v3.9 update (#2203)
* v3.9 update * voidD --------- Co-authored-by: yuzhai <yuzhai@nvidia.com>
This commit is contained in:
@@ -499,3 +499,40 @@ TEST(SM80_CuTe_Ampere, CooperativeGemm2_Double_MMA_Predicated_Reg) {
|
||||
|
||||
test_cooperative_gemm_col_major_layout_rmem_c<thread_block_size, value_type>(shape_mnk, tiled_mma);
|
||||
}
|
||||
|
||||
TEST(SM80_CuTe_Ampere, CooperativeGemmLDSMx2) {
|
||||
|
||||
constexpr uint32_t thread_block_size = 128;
|
||||
constexpr int MaxVecBits = 128;
|
||||
using TA = cute::half_t;
|
||||
using TB = cute::half_t;
|
||||
using TC = float;
|
||||
|
||||
auto tiled_mma =
|
||||
TiledMMA<
|
||||
MMA_Atom<SM80_16x8x16_F32F16F16F32_TN>,
|
||||
Layout<Shape<_2, _2, _1>, Stride<_1, _2, _0>>,
|
||||
Tile<_32, _16, _16>
|
||||
>{};
|
||||
|
||||
auto global_a_layout = make_layout(Shape<_32, _32>{}, LayoutRight{});
|
||||
auto global_b_layout = make_layout(Shape<_16, _32>{}, LayoutRight{});
|
||||
auto global_c_layout = make_layout(Shape<_32, _16>{}, LayoutRight{});
|
||||
|
||||
test_cooperative_gemm<thread_block_size,
|
||||
MaxVecBits,
|
||||
TA, TB, TC>
|
||||
(global_a_layout,
|
||||
global_b_layout,
|
||||
global_c_layout,
|
||||
global_a_layout,
|
||||
global_b_layout,
|
||||
global_c_layout,
|
||||
tiled_mma,
|
||||
identity{},
|
||||
identity{},
|
||||
identity{},
|
||||
identity{},
|
||||
SM75_U32x4_LDSM_N{},
|
||||
SM75_U32x2_LDSM_N{});
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user