ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-02 12:41:26 +00:00
Code Issues Packages Projects Releases Wiki Activity

ck_tile kernel for gemm with groupwise quantized B tensor. (#2663)

Browse Source 
* This change introduces new pipelines with Intrawave scheduler and block gemm primitives that loads the scale tensor to registers to perform dequantization post MFMA on C tensor in registers.

Scale tensor data, BQ is spliced across threads in registers and not stored in LDS.

Current support is for the following combinations, but it should be fairly straightforward to extend support to more formats.

fp8, fp8 -> f32
bf8, bf8 -> f32
fp8, i4 -> f32
bf8, i4 -> f32
Group size can go down to as low as K length of underlying WarpGemm primitive.

* Solve merge conflict

* [CK TILE] Update CHANGELOG.md

---------

Co-authored-by: Vijay Krishnamoorthy <vjkrish@fb.com>
Co-authored-by: ThomasNing <thomas.ning@amd.com>
Co-authored-by: Cong Ma <congma13@amd.com>
This commit is contained in:
Vijay Krish
2025-08-28 23:43:02 -07:00
committed by GitHub
parent 428090f749
commit 4208e28988
20 changed files with 2471 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

16
include/ck_tile/ops/gemm_group_quant/block/block_universal_gemm_as_aquant_bs_cr.hpp
View File

@@ -12,7 +12,7 @@
namespace ck_tile {
template <typename Problem, index_t UnaryOpSize_ = 8>
struct BlockGemmQuantBase
struct BlockGemmAQuantBase
{
using AQDataType = remove_cvref_t<typename Problem::AQDataType>;
using ComputeDataType = remove_cvref_t<typename Problem::ComputeDataType>;
@@ -67,7 +67,7 @@ struct BlockGemmQuantBase
// B is block window on shared memory
// C is block distributed tensor
template <typename Problem_, typename Policy_ = BlockGemmASmemBSmemCRegV1DefaultPolicy>
struct AQuantBlockUniversalGemmAsBsCr : public BlockGemmQuantBase<Problem_>
struct AQuantBlockUniversalGemmAsBsCr : public BlockGemmAQuantBase<Problem_>
{
private:
template <typename PipelineProblem_, typename GemmPolicy_>
@@ -103,13 +103,13 @@ struct AQuantBlockUniversalGemmAsBsCr : public BlockGemmQuantBase<Problem_>
using I1 = number<1>;
static_assert(MWarp == BlockGemmShape::BlockWarps::at(I0{}),
"Error! WarpGemm's MWarp is not consisten with BlockGemmShape!");
"Error! WarpGemm's MWarp is not consistent with BlockGemmShape!");
static_assert(NWarp == BlockGemmShape::BlockWarps::at(I1{}),
"Error! WarpGemm's NWarp is not consisten with BlockGemmShape!");
"Error! WarpGemm's NWarp is not consistent with BlockGemmShape!");
static_assert(WarpGemm::kM == BlockGemmShape::WarpTile::at(I0{}),
"Error! WarpGemm's M is not consisten with BlockGemmShape!");
"Error! WarpGemm's M is not consistent with BlockGemmShape!");
static_assert(WarpGemm::kN == BlockGemmShape::WarpTile::at(I1{}),
"Error! WarpGemm's N is not consisten with BlockGemmShape!");
"Error! WarpGemm's N is not consistent with BlockGemmShape!");
static constexpr index_t MIterPerWarp = MPerBlock / (MWarp * WarpGemm::kM);
static constexpr index_t NIterPerWarp = NPerBlock / (NWarp * WarpGemm::kN);
@@ -170,7 +170,7 @@ struct AQuantBlockUniversalGemmAsBsCr : public BlockGemmQuantBase<Problem_>
using ComputeDataType = remove_cvref_t<typename Traits::ComputeDataType>;
using CDataType = remove_cvref_t<typename Traits::CDataType>;
using Base = BlockGemmQuantBase<Problem_>;
using Base = BlockGemmAQuantBase<Problem_>;
using WarpGemm = remove_cvref_t<typename Traits::WarpGemm>;
@@ -323,7 +323,7 @@ struct AQuantBlockUniversalGemmAsBsCr : public BlockGemmQuantBase<Problem_>
[[maybe_unused]] BSmemBlockWindow& b_block_window)
{
static_assert(std::is_same_v<CDataType, typename CBlockTensor::DataType>,
"The CDataType as defined in traits should be the same as correspoinding "
"The CDataType as defined in traits should be the same as corresponding "
"C block tensor data type!");
constexpr auto warp_size = get_warp_size();

