support blockwise fp8 matmul kernel (#3267)

sgl-kernel/benchmark/bench_fp8_blockwise_gemm.py

+import argparse
+import copy
+import itertools
+
+import torch
+import triton
+from sgl_kernel import fp8_blockwise_scaled_mm
+from vllm._custom_ops import cutlass_scaled_mm as vllm_scaled_mm
+
+
+def get_weight_shapes(args):
+    models_tps = list(itertools.product(args.models, args.tp_sizes))
+    # NOTE(HandH1998): The weight shapes only works for DeepSeek-V3. Modify them, if you tune for another different model.
+    # cannot TP
+    total = [
+        # (512 + 64, 7168), # this weight is not supported by current kernel
+        ((128 + 64) * 128, 7168),
+        (128 * (128 + 128), 512),
+        (7168, 16384),
+        (7168, 18432),
+    ]
+    # N can TP
+    n_tp = [
+        (18432 * 2, 7168),
+        ((128 + 64) * 128, 7168),
+        (128 * (128 + 128), 512),
+        (24576, 1536),
+        (4096, 7168),
+    ]
+    # K can TP
+    k_tp = [(7168, 18432), (7168, 16384), (7168, 2048)]
+    # only support Deepseek-V3
+    SUPPORT_MODEL = ["deepseek-ai/DeepSeek-V3"]
+
+    weight_shapes = []
+    for model, tp_size in models_tps:
+        assert model in SUPPORT_MODEL
+        for t in total:
+            new_t = [t[0], t[1], model]
+            weight_shapes.append(new_t)
+        for n_t in n_tp:
+            new_t = [n_t[0] // tp_size, n_t[1], model]
+            weight_shapes.append(new_t)
+        for k_t in k_tp:
+            new_t = [k_t[0], k_t[1] // tp_size, model]
+            weight_shapes.append(new_t)
+    return weight_shapes
+
+
+def cdiv(a: int, b: int) -> int:
+    """Ceiling division."""
+    return -(a // -b)
+
+
+def scale_shape(shape, group_shape):
+    assert len(shape) == len(group_shape)
+    return tuple(cdiv(shape[i], group_shape[i]) for i in range(len(group_shape)))
+
+
+@triton.testing.perf_report(
+    triton.testing.Benchmark(
+        x_names=["batch_size"],
+        x_vals=[1, 16, 32, 64, 128, 256, 512, 1024, 2048],
+        x_log=False,
+        line_arg="provider",
+        line_vals=["vllm", "sgl-kernel"],
+        line_names=["vllm fp8 blockwise gemm", "sgl-kernel fp8 blockwise gemm"],
+        styles=[("blue", "-"), ("orange", "-")],
+        ylabel="GB/s",
+        plot_name="fp8 blockwise scaled matmul",
+        args={},
+    )
+)
+def benchmark(batch_size, provider, N, K):
+    M = batch_size
+    fp8_info = torch.finfo(torch.float8_e4m3fn)
+    fp8_max, fp8_min = fp8_info.max, fp8_info.min
+
+    a_fp32 = (torch.rand(M, K, dtype=torch.float32, device="cuda") - 0.5) * 2 * fp8_max
+    a_fp8 = a_fp32.clamp(min=fp8_min, max=fp8_max).to(torch.float8_e4m3fn)
+
+    b_fp32 = (torch.rand(N, K, dtype=torch.float32, device="cuda") - 0.5) * 2 * fp8_max
+    b_fp8 = b_fp32.clamp(min=fp8_min, max=fp8_max).to(torch.float8_e4m3fn).t()
+
+    scale_a_group_shape = (1, 128)
+    scale_b_group_shape = (128, 128)
+    scale_a_shape = scale_shape(a_fp8.shape, scale_a_group_shape)
+    scale_b_shape = scale_shape(b_fp8.shape, scale_b_group_shape)
+
+    scale_a = torch.randn(scale_a_shape, device="cuda", dtype=torch.float32)
+    scale_b = torch.randn(scale_b_shape, device="cuda", dtype=torch.float32)
+    scale_a = scale_a.t().contiguous().t()
+    scale_b = scale_b.t().contiguous().t()
+
+    quantiles = [0.5, 0.2, 0.8]
+    if provider == "sgl-kernel":
+        ms, min_ms, max_ms = triton.testing.do_bench(
+            lambda: fp8_blockwise_scaled_mm(
+                a_fp8, b_fp8, scale_a, scale_b, torch.float16
+            ),
+            quantiles=quantiles,
+        )
+    if provider == "vllm":
+        ms, min_ms, max_ms = triton.testing.do_bench(
+            lambda: vllm_scaled_mm(a_fp8, b_fp8, scale_a, scale_b, torch.float16),
+            quantiles=quantiles,
+        )
+    gbps = (
+        lambda ms: (
+            (2 * M * N * K - M * N) * a_fp8.element_size()
+            + (3 * M * N) * scale_a.element_size()
+        )
+        * 1e-9
+        / (ms * 1e-3)
+    )
+    return gbps(ms), gbps(max_ms), gbps(min_ms)
+
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "--models",
+        nargs="+",
+        type=str,
+        default=["deepseek-ai/DeepSeek-V3"],
+        help="List of models to benchmark",
+    )
+    parser.add_argument(
+        "--tp-sizes",
+        nargs="+",
+        type=int,
+        default=[1],
+        help="List of tensor parallel sizes",
+    )
+    args = parser.parse_args()
+
+    NK_model_names = get_weight_shapes(args)
+    for N, K, model_name in NK_model_names:
+        print(f"{model_name} N={N} K={K}: ")
+        benchmark.run(
+            print_data=True,
+            show_plots=True,
+            save_path="bench_fp8_blockwise_res",
+            N=N,
+            K=K,
+        )
+
+    print("Benchmark finished!")

