Skip to content

GitLab

Commit b560db68 authored by ltqin's avatar ltqin
Browse files

add v2r3r1

parent fd3d907a
Hide whitespace changes
Inline Side-by-side
Showing with 678 additions and 0 deletions
composable_kernel/include/tensor_operation/gridwise_gemm_xdlops_v2r3r1.hpp 0 → 100644
View file @ b560db68
#ifndef CK_GRIDWISE_GEMM_XDLOPS_V2R3R1_HPP
#define CK_GRIDWISE_GEMM_XDLOPS_V2R3R1_HPP
#include "common_header.hpp"
#include "multi_index_transform_helper.hpp"
#include "tensor_descriptor.hpp"
#include "tensor_descriptor_helper.hpp"
#include "blockwise_gemm_xdlops.hpp"
#include "blockwise_tensor_slice_transfer.hpp"
#include "threadwise_tensor_slice_transfer.hpp"
#include "threadwise_tensor_slice_set.hpp"
namespace ck {
#if CK_EXPERIMENTAL_PASS_TENSOR_DESCRIPTOR_BY_VALUE
template <typename GridwiseGemm,
typename FloatAB,
typename FloatC,
typename AGridDesc_K0_M_K1,
typename BGridDesc_K0_N_K1,
typename CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
typename Block2CTileMap,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_gemm_xdlops_v2r3r1(
const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
FloatC* __restrict__ p_c_grid,
const AGridDesc_K0_M_K1 a_grid_desc_k0_m_k1,
const BGridDesc_K0_N_K1 b_grid_desc_k0_n_k1,
const CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2 c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const CElementwiseOperation c_element_op,
const Block2CTileMap block_2_ctile_map)
{
constexpr index_t shared_block_size =
GridwiseGemm::GetSharedMemoryNumberOfByte() / sizeof(FloatAB);
__shared__ FloatAB p_shared_block[shared_block_size];
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid,
p_b_grid,
p_c_grid,
p_shared_block,
a_grid_desc_k0_m_k1,
b_grid_desc_k0_n_k1,
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
a_element_op,
b_element_op,
c_element_op,
block_2_ctile_map);
}
#elif CK_EXPERIMENTAL_PASS_TENSOR_DESCRIPTOR_BY_VOID_POINTER
template <typename GridwiseGemm,
typename FloatAB,
typename FloatC,
typename AGridDesc_K0_M_K1,
typename BGridDesc_K0_N_K1,
typename CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
typename Block2CTileMap>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_gemm_xdlops_v2r3r1(const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
FloatC* __restrict__ p_c_grid,
const void CONSTANT* p_a_grid_desc_k0_m_k1,
const void CONSTANT* p_b_grid_desc_k0_n_k1,
const void CONSTANT* p_c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
const void CONSTANT* p_a_element_op,
const void CONSTANT* p_b_element_op,
const void CONSTANT* p_c_element_op,
const void CONSTANT* p_block_2_ctile_map)
{
constexpr index_t shared_block_size =
GridwiseGemm::GetSharedMemoryNumberOfByte() / sizeof(FloatAB);
const auto a_grid_desc_k0_m_k1 = *reinterpret_cast<const AGridDesc_K0_M_K1*>(
cast_pointer_to_generic_address_space(p_a_grid_desc_k0_m_k1));
const auto b_grid_desc_k0_n_k1 = *reinterpret_cast<const BGridDesc_K0_N_K1*>(
cast_pointer_to_generic_address_space(p_b_grid_desc_k0_n_k1));
