Run sycl ampere example on cuda

codeplaysoftware · May 27, 2024 · 26d4b4a · 26d4b4a
1 parent 7faa4d3
commit 26d4b4a
Show file tree

Hide file tree

Showing 10 changed files with 13 additions and 788 deletions.
diff --git a/benchmarks/CMakeLists.txt b/benchmarks/CMakeLists.txt
@@ -43,20 +43,22 @@ function(cutlass_benchmark_add_executable NAME)
 
   add_dependencies(cutlass_benchmarks ${NAME})
 
+  if (NOT CUTLASS_ENABLE_SYCL)
+    SET(ADD_CUDA ON)
+  endif()
+
   target_link_libraries(
     ${NAME}
     PRIVATE
     CUTLASS
     cutlass_tools_util_includes
-    )
-
-  target_include_directories(
-    ${NAME}
-    PRIVATE
-    ${CUTLASS_BENCHMARKS_COMMON_SOURCE_DIR}
-    )
+    $<$<BOOL:${CUTLASS_ENABLE_CUBLAS}>:nvidia::cublas>
+    $<$<BOOL:${ADD_CUDA}>:cuda>
+  )
 
-  add_sycl_to_target(TARGET ${NAME})
+  if (CUTLASS_ENABLE_SYCL)
+    add_sycl_to_target(TARGET ${NAME})
+  endif()
 
   install(
     TARGETS ${NAME}
@@ -66,6 +68,6 @@ endfunction()
 
 if(SYCL_INTEL_TARGET)
   add_subdirectory(pvc)
-else(SYCL_NVIDIA_TARGET)
+else(SYCL_NVIDIA_TARGET OR NOT CUTLASS_ENABLE_SYCL)
   add_subdirectory(ampere)
 endif()
diff --git a/benchmarks/ampere/CMakeLists.txt b/benchmarks/ampere/CMakeLists.txt
@@ -29,10 +29,10 @@
 
 cutlass_benchmark_add_executable(
   bench_ampere_gemm_fp16_fp16_fp32_tensor_op_fp32
-  bench_ampere_gemm_fp16_fp16_fp32_tensor_op_fp32.cpp
+  bench_ampere_gemm_fp16_fp16_fp32_tensor_op_fp32.cu
 )
 
 cutlass_benchmark_add_executable(
   bench_ampere_gemm_bf16_bf16_fp32_tensor_op_fp32
-  bench_ampere_gemm_bf16_bf16_fp32_tensor_op_fp32.cpp
+  bench_ampere_gemm_bf16_bf16_fp32_tensor_op_fp32.cu
 )
diff --git a/...re_gemm_bf16_bf16_fp32_tensor_op_fp32.cpp → ...ere_gemm_bf16_bf16_fp32_tensor_op_fp32.cu b/...re_gemm_bf16_bf16_fp32_tensor_op_fp32.cpp → ...ere_gemm_bf16_bf16_fp32_tensor_op_fp32.cu
diff --git a/...re_gemm_fp16_fp16_fp32_tensor_op_fp32.cpp → ...ere_gemm_fp16_fp16_fp32_tensor_op_fp32.cu b/...re_gemm_fp16_fp16_fp32_tensor_op_fp32.cpp → ...ere_gemm_fp16_fp16_fp32_tensor_op_fp32.cu
diff --git a/examples/sycl/CMakeLists.txt b/examples/sycl/CMakeLists.txt
@@ -29,6 +29,4 @@
 
 if(SYCL_INTEL_TARGET)
   add_subdirectory(pvc)
-else(SYCL_NVIDIA_TARGET)
-  add_subdirectory(ampere)
 endif()
diff --git a/examples/sycl/ampere/CMakeLists.txt b/examples/sycl/ampere/CMakeLists.txt
diff --git a/examples/sycl/ampere/ampere_gemm_bf16_bf16_fp32_tensor_op_fp32.cpp b/examples/sycl/ampere/ampere_gemm_bf16_bf16_fp32_tensor_op_fp32.cpp
diff --git a/examples/sycl/ampere/ampere_gemm_fp16_fp16_fp32_tensor_op_fp32.cpp b/examples/sycl/ampere/ampere_gemm_fp16_fp16_fp32_tensor_op_fp32.cpp