/home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-composable-kernel/checkouts/develop/include/ck_tile/ops/gemm/warp/warp_gemm_dispatcher.hpp Source File

/home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-composable-kernel/checkouts/develop/include/ck_tile/ops/gemm/warp/warp_gemm_dispatcher.hpp Source File#

Composable Kernel: /home/docs/checkouts/readthedocs.org/user_builds/advanced-micro-devices-composable-kernel/checkouts/develop/include/ck_tile/ops/gemm/warp/warp_gemm_dispatcher.hpp Source File
Go to the documentation of this file.
 // SPDX-License-Identifier: MIT
 // Copyright (c) 2018-2025, Advanced Micro Devices, Inc. All rights reserved.
  
 #pragma once
  
 #include "ck_tile/core.hpp"
 #include "ck_tile/ops/gemm/warp/warp_gemm.hpp"
 #include "ck_tile/ops/gemm/warp/warp_wmma_gemm.hpp"
  
 namespace ck_tile {
  
 namespace impl {
 template <typename AType,
           typename BType,
           typename AccType,
           index_t MPerWave,
           index_t NPerWave,
           index_t KPerWave,
           bool TransposeC,
           bool SwizzleA                     = false,
           bool UseStructuredSparsity        = false,
           WGAttrNumAccessEnum AttrNumAccess = WGAttrNumAccessEnum::Single>
 struct WarpGemmDispatcher;
  
 // clang-format off
 // fp16
 // ADataType, BDataType, AccDataType, MPerWave, NPerWave, KPerWave, TransposeC, SwizzleA, UseStructuredSparsity
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32,  8, false> { using Type = WarpGemmMfmaF16F16F32M32N32K8; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32,  8, true>  { using Type = WarpGemmMfmaF16F16F32M32N32K8TransposedCDistribution; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32, 16, false> { using Type = WarpGemmMfmaF16F16F32M32N32K16<>; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32, 16, true>  { using Type = WarpGemmMfmaF16F16F32M32N32K16TransposedCDistribution<>; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32, 16, false, false, false, WGAttrNumAccessEnum::Double> {
     using Type = WarpGemmMfmaF16F16F32M32N32K16<WGAttrNumAccessEnum::Double>; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32, 16, true, false, false, WGAttrNumAccessEnum::Double> {
     using Type = WarpGemmMfmaF16F16F32M32N32K16TransposedCDistribution<WGAttrNumAccessEnum::Double>; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 16, 16, 32, false> { using Type = WarpGemmMfmaF16F16F32M16N16K32<>; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 16, 16, 32, true>  { using Type = WarpGemmMfmaF16F16F32M16N16K32TransposedCDistribution<>; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 16, 16, 32, false, false, false, WGAttrNumAccessEnum::Double> {
     using Type = WarpGemmMfmaF16F16F32M16N16K32<WGAttrNumAccessEnum::Double>; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 16, 16, 32, true, false, false, WGAttrNumAccessEnum::Double> {
     using Type = WarpGemmMfmaF16F16F32M16N16K32TransposedCDistribution<WGAttrNumAccessEnum::Double>; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float,  4, 64, 16, false> { using Type = WarpGemmMfmaF16F16F32M4N64K16; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 64,  4, 16, false> { using Type = WarpGemmMfmaF16F16F32M64N4K16; };
 // WMMA cases
 #if defined(__gfx11__) || defined(__gfx12__)
 template<bool TransposeC> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 16, 16, 16, TransposeC, false> { using Type = WarpGemmWmma_f32_16x16x16_f16_f16<TransposeC>;};
 #else
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 16, 16, 16, false> { using Type = WarpGemmMfmaF16F16F32M16N16K16; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 16, 16, 16, true>  { using Type = WarpGemmMfmaF16F16F32M16N16K16TransposedCDistribution; };
 #endif
  
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32,  8, false, true> { using Type = WarpGemmMfmaF16F16F32M32N32K8SwizzleA; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32, 16, false, true> { using Type = WarpGemmMfmaF16F16F32M32N32K16SwizzleA; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32,  8,  true, true> { using Type = WarpGemmMfmaF16F16F32M32N32K8SwizzleBTransposedCDistribution; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32, 16,  true, true> { using Type = WarpGemmMfmaF16F16F32M32N32K16SwizzleBTransposedCDistribution; };
  
