ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-14 02:02:46 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add the missing archs (#785)

Browse Source 
[ROCm/composable_kernel commit: 61dc9aa932]
This commit is contained in:
Rostyslav Geyyer
2023-07-05 18:29:56 -05:00
committed by GitHub
parent 3c1b791968
commit eb30728cd2
1 changed files with 2 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
include/ck/utility/amd_xdlops.hpp
View File

@@ -364,7 +364,7 @@ struct intrin_mfma_f32_32x32x16f8f8<32, 32>
template <class FloatC>
__device__ static void Run(const f8x8_t& reg_a, const f8x8_t& reg_b, FloatC& reg_c)
{
#if defined(__gfx940__)
#if defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__)
reg_c.template AsType<float16_t>()(Number<0>{}) =
__builtin_amdgcn_mfma_f32_32x32x16_fp8_fp8(
bit_cast<long>(reg_a),
@@ -396,7 +396,7 @@ struct intrin_mfma_f32_16x16x32f8f8<16, 16>
template <class FloatC>
__device__ static void Run(const f8x8_t& reg_a, const f8x8_t& reg_b, FloatC& reg_c)
{
#if defined(__gfx940__)
#if defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__)
reg_c.template AsType<float4_t>()(Number<0>{}) = __builtin_amdgcn_mfma_f32_16x16x32_fp8_fp8(
bit_cast<long>(reg_a),
bit_cast<long>(reg_b),