Skip to content

GitLab

Unverified Commit 8f84a012 authored by Bartlomiej Wroblewski's avatar Bartlomiej Wroblewski Committed by GitHub
Browse files

Enable DPP8 GEMM on Navi3 (#892)

parent 562b4cec
Show whitespace changes
Inline Side-by-side
Showing with 4 additions and 2 deletions
include/ck/tensor_operation/gpu/device/impl/device_gemm_dpp.hpp
View file @ 8f84a012
...@@ -168,7 +168,8 @@ struct DeviceGemmDpp : public DeviceGemm<ALayout, ...@@ -168,7 +168,8 @@ struct DeviceGemmDpp : public DeviceGemm<ALayout,
static bool IsSupportedArgument(const Argument& karg) static bool IsSupportedArgument(const Argument& karg)
{ {
if(ck::get_device_name() == "gfx1030") if(ck::get_device_name() == "gfx1030" || ck::get_device_name() == "gfx1100" ||
ck::get_device_name() == "gfx1101" || ck::get_device_name() == "gfx1102")
{ {
return GridwiseGemm::CheckValidity(karg); return GridwiseGemm::CheckValidity(karg);
} }
......
include/ck/tensor_operation/gpu/grid/gridwise_gemm_dpp.hpp
View file @ 8f84a012
...@@ -28,7 +28,8 @@ __global__ void ...@@ -28,7 +28,8 @@ __global__ void
#endif #endif
kernel_gemm_dpp(const typename GridwiseGemm::Argument karg) kernel_gemm_dpp(const typename GridwiseGemm::Argument karg)
{ {
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx1030__)) #if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx1030__) || defined(__gfx1100__) || \
defined(__gfx1101__) || defined(__gfx1102__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()]; __shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
const auto a_grid_desc_ak0_m_ak1 = amd_wave_read_first_lane( const auto a_grid_desc_ak0_m_ak1 = amd_wave_read_first_lane(
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment