Fixing format (#560)

* fixing format * Adding other unformatted files
triton-inference-server · Aug 16, 2019 · 4d5a55c · 4d5a55c
1 parent 664625e
commit 4d5a55c
Show file tree

Hide file tree

Showing 4 changed files with 44 additions and 33 deletions.
diff --git a/qa/common/busy_op_kernel.cc b/qa/common/busy_op_kernel.cc
@@ -26,43 +26,47 @@
 
 #include <time.h>
 
+#include "tensorflow/core/framework/device_base.h"
 #include "tensorflow/core/framework/op.h"
 #include "tensorflow/core/framework/op_kernel.h"
 #include "tensorflow/core/framework/tensor_shape.h"
-#include "tensorflow/core/framework/device_base.h"
 
 using namespace tensorflow;  // NOLINT(build/namespaces)
 
-REGISTER_OP("BusyLoop")
-    .Input("input: int32")
-    .Output("output: int32")
-    .Doc(R"doc(
+REGISTER_OP("BusyLoop").Input("input: int32").Output("output: int32").Doc(R"doc(
 Busy waits for input number of clock cycles
 )doc");
 
-void BusyLoopKernelLauncher(const Eigen::GpuDevice& device, const int* num_delay_cycles, int* out);
+void BusyLoopKernelLauncher(
+    const Eigen::GpuDevice& device, const int* num_delay_cycles, int* out);
 
 class BusyLoopOp : public OpKernel {
  public:
   explicit BusyLoopOp(OpKernelConstruction* context) : OpKernel(context) {}
 
-  void Compute(OpKernelContext* context) override {
+  void Compute(OpKernelContext* context) override
+  {
     // Grab the input
     const Tensor& input_tensor = context->input(0);
     auto num_delay_cycles = input_tensor.flat<int32>();
 
     // Create dummy output
     Tensor* output_tensor = nullptr;
-    OP_REQUIRES_OK(context, context->allocate_output(0, input_tensor.shape(),
-                                                     &output_tensor));
+    OP_REQUIRES_OK(
+        context,
+        context->allocate_output(0, input_tensor.shape(), &output_tensor));
     auto output = output_tensor->template flat<int32>();
 
     // Verify input dimension
-    OP_REQUIRES(context, TensorShapeUtils::IsVector(input_tensor.shape()),
-        errors::InvalidArgument("BusyLoop expects a single value as a 1-D Vector"));
+    OP_REQUIRES(
+        context, TensorShapeUtils::IsVector(input_tensor.shape()),
+        errors::InvalidArgument(
+            "BusyLoop expects a single value as a 1-D Vector"));
 
     // Call the cuda kernel launcher
-    BusyLoopKernelLauncher(context->eigen_device<Eigen::GpuDevice>(), num_delay_cycles.data(), output.data());
+    BusyLoopKernelLauncher(
+        context->eigen_device<Eigen::GpuDevice>(), num_delay_cycles.data(),
+        output.data());
   }
 };
 

diff --git a/qa/common/busy_op_kernel.cu.cc b/qa/common/busy_op_kernel.cu.cc
@@ -27,30 +27,36 @@
 #if GOOGLE_CUDA
 #define EIGEN_USE_GPU
 
-#include <time.h>
 #include <cuda_runtime.h>
+#include <time.h>
 
 #include "third_party/eigen3/unsupported/Eigen/CXX11/Tensor"
 
 __device__ long store_now[1];
 
-__global__ void BusyLoopKernel(const int* num_delay_cycles, int* out) {
-    // As shown in https://stackoverflow.com/questions/11217117/equivalent-of-usleep-in-cuda-kernel
-    clock_t start = clock();
-    
-    for(;;) {
-        clock_t now = clock();
-        // Adjust for overflow
-        clock_t cycles = now > start ? now - start : now + (0xffffffff - start);
-        if (cycles >= num_delay_cycles[0]) {
-            break;
-        }
-        // Prevent nvcc optimizations 
-        store_now[0] = cycles;
+__global__ void
+BusyLoopKernel(const int* num_delay_cycles, int* out)
+{
+  // As shown in
+  // https://stackoverflow.com/questions/11217117/equivalent-of-usleep-in-cuda-kernel
+  clock_t start = clock();
+
+  for (;;) {
+    clock_t now = clock();
+    // Adjust for overflow
+    clock_t cycles = now > start ? now - start : now + (0xffffffff - start);
+    if (cycles >= num_delay_cycles[0]) {
+      break;
     }
+    // Prevent nvcc optimizations
+    store_now[0] = cycles;
+  }
 }
 
-void BusyLoopKernelLauncher(const Eigen::GpuDevice& device, const int* num_delay_cycles, int* out) {
+void
+BusyLoopKernelLauncher(
+    const Eigen::GpuDevice& device, const int* num_delay_cycles, int* out)
+{
   auto stream = device.stream();
   BusyLoopKernel<<<1, 256, 0, stream>>>(num_delay_cycles, out);
 }

diff --git a/src/clients/c++/inference_profiler.h b/src/clients/c++/inference_profiler.h
@@ -170,7 +170,8 @@ class InferenceProfiler {
       std::unique_ptr<LoadManager> manager);
 
   /// A helper function to construct the map of ensemble models to its composing
-  /// models \param model_name The ensemble model to be added into the map
+  /// models.
+  /// \param model_name The ensemble model to be added into the map
   /// \param model_version The version of the model to be added
   /// \server_status The server status response from TRTIS.
   /// \return Error object indicating success or failure

diff --git a/src/core/ensemble_scheduler.cc b/src/core/ensemble_scheduler.cc
@@ -281,11 +281,11 @@ EnsembleContext::ResponseAlloc(
                    << memory_type << " for " << tensor_name;
     return nullptr;  // Success
   } else {
-    auto it = tensor_data_map
-                  ->emplace(
-                      tensor_name,
-                      std::make_shared<AllocatedSystemMemory>(byte_size))
-                  .first;
+    auto it =
+        tensor_data_map
+            ->emplace(
+                tensor_name, std::make_shared<AllocatedSystemMemory>(byte_size))
+            .first;
     if (byte_size > 0) {
       *buffer = static_cast<void*>(it->second->MutableBuffer());
     }