439
include/ck_tile/ops/gemm_group_quant/block/block_universal_gemm_as_bs_bquant_cr.hpp Normal file
View File

@@ -0,0 +1,439 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck_tile/core.hpp"
#include "ck_tile/core/arch/arch.hpp"
#include "ck_tile/ops/gemm/block/block_gemm_asmem_bsmem_creg_v1_default_policy.hpp"
#include "ck_tile/ops/gemm/pipeline/gemm_pipeline_ag_bg_cr_scheduler.hpp"
#include "ck_tile/ops/elementwise.hpp"
namespace ck_tile {
template <typename Problem, index_t UnaryOpSize_ = 8>
struct BlockGemmBQuantBase
{
using BQDataType = remove_cvref_t<typename Problem::BQDataType>;
using ComputeDataType = remove_cvref_t<typename Problem::ComputeDataType>;
static constexpr index_t UnaryOpSize = UnaryOpSize_;
template <typename T>
CK_TILE_DEVICE static float cvt_scale_to_fp32(T scale)
{
float scale_reg_f = 0.f;
if constexpr(std::is_same_v<BQDataType, ck_tile::fp8_t>)
{
scale_reg_f =
ck_tile::element_wise::amd_assembly_fp8_to_fp32(static_cast<uint32_t>(scale));
}
else if constexpr(std::is_same_v<BQDataType, ck_tile::bf8_t>)
{
scale_reg_f =
ck_tile::element_wise::amd_assembly_bf8_to_fp32(static_cast<uint32_t>(scale));
}
else if constexpr(std::is_same_v<BQDataType, float>)
{
scale_reg_f = ck_tile::bit_cast<float>(scale);
}
else
{
static_assert(false, "BQDataType must be float, fp8_t or bf8_t.");
}
return scale_reg_f;
}
// can be inherited from A
template <typename WarpWindow, typename WarpTile>
CK_TILE_DEVICE static void load_interleaved_pk_type(WarpTile& warp_tile,
const WarpWindow& warp_window)
{
const element_wise::PassThroughPack8 elementwise_op{};
static_assert(WarpTile::get_thread_buffer_size() % UnaryOpSize == 0);
constexpr index_t thread_buffer_size = WarpTile::get_thread_buffer_size() / UnaryOpSize;
const auto in_dstr_tensors = load_tile(warp_window);
using ComputeVectorType = ComputeDataType __attribute__((ext_vector_type(UnaryOpSize)));
static_for<0, thread_buffer_size, 1>{}([&](auto i) {
elementwise_op(warp_tile.get_thread_buffer().template get_as<ComputeVectorType>()(i),
in_dstr_tensors.get_thread_buffer().template get_as<pk_int4x4_t>()[i]);
});
}
};
// A is block window on shared memory
// BQ (scale tensor) is block distributed tensor.
// Consecutive kQuantGroupSize elements of B are quantized with a separate scale.
// B is block window on shared memory
// C is block distributed tensor
template <typename Problem_, typename Policy_ = BlockGemmASmemBSmemCRegV1DefaultPolicy>
struct BQuantBlockUniversalGemmAsBsCr : public BlockGemmBQuantBase<Problem_>
{
private:
template <typename PipelineProblem_, typename GemmPolicy_>
struct GemmTraits_
{
using Problem = remove_cvref_t<PipelineProblem_>;
using Policy = remove_cvref_t<GemmPolicy_>;
using ADataType = remove_cvref_t<typename Problem::ADataType>;
using BDataType = remove_cvref_t<typename Problem::BDataType>;
using BQDataType = remove_cvref_t<typename Problem::BQDataType>;
using ComputeDataType = remove_cvref_t<typename Problem::ComputeDataType>;
using CDataType = remove_cvref_t<typename Problem::CDataType>;
using BlockGemmShape = remove_cvref_t<typename Problem::BlockGemmShape>;
static constexpr index_t kQuantGroupSize = Problem::kQuantGroupSize;