sgl-kernel/setup.py

@@ -96,6 +96,7 @@ sources = [
     "src/sgl-kernel/csrc/moe_align_kernel.cu",
     "src/sgl-kernel/csrc/int8_gemm_kernel.cu",
     "src/sgl-kernel/csrc/fp8_gemm_kernel.cu",
+    "src/sgl-kernel/csrc/fp8_blockwise_gemm_kernel.cu",
     "src/sgl-kernel/csrc/lightning_attention_decode_kernel.cu",
     "src/sgl-kernel/csrc/fused_add_rms_norm_kernel.cu",
     "src/sgl-kernel/csrc/eagle_utils.cu",

sgl-kernel/src/sgl-kernel/__init__.py

@@ -14,6 +14,7 @@ from sgl_kernel.ops import (
     build_tree_kernel_efficient,
     custom_dispose,
     custom_reduce,
+    fp8_blockwise_scaled_mm,
     fp8_scaled_mm,
     fused_add_rmsnorm,
     gelu_and_mul,
@@ -44,6 +45,7 @@ __all__ = [
     "bmm_fp8",
     "custom_dispose",
     "custom_reduce",
+    "fp8_blockwise_scaled_mm",
     "fp8_scaled_mm",
     "fused_add_rmsnorm",
     "gelu_and_mul",

sgl-kernel/src/sgl-kernel/csrc/cutlass_extensions/gemm/collective/collective_builder.hpp

