universal streamk files and ckprofiler files for same

a4d67230 · Harisankar Sadasivan · cb138394 · a4d67230 · a4d67230 · a4d67230
Commit a4d67230 authored Jun 24, 2024 by Harisankar Sadasivan
18 changed files
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_mem_v1_mnkpadding_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_mem_v2_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_mem_v2_default_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_mem_v2_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_mem_v2_kpadding_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_kn_mn_mem_v2_mnkpadding_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn.hpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_default_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_kpadding_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnkpadding_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_comp_mnpadding_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v1_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v1_default_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v1_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v1_kpadding_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v1_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v1_mnkpadding_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_default_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_default_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_kpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_kpadding_instance.cpp
--- a/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_mnkpadding_instance.cpp
+++ b/library/src/tensor_operation_instance/gpu/gemm_universal_streamk/gemm_universal_streamk/device_gemm_xdl_universal_streamk_f16_f16_f16/device_gemm_xdl_universal_streamk_f16_f16_f16_mk_nk_mn_mem_v2_mnkpadding_instance.cpp
--- a/profiler/include/profiler/profile_gemm_universal_streamk_impl.hpp
+++ b/profiler/include/profiler/profile_gemm_universal_streamk_impl.hpp
--- a/profiler/src/CMakeLists.txt
+++ b/profiler/src/CMakeLists.txt
@@ -52,6 +52,7 @@ if(GPU_TARGETS MATCHES "gfx9")
  list(APPEND PROFILER_SOURCES profile_gemm_bias_add_reduce.cpp)
  list(APPEND PROFILER_SOURCES profile_gemm_splitk.cpp)
  list(APPEND PROFILER_SOURCES profile_gemm_universal.cpp)
+  list(APPEND PROFILER_SOURCES profile_gemm_universal_streamk.cpp)
  list(APPEND PROFILER_SOURCES profile_conv_fwd_bias_relu.cpp)
  list(APPEND PROFILER_SOURCES profile_conv_fwd_bias_relu_add.cpp)
  list(APPEND PROFILER_SOURCES profile_conv_bwd_data.cpp)
@@ -120,6 +121,7 @@ if(GPU_TARGETS MATCHES "gfx9")
  target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_multiply_add_instance)
  target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_splitk_instance)
  target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_universal_instance)
+  target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_universal_streamk_instance)
  target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_add_multiply_instance)
  target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_reduce_instance)
  target_link_libraries(${PROFILER_EXECUTABLE} PRIVATE device_gemm_bias_add_reduce_instance)

--- a/profiler/src/profile_gemm_universal_streamk.cpp
+++ b/profiler/src/profile_gemm_universal_streamk.cpp