 // fp16 2:4 structural sparsity
 // ADataType, BDataType, AccDataType, MPerWave, NPerWave, KPerWave, TransposeC, SwizzleA, UseStructuredSparsity
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 32, 32, 16, false, false, true> { using Type = WarpGemmSmfmacF16F16F32M32N32K16; };
 template<> struct WarpGemmDispatcher<ck_tile::half_t, ck_tile::half_t, float, 16, 16, 32, false, false, true> { using Type = WarpGemmSmfmacF16F16F32M16N16K32; };
  
 // bf16
 // ADataType, BDataType, AccDataType, MPerWave, NPerWave, KPerWave, TransposeC, SwizzleA, UseStructuredSparsity
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 32, 32,  8, false> { using Type = WarpGemmMfmaBf16Bf16F32M32N32K8; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 32, 32,  8, true>  { using Type = WarpGemmMfmaBf16Bf16F32M32N32K8TransposedCDistribution; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 32, 32, 16, false> { using Type = WarpGemmMfmaBf16Bf16F32M32N32K16<>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 32, 32, 16, true>  { using Type = WarpGemmMfmaBf16Bf16F32M32N32K16TransposedCDistribution<>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 32, 32, 16, false, false, false, WGAttrNumAccessEnum::Double> {
     using Type = WarpGemmMfmaBf16Bf16F32M32N32K16<WGAttrNumAccessEnum::Double>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 32, 32, 16, true, false, false, WGAttrNumAccessEnum::Double> {
     using Type = WarpGemmMfmaBf16Bf16F32M32N32K16TransposedCDistribution<WGAttrNumAccessEnum::Double>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 16, 16, 32, false> { using Type = WarpGemmMfmaBf16Bf16F32M16N16K32<>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 16, 16, 32, true>  { using Type = WarpGemmMfmaBf16Bf16F32M16N16K32TransposedCDistribution<>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 16, 16, 32, false, false, false, WGAttrNumAccessEnum::Double> {
     using Type = WarpGemmMfmaBf16Bf16F32M16N16K32<WGAttrNumAccessEnum::Double>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 16, 16, 32, true, false, false, WGAttrNumAccessEnum::Double> {
     using Type = WarpGemmMfmaBf16Bf16F32M16N16K32TransposedCDistribution<WGAttrNumAccessEnum::Double>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float,  4, 64, 16, false> { using Type = WarpGemmMfmaBf16Bf16F32M4N64K16; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 64,  4, 16, false> { using Type = WarpGemmMfmaBf16Bf16F32M64N4K16; };
 // WMMA cases
 #if defined(__gfx11__) || defined(__gfx12__)
 template<bool TransposeC> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 16, 16, 16, TransposeC, false> { using Type = WarpGemmWmma_f32_16x16x16_bf16_bf16<TransposeC>; };
 #else
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 16, 16, 16, false> { using Type = WarpGemmMfmaBf16Bf16F32M16N16K16; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 16, 16, 16, true> { using Type = WarpGemmMfmaBf16Bf16F32M16N16K16TransposedCDistribution; };
 #endif
  
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 32, 32,  8, false, true> { using Type = WarpGemmMfmaBf16Bf16F32M32N32K8SwizzleA; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 32, 32, 16, false, true> { using Type = WarpGemmMfmaBf16Bf16F32M32N32K16SwizzleA; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 32, 32,  8, true, true> { using Type = WarpGemmMfmaBf16Bf16F32M32N32K8SwizzleBTransposedCDistribution; };
 template<> struct WarpGemmDispatcher<ck_tile::bf16_t, ck_tile::bf16_t, float, 32, 32, 16, true, true> { using Type = WarpGemmMfmaBf16Bf16F32M32N32K16SwizzleBTransposedCDistribution; };
  