+// Adapt from
+// https://github.com/vllm-project/vllm/blob/v0.7.1/csrc/cutlass_extensions/gemm/collective/collective_buildler.hpp
+// Modified from: cutlass/gemm/collective/builders/sm90_gmma_builder.inl
+// clang-format off
+#pragma once
+
+#include <cutlass/gemm/collective/builders/sm90_gmma_builder.inl>
+#include "cutlass_extensions/gemm/dispatch_policy.hpp"
+#include "cutlass_extensions/gemm/collective/sm90_mma_tma_gmma_ss_warpspecialized_fp8_blockwise_scaling.hpp"
+
+
+/////////////////////////////////////////////////////////////////////////////////////////////////
+
+namespace cutlass::gemm::collective {
+
+/////////////////////////////////////////////////////////////////////////////////////////////////
+
+// GMMA_TMA_WS_SS (BlockScaled Builders)
+template <
+  class ElementA,
+  class GmemLayoutATag,
+  int AlignmentA,
+  class ElementB,
+  class GmemLayoutBTag,
+  int AlignmentB,
+  class ElementAccumulator,
+  class TileShape_MNK,
+  class ClusterShape_MNK,
+  class StageCountType,
+  int ScaleGranularityM
+>
+struct CollectiveBuilder<
+    arch::Sm90,
+    arch::OpClassTensorOp,
+    ElementA,
+    GmemLayoutATag,
+    AlignmentA,
+    ElementB,
+    GmemLayoutBTag,
+    AlignmentB,
+    ElementAccumulator,
+    TileShape_MNK,
+    ClusterShape_MNK,
+    StageCountType,
+    KernelTmaWarpSpecializedCooperativeFP8BlockScaledSubGroupMAccum<ScaleGranularityM>,
+    cute::enable_if_t<
+      not detail::is_use_rmem_A<ElementA, GmemLayoutATag, ElementB, GmemLayoutBTag>()>
+> {
+  using KernelScheduleType = KernelTmaWarpSpecializedCooperativeFP8BlockScaledSubGroupMAccum<ScaleGranularityM>;
+
+  static_assert(is_static<TileShape_MNK>::value);
+  static_assert(is_static<ClusterShape_MNK>::value);
+#ifndef CUTLASS_SM90_COLLECTIVE_BUILDER_SUPPORTED
+  static_assert(cutlass::detail::dependent_false<ElementA>, "Unsupported Toolkit for SM90 Collective Builder\n");
+#endif
+  static_assert(detail::is_aligned<ElementA, AlignmentA, ElementB, AlignmentB, detail::tma_alignment_bytes>(),
+                "Should meet TMA alignment requirement\n");
+
+  static constexpr bool IsArrayOfPointersGemm = (cute::is_any_of_v<KernelScheduleType,
+                                                                   KernelPtrArrayTmaWarpSpecializedCooperative,
+                                                                   KernelPtrArrayTmaWarpSpecializedPingpong>);
+  static constexpr bool IsFP8Input = detail::is_input_fp8<ElementA, ElementB>();
+  static_assert((!IsFP8Input || !IsArrayOfPointersGemm),
+                "KernelTmaWarpSpecializedCooperativeFP8BlockScaledAccum is only compatible with FP8 Blocked Scaled version right now.");
+
+  // For fp32 types, map to tf32 MMA value type
+  using ElementAMma = cute::conditional_t<cute::is_same_v<ElementA, float>, tfloat32_t, ElementA>;
+  using ElementBMma = cute::conditional_t<cute::is_same_v<ElementB, float>, tfloat32_t, ElementB>;
+
+  static constexpr cute::GMMA::Major GmmaMajorA = detail::gmma_ss_tag_to_major_A<ElementAMma, GmemLayoutATag>();
+  static constexpr cute::GMMA::Major GmmaMajorB = detail::gmma_ss_tag_to_major_B<ElementBMma, GmemLayoutBTag>();
+
+  static constexpr bool IsCooperative = cute::is_any_of_v<KernelScheduleType,
+                                                          KernelTmaWarpSpecializedCooperative,
+                                                          KernelPtrArrayTmaWarpSpecializedCooperative,
+                                                          KernelTmaWarpSpecializedCooperativeFP8BlockScaledSubGroupMAccum<ScaleGranularityM>>;
+  using AtomLayoutMNK = cute::conditional_t<IsCooperative,
+      Layout<Shape<_2,_1,_1>>, Layout<Shape<_1,_1,_1>>>;
+
+  using TiledMma = decltype(cute::make_tiled_mma(cute::GMMA::ss_op_selector<
+      ElementAMma, ElementBMma, ElementAccumulator, TileShape_MNK, GmmaMajorA, GmmaMajorB>(), AtomLayoutMNK{}));
+
+  using GmemTiledCopyA = decltype(detail::sm90_cluster_shape_to_tma_atom(shape<1>(ClusterShape_MNK{})));
+  using GmemTiledCopyB = decltype(detail::sm90_cluster_shape_to_tma_atom(shape<0>(ClusterShape_MNK{})));
+
+  using SmemLayoutAtomA = decltype(detail::ss_smem_selector<
+      GmmaMajorA, ElementAMma, decltype(cute::get<0>(TileShape_MNK{})), decltype(cute::get<2>(TileShape_MNK{}))>());
+  using SmemLayoutAtomB = decltype(detail::ss_smem_selector<
+      GmmaMajorB, ElementBMma, decltype(cute::get<1>(TileShape_MNK{})), decltype(cute::get<2>(TileShape_MNK{}))>());
+
+  static constexpr size_t TensorMapStorage = IsArrayOfPointersGemm ? sizeof(cute::TmaDescriptor) * 2 /* for A and B */ : 0;
+  static constexpr int KernelSmemCarveout = static_cast<int>(TensorMapStorage);
+
+  static constexpr int PipelineStages = detail::compute_stage_count_or_override<detail::sm90_smem_capacity_bytes - KernelSmemCarveout,
+      ElementAMma, ElementBMma, TileShape_MNK>(StageCountType{});
+  using DispatchPolicy = MainloopSm90TmaGmmaWarpSpecializedBlockScalingSubGroupMFP8<PipelineStages, ClusterShape_MNK, KernelScheduleType, ScaleGranularityM>;
+
+  using SmemCopyAtomA = void;
+  using SmemCopyAtomB = void;
+
+  using CollectiveOp = CollectiveMma<
+      DispatchPolicy,
+      TileShape_MNK,
+      ElementA,
+      TagToStrideA_t<GmemLayoutATag>,
+      ElementB,
+      TagToStrideB_t<GmemLayoutBTag>,
+      TiledMma,
+      GmemTiledCopyA,
+      SmemLayoutAtomA,
+      SmemCopyAtomA,
+      cute::identity,
+      GmemTiledCopyB,
+      SmemLayoutAtomB,
+      SmemCopyAtomB,
+      cute::identity
+    >;
+};
+
+
+/////////////////////////////////////////////////////////////////////////////////////////////////
+
+}  // namespace cutlass::gemm::collective
+
+/////////////////////////////////////////////////////////////////////////////////////////////////