const auto c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2 =
*reinterpret_cast<const CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2*>(
cast_pointer_to_generic_address_space(p_c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2));
const auto block_2_ctile_map = *reinterpret_cast<const Block2CTileMap*>(
cast_pointer_to_generic_address_space(p_block_2_ctile_map));
const auto a_element_op = *reinterpret_cast<const AElementwiseOperation*>(
cast_pointer_to_generic_address_space(p_a_element_op));
const auto b_element_op = *reinterpret_cast<const BElementwiseOperation*>(
cast_pointer_to_generic_address_space(p_b_element_op));
const auto c_element_op = *reinterpret_cast<const CElementwiseOperation*>(
cast_pointer_to_generic_address_space(p_c_element_op));
__shared__ FloatAB p_shared_block[shared_block_size];
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid,
p_b_grid,
p_c_grid,
p_shared_block,
a_grid_desc_k0_m_k1,
b_grid_desc_k0_n_k1,
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
a_element_op,
b_element_op,
c_element_op,
block_2_ctile_map);
}
#endif
template <index_t BlockSize,
typename FloatAB,
typename FloatAcc,
typename FloatC,
InMemoryDataOperationEnum_t CGlobalMemoryDataOperation,
typename AGridDesc_K0_M_K1,
typename BGridDesc_K0_N_K1,
typename CGridDesc_M_N,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename CElementwiseOperation,
index_t MPerBlock,
index_t NPerBlock,
index_t K0PerBlock,
index_t MPerXDL,
index_t NPerXDL,
index_t K1Value,
index_t MRepeat,
index_t NRepeat,
typename ABlockTransferThreadSliceLengths_K0_M_K1,
typename ABlockTransferThreadClusterLengths_K0_M_K1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_K1,
bool AThreadTransferSrcResetCoordinateAfterRun,
typename BBlockTransferThreadSliceLengths_K0_N_K1,
typename BBlockTransferThreadClusterLengths_K0_N_K1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_K1,
bool BThreadTransferSrcResetCoordinateAfterRun,
typename CThreadTransferSrcDstAccessOrder,
index_t CThreadTransferSrcDstVectorDim,
index_t CThreadTransferDstScalarPerVector,
typename AGridStepHacks,
typename BGridStepHacks,
typename CGridStepHacks,
typename AGridMoveSliceWindowStepHacks,
typename BGridMoveSliceWindowStepHacks,
bool CAccessOrderMRepeatNRepeat,
bool ABlockLdsExtraM,
bool BBlockLdsExtraN>
struct GridwiseGemm_k0mk1_k0nk1_mn_xdlops_v2r3r1
{
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
static constexpr auto I3 = Number<3>{};
static constexpr auto I4 = Number<4>{};
static constexpr auto I5 = Number<5>{};
static constexpr auto I6 = Number<6>{};
static constexpr auto I7 = Number<7>{};
// K1 should be Number<...>
static constexpr auto K1 = Number<K1Value>{};
__host__ __device__ static constexpr index_t GetSharedMemoryNumberOfByte()
{
constexpr auto max_lds_align = K1;
// A matrix in LDS memory, dst of blockwise copy
constexpr auto a_block_desc_k0_m_k1 = [&]() {
if constexpr(ABlockLdsExtraM)
{
return make_naive_tensor_descriptor(
make_tuple(Number<K0PerBlock>{}, Number<MPerBlock>{}, K1),
make_tuple(Number<MPerBlock + 1>{} * K1, K1, I1));
