CURRENT_SOURCE_DIR}/build-info.h"
+        COMMENT "Generating build details from Git"
+        COMMAND ${CMAKE_COMMAND} -P "${CMAKE_CURRENT_SOURCE_DIR}/scripts/build-info.cmake"
+        WORKING_DIRECTORY ${CMAKE_CURRENT_SOURCE_DIR}
+        DEPENDS "${CMAKE_CURRENT_SOURCE_DIR}/.git/index"
+        VERBATIM
+    )
+else()
+    message(WARNING "Git repository not found; to enable automatic generation of build info, make sure Git is installed and the project is a Git repository.")
+endif()
+
 #
 # Compile flags
 #
@@ -288,9 +323,22 @@ if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm" OR ${CMAKE_SYSTEM_PROCESSOR} MATCHES
         # TODO: arm msvc?
     else()
         if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64")
+            # Apple M1, M2, etc.
+            # Raspberry Pi 3, 4, Zero 2 (64-bit)
             add_compile_options(-mcpu=native)
         endif()
-        # TODO: armv6,7,8 version specific flags
+        if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "armv6")
+            # Raspberry Pi 1, Zero
+            add_compile_options(-mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access)
+        endif()
+        if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "armv7")
+            # Raspberry Pi 2
+            add_compile_options(-mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations)
+        endif()
+        if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "armv8")
+            # Raspberry Pi 3, 4, Zero 2 (32-bit)
+            add_compile_options(-mfp16-format=ieee -mno-unaligned-access)
+        endif()
     endif()
 elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^(x86_64|i686|AMD64)$")
     message(STATUS "x86 detected")

Makefile

@@ -148,19 +148,21 @@ ifdef LLAMA_PERF
 	CXXFLAGS += -DGGML_PERF
 endif
 ifneq ($(filter aarch64%,$(UNAME_M)),)
+	# Apple M1, M2, etc.
+	# Raspberry Pi 3, 4, Zero 2 (64-bit)
 	CFLAGS   += -mcpu=native
 	CXXFLAGS += -mcpu=native
 endif
 ifneq ($(filter armv6%,$(UNAME_M)),)
-	# Raspberry Pi 1, 2, 3
+	# Raspberry Pi 1, Zero
 	CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access
 endif
 ifneq ($(filter armv7%,$(UNAME_M)),)
-	# Raspberry Pi 4
+	# Raspberry Pi 2
 	CFLAGS += -mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations
 endif
 ifneq ($(filter armv8%,$(UNAME_M)),)
-	# Raspberry Pi 4
+	# Raspberry Pi 3, 4, Zero 2 (32-bit)
 	CFLAGS += -mfp16-format=ieee -mno-unaligned-access
 endif
 
@@ -192,41 +194,56 @@ llama.o: llama.cpp ggml.h ggml-cuda.h llama.h llama-util.h
 common.o: examples/common.cpp examples/common.h
 	$(CXX) $(CXXFLAGS) -c $< -o $@
 
+libllama.so: llama.o ggml.o $(OBJS)
+	$(CXX) $(CXXFLAGS) -shared -fPIC -o $@ $^ $(LDFLAGS)
+
 clean:
-	rm -vf *.o main quantize quantize-stats perplexity embedding benchmark-matmult
+	rm -vf *.o main quantize quantize-stats perplexity embedding benchmark-matmult save-load-state build-info.h
 
-main: examples/main/main.cpp ggml.o llama.o common.o $(OBJS)
-	$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
+#
+# Examples
+#
+
+main: examples/main/main.cpp build-info.h ggml.o llama.o common.o $(OBJS)
+	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
 	@echo
 	@echo '====  Run ./main -h for help.  ===='
 	@echo
 
-quantize: examples/quantize/quantize.cpp ggml.o llama.o $(OBJS)
-	$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
+quantize: examples/quantize/quantize.cpp build-info.h ggml.o llama.o $(OBJS)
+	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
 