sgl-kernel/src/sgl-kernel/csrc/cutlass_extensions/gemm/dispatch_policy.hpp

+// Adapt from https://github.com/vllm-project/vllm/blob/v0.7.1/csrc/cutlass_extensions/gemm/dispatch_policy.hpp
+#pragma once
+
+#include <cutlass/gemm/dispatch_policy.hpp>
+
+namespace cutlass::gemm {
+
+//////////////////////////////////////////////////////////////////////////////
+
+// FP8 related policies (including Blocked Scaled Accumulation)
+//  `ScaleGranularityM` specifies scaling granularity along M, while zero-value
+//  `ScaleGranularityM` indicates that scaling granularity is
+//  `size<0>(TileShape_MNK{})` along M.
+template <int ScaleGranularityM = 0>
+struct KernelTmaWarpSpecializedCooperativeFP8BlockScaledSubGroupMAccum : KernelTmaWarpSpecializedCooperative {};
+
+// n-buffer in smem (Hopper TMA), pipelined with Hopper GMMA and TMA, Warp
+// specialized dynamic schedule For FP8 kernels with Block Scaling
+template <int Stages_, class ClusterShape_ = Shape<_1, _1, _1>, class KernelSchedule = KernelTmaWarpSpecialized,
+          int ScaleGranularityM = 0  // `ScaleGranularityM` specifies scaling granularity along M,
+                                     // while zero-value `ScaleGranularityM` indicates that scaling
+                                     // granularity is `size<0>(TileShape_MNK{})` along M.
+          >
+struct MainloopSm90TmaGmmaWarpSpecializedBlockScalingSubGroupMFP8
+    : MainloopSm90TmaGmmaWarpSpecialized<Stages_, ClusterShape_, KernelSchedule> {
+  static_assert(cute::is_same_v<KernelSchedule,
+                                KernelTmaWarpSpecializedCooperativeFP8BlockScaledSubGroupMAccum<ScaleGranularityM>>,
+                "KernelSchedule must be one of the warp specialized policies");
+};
+
+//////////////////////////////////////////////////////////////////////////////
+
+}  // namespace cutlass::gemm