static constexpr index_t kBlockSize = Problem::kBlockSize;
static constexpr auto Scheduler = Problem::Scheduler;
// Threadblock GEMM tile size
static constexpr index_t MPerBlock = BlockGemmShape::kM;
static constexpr index_t NPerBlock = BlockGemmShape::kN;
static constexpr index_t KPerBlock = BlockGemmShape::kK;
static constexpr index_t BQPerBlock = KPerBlock / kQuantGroupSize;
static constexpr auto config = Policy::template GetWarpGemmMWarpNWarp<Problem>();
using WarpGemm = remove_cvref_t<decltype(config.template at<0>())>;
// number of warps along M and N for threadblock's GEMM problem size
static constexpr index_t MWarp = config.template at<1>();
static constexpr index_t NWarp = config.template at<2>();
using I0 = number<0>;
using I1 = number<1>;
static_assert(MWarp == BlockGemmShape::BlockWarps::at(I0{}),
"Error! WarpGemm's MWarp is not consistent with BlockGemmShape!");
static_assert(NWarp == BlockGemmShape::BlockWarps::at(I1{}),
"Error! WarpGemm's NWarp is not consistent with BlockGemmShape!");
static_assert(WarpGemm::kM == BlockGemmShape::WarpTile::at(I0{}),
"Error! WarpGemm's M is not consistent with BlockGemmShape!");
static_assert(WarpGemm::kN == BlockGemmShape::WarpTile::at(I1{}),
"Error! WarpGemm's N is not consistent with BlockGemmShape!");
static constexpr index_t MIterPerWarp = MPerBlock / (MWarp * WarpGemm::kM);
static constexpr index_t NIterPerWarp = NPerBlock / (NWarp * WarpGemm::kN);
static constexpr index_t KIterPerWarp = KPerBlock / WarpGemm::kK;
static constexpr index_t QScalesPerBlockRow =
(KPerBlock + kQuantGroupSize - 1) / kQuantGroupSize;
static constexpr index_t QScalesPerWarpGemmRow =
(WarpGemm::kK + kQuantGroupSize - 1) / kQuantGroupSize;
static constexpr index_t KIterPerQScale = KIterPerWarp / QScalesPerBlockRow;
static_assert(kQuantGroupSize % WarpGemm::kK == 0,
"Error! WarpGemm::kK should be a multiple of kQuantGroupSize");
static_assert(QScalesPerWarpGemmRow == 1,
"Error! kQuantGroupSize shouldn't be smaller than WarpGemm::kK");
static_assert(KIterPerWarp % QScalesPerBlockRow == 0,
"Error! KItersPerWarp should be a multiple of QscalesPerBlockRow");
static_assert(KPerBlock / kQuantGroupSize > 0,
"Error! Each row of blockgemm should have a separate scale");
static_assert(MIterPerWarp * MWarp * WarpGemm::kM == MPerBlock,
"Error! Warps should cover all Block tile!");
static_assert(NIterPerWarp * NWarp * WarpGemm::kN == NPerBlock,
"Error! Warps should cover all Block tile!");
// Currently tested combinations (A, B, BQ)
// 1. fp8, fp8, fp32 -> f32
// 2. bf8, bf8, fp32 -> f32
// 3. i4, fp8, (fp8/fp32) -> f32
// 4. i4, bf8, (fp8/fp32) -> f32
static_assert((std::is_same_v<ADataType, fp8_t> || std::is_same_v<ADataType, bf8_t>) &&
(std::is_same_v<BDataType, fp8_t> || std::is_same_v<BDataType, bf8_t> ||
std::is_same_v<BDataType, ck_tile::pk_int4_t>) &&
(std::is_same_v<BQDataType, float> ||
std::is_same_v<BQDataType, ck_tile::fp8_t> ||
std::is_same_v<BQDataType, ck_tile::bf8_t>) &&
(std::is_same_v<ComputeDataType, fp8_t> ||
std::is_same_v<ComputeDataType, bf8_t>) &&
std::is_same_v<CDataType, fp32_t>);
static constexpr index_t InterWaveSchedulingMacClusters = 1;
static constexpr index_t KPack = WarpGemm::kKPerThread;