-quantize-stats: examples/quantize-stats/quantize-stats.cpp ggml.o llama.o $(OBJS)
-	$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
+quantize-stats: examples/quantize-stats/quantize-stats.cpp build-info.h ggml.o llama.o $(OBJS)
+	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
 
-perplexity: examples/perplexity/perplexity.cpp ggml.o llama.o common.o $(OBJS)
-	$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
+perplexity: examples/perplexity/perplexity.cpp build-info.h ggml.o llama.o common.o $(OBJS)
+	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
 
-embedding: examples/embedding/embedding.cpp ggml.o llama.o common.o $(OBJS)
-	$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
+embedding: examples/embedding/embedding.cpp build-info.h ggml.o llama.o common.o $(OBJS)
+	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
 
-vdot: pocs/vdot/vdot.cpp ggml.o $(OBJS)
-	$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
+save-load-state: examples/save-load-state/save-load-state.cpp build-info.h ggml.o llama.o common.o $(OBJS)
+	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
 
-libllama.so: llama.o ggml.o $(OBJS)
-	$(CXX) $(CXXFLAGS) -shared -fPIC -o $@ $^ $(LDFLAGS)
+build-info.h: $(wildcard .git/index) scripts/build-info.sh
+	@scripts/build-info.sh > $@.tmp
+	@if ! cmp -s $@.tmp $@; then \
+		mv $@.tmp $@; \
+	else \
+		rm $@.tmp; \
+	fi
 
 #
 # Tests
 #
 
-benchmark-matmult: examples/benchmark/benchmark-matmult.cpp ggml.o $(OBJS)
-	$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
+benchmark-matmult: examples/benchmark/benchmark-matmult.cpp build-info.h ggml.o $(OBJS)
+	$(CXX) $(CXXFLAGS) $(filter-out %.h,$^) -o $@ $(LDFLAGS)
 	./$@
 
+vdot: pocs/vdot/vdot.cpp ggml.o $(OBJS)
+	$(CXX) $(CXXFLAGS) $^ -o $@ $(LDFLAGS)
+
 .PHONY: tests
 tests:
 	bash ./tests/run-tests.sh

examples/benchmark/CMakeLists.txt

@@ -2,3 +2,6 @@ set(TARGET benchmark)
 add_executable(${TARGET} benchmark-matmult.cpp)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_11)
+if(TARGET BUILD_INFO)
+  add_dependencies(${TARGET} BUILD_INFO)
+endif()

examples/benchmark/benchmark-matmult.cpp

@@ -1,5 +1,6 @@
 #include <locale.h>
 #include "ggml.h"
+#include "build-info.h"
 #include <assert.h>
 #include <math.h>
 #include <cstring>
@@ -90,9 +91,10 @@ int main(int argc, char ** argv)  {
         }
     }
 
-    // create the ggml context
+    fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT);
     printf("Starting Test\n");
 
+    // create the ggml context
     struct ggml_context * ctx;
     //const int sizex = 4096;
     //const int sizey = 11008;

examples/common.cpp

@@ -1,13 +1,18 @@
 #include "common.h"
 
 #include <cassert>
+#include <iostream>
 #include <cstring>
 #include <fstream>
 #include <string>
 #include <iterator>
 #include <algorithm>
 #include <sstream>
-#include <iostream>
+
+#if defined(__APPLE__) && defined(__MACH__)
+#include <sys/types.h>
+#include <sys/sysctl.h>
+#endif
 
 #if defined (_WIN32)
 #include <fcntl.h>
@@ -25,19 +30,43 @@ extern "C" __declspec(dllimport) int __stdcall WideCharToMultiByte(unsigned int
 #define CP_UTF8 65001
 #endif
 
-bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
-    // determine sensible default number of threads.
-    // std::thread::hardware_concurrency may not be equal to the number of cores, or may return 0.
+int32_t get_num_physical_cores() {
 #ifdef __linux__
     std::ifstream cpuinfo("/proc/cpuinfo");
-    params.n_threads = std::count(std::istream_iterator<std::string>(cpuinfo),
-                                  std::istream_iterator<std::string>(),
-                                  std::string("processor"));
-#endif
-    if (params.n_threads == 0) {
-        params.n_threads = std::max(1, (int32_t) std::thread::hardware_concurrency());
+    std::string line;
+    while (std::getline(cpuinfo, line)) {
+        std::size_t pos = line.find("cpu cores");
+        if (pos != std::string::npos) {
+            pos = line.find(": ", pos);
+            if (pos != std::string::npos) {
+                try {
+                    // Extract the number and return it
+                    return static_cast<int32_t>(std::stoul(line.substr(pos + 2)));
+                } catch (const std::invalid_argument &) {
+                    // Ignore if we could not parse
+                }
+            }
+        }
+    }
+#elif defined(__APPLE__) && defined(__MACH__)
+    int32_t num_physical_cores;
+    size_t len = sizeof(num_physical_cores);
+    int result = sysctlbyname("hw.perflevel0.physicalcpu", &num_physical_cores, &len, NULL, 0);
+    if (result == 0) {
+        return num_physical_cores;
+    }
+    result = sysctlbyname("hw.physicalcpu", &num_physical_cores, &len, NULL, 0);
+    if (result == 0) {
+        return num_physical_cores;
     }
+#elif defined(_WIN32)
+    //TODO: Implement
+#endif
+    unsigned int n_threads = std::thread::hardware_concurrency();
+    return n_threads > 0 ? (n_threads <= 4 ? n_threads : n_threads / 2) : 4;
+}
 
+bool gpt_params_parse(int argc, char ** argv, gpt_params & params) {
     bool invalid_param = false;
     std::string arg;
     gpt_params default_params;

examples/common.h

@@ -13,11 +13,12 @@
 //
 // CLI argument parsing
 //
+int32_t get_num_physical_cores();
 
 struct gpt_params {
     int32_t seed          = -1;   // RNG seed
-    int32_t n_threads     = std::min(4, (int32_t) std::thread::hardware_concurrency());
-    int32_t n_predict     = -1;   // new tokens to predict
+    int32_t n_threads     = get_num_physical_cores();
+    int32_t n_predict     = -1;  // new tokens to predict
     int32_t n_parts       = -1;   // amount of model parts (-1 = determine from model dimensions)
     int32_t n_ctx         = 512;  // context size
     int32_t n_batch       = 512;  // batch size for prompt processing (must be >=32 to use BLAS)

examples/embedding/CMakeLists.txt

@@ -2,3 +2,6 @@ set(TARGET embedding)
 add_executable(${TARGET} embedding.cpp)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_11)
+if(TARGET BUILD_INFO)
+  add_dependencies(${TARGET} BUILD_INFO)
+endif()

examples/embedding/embedding.cpp

@@ -1,5 +1,6 @@
 #include "common.h"
 #include "llama.h"
+#include "build-info.h"
 
 #include <ctime>
 
@@ -18,11 +19,13 @@ int main(int argc, char ** argv) {
                 "expect poor results\n", __func__, params.n_ctx);
     }
 
+    fprintf(stderr, "%s: build = %d (%s)\n", __func__, BUILD_NUMBER, BUILD_COMMIT);
+
     if (params.seed <= 0) {
         params.seed = time(NULL);
     }
 
-    fprintf(stderr, "%s: seed = %d\n", __func__, params.seed);
+    fprintf(stderr, "%s: seed  = %d\n", __func__, params.seed);
 
     std::mt19937 rng(params.seed);
     if (params.random_prompt) {

examples/main/CMakeLists.txt

@@ -2,3 +2,6 @@ set(TARGET main)
 add_executable(${TARGET} main.cpp)
 target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
 target_compile_features(${TARGET} PRIVATE cxx_std_11)
+if(TARGET BUILD_INFO)
+  add_dependencies(${TARGET} BUILD_INFO)
+endif()