sgl-kernel/src/sgl-kernel/csrc/fp8_blockwise_gemm_kernel.cu

+#include <ATen/cuda/CUDAContext.h>
+#include <cudaTypedefs.h>
+#include <cutlass/arch/arch.h>
+#include <cutlass/arch/memory.h>
+#include <cutlass/arch/mma.h>
+#include <cutlass/array.h>
+#include <cutlass/cutlass.h>
+#include <cutlass/epilogue/thread/activation.h>
+#include <cutlass/epilogue/thread/linear_combination.h>
+#include <cutlass/epilogue/threadblock/default_thread_map_tensor_op.h>
+#include <cutlass/gemm/device/gemm.h>
+#include <cutlass/gemm/device/gemm_universal_adapter.h>
+#include <cutlass/gemm/gemm.h>
+#include <cutlass/gemm/kernel/default_gemm_universal_with_visitor.h>
+#include <cutlass/gemm/thread/mma.h>
+#include <cutlass/layout/matrix.h>
+#include <cutlass/matrix_coord.h>
+#include <cutlass/numeric_types.h>
+#include <cutlass/tensor_ref.h>
+#include <torch/all.h>
+
+#include <cute/tensor.hpp>
+#include <cutlass/epilogue/collective/collective_builder.hpp>
+#include <cutlass/epilogue/collective/default_epilogue.hpp>
+#include <cutlass/epilogue/threadblock/fusion/visitors.hpp>
+#include <cutlass/gemm/collective/collective_builder.hpp>
+#include <cutlass/gemm/dispatch_policy.hpp>
+#include <cutlass/gemm/kernel/gemm_universal.hpp>
+#include <cutlass/util/packed_stride.hpp>
+
+#include "cutlass_extensions/gemm/collective/collective_builder.hpp"
+#include "cutlass_extensions/gemm/dispatch_policy.hpp"
+#include "utils.h"
+
+using namespace cute;
+
+template <typename OutType, typename TileShape, typename ClusterShape, int ScaleGranularityM = 1>
+void launch_sm90_fp8_blockwise_scaled_mm(torch::Tensor& out, const torch::Tensor& a, const torch::Tensor& b,
+                                         const torch::Tensor& scales_a, const torch::Tensor& scales_b) {
+  using ElementAccumulator = float;
+  using ElementCompute = float;
+  using ElementBlockScale = float;
+
+  using ElementA = cutlass::float_e4m3_t;
+  using LayoutA = cutlass::layout::RowMajor;
+  constexpr int AlignmentA = 128 / cutlass::sizeof_bits<ElementA>::value;
+
+  using ElementB = cutlass::float_e4m3_t;
+  using LayoutB = cutlass::layout::ColumnMajor;
+  constexpr int AlignmentB = 128 / cutlass::sizeof_bits<ElementB>::value;
+
+  using ElementC = void;
+  using LayoutC = cutlass::layout::RowMajor;
+  constexpr int AlignmentC = 128 / cutlass::sizeof_bits<OutType>::value;
+
+  using ElementD = OutType;
+  using LayoutD = cutlass::layout::RowMajor;
+  constexpr int AlignmentD = AlignmentC;
+
+  using ArchTag = cutlass::arch::Sm90;
+  using OperatorClass = cutlass::arch::OpClassTensorOp;
+  using EpilogueSchedule = cutlass::epilogue::TmaWarpSpecializedCooperative;
+  using EpilogueTileType = cutlass::epilogue::collective::EpilogueTileAuto;
+  using StoreEpilogueCompute = typename cutlass::epilogue::fusion::Sm90EVT<cutlass::epilogue::fusion::Sm90AccFetch>;
+
+  using KernelSchedule =
+      cutlass::gemm::KernelTmaWarpSpecializedCooperativeFP8BlockScaledSubGroupMAccum<ScaleGranularityM>;
+  using CollectiveEpilogue = typename cutlass::epilogue::collective::CollectiveBuilder<
+      ArchTag, OperatorClass, TileShape, ClusterShape, EpilogueTileType, ElementAccumulator, ElementCompute, ElementC,
+      LayoutC, AlignmentC, ElementD, LayoutD, AlignmentD, EpilogueSchedule, StoreEpilogueCompute>::CollectiveOp;
+
+  using CollectiveMainloop = typename cutlass::gemm::collective::CollectiveBuilder<
+      ArchTag, OperatorClass, ElementA, LayoutA, AlignmentA, ElementB, LayoutB, AlignmentB, ElementAccumulator,
+      TileShape, ClusterShape,
+      cutlass::gemm::collective::StageCountAutoCarveout<static_cast<int>(
+          sizeof(typename CollectiveEpilogue::SharedStorage))>,
+      KernelSchedule>::CollectiveOp;
+
+  using GemmKernel =
+      cutlass::gemm::kernel::GemmUniversal<Shape<int, int, int, int>,  // Indicates ProblemShape
+                                           CollectiveMainloop, CollectiveEpilogue, cutlass::gemm::PersistentScheduler>;
+  using Gemm = cutlass::gemm::device::GemmUniversalAdapter<GemmKernel>;
+
+  Gemm gemm_op;
+
+  int m = a.size(0);
+  int k = a.size(1);
+  int n = b.size(1);
+
+  auto a_ptr = static_cast<ElementA*>(a.data_ptr());
+  auto b_ptr = static_cast<ElementB*>(b.data_ptr());
+  auto o_ptr = static_cast<ElementD*>(out.data_ptr());
+
+  auto a_s_ptr = static_cast<ElementBlockScale*>(scales_a.data_ptr());
+  auto b_s_ptr = static_cast<ElementBlockScale*>(scales_b.data_ptr());
+
+  using StrideA = typename Gemm::GemmKernel::StrideA;
+  using StrideB = typename Gemm::GemmKernel::StrideB;
+  using StrideC = typename Gemm::GemmKernel::StrideC;
+  using StrideD = typename Gemm::GemmKernel::StrideD;
+
+  StrideA stride_a = cutlass::make_cute_packed_stride(StrideA{}, cute::make_shape(m, k, 1));