}
else
{
return make_naive_tensor_descriptor_aligned(
make_tuple(Number<K0PerBlock>{}, Number<MPerBlock>{}, K1), max_lds_align);
}
}();
// B matrix in LDS memory, dst of blockwise copy
constexpr auto b_block_desc_k0_n_k1 = [&]() {
if constexpr(BBlockLdsExtraN)
{
return make_naive_tensor_descriptor(
make_tuple(Number<K0PerBlock>{}, Number<NPerBlock>{}, K1),
make_tuple(Number<NPerBlock + 1>{} * K1, K1, I1));
}
else
{
return make_naive_tensor_descriptor_aligned(
make_tuple(Number<K0PerBlock>{}, Number<NPerBlock>{}, K1), max_lds_align);
}
}();
// LDS allocation for A and B: be careful of alignment
constexpr auto a_block_space_size =
math::integer_least_multiple(a_block_desc_k0_m_k1.GetElementSpaceSize(), max_lds_align);
constexpr auto b_block_space_size =
math::integer_least_multiple(b_block_desc_k0_n_k1.GetElementSpaceSize(), max_lds_align);
return (a_block_space_size + b_block_space_size) * sizeof(FloatAB);
}
// block_id to matrix tile idx (m0, n0) mapping are controlled by {M01, N01}
__host__ __device__ static constexpr bool
CheckValidity(const AGridDesc_K0_M_K1& a_grid_desc_k0_m_k1,
const BGridDesc_K0_N_K1& b_grid_desc_k0_n_k1,
const CGridDesc_M_N& c_grid_desc_m_n,
index_t M01,
index_t N01)
{
static_assert(is_known_at_compile_time<remove_cv_t<decltype(K1)>>::value,
"wrong! K1 need to be known at compile-time");
static_assert((MPerBlock % (MPerXDL * MRepeat) == 0) &&
(NPerBlock % (NRepeat * NPerXDL)) == 0,
"Invalid tuning param!");
const auto M = a_grid_desc_k0_m_k1.GetLength(I1);
const auto N = b_grid_desc_k0_n_k1.GetLength(I1);
const auto K0 = a_grid_desc_k0_m_k1.GetLength(I0);
if(!(M == c_grid_desc_m_n.GetLength(I0) && N == c_grid_desc_m_n.GetLength(I1) &&
K0 == b_grid_desc_k0_n_k1.GetLength(I0) && K1 == a_grid_desc_k0_m_k1.GetLength(I2) &&
K1 == b_grid_desc_k0_n_k1.GetLength(I2)))
return false;
if(!(M % MPerBlock == 0 && N % NPerBlock == 0 && K0 % K0PerBlock == 0))
return false;
// check M01, N01
constexpr auto M1 = Number<MPerBlock>{};
constexpr auto N1 = Number<NPerBlock>{};
const auto M0 = M / M1;
const auto N0 = N / N1;
if(!(M0 % M01 == 0 && N0 % N01 == 0))
return false;
// TODO: also check validity of all components (blockwise-copy, threadwise-copy, etc)
return true;
}
__host__ __device__ static constexpr index_t
CalculateGridSize(const CGridDesc_M_N& c_grid_desc_m_n)
{
const auto M = c_grid_desc_m_n.GetLength(I0);
const auto N = c_grid_desc_m_n.GetLength(I1);
const index_t grid_size = (M / MPerBlock) * (N / NPerBlock);
return grid_size;
}
__host__ __device__ static constexpr bool CalculateHasMainK0BlockLoop(index_t K0)
{
const bool has_main_k0_block_loop = (K0 / K0PerBlock) > 1;
return has_main_k0_block_loop;
}
__host__ __device__ static constexpr auto
MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(const CGridDesc_M_N& c_grid_desc_m_n)
{
constexpr auto max_lds_align = K1;
// A matrix in LDS memory, dst of blockwise copy
constexpr auto a_block_desc_k0_m_k1 = [&]() {
if constexpr(ABlockLdsExtraM)
{
return make_naive_tensor_descriptor(
