leavelet/cutlass
1
0
Fork 0
Code Issues Pull Requests Actions 26 Packages Projects Releases Wiki Activity

update 3.8 v2 (#2112)

Browse Source 
* update 3.8 v2

* update 3.8

---------

Co-authored-by: yuzhai <yuzhai@nvidia.com>
This commit is contained in:
Yujia Zhai
2025-02-19 22:03:14 -05:00
committed by GitHub
co-authored by yuzhai
parent e9627ce55b
commit b84e9802d8
166 changed files with 3986 additions and 4037 deletions
Download Patch File Download Diff File Expand all files Collapse all files
examples/70_blackwell_gemm/70_blackwell_fp16_gemm.cu
+1 -5
View File
@@ -115,15 +115,11 @@ using OperatorClass = cutlass::arch::OpClassTensorOp; // O
using MmaTileShape_MNK = Shape<_256,_128,_64>;
// Shape of the threadblocks in a cluster
using ClusterShape_MNK = Shape<_2,_2,_1>;
// Shape of the threadblocks participating in a tcgen05 MMA. <1, 1, 1> for cta_group = 1, <2, 1, 1> for cta_group = 2
using AtomThrShape_MNK = Shape<_2, _1, _1>;
// Shape of the tile computed by each SM
using PerSmTileShape_MNK = decltype(shape_div(MmaTileShape_MNK{}, AtomThrShape_MNK{}));
// Build the epilogue
using CollectiveEpilogue = typename cutlass::epilogue::collective::CollectiveBuilder<
ArchTag, OperatorClass,
PerSmTileShape_MNK, ClusterShape_MNK,
MmaTileShape_MNK, ClusterShape_MNK,
cutlass::epilogue::collective::EpilogueTileAuto,
ElementAccumulator, ElementAccumulator,
ElementC, LayoutC, AlignmentC,
examples/70_blackwell_gemm/70_blackwell_fp8_gemm.cu
+1 -5
View File
@@ -131,17 +131,13 @@ using ElementAmax = float;
using MmaTileShape_MNK = Shape<_256,_128,_64>;
// Shape of the threadblocks in a cluster
using ClusterShape_MNK = Shape<_2,_2,_1>;
// Shape of the threadblocks participating in a tcgen05 MMA. <1, 1, 1> for cta_group = 1, <2, 1, 1> for cta_group = 2
using AtomThrShape_MNK = Shape<_2, _1, _1>;
// Shape of the tile computed by each SM
using PerSmTileShape_MNK = decltype(shape_div(MmaTileShape_MNK{}, AtomThrShape_MNK{}));
using FusionOp = cutlass::epilogue::fusion::ScaledLinCombPerRowBiasEltActAmaxAux<
LayoutC, cutlass::epilogue::thread::ReLU, ElementD, ElementCompute, ElementAux, ElementAmax, ElementBias>;
using CollectiveEpilogue = typename cutlass::epilogue::collective::CollectiveBuilder<
cutlass::arch::Sm100, cutlass::arch::OpClassTensorOp,
PerSmTileShape_MNK, ClusterShape_MNK,
MmaTileShape_MNK, ClusterShape_MNK,
cutlass::epilogue::collective::EpilogueTileAuto,
ElementAccumulator, ElementCompute,
ElementC, LayoutC, AlignmentC,
examples/70_blackwell_gemm/CMakeLists.txt
+1 -1
View File
@@ -28,7 +28,7 @@
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
if(NOT CUTLASS_NVCC_ARCHS STREQUAL "100")
if (CUTLASS_NVCC_ARCHS MATCHES 100a)
cutlass_example_add_executable(
70_blackwell_fp16_gemm
70_blackwell_fp16_gemm.cu