diff --git a/profiler/CMakeLists.txt b/profiler/CMakeLists.txt index 861dcaecea..a2cf6eeb62 100644 --- a/profiler/CMakeLists.txt +++ b/profiler/CMakeLists.txt @@ -24,38 +24,38 @@ include_directories(BEFORE set(PROFILER_SOURCE src/profiler.cpp src/profile_gemm.cpp -# src/profile_gemm_bias_2d.cpp -# src/profile_gemm_bias_relu.cpp -# src/profile_gemm_bias_relu_add.cpp -# src/profile_gemm_reduce.cpp -# src/profile_batched_gemm.cpp -# src/profile_conv_fwd.cpp -# src/profile_conv_fwd_bias_relu.cpp -# src/profile_conv_fwd_bias_relu_add.cpp -# src/profile_conv_fwd_bias_relu_atomic_add.cpp -# src/profile_convnd_bwd_data.cpp -# src/profile_reduce.cpp -# src/profile_grouped_gemm.cpp -# src/profile_conv_bwd_weight.cpp -# src/profile_batched_gemm_reduce.cpp + src/profile_gemm_bias_2d.cpp + src/profile_gemm_bias_relu.cpp + src/profile_gemm_bias_relu_add.cpp + src/profile_gemm_reduce.cpp + src/profile_batched_gemm.cpp + src/profile_conv_fwd.cpp + src/profile_conv_fwd_bias_relu.cpp + src/profile_conv_fwd_bias_relu_add.cpp + src/profile_conv_fwd_bias_relu_atomic_add.cpp + src/profile_convnd_bwd_data.cpp + src/profile_reduce.cpp + src/profile_grouped_gemm.cpp + src/profile_conv_bwd_weight.cpp + src/profile_batched_gemm_reduce.cpp ) add_executable(ckProfiler ${PROFILER_SOURCE}) target_link_libraries(ckProfiler PRIVATE host_tensor) -#target_link_libraries(ckProfiler PRIVATE device_gemm_reduce_instance) +target_link_libraries(ckProfiler PRIVATE device_gemm_reduce_instance) target_link_libraries(ckProfiler PRIVATE device_gemm_instance) -#target_link_libraries(ckProfiler PRIVATE device_gemm_bias2d_instance) -#target_link_libraries(ckProfiler PRIVATE device_gemm_bias_relu_instance) -#target_link_libraries(ckProfiler PRIVATE device_gemm_bias_relu_add_instance) -#target_link_libraries(ckProfiler PRIVATE device_batched_gemm_instance) -#target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_instance) -#target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_bias_relu_instance) -#target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_bias_relu_add_instance) -#target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_bias_relu_atomic_add_instance) -#target_link_libraries(ckProfiler PRIVATE device_convnd_bwd_data_instance) -#target_link_libraries(ckProfiler PRIVATE device_reduce_instance) -#target_link_libraries(ckProfiler PRIVATE device_reduce_instance) -#target_link_libraries(ckProfiler PRIVATE device_grouped_gemm_instance) -#target_link_libraries(ckProfiler PRIVATE device_conv2d_bwd_weight_instance) -#target_link_libraries(ckProfiler PRIVATE device_batched_gemm_reduce_instance) +target_link_libraries(ckProfiler PRIVATE device_gemm_bias2d_instance) +target_link_libraries(ckProfiler PRIVATE device_gemm_bias_relu_instance) +target_link_libraries(ckProfiler PRIVATE device_gemm_bias_relu_add_instance) +target_link_libraries(ckProfiler PRIVATE device_batched_gemm_instance) +target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_instance) +target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_bias_relu_instance) +target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_bias_relu_add_instance) +target_link_libraries(ckProfiler PRIVATE device_conv2d_fwd_bias_relu_atomic_add_instance) +target_link_libraries(ckProfiler PRIVATE device_convnd_bwd_data_instance) +target_link_libraries(ckProfiler PRIVATE device_reduce_instance) +target_link_libraries(ckProfiler PRIVATE device_reduce_instance) +target_link_libraries(ckProfiler PRIVATE device_grouped_gemm_instance) +target_link_libraries(ckProfiler PRIVATE device_conv2d_bwd_weight_instance) +target_link_libraries(ckProfiler PRIVATE device_batched_gemm_reduce_instance) diff --git a/profiler/src/profiler.cpp b/profiler/src/profiler.cpp index 64344e8f8d..3cd454e351 