make_tuple(Number<K0PerBlock>{}, Number<MPerBlock>{}, K1),
make_tuple(Number<MPerBlock + 1>{} * K1, K1, I1));
}
else
{
return make_naive_tensor_descriptor_aligned(
make_tuple(Number<K0PerBlock>{}, Number<MPerBlock>{}, K1), max_lds_align);
}
}();
// B matrix in LDS memory, dst of blockwise copy
constexpr auto b_block_desc_k0_n_k1 = [&]() {
if constexpr(BBlockLdsExtraN)
{
return make_naive_tensor_descriptor(
make_tuple(Number<K0PerBlock>{}, Number<NPerBlock>{}, K1),
make_tuple(Number<NPerBlock + 1>{} * K1, K1, I1));
}
else
{
return make_naive_tensor_descriptor_aligned(
make_tuple(Number<K0PerBlock>{}, Number<NPerBlock>{}, K1), max_lds_align);
}
}();
using BlockwiseGemm =
BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1<BlockSize,
FloatAB,
FloatAcc,
decltype(a_block_desc_k0_m_k1),
decltype(b_block_desc_k0_n_k1),
MPerXDL,
NPerXDL,
MRepeat,
NRepeat,
K1>;
return BlockwiseGemm::MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(c_grid_desc_m_n);
}
// return block_id to C matrix tile idx (m0, n0) mapping
__host__ __device__ static constexpr auto
MakeBlock2CTileMap(const CGridDesc_M_N& c_grid_desc_m_n, index_t M01, index_t N01)
{
const auto M = c_grid_desc_m_n.GetLength(I0);
const auto N = c_grid_desc_m_n.GetLength(I1);
constexpr auto M1 = Number<MPerBlock>{};
constexpr auto N1 = Number<NPerBlock>{};
const auto M0 = M / M1;
const auto N0 = N / N1;
const auto M00 = M0 / M01;
const auto N00 = N0 / N01;
const auto m00_m01_n00_n01_to_m0_n0_block_cluster_adaptor =
make_single_stage_tensor_adaptor(
make_tuple(make_unmerge_transform(make_tuple(M00, M01)),
make_unmerge_transform(make_tuple(N00, N01))),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1, 3>{}));
const auto c_blockid_to_m00_m01_n00_n01_block_cluster_adaptor =
make_single_stage_tensor_adaptor(
make_tuple(make_merge_transform(make_tuple(M00, N00, M01, N01))),
make_tuple(Sequence<0, 1, 2, 3>{}),
make_tuple(Sequence<0>{}));
const auto c_blockid_to_m0_n0_block_cluster_adaptor =
chain_tensor_adaptors(m00_m01_n00_n01_to_m0_n0_block_cluster_adaptor,
c_blockid_to_m00_m01_n00_n01_block_cluster_adaptor);
return c_blockid_to_m0_n0_block_cluster_adaptor;
}
using CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2 =
decltype(MakeCGridDescriptor_M0_N0_M1_N1_M2_M3_M4_N2(CGridDesc_M_N{}));
using Block2CTileMap = decltype(MakeBlock2CTileMap(CGridDesc_M_N{}, 1, 1));
template <bool HasMainKBlockLoop>
__device__ static void
Run(const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
FloatC* __restrict__ p_c_grid,
FloatAB* __restrict__ p_shared_block,
const AGridDesc_K0_M_K1& a_grid_desc_k0_m_k1,
const BGridDesc_K0_N_K1& b_grid_desc_k0_n_k1,
const CGridDesc_M0_N0_M1_N1_M2_M3_M4_N2& c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
const AElementwiseOperation& a_element_op,
const BElementwiseOperation& b_element_op,
const CElementwiseOperation& c_element_op,
const Block2CTileMap& block_2_ctile_map)
{
const auto a_grid_buf = make_dynamic_buffer<AddressSpaceEnum_t::Global>(
p_a_grid, a_grid_desc_k0_m_k1.GetElementSpaceSize());
