enable gfx940

include/ck/tensor_operation/gpu/grid/gridwise_gemm_xdlops_v3r3.hpp

@@ -53,7 +53,7 @@ __global__ void
             const CElementwiseOperation c_element_op,
             const Block2CTileMap block_2_ctile_map)
 {
-#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__))
+#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__) || defined(__gfx940__))
     __shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
 
     GridwiseGemm::template Run<HasMainKBlockLoop>(

include/ck/utility/amd_xdlops.hpp

@@ -306,7 +306,7 @@ struct intrin_mfma_f64_16x16x4f64<16, 16>
     template <class FloatC>
     __device__ static void Run(const double& reg_a, const double& reg_b, FloatC& reg_c)
     {
-#ifdef __gfx90a__
+#if defined(__gfx90a__) || defined(__gfx940__)
         reg_c.template AsType<double4_t>()(Number<0>{}) = __builtin_amdgcn_mfma_f64_16x16x4f64(
             reg_a, reg_b, reg_c.template AsType<double4_t>()[Number<0>{}], 0, 0, 0);
 #else

script/cmake-ck-dev.sh

@@ -11,9 +11,8 @@ cmake
 -D CMAKE_CXX_FLAGS="-O3 -ftemplate-backtrace-limit=0 -gline-tables-only -save-temps=$PWD"         \
 -D CMAKE_BUILD_TYPE=Release                                                                       \
 -D BUILD_DEV=ON                                                                                   \
--D GPU_TARGETS="gfx908;gfx90a"                                                                    \
+-D GPU_TARGETS="gfx90a"                                                                           \
 -D CMAKE_VERBOSE_MAKEFILE:BOOL=ON                                                                 \
 -D USE_BITINT_EXTENSION_INT4=OFF                                                                  \
 ${MY_PROJECT_SOURCE}
 
-#-D AMDGPU_TARGETS=gfx90a;gfx908

script/cmake-ck-release.sh

@@ -11,9 +11,8 @@ cmake
 -D CMAKE_CXX_FLAGS="-O3"                                                                          \
 -D CMAKE_BUILD_TYPE=Release                                                                       \
 -D BUILD_DEV=OFF                                                                                  \
--D GPU_TARGETS="gfx908;gfx90a"                                                                      \
+-D GPU_TARGETS="gfx908;gfx90a;gfx940"                                                             \
 -D CMAKE_VERBOSE_MAKEFILE:BOOL=ON                                                                 \
 -D USE_BITINT_EXTENSION_INT4=OFF                                                                  \
 ${MY_PROJECT_SOURCE}
 
-#-D AMDGPU_TARGETS=gfx90a;gfx908