 // fp8
 // ADataType, BDataType, AccDataType, MPerWave, NPerWave, KPerWave, TransposeC, SwizzleA, UseStructuredSparsity
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 32, 32,  16, false> { using Type = WarpGemmMfma_f32_32x32x16_fp8_fp8; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 32, 32,  32, false> { using Type = WarpGemmMfma_f32_32x32x32_fp8_fp8; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 16, 16,  32, false> { using Type = WarpGemmMfma_f32_16x16x32_fp8_fp8; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 16, 16,  64, false> { using Type = WarpGemmMfma_f32_16x16x64_fp8_fp8; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 32, 32,  16, true> { using Type = WarpGemmMfma_f32_32x32x16_fp8_fp8_CTransposed; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 16, 16,  32, true> { using Type = WarpGemmMfma_f32_16x16x32_fp8_fp8_CTransposed; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::bf8_t, float, 32, 32,  16, false> { using Type = WarpGemmMfma_f32_32x32x16_fp8_bf8; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::bf8_t, float, 32, 32,  16, true> { using Type = WarpGemmMfma_f32_32x32x16_fp8_bf8_CTransposed; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::fp8_t, float, 32, 32,  16, false> { using Type = WarpGemmMfma_f32_32x32x16_bf8_fp8; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::fp8_t, float, 32, 32,  16, true> { using Type = WarpGemmMfma_f32_32x32x16_bf8_fp8_CTransposed; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 32, 32,  16, false> { using Type = WarpGemmMfma_f32_32x32x16_bf8_bf8; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 32, 32,  32, false> { using Type = WarpGemmMfma_f32_32x32x32_bf8_bf8; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 16, 16,  32, false> { using Type = WarpGemmMfma_f32_16x16x32_bf8_bf8; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 16, 16,  32, true> { using Type = WarpGemmMfma_f32_16x16x32_bf8_bf8_CTransposed; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 16, 16,  64, false> { using Type = WarpGemmMfma_f32_16x16x64_bf8_bf8; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 32, 32,  16, true> { using Type = WarpGemmMfma_f32_32x32x16_bf8_bf8_CTransposed; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 16, 16,  128, false> { using Type = WarpGemmMfma_f32_16x16x128_fp8_fp8<>; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::bf8_t, float, 16, 16,  128, false> { using Type = WarpGemmMfma_f32_16x16x128_fp8_bf8<>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::fp8_t, float, 16, 16,  128, false> { using Type = WarpGemmMfma_f32_16x16x128_bf8_fp8<>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 16, 16,  128, false> { using Type = WarpGemmMfma_f32_16x16x128_bf8_bf8<>; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 16, 16,  128, true> { using Type = WarpGemmMfma_f32_16x16x128_fp8_fp8_CTransposed<>; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::bf8_t, float, 16, 16,  128, true> { using Type = WarpGemmMfma_f32_16x16x128_fp8_bf8_CTransposed<>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::fp8_t, float, 16, 16,  128, true> { using Type = WarpGemmMfma_f32_16x16x128_bf8_fp8_CTransposed<>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 16, 16,  128, true> { using Type = WarpGemmMfma_f32_16x16x128_bf8_bf8_CTransposed<>; };
  
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 32, 32,  64, false> { using Type = WarpGemmMfma_f32_32x32x64_fp8_fp8<>; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::bf8_t, float, 32, 32,  64, false> { using Type = WarpGemmMfma_f32_32x32x64_fp8_bf8<>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::fp8_t, float, 32, 32,  64, false> { using Type = WarpGemmMfma_f32_32x32x64_bf8_fp8<>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 32, 32,  64, false> { using Type = WarpGemmMfma_f32_32x32x64_bf8_bf8<>; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 32, 32,  64, false, false, false, WGAttrNumAccessEnum::Quad> {
     using Type = WarpGemmMfma_f32_32x32x64_fp8_fp8<WGAttrNumAccessEnum::Quad>; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::bf8_t, float, 32, 32,  64, false, false, false, WGAttrNumAccessEnum::Quad> {
     using Type = WarpGemmMfma_f32_32x32x64_fp8_bf8<WGAttrNumAccessEnum::Quad>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::fp8_t, float, 32, 32,  64, false, false, false, WGAttrNumAccessEnum::Quad> {
     using Type = WarpGemmMfma_f32_32x32x64_bf8_fp8<WGAttrNumAccessEnum::Quad>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 32, 32,  64, false, false, false, WGAttrNumAccessEnum::Quad> {
     using Type = WarpGemmMfma_f32_32x32x64_bf8_bf8<WGAttrNumAccessEnum::Quad>; };
  