100644 --- a/profiler/src/profiler.cpp +++ b/profiler/src/profiler.cpp @@ -26,70 +26,70 @@ int main(int argc, char* argv[]) { return profile_gemm(argc, argv); } - // else if(strcmp(argv[1], "gemm_bias_2d") == 0) - // { - // return profile_gemm_bias_2d(argc, argv); - // } - // else if(strcmp(argv[1], "gemm_bias_relu") == 0) - // { - // return profile_gemm_bias_relu(argc, argv); - // } - // else if(strcmp(argv[1], "gemm_bias_relu_add") == 0) - // { - // return profile_gemm_bias_relu_add(argc, argv); - // } - // else if(strcmp(argv[1], "gemm_reduce") == 0) - // { - // return profile_gemm_reduce(argc, argv); - // } - // else if(strcmp(argv[1], "batched_gemm") == 0) - // { - // return profile_batched_gemm(argc, argv); - // } - // else if(strcmp(argv[1], "batched_gemm_reduce") == 0) - // { - // return profile_batched_gemm_reduce(argc, argv); - // } - // else if(strcmp(argv[1], "grouped_gemm") == 0) - // { - // profile_grouped_gemm(argc, argv); - // } - // else if(strcmp(argv[1], "conv_fwd") == 0) - // { - // return profile_conv_fwd(argc, argv); - // } - // else if(strcmp(argv[1], "conv_fwd_bias_relu") == 0) - // { - // return profile_conv_fwd_bias_relu(argc, argv); - // } - // else if(strcmp(argv[1], "conv_fwd_bias_relu_add") == 0) - // { - // return profile_conv_fwd_bias_relu_add(argc, argv); - // } - // else if(strcmp(argv[1], "conv_fwd_bias_relu_atomic_add") == 0) - // { - // return profile_conv_fwd_bias_relu_atomic_add(argc, argv); - // } - // else if(strcmp(argv[1], "conv1d_bwd_data") == 0) - // { - // return profile_convnd_bwd_data(argc, argv, 1); - // } - // else if(strcmp(argv[1], "conv2d_bwd_data") == 0) - // { - // return profile_convnd_bwd_data(argc, argv, 2); - // } - // else if(strcmp(argv[1], "conv3d_bwd_data") == 0) - // { - // return profile_convnd_bwd_data(argc, argv, 3); - // } - // else if(strcmp(argv[1], "reduce") == 0) - // { - // return profile_reduce(argc, argv); - // } - // else if(strcmp(argv[1], "conv2d_bwd_weight") == 0) - // { - // return profile_conv_bwd_weight(argc, argv); - // } + else if(strcmp(argv[1], "gemm_bias_2d") == 0) + { + return profile_gemm_bias_2d(argc, argv); + } + else if(strcmp(argv[1], "gemm_bias_relu") == 0) + { + return profile_gemm_bias_relu(argc, argv); + } + else if(strcmp(argv[1], "gemm_bias_relu_add") == 0) + { + return profile_gemm_bias_relu_add(argc, argv); + } + else if(strcmp(argv[1], "gemm_reduce") == 0) + { + return profile_gemm_reduce(argc, argv); + } + else if(strcmp(argv[1], "batched_gemm") == 0) + { + return profile_batched_gemm(argc, argv); + } + else if(strcmp(argv[1], "batched_gemm_reduce") == 0) + { + return profile_batched_gemm_reduce(argc, argv); + } + else if(strcmp(argv[1], "grouped_gemm") == 0) + { + profile_grouped_gemm(argc, argv); + } + else if(strcmp(argv[1], "conv_fwd") == 0) + { + return profile_conv_fwd(argc, argv); + } + else if(strcmp(argv[1], "conv_fwd_bias_relu") == 0) + { + return profile_conv_fwd_bias_relu(argc, argv); + } + else if(strcmp(argv[1], "conv_fwd_bias_relu_add") == 0) + { + return profile_conv_fwd_bias_relu_add(argc, argv); + } + else if(strcmp(argv[1], "conv_fwd_bias_relu_atomic_add") == 0) + { + return profile_conv_fwd_bias_relu_atomic_add(argc, argv); + } + else if(strcmp(argv[1], "conv1d_bwd_data") == 0) + { + return profile_convnd_bwd_data(argc, argv, 1); + } + else if(strcmp(argv[1], "conv2d_bwd_data") == 0) + { + return profile_convnd_bwd_data(argc, argv, 2); + } + else if(strcmp(argv[1], "conv3d_bwd_data") == 0) + { + return profile_convnd_bwd_data(argc, argv, 3); + } + else if(strcmp(argv[1], "reduce") == 0) + { + return profile_reduce(argc, argv); + } + else if(strcmp(argv[1], "conv2d_bwd_weight") == 0) + { + return profile_conv_bwd_weight(argc, argv); + } else { // clang-format off