Skip to content

[Fix][Ready]Fix register spilling in cutlass nvfp4 gemm kernel on Blackwell #8127

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
zhyncs merged 3 commits into from
Jul 18, 2025
Merged

[Fix][Ready]Fix register spilling in cutlass nvfp4 gemm kernel on Blackwell #8127

Changes from 1 commit
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
ef5e248
Align MmaTileShape, ClusterShape, EpilogueTile to TRT-LLM.
HydraQYH Jul 17, 2025
7f47803
Fix fp32 schedule.
HydraQYH Jul 17, 2025
1dab27e
Refix fp32 schedule.
HydraQYH Jul 18, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
47 changes: 23 additions & 24 deletions sgl-kernel/csrc/gemm/nvfp4_scaled_mm_kernels.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,27 +40,21 @@ using namespace cute;
#if defined(CUTLASS_ARCH_MMA_SM100_SUPPORTED)
// Kernel Perf config
template <typename T>
struct KernelTraits;

template <>
struct KernelTraits<float> {
using MmaTileShape = Shape<_128, _128, _256>;
using ClusterShape = Shape<_1, _1, _1>;
using PerSmTileShape_MNK = Shape<_128, _128, _256>;
};

template <>
struct KernelTraits<cutlass::half_t> {
struct KernelTraits {
using MmaTileShape = Shape<_256, _256, _256>;
using ClusterShape = Shape<_4, _4, _1>;
using PerSmTileShape_MNK = Shape<_128, _256, _256>;
using ClusterShape = Shape<int, int, _1>;
using EpilogueTile = Shape<_128, _64>;
using EpilogueSchedule = cutlass::epilogue::TmaWarpSpecialized2Sm;
using MainloopSchedule = cutlass::gemm::KernelTmaWarpSpecialized2SmNvf4Sm100;
};

template <>
struct KernelTraits<cutlass::bfloat16_t> {
using MmaTileShape = Shape<_256, _256, _256>;
using ClusterShape = Shape<_4, _4, _1>;
using PerSmTileShape_MNK = Shape<_128, _256, _256>;
struct KernelTraits<float> {
using MmaTileShape = Shape<_128, _128, _256>;
using ClusterShape = Shape<int, int, _1>;
using EpilogueTile = cutlass::epilogue::collective::EpilogueTileAuto;
using EpilogueSchedule = cutlass::epilogue::TmaWarpSpecialized1Sm;
using MainloopSchedule = cutlass::gemm::KernelTmaWarpSpecialized1SmNvf4Sm100;
};

template <typename T>
Expand Down Expand Up @@ -90,23 +84,26 @@ struct Fp4GemmSm100 {
// Kernel Perf config
using MmaTileShape = typename KernelTraits<T>::MmaTileShape;
using ClusterShape = typename KernelTraits<T>::ClusterShape;
using PerSmTileShape_MNK = typename KernelTraits<T>::PerSmTileShape_MNK;
using EpilogueTile = typename KernelTraits<T>::EpilogueTile;
using EpilogueSchedule = typename KernelTraits<T>::EpilogueSchedule;
using MainloopSchedule = typename KernelTraits<T>::MainloopSchedule;

using CollectiveEpilogue = typename cutlass::epilogue::collective::CollectiveBuilder<
ArchTag,
OperatorClass,
PerSmTileShape_MNK,
cutlass::arch::OpClassTensorOp,
MmaTileShape,
ClusterShape,
cutlass::epilogue::collective::EpilogueTileAuto,
EpilogueTile,
ElementAccumulator,
ElementAccumulator,
ElementC,
void,
LayoutCTag,
AlignmentC,
ElementD,
LayoutDTag,
AlignmentD,
cutlass::epilogue::collective::EpilogueScheduleAuto>::CollectiveOp;
EpilogueSchedule,
cutlass::epilogue::fusion::LinearCombination<ElementD, float, void, float>>::CollectiveOp;

using CollectiveMainloop = typename cutlass::gemm::collective::CollectiveBuilder<
ArchTag,
Expand All @@ -122,7 +119,7 @@ struct Fp4GemmSm100 {
ClusterShape,
cutlass::gemm::collective::StageCountAutoCarveout<static_cast<int>(
sizeof(typename CollectiveEpilogue::SharedStorage))>,
cutlass::gemm::collective::KernelScheduleAuto>::CollectiveOp;
MainloopSchedule>::CollectiveOp;

using GemmKernel =
cutlass::gemm::kernel::GemmUniversal<Shape<int, int, int, int>, CollectiveMainloop, CollectiveEpilogue, void>;
Expand Down Expand Up @@ -191,6 +188,8 @@ typename T::Gemm::Arguments args_from_options(
stride_D}};
auto& fusion_args = arguments.epilogue.thread;
fusion_args.alpha_ptr = static_cast<ElementCompute const*>(alpha.data_ptr());
arguments.hw_info.cluster_shape = dim3(4, 4, 1);
arguments.hw_info.cluster_shape_fallback = dim3(2, 1, 1);

@gemini-code-assist gemini-code-assist Bot Jul 17, 2025

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

The hardcoded cluster shapes are applied to all data types. However, the float specialization uses TmaWarpSpecialized1Sm, which is a single-SM schedule and likely does not support or require clustering. The previous configuration for float used a 1x1x1 cluster. Applying a 4x4x1 cluster shape might be incorrect or confusing for the float kernel. It would be safer to apply these cluster shapes only for the data types that use a multi-SM schedule (like half_t).

// Set cluster shape for multi-SM schedules. These values are tuned for Blackwell performance.
if constexpr (!std::is_same_v<typename T::ElementD, float>) {
  arguments.hw_info.cluster_shape = dim3(4, 4, 1);
  arguments.hw_info.cluster_shape_fallback = dim3(2, 1, 1);
}

return arguments;
}

Expand Down
Loading