+  StrideB stride_b = cutlass::make_cute_packed_stride(StrideB{}, cute::make_shape(n, k, 1));
+  StrideC stride_c;
+  StrideD stride_d = cutlass::make_cute_packed_stride(StrideD{}, cute::make_shape(m, n, 1));
+
+  typename GemmKernel::MainloopArguments mainloop_args{a_ptr, stride_a, b_ptr, stride_b, 4, a_s_ptr, b_s_ptr};
+  typename GemmKernel::EpilogueArguments epilogue_args{{}, nullptr, stride_d, o_ptr, stride_d};
+
+  typename Gemm::Arguments args = {
+      cutlass::gemm::GemmUniversalMode::kGemm,
+      {m, n, k, 1},
+      mainloop_args,
+      epilogue_args,
+  };
+
+  size_t workspace_size = gemm_op.get_workspace_size(args);
+  auto const workspace_options = torch::TensorOptions().dtype(torch::kUInt8).device(a.device());
+  auto workspace = torch::empty(workspace_size, workspace_options);
+  auto stream = at::cuda::getCurrentCUDAStream(a.get_device());
+
+  auto can_implement = gemm_op.can_implement(args);
+  TORCH_CHECK(can_implement == cutlass::Status::kSuccess, cutlassGetStatusString(can_implement))
+
+  auto status = gemm_op.run(args, workspace.data_ptr(), stream);
+  TORCH_CHECK(status == cutlass::Status::kSuccess, cutlassGetStatusString(status))
+}
+
+template <typename OutType>
+void sm90_fp8_blockwise_dispatch_shape(torch::Tensor& out, const torch::Tensor& a, const torch::Tensor& b,
+                                       const torch::Tensor& scales_a, const torch::Tensor& scales_b) {
+  using TileShape = Shape<_128, _128, _128>;
+  using ClusterShape = Shape<_1, _1, _1>;
+  launch_sm90_fp8_blockwise_scaled_mm<OutType, TileShape, ClusterShape>(out, a, b, scales_a, scales_b);
+}
+
+torch::Tensor fp8_blockwise_scaled_mm(const torch::Tensor& mat_a, const torch::Tensor& mat_b,
+                                      const torch::Tensor& scales_a, const torch::Tensor& scales_b,
+                                      const torch::Dtype& out_dtype) {
+  TORCH_CHECK(mat_a.is_cuda(), "mat_a must be a CUDA tensor");
+  TORCH_CHECK(mat_b.is_cuda(), "mat_b must be a CUDA tensor");
+  TORCH_CHECK(mat_a.dim() == 2, "mat_a must be a 2D tensor");
+  TORCH_CHECK(mat_b.dim() == 2, "mat_b must be a 2D tensor");
+  TORCH_CHECK(mat_a.stride(1) == 1, "mat_a must be a row major tensor");
+  TORCH_CHECK(mat_b.stride(0) == 1, "mat_a must be a column major tensor");
+  TORCH_CHECK(mat_a.size(1) == mat_b.size(0), "mat_a and mat_b shapes cannot be multiplied");
+
+  TORCH_CHECK((mat_a.size(1) * mat_a.element_size()) % 16 == 0,
+              "mat_a must be multiple of 16 bytes for memory alignment");
+  TORCH_CHECK((mat_b.size(0) * mat_b.element_size()) % 16 == 0,
+              "mat_b must be multiple of 16 bytes for memory alignment");
+  TORCH_CHECK(mat_a.scalar_type() == torch::kFloat8_e4m3fn, "mat_a must be Float8_e4m3fn");
+  TORCH_CHECK(mat_b.scalar_type() == torch::kFloat8_e4m3fn, "mat_b must be Float8_e4m3fn");
+  TORCH_CHECK(out_dtype == torch::kHalf || out_dtype == torch::kBFloat16, "out_dtype must be Half or BFloat16");
+
+  auto is_contiguous_vector = [](const torch::Tensor& t) {
+    auto t_sizes = t.sizes();
+    return t.is_contiguous() &&
+           (t.dim() == 1 || (t.dim() == 2 && *std::min_element(t_sizes.begin(), t_sizes.end()) == 1));
+  };
+
+  TORCH_CHECK(mat_a.size(0) == scales_a.size(0), "size of scales_a is not matched");
+  TORCH_CHECK(mat_a.size(1) / 128 == scales_a.size(1), "size of scales_a is not matched");
+  TORCH_CHECK(scales_a.stride(0) == 1 || is_contiguous_vector(scales_a), "scales_a must be M major");
+  TORCH_CHECK(mat_b.size(0) / 128 == scales_b.size(0), "size of scales_b is not matched");
+  TORCH_CHECK(mat_b.size(1) / 128 == scales_b.size(1), "size of scales_b is not matched");
+  TORCH_CHECK(scales_b.stride(0) == 1 || is_contiguous_vector(scales_b), "scales_b must be K major");
+  TORCH_CHECK(scales_a.scalar_type() == torch::kFloat32, "scales_a must be Float32");
+  TORCH_CHECK(scales_b.scalar_type() == torch::kFloat32, "scales_b must be Float32");
+
+  torch::Tensor out = torch::empty({mat_a.size(0), mat_b.size(1)}, mat_a.options().dtype(out_dtype));
+  TORCH_CHECK((out.size(1) * out.element_size()) % 16 == 0, "out must be multiple of 16 bytes for memory alignment");
+
+  auto sm_version = getSMVersion();
+
+#if defined(CUTLASS_ARCH_MMA_SM90_SUPPORTED)
+#if defined CUDA_VERSION && CUDA_VERSION >= 12000
+  if (sm_version >= 90) {
+    if (out_dtype == torch::kBFloat16) {
+      sm90_fp8_blockwise_dispatch_shape<cutlass::bfloat16_t>(out, mat_a, mat_b, scales_a, scales_b);
+    } else {
+      sm90_fp8_blockwise_dispatch_shape<cutlass::half_t>(out, mat_a, mat_b, scales_a, scales_b);
+    }
+    return out;
+  }
+#endif
+#endif
+
+  TORCH_CHECK_NOT_IMPLEMENTED(false,
+                              "No implemented fp8_blockwise_scaled_mm for current compute capability: ", sm_version);
+}