const auto b_grid_buf = make_dynamic_buffer<AddressSpaceEnum_t::Global>(
p_b_grid, b_grid_desc_k0_n_k1.GetElementSpaceSize());
auto c_grid_buf = make_dynamic_buffer<AddressSpaceEnum_t::Global>(
p_c_grid, c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2.GetElementSpaceSize());
const auto K0 = a_grid_desc_k0_m_k1.GetLength(I0);
// divide block work by [M, N]
const auto block_work_idx =
block_2_ctile_map.CalculateBottomIndex(make_multi_index(get_block_1d_id()));
// HACK: this force m/n_block_data_idx_on_grid into SGPR
const index_t m_block_data_idx_on_grid =
__builtin_amdgcn_readfirstlane(block_work_idx[I0] * MPerBlock);
const index_t n_block_data_idx_on_grid =
__builtin_amdgcn_readfirstlane(block_work_idx[I1] * NPerBlock);
// lds max alignment
constexpr auto max_lds_align = K1;
// A matrix in LDS memory, dst of blockwise copy
constexpr auto a_block_desc_k0_m_k1 = [&]() {
if constexpr(ABlockLdsExtraM)
{
return make_naive_tensor_descriptor(
make_tuple(Number<K0PerBlock>{}, Number<MPerBlock>{}, K1),
make_tuple(Number<MPerBlock + 1>{} * K1, K1, I1));
}
else
{
return make_naive_tensor_descriptor_aligned(
make_tuple(Number<K0PerBlock>{}, Number<MPerBlock>{}, K1), max_lds_align);
}
}();
// B matrix in LDS memory, dst of blockwise copy
constexpr auto b_block_desc_k0_n_k1 = [&]() {
if constexpr(BBlockLdsExtraN)
{
return make_naive_tensor_descriptor(
make_tuple(Number<K0PerBlock>{}, Number<NPerBlock>{}, K1),
make_tuple(Number<NPerBlock + 1>{} * K1, K1, I1));
}
else
{
return make_naive_tensor_descriptor_aligned(
make_tuple(Number<K0PerBlock>{}, Number<NPerBlock>{}, K1), max_lds_align);
}
}();
// A matrix blockwise copy
auto a_blockwise_copy =
BlockwiseTensorSliceTransfer_v4<BlockSize,
AElementwiseOperation,
InMemoryDataOperationEnum_t::Set,
Sequence<K0PerBlock, MPerBlock, K1>,
ABlockTransferThreadSliceLengths_K0_M_K1,
ABlockTransferThreadClusterLengths_K0_M_K1,
ABlockTransferThreadClusterArrangeOrder,
FloatAB,
FloatAB,
decltype(a_grid_desc_k0_m_k1),
decltype(a_block_desc_k0_m_k1),
ABlockTransferSrcAccessOrder,
Sequence<1, 0, 2>,
ABlockTransferSrcVectorDim,
2,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_K1,
1,
1,
AThreadTransferSrcResetCoordinateAfterRun,
true>(a_grid_desc_k0_m_k1,
make_multi_index(0, m_block_data_idx_on_grid, 0),
a_block_desc_k0_m_k1,
make_multi_index(0, 0, 0),
a_element_op);
// B matrix blockwise copy
auto b_blockwise_copy =
BlockwiseTensorSliceTransfer_v4<BlockSize,
BElementwiseOperation,
InMemoryDataOperationEnum_t::Set,
Sequence<K0PerBlock, NPerBlock, K1>,
BBlockTransferThreadSliceLengths_K0_N_K1,
BBlockTransferThreadClusterLengths_K0_N_K1,
BBlockTransferThreadClusterArrangeOrder,
FloatAB,
FloatAB,
decltype(b_grid_desc_k0_n_k1),
decltype(b_block_desc_k0_n_k1),
BBlockTransferSrcAccessOrder,
Sequence<1, 0, 2>,
BBlockTransferSrcVectorDim,
2,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_K1,
1,
1,
BThreadTransferSrcResetCoordinateAfterRun,
true>(b_grid_desc_k0_n_k1,
make_multi_index(0, n_block_data_idx_on_grid, 0),
b_block_desc_k0_n_k1,
make_multi_index(0, 0, 0),