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 16, 16,  128, false, false, false, WGAttrNumAccessEnum::Quad> {
     using Type = WarpGemmMfma_f32_16x16x128_fp8_fp8<WGAttrNumAccessEnum::Quad>; };
 template<> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::bf8_t, float, 16, 16,  128, false, false, false, WGAttrNumAccessEnum::Quad> {
     using Type = WarpGemmMfma_f32_16x16x128_fp8_bf8<WGAttrNumAccessEnum::Quad>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::fp8_t, float, 16, 16,  128, false, false, false, WGAttrNumAccessEnum::Quad> {
     using Type = WarpGemmMfma_f32_16x16x128_bf8_fp8<WGAttrNumAccessEnum::Quad>; };
 template<> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 16, 16,  128, false, false, false, WGAttrNumAccessEnum::Quad> {
     using Type = WarpGemmMfma_f32_16x16x128_bf8_bf8<WGAttrNumAccessEnum::Quad>; };
 //WMMA cases
 template<bool TransposeC> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::fp8_t, float, 16, 16,  16, TransposeC, false> { using Type =WarpGemmWmma_f32_16x16x16_f8_f8<TransposeC>; };
 template<bool TransposeC> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::bf8_t, float, 16, 16,  16, TransposeC, false> { using Type =WarpGemmWmma_f32_16x16x16_bf8_bf8<TransposeC>; };
 template<bool TransposeC> struct WarpGemmDispatcher<ck_tile::fp8_t, ck_tile::bf8_t, float, 16, 16,  16, TransposeC, false> { using Type =WarpGemmWmma_f32_16x16x16_f8_bf8<TransposeC>; };
 template<bool TransposeC> struct WarpGemmDispatcher<ck_tile::bf8_t, ck_tile::fp8_t, float, 16, 16,  16, TransposeC, false> { using Type =WarpGemmWmma_f32_16x16x16_bf8_f8<TransposeC>; };
  
 // int8
 // ADataType, BDataType, AccDataType, MPerWave, NPerWave, KPerWave, TransposeC, SwizzleA, UseStructuredSparsity
 template<> struct WarpGemmDispatcher<ck_tile::int8_t, ck_tile::int8_t, ck_tile::int32_t, 32, 32,  16, false> { using Type = WarpGemmMfma_i32_32x32x16_i8_i8; };
 template<> struct WarpGemmDispatcher<ck_tile::int8_t, ck_tile::int8_t, ck_tile::int32_t, 32, 32,  16, true>  { using Type = WarpGemmMfma_i32_32x32x16_i8_i8_CTransposed; };
 template<> struct WarpGemmDispatcher<ck_tile::int8_t, ck_tile::int8_t, ck_tile::int32_t, 16, 16,  32, false> { using Type = WarpGemmMfma_i32_16x16x32_i8_i8; };
 template<> struct WarpGemmDispatcher<ck_tile::int8_t, ck_tile::int8_t, ck_tile::int32_t, 16, 16,  32, true>  { using Type = WarpGemmMfma_i32_16x16x32_i8_i8_CTransposed; };
 // WMMA cases
 template<bool TransposeC> struct WarpGemmDispatcher<ck_tile::int8_t, ck_tile::int8_t, int32_t, 16, 16, 16, TransposeC, false> { using Type = WarpGemmWmma_i32_16x16x16_i8_i8<TransposeC>;};
  
 // clang-format on
 } // namespace impl
  
 template <typename AType,
           typename BType,
           typename AccType,
           index_t MPerWave,
           index_t NPerWave,
           index_t KPerWave,
           bool TransposeC,
           bool SwizzleA                     = false,
           bool UseStructuredSparsity        = false,
           WGAttrNumAccessEnum AttrNumAccess = WGAttrNumAccessEnum::Single>
 using WarpGemmDispatcher = typename impl::WarpGemmDispatcher<AType,
                                                              BType,
                                                              AccType,
                                                              MPerWave,
                                                              NPerWave,
                                                              KPerWave,
                                                              TransposeC,
                                                              SwizzleA,
                                                              UseStructuredSparsity,
                                                              AttrNumAccess>::Type;
  
 } // namespace ck_tile