sgl-kernel/src/sgl-kernel/include/sgl_kernels_ops.h

@@ -60,6 +60,11 @@ torch::Tensor fp8_scaled_mm(const torch::Tensor& mat_a, const torch::Tensor& mat
                             const torch::Tensor& scales_b, const torch::Dtype& out_dtype,
                             const c10::optional<torch::Tensor>& bias);
 
+// fp8_blockwise_scaled_mm
+torch::Tensor fp8_blockwise_scaled_mm(const torch::Tensor& mat_a, const torch::Tensor& mat_b,
+                                      const torch::Tensor& scales_a, const torch::Tensor& scales_b,
+                                      const torch::Dtype& out_dtype);
+
 // lightning_attention_decode
 void lightning_attention_decode(const torch::Tensor& q, const torch::Tensor& k, const torch::Tensor& v,
                                 const torch::Tensor& past_kv, const torch::Tensor& slope, torch::Tensor output,

sgl-kernel/src/sgl-kernel/ops/__init__.py

@@ -125,6 +125,16 @@ def int8_scaled_mm(mat_a, mat_b, scales_a, scales_b, out_dtype, bias=None):
     )
 
 
+def fp8_blockwise_scaled_mm(mat_a, mat_b, scales_a, scales_b, out_dtype):
+    return torch.ops.sgl_kernels.fp8_blockwise_scaled_mm(
+        mat_a,
+        mat_b,
+        scales_a,
+        scales_b,
+        out_dtype,
+    )
+
+
 def fp8_scaled_mm(mat_a, mat_b, scales_a, scales_b, out_dtype, bias=None):
     return torch.ops.sgl_kernels.fp8_scaled_mm(
         mat_a,

sgl-kernel/src/sgl-kernel/torch_extension.cc

@@ -54,6 +54,12 @@ TORCH_LIBRARY_EXPAND(sgl_kernels, m) {
       "bias) -> Tensor");
   m.impl("fp8_scaled_mm", torch::kCUDA, &fp8_scaled_mm);
 
+  // fp8_blockwise_scaled_mm
+  m.def(
+      "fp8_blockwise_scaled_mm(Tensor mat_a, Tensor mat_b, Tensor scales_a, Tensor scales_b, ScalarType out_dtype) -> "
+      "Tensor");
+  m.impl("fp8_blockwise_scaled_mm", torch::kCUDA, &fp8_blockwise_scaled_mm);
+
   // lightning_attention_decode
   m.def(
       "lightning_attention_decode(Tensor q, Tensor k, Tensor v, Tensor past_kv, Tensor slope, Tensor! output, Tensor! "