b_element_op);
// GEMM definition
// c_mtx += transpose(a_mtx) * b_mtx
// a_mtx[K0PerBlock, MPerBlock] is in LDS
// b_mtx[K0PerBlock, NPerBlock] is in LDS
// c_mtx[MPerBlock, NPerBlock] is distributed among threads, and saved in
// register
// sanity check
auto blockwise_gemm =
BlockwiseGemmXdlops_k0mk1_k0nk1_m0n0m1n1m2m3m4n2_v1<BlockSize,
FloatAB,
FloatAcc,
decltype(a_block_desc_k0_m_k1),
decltype(b_block_desc_k0_n_k1),
MPerXDL,
NPerXDL,
MRepeat,
NRepeat,
K1>{};
auto c_thread_buf = blockwise_gemm.GetCThreadBuffer();
// LDS allocation for A and B: be careful of alignment
constexpr auto a_block_space_size =
math::integer_least_multiple(a_block_desc_k0_m_k1.GetElementSpaceSize(), max_lds_align);
FloatAB* p_a_block = p_shared_block;
FloatAB* p_b_block = p_shared_block + a_block_space_size;
constexpr auto a_block_slice_copy_step = make_multi_index(K0PerBlock, 0, 0);
constexpr auto b_block_slice_copy_step = make_multi_index(K0PerBlock, 0, 0);
// hack to control index calculation when iterating over A and B matrix for threadwise copy
constexpr auto a_k0_m_k1_grid_step_hacks = AGridStepHacks{};
constexpr auto b_k0_n_k1_grid_step_hacks = BGridStepHacks{};
// hack to control index calculation when move slice window for A and B matrix for
// threadwise copy
constexpr auto a_k0_m_k1_grid_move_slice_window_step_hack = AGridMoveSliceWindowStepHacks{};
constexpr auto b_k0_n_k1_grid_move_slice_window_step_hack = BGridMoveSliceWindowStepHacks{};
auto a_block_buf = make_dynamic_buffer<AddressSpaceEnum_t::Lds>(
p_a_block, a_block_desc_k0_m_k1.GetElementSpaceSize());
auto b_block_buf = make_dynamic_buffer<AddressSpaceEnum_t::Lds>(
p_b_block, b_block_desc_k0_n_k1.GetElementSpaceSize());
// preload data into LDS
{
a_blockwise_copy.RunRead(a_grid_desc_k0_m_k1, a_grid_buf, a_k0_m_k1_grid_step_hacks);
b_blockwise_copy.RunRead(b_grid_desc_k0_n_k1, b_grid_buf, b_k0_n_k1_grid_step_hacks);
a_blockwise_copy.RunWrite(a_block_desc_k0_m_k1, a_block_buf);
b_blockwise_copy.RunWrite(b_block_desc_k0_n_k1, b_block_buf);
}
// main body
index_t k0_block_data_begin = 0;
c_thread_buf.Clear();
if constexpr(HasMainKBlockLoop)
{
do
{
a_blockwise_copy.MoveSrcSliceWindow(a_grid_desc_k0_m_k1,
a_block_slice_copy_step,
a_k0_m_k1_grid_move_slice_window_step_hack);
b_blockwise_copy.MoveSrcSliceWindow(b_grid_desc_k0_n_k1,
b_block_slice_copy_step,
b_k0_n_k1_grid_move_slice_window_step_hack);
a_blockwise_copy.RunRead(
a_grid_desc_k0_m_k1, a_grid_buf, a_k0_m_k1_grid_step_hacks);
block_sync_lds();
b_blockwise_copy.RunRead(
b_grid_desc_k0_n_k1, b_grid_buf, b_k0_n_k1_grid_step_hacks);
blockwise_gemm.Run(a_block_buf, b_block_buf, c_thread_buf);
block_sync_lds();
a_blockwise_copy.RunWrite(a_block_desc_k0_m_k1, a_block_buf);
b_blockwise_copy.RunWrite(b_block_desc_k0_n_k1, b_block_buf);
k0_block_data_begin += K0PerBlock;
} while(k0_block_data_begin < (K0 - K0PerBlock));
}
// tail
{
block_sync_lds();
blockwise_gemm.Run(a_block_buf, b_block_buf, c_thread_buf);
}
// output: register to global memory
{
constexpr auto c_thread_desc_m0_n0_m1_n1_m2_m3_m4_n2 =