static constexpr index_t KPerThread = KIterPerWarp * WarpGemm::kKPerThread;
};
public:
using Traits = GemmTraits_<Problem_, Policy_>;
using ADataType = remove_cvref_t<typename Traits::ADataType>;
using BDataType = remove_cvref_t<typename Traits::BDataType>;
using BQDataType = remove_cvref_t<typename Traits::BQDataType>;
using ComputeDataType = remove_cvref_t<typename Traits::ComputeDataType>;
using CDataType = remove_cvref_t<typename Traits::CDataType>;
using Base = BlockGemmBQuantBase<Problem_>;
using WarpGemm = remove_cvref_t<typename Traits::WarpGemm>;
static constexpr index_t KIterPerWarp = Traits::KIterPerWarp;
static constexpr index_t MIterPerWarp = Traits::MIterPerWarp;
static constexpr index_t NIterPerWarp = Traits::NIterPerWarp;
static constexpr index_t MWarp = Traits::MWarp;
static constexpr index_t NWarp = Traits::NWarp;
static constexpr auto Scheduler = Traits::Scheduler;
static constexpr uint8_t kA_cvt_scale = std::is_same_v<ADataType, pk_int4_t> ? 16 : 1;
static constexpr uint8_t kB_cvt_scale = std::is_same_v<BDataType, pk_int4_t> ? 16 : 1;
using AWarpDstr = typename WarpGemm::AWarpDstr;
using BWarpDstr = typename WarpGemm::BWarpDstr;
using CWarpDstr = typename WarpGemm::CWarpDstr;
using AWarpTensor = typename WarpGemm::AWarpTensor;
using BWarpTensor = typename WarpGemm::BWarpTensor;
using CWarpTensor = typename WarpGemm::CWarpTensor;
static_assert(std::is_same_v<typename WarpGemm::CDataType, float>);
static constexpr auto a_warp_y_lengths =
to_sequence(AWarpDstr{}.get_ys_to_d_descriptor().get_lengths());
static constexpr auto b_warp_y_lengths =
to_sequence(BWarpDstr{}.get_ys_to_d_descriptor().get_lengths());
static constexpr auto c_warp_y_lengths =
to_sequence(CWarpDstr{}.get_ys_to_d_descriptor().get_lengths());
static constexpr auto a_warp_y_index_zeros = uniform_sequence_gen_t<AWarpDstr::NDimY, 0>{};
static constexpr auto b_warp_y_index_zeros = uniform_sequence_gen_t<BWarpDstr::NDimY, 0>{};
static constexpr auto c_warp_y_index_zeros = uniform_sequence_gen_t<CWarpDstr::NDimY, 0>{};
static constexpr index_t APackedSize =
ck_tile::numeric_traits<remove_cvref_t<ADataType>>::PackedSize;
static constexpr index_t BPackedSize =
ck_tile::numeric_traits<remove_cvref_t<BDataType>>::PackedSize;
using I0 = number<0>;
using I1 = number<1>;
CK_TILE_DEVICE static constexpr auto MakeABlockDistributionEncode()
{
constexpr index_t KPerThread = Traits::KPerThread;
constexpr index_t NumMacClusters = Traits::InterWaveSchedulingMacClusters;
constexpr index_t KPerInnerLoop =
ck_tile::max(KPerThread / NumMacClusters, WarpGemm::kKPerThread);
constexpr index_t KIterInterwave = KPerInnerLoop / WarpGemm::kKPerThread;
using KIterSeq = std::conditional_t<Scheduler == GemmPipelineScheduler::Interwave,
sequence<KIterInterwave>,
sequence<KIterPerWarp>>;
constexpr auto a_block_outer_dstr_encoding =
tile_distribution_encoding<sequence<NWarp>,
tuple<sequence<MIterPerWarp, MWarp>, KIterSeq>,
tuple<sequence<1, 0>>,
tuple<sequence<1, 0>>,
sequence<1, 2>,
sequence<0, 0>>{};
constexpr auto a_block_dstr_encode = detail::make_embed_tile_distribution_encoding(
a_block_outer_dstr_encoding, typename WarpGemm::AWarpDstrEncoding{});
return a_block_dstr_encode;
}
CK_TILE_DEVICE static constexpr auto MakeBBlockDistributionEncode()
{
constexpr index_t KPerThread = Traits::KPerThread;
constexpr index_t NumMacClusters = Traits::InterWaveSchedulingMacClusters;
constexpr index_t KPerInnerLoop =
ck_tile::max(KPerThread / NumMacClusters, WarpGemm::kKPerThread);
constexpr index_t KIterInterwave = KPerInnerLoop / WarpGemm::kKPerThread;
using KIterSeq = std::conditional_t<Scheduler == GemmPipelineScheduler::Interwave,
sequence<KIterInterwave>,
sequence<KIterPerWarp>>;
constexpr auto b_block_outer_dstr_encoding =
tile_distribution_encoding<sequence<MWarp>,
tuple<sequence<NIterPerWarp, NWarp>, KIterSeq>,
tuple<sequence<0, 1>>,
tuple<sequence<0, 1>>,
sequence<1, 2>,
sequence<0, 0>>{};
constexpr auto b_block_dstr_encode = detail::make_embed_tile_distribution_encoding(
b_block_outer_dstr_encoding, typename WarpGemm::BWarpDstrEncoding{});
return b_block_dstr_encode;
}
private:
template <GemmPipelineScheduler Scheduler, typename GemmTraits>
struct BlockGemmImpl
{
};
template <typename GemmTraits>
struct BlockGemmImpl<GemmPipelineScheduler::Intrawave, GemmTraits>
{
static constexpr auto ALdsTileDistr =
decltype(make_static_tile_distribution(MakeABlockDistributionEncode())){};
static constexpr auto BLdsTileDistr =
decltype(make_static_tile_distribution(MakeBBlockDistributionEncode())){};
using ALdsTile = decltype(make_static_distributed_tensor<ComputeDataType>(ALdsTileDistr));
using BLdsTile = decltype(make_static_distributed_tensor<ComputeDataType>(BLdsTileDistr));
ALdsTile a_warp_tile_;
BLdsTile b_warp_tile_;
template <typename ASmemBlockWindow, typename BSmemBlockWindow>
CK_TILE_DEVICE void LocalPrefetch(const ASmemBlockWindow& a_block_window,
const BSmemBlockWindow& b_block_window)
{
if constexpr(std::is_same_v<ADataType, pk_int4_t>)
{
static_assert(std::is_same_v<ComputeDataType, fp8_t> ||
std::is_same_v<ComputeDataType, bf8_t>);
Base::load_interleaved_pk_type(a_warp_tile_, a_block_window);
}
else
{
load_tile(a_warp_tile_, a_block_window);
}
if constexpr(std::is_same_v<BDataType, pk_int4_t>)
{
static_assert(std::is_same_v<ComputeDataType, fp8_t> ||
std::is_same_v<ComputeDataType, bf8_t>);
Base::load_interleaved_pk_type(b_warp_tile_, b_block_window);
}
else
{
load_tile(b_warp_tile_, b_block_window);
}
}
// C += A * B
template <typename CBlockTensor,
typename BQBlockTensor,
typename ASmemBlockWindow,
typename BSmemBlockWindow>
CK_TILE_DEVICE void operator()(CBlockTensor& c_block_tensor,
BQBlockTensor& bq_block_tensor,
[[maybe_unused]] ASmemBlockWindow& a_block_window,
[[maybe_unused]] BSmemBlockWindow& b_block_window)
{
static_assert(std::is_same_v<CDataType, typename CBlockTensor::DataType>,
"The CDataType as defined in traits should be the same as corresponding "
"C block tensor data type!");
// hot loop:
static_for<0, MIterPerWarp, 1>{}([&](auto mIter) {
static_for<0, NIterPerWarp, 1>{}([&](auto nIter) {
CWarpTensor c_warp_tensor;
static_for<0, Traits::QScalesPerBlockRow, 1>{}([&](auto kQScale) {
static_for<0, Traits::KIterPerQScale, 1>{}([&](auto kIterInQScale) {
constexpr auto kIter = kQScale * Traits::KIterPerQScale + kIterInQScale;
AWarpTensor a_warp_tensor;
a_warp_tensor.get_thread_buffer() =
a_warp_tile_.get_y_sliced_thread_data(
merge_sequences(sequence<mIter, kIter>{}, a_warp_y_index_zeros),
merge_sequences(sequence<1, 1>{}, a_warp_y_lengths));
BWarpTensor b_warp_tensor;
b_warp_tensor.get_thread_buffer() =
b_warp_tile_.get_y_sliced_thread_data(
merge_sequences(sequence<nIter, kIter>{}, b_warp_y_index_zeros),
merge_sequences(sequence<1, 1>{}, b_warp_y_lengths));
if constexpr(kIterInQScale == 0)
{
c_warp_tensor = WarpGemm{}(a_warp_tensor, b_warp_tensor);
}
else
{
WarpGemm{}(c_warp_tensor, a_warp_tensor, b_warp_tensor);
}
});
// Need to multiply bquant with accumulated C
//
// The accumulated C tile has the standard distribution. For example
// lane 0 holds elements [0,0], [1,0], [2,0], [3,0], [8,0], [9,0],
// [10,0], [11,0], [16,0], [17,0], [18,0], [19,0], [24,0], [25,0],
// [26,0], [27,0].
//
// These elements are in different rows, need to get the scale value
// for the corresponding row.
// Based on bquant's tile distribution, it can be inferred which
// lane holds the relevant scale. For example, the scales corresponding
// to the 16 elements held by lane 0 are held by lanes 0, 1, 2, 3, 8, 9,
// 10, 11, 16, 17, 18, 19, 24, 25, 26, 27 respectively.
//
// These scales can be obtained using __builtin_amdgcn_ds_bpermute.
constexpr index_t reg_offset = nIter * Traits::BQPerBlock + kQScale;
constexpr auto tbuf_offset =
number<typename CBlockTensor::ThreadTensorDesc{}.calculate_offset(
merge_sequences(sequence<mIter, nIter>{},
c_warp_y_index_zeros)) /
CBlockTensor::PackedSize>{};
auto& scale_reg = bq_block_tensor.get_thread_buffer()[reg_offset];
float scale_reg_f = Base::cvt_scale_to_fp32(scale_reg);
static_for<0, WarpGemm::kM / 2, 1>{}([&](auto c_row) {
c_block_tensor.get_thread_buffer()[tbuf_offset + c_row] +=
(c_warp_tensor.get_thread_buffer()[c_row] * scale_reg_f *
kA_cvt_scale * kB_cvt_scale);
});
});
});
});
}
};
public:
CK_TILE_DEVICE static constexpr auto MakeCBlockTile()
{
constexpr auto c_block_outer_dstr_encoding = tile_distribution_encoding<
sequence<>,
tuple<sequence<MIterPerWarp, MWarp>, sequence<NIterPerWarp, NWarp>>,
tuple<sequence<1, 2>>,
tuple<sequence<1, 1>>,
sequence<1, 2>,
sequence<0, 0>>{};
constexpr auto c_block_dstr_encode = detail::make_embed_tile_distribution_encoding(
c_block_outer_dstr_encoding, typename WarpGemm::CWarpDstrEncoding{});
constexpr auto c_block_dstr = make_static_tile_distribution(c_block_dstr_encode);
auto c_block_tensor = make_static_distributed_tensor<CDataType>(c_block_dstr);
return c_block_tensor;
}
template <typename ASmemBlockWindow, typename BSmemBlockWindow>
CK_TILE_DEVICE void LocalPrefetch(const ASmemBlockWindow& a_block_window,
const BSmemBlockWindow& b_block_window)
{
block_gemm_impl_.LocalPrefetch(a_block_window, b_block_window);
}
// C += A * B
template <typename CBlockTensor,
typename BQBlockTensor,
typename ASmemBlockWindow,
typename BSmemBlockWindow>
CK_TILE_DEVICE void operator()(CBlockTensor& c_block_tensor,
BQBlockTensor& bq_block_tensor,
const ASmemBlockWindow& a_block_window,
const BSmemBlockWindow& b_block_window)
{
block_gemm_impl_(c_block_tensor, bq_block_tensor, a_block_window, b_block_window);
}
private:
BlockGemmImpl<Scheduler, Traits> block_gemm_impl_{};
};
} // namespace ck_tile