sgl-kernel/tests/test_fp8_blockwise_gemm.py

+import unittest
+from typing import Optional, Type
+
+import torch
+from sgl_kernel import fp8_blockwise_scaled_mm
+
+
+def cdiv(a: int, b: int) -> int:
+    """Ceiling division."""
+    return -(a // -b)
+
+
+def scale_shape(shape, group_shape):
+    assert len(shape) == len(group_shape)
+    return tuple(cdiv(shape[i], group_shape[i]) for i in range(len(group_shape)))
+
+
+def baseline_scaled_mm(
+    a: torch.Tensor,
+    b: torch.Tensor,
+    scale_a: torch.Tensor,
+    scale_b: torch.Tensor,
+    out_dtype: Type[torch.dtype],
+    bias: Optional[torch.Tensor] = None,
+) -> torch.Tensor:
+
+    # We treat N-dimensional group scaling as extended numpy-style broadcasting
+    # in numpy simply stretches dimensions with an extent of 1 to match the
+    # the target shape by repeating the data along that dimension (broadcasting)
+    # , we extend these semantics to say if the extent of a dimension in the
+    # source shape is not 1 and does not match the target shape we repeat each
+    # element along that dimension src_shape[dim] // target_shape[dim] times
+    # example if we have:
+    #       a = [[1, 2], and target_shape = (2, 4)
+    #            [3, 4]]
+    # then we would expand a to:
+    #       a = [[1, 1, 2, 2],
+    #            [3, 3, 4, 4]]
+    # NOTE this function this function does not explicitly broadcast dimensions
+    # with an extent of 1, since this can be done implicitly by pytorch
+    def group_broadcast(t, shape):
+        for i, s in enumerate(shape):
+            if t.shape[i] != s and t.shape[i] != 1:
+                assert s % t.shape[i] == 0
+                t = (
+                    t.unsqueeze(i + 1)
+                    .expand(*t.shape[: i + 1], s // t.shape[i], *t.shape[i + 1 :])
+                    .flatten(i, i + 1)
+                )
+        return t
+
+    scale_a = group_broadcast(scale_a, a.shape)
+    scale_b = group_broadcast(scale_b, b.shape)
+
+    output = torch.mm(
+        (scale_a * a.to(dtype=torch.float32)), (scale_b * b.to(dtype=torch.float32))
+    ).to(out_dtype)
+
+    if bias is not None:
+        output = output + bias
+
+    return output
+
+
+class TestFp8Gemm(unittest.TestCase):
+    def _test_accuracy_once(self, M, N, K, out_dtype, device):
+        fp8_info = torch.finfo(torch.float8_e4m3fn)
+        fp8_max, fp8_min = fp8_info.max, fp8_info.min
+
+        a_fp32 = (
+            (torch.rand(M, K, dtype=torch.float32, device=device) - 0.5) * 2 * fp8_max
+        )
+        a_fp8 = a_fp32.clamp(min=fp8_min, max=fp8_max).to(torch.float8_e4m3fn)
+
+        b_fp32 = (
+            (torch.rand(N, K, dtype=torch.float32, device=device) - 0.5) * 2 * fp8_max
+        )
+        b_fp8 = b_fp32.clamp(min=fp8_min, max=fp8_max).to(torch.float8_e4m3fn).t()
+
+        scale_a_group_shape = (1, 128)
+        scale_b_group_shape = (128, 128)
+        scale_a_shape = scale_shape(a_fp8.shape, scale_a_group_shape)
+        scale_b_shape = scale_shape(b_fp8.shape, scale_b_group_shape)
+
+        scale_a = torch.randn(scale_a_shape, device=device, dtype=torch.float32) * 0.001
+        scale_b = torch.randn(scale_b_shape, device=device, dtype=torch.float32) * 0.001
+        scale_a = scale_a.t().contiguous().t()
+        scale_b = scale_b.t().contiguous().t()
+
+        o1 = torch.empty((M, N), device=device, dtype=torch.bfloat16)
+        o = baseline_scaled_mm(a_fp8, b_fp8, scale_a, scale_b, out_dtype)
+        o1 = fp8_blockwise_scaled_mm(a_fp8, b_fp8, scale_a, scale_b, out_dtype)
+
+        rtol = 0.02
+        atol = 1
+        torch.testing.assert_close(o, o1, rtol=rtol, atol=atol)
+        print(f"M={M}, N={N}, K={K}, out_dtype={out_dtype}: OK")
+
+    def test_accuracy(self):
+        Ms = [1, 128, 512, 1024, 4096]
+        Ns = [128, 512, 1024, 4096]
+        Ks = [512, 1024, 4096, 8192, 16384]
+        out_dtypes = [torch.bfloat16, torch.float16]
+        for M in Ms:
+            for N in Ns:
+                for K in Ks:
+                    for out_dtype in out_dtypes:
+                        self._test_accuracy_once(M, N, K, out_dtype, "cuda")
+
+
+if __name__ == "__main__":
+    unittest.main()