blockwise_gemm.GetCThreadDescriptor_M0_N0_M1_N1_M2_M3_M4_N2();
constexpr auto c_block_desc_m0_n0_m1_n1_m2_m3_m4_n2 =
blockwise_gemm.GetCBlockDescriptor_M0_N0_M1_N1_M2_M3_M4_N2();
constexpr auto M0 = c_block_desc_m0_n0_m1_n1_m2_m3_m4_n2.GetLength(I0);
constexpr auto N0 = c_block_desc_m0_n0_m1_n1_m2_m3_m4_n2.GetLength(I1);
constexpr auto M1 = c_block_desc_m0_n0_m1_n1_m2_m3_m4_n2.GetLength(I2);
constexpr auto N1 = c_block_desc_m0_n0_m1_n1_m2_m3_m4_n2.GetLength(I3);
constexpr auto M2 = c_block_desc_m0_n0_m1_n1_m2_m3_m4_n2.GetLength(I4);
constexpr auto M3 = c_block_desc_m0_n0_m1_n1_m2_m3_m4_n2.GetLength(I5);
constexpr auto M4 = c_block_desc_m0_n0_m1_n1_m2_m3_m4_n2.GetLength(I6);
constexpr auto N2 = c_block_desc_m0_n0_m1_n1_m2_m3_m4_n2.GetLength(I7);
// calculate origin of thread output tensor on global memory
// blockwise GEMM c matrix starting index
const auto c_thread_mtx_on_block =
blockwise_gemm.CalculateCThreadOriginDataIndex(I0, I0, I0, I0);
const index_t m_thread_data_on_grid =
m_block_data_idx_on_grid + c_thread_mtx_on_block[I0];
const index_t n_thread_data_on_grid =
n_block_data_idx_on_grid + c_thread_mtx_on_block[I1];
constexpr auto c_m0_n0_m1_n1_m2_m3_m4_n2_grid_tensor_step_hacks = CGridStepHacks{};
const auto m_thread_data_on_grid_to_m0_m1_m2_m3_m4_adaptor =
make_single_stage_tensor_adaptor(
make_tuple(make_merge_transform(make_tuple(M0, M1, M2, M3, M4))),
make_tuple(Sequence<0, 1, 2, 3, 4>{}),
make_tuple(Sequence<0>{}));
const auto m_thread_data_on_grid_idx =
m_thread_data_on_grid_to_m0_m1_m2_m3_m4_adaptor.CalculateBottomIndex(
make_multi_index(m_thread_data_on_grid));
const auto n_thread_data_on_grid_to_n0_n1_n2_adaptor = make_single_stage_tensor_adaptor(
make_tuple(make_merge_transform(make_tuple(N0, N1, N2))),
make_tuple(Sequence<0, 1, 2>{}),
make_tuple(Sequence<0>{}));
const auto n_thread_data_on_grid_idx =
n_thread_data_on_grid_to_n0_n1_n2_adaptor.CalculateBottomIndex(
make_multi_index(n_thread_data_on_grid));
auto c_thread_copy =
ThreadwiseTensorSliceTransfer_v1r3<FloatAcc,
FloatC,
decltype(c_thread_desc_m0_n0_m1_n1_m2_m3_m4_n2),
decltype(c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2),
CElementwiseOperation,
Sequence<M0, N0, I1, I1, M2, I1, M4, I1>,
CThreadTransferSrcDstAccessOrder,
CThreadTransferSrcDstVectorDim,
CThreadTransferDstScalarPerVector,
CGlobalMemoryDataOperation,
1,
true>{
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
make_multi_index(m_thread_data_on_grid_idx[I0],
n_thread_data_on_grid_idx[I0],
m_thread_data_on_grid_idx[I1],
n_thread_data_on_grid_idx[I1],
m_thread_data_on_grid_idx[I2],
m_thread_data_on_grid_idx[I3],
m_thread_data_on_grid_idx[I4],
n_thread_data_on_grid_idx[I2]),
c_element_op};
c_thread_copy.Run(c_thread_desc_m0_n0_m1_n1_m2_m3_m4_n2,
make_tuple(I0, I0, I0, I0, I0, I0, I0, I0),
c_thread_buf,
c_grid_desc_m0_n0_m1_n1_m2_m3_m4_n2,
c_grid_buf,
c_m0_n0_m1_n1_m2_m3_m4_n2_grid_tensor_step_hacks);
}
}
}; // namespace ck
} // namespace ck
#endif
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment