Skip to content

GitLab

Unverified Commit b23b3d71 authored by Dan Yao's avatar Dan Yao Committed by GitHub
Browse files

Merge pull request #943 from ROCmSoftwarePlatform/mha-train-develop-inference-forward 

Add new flash attention forward for inference
parents 04c206da 1af0a0a4
Show whitespace changes
Inline Side-by-side
Showing with 4771 additions and 15 deletions
example/52_flash_atten_bias/CMakeLists.txt
View file @ b23b3d71
add_example_executable(example_batched_multihead_attention_infer batched_gemm_multihead_attention_infer.cpp)
add_example_executable(example_batched_multihead_attention_bias_infer batched_gemm_multihead_attention_bias_infer.cpp)
add_example_executable(example_grouped_multihead_attention_bias_infer grouped_mutihead_attention_bias_infer.cpp)
add_example_executable(example_batched_multihead_attention_bias_forward_v2 batched_multihead_attention_bias_forward_v2.cpp) add_example_executable(example_batched_multihead_attention_bias_forward_v2 batched_multihead_attention_bias_forward_v2.cpp)
add_example_executable(example_grouped_multihead_attention_bias_forward_v2 grouped_multihead_attention_bias_forward_v2.cpp) add_example_executable(example_grouped_multihead_attention_bias_forward_v2 grouped_multihead_attention_bias_forward_v2.cpp)
......
example/52_flash_atten_bias/batched_gemm_multihead_attention_bias_infer.cpp 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
/*
Gemm + Softmax + Gemm fused operation. Computes C_g_m_o = Softmax(A_g_m_k * B0_g_k_n) * B1_g_n_o
|-----------------|
Gemm0
|-------------------------------------|
Gemm1
*/
#define DIM 128 // DIM should be a multiple of 8.
#include <iostream>
#include <numeric>
#include <initializer_list>
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/tensor_specialization.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_batched_mha_infer_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/utility/check_err.hpp"
#include "ck/library/utility/device_memory.hpp"
#include "ck/library/utility/host_tensor.hpp"
#include "ck/library/utility/host_tensor_generator.hpp"
#include "ck/library/utility/literals.hpp"
#include "ck/library/reference_tensor_operation/cpu/reference_batched_gemm.hpp"
#include "ck/library/reference_tensor_operation/cpu/reference_softmax.hpp"
template <ck::index_t... Is>
using S = ck::Sequence<Is...>;
using F16 = ck::half_t;
using F32 = float;
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
using ADataType = F16;
using B0DataType = F16;
using B1DataType = F16;
using AccDataType = F32;
using CShuffleDataType = F32;
using CDataType = F16;
using Acc0BiasDataType = F16;
using Acc1BiasDataType = void;
static constexpr ck::index_t NumDimG = 2;
static constexpr ck::index_t NumDimM = 1;
static constexpr ck::index_t NumDimN = 1;
static constexpr ck::index_t NumDimK = 1;
static constexpr ck::index_t NumDimO = 1;
using AElementOp = PassThrough;
using B0ElementOp = PassThrough;
using Acc0ElementOp = ck::tensor_operation::element_wise::Scale;
using B1ElementOp = PassThrough;
using CElementOp = PassThrough;
static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKOPadding;
static constexpr auto MaskingSpec =
ck::tensor_operation::device::MaskingSpecialization::MaskDisabled;
static constexpr auto TensorSpecA = ck::tensor_operation::device::TensorSpecialization::Default;
static constexpr auto TensorSpecB0 = ck::tensor_operation::device::TensorSpecialization::Default;
static constexpr auto TensorSpecB1 = ck::tensor_operation::device::TensorSpecialization::Default;
static constexpr auto TensorSpecC = ck::tensor_operation::device::TensorSpecialization::Default;
using DeviceGemmInstance =
ck::tensor_operation::device::DeviceBatchedMultiheadAttentionInfer_Xdl_CShuffle<
NumDimG,
NumDimM,
NumDimN,
NumDimK,
NumDimO,
ADataType,
B0DataType,
B1DataType,
CDataType,
Acc0BiasDataType,
Acc1BiasDataType,
AccDataType,
CShuffleDataType,
AElementOp,
B0ElementOp,
Acc0ElementOp,
B1ElementOp,
CElementOp,
GemmSpec,
TensorSpecA,
TensorSpecB0,
TensorSpecB1,
TensorSpecC,
1,
256,
128, // MPerBlock
128, // NPerBlock
32, // KPerBlock
DIM, // Gemm1NPerBlock
32, // Gemm1KPerBlock
8, // AK1
8, // BK1
2, // B1K1
32, // MPerXDL
32, // NPerXDL
1, // MXdlPerWave
4, // NXdlPerWave
DIM / 32, // Gemm1NXdlPerWave
S<4, 64, 1>, // ABlockTransfer
S<1, 0, 2>,
S<1, 0, 2>,
2,
8,
8,
true,
S<4, 64, 1>, // BBlockTransfer
S<1, 0, 2>,
S<1, 0, 2>,
2,
8,
8,
true,
4,
S<16, 16, 1>, // B1BlockTransfer
S<0, 2, 1>,
S<0, 2, 1>,
1,
4,
2,
false,
1, // CShuffleMXdlPerWavePerShuffle
2, // CShuffleNXdlPerWavePerShuffle
S<1, 32, 1, 8>, // CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
8, // CShuffleBlockTransferScalarPerVector_NPerBlock
MaskingSpec>; // MaskingSpecialization
// Ref Gemm0: fp16 in, fp32 out
using ReferenceGemm0Instance = ck::tensor_operation::host::ReferenceBatchedGemm<ADataType,
B0DataType,
AccDataType,
AccDataType,
AElementOp,
B0ElementOp,
Acc0ElementOp>;
// Ref Softmax: fp32 in, fp16 out
using ReferenceSoftmaxInstance =
ck::tensor_operation::host::ReferenceSoftmax<AccDataType, ADataType, AccDataType>;
// Ref Gemm1: fp16 in, fp16 out
using ReferenceGemm1Instance = ck::tensor_operation::host::ReferenceBatchedGemm<ADataType,
B1DataType,
CDataType,
AccDataType,
AElementOp,
B1ElementOp,
CElementOp>;
#include "run_batched_multihead_attention_bias_infer.inc"
int main(int argc, char* argv[]) { return run(argc, argv); }
example/52_flash_atten_bias/batched_gemm_multihead_attention_infer.cpp 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
/*
Gemm + Softmax + Gemm fused operation. Computes C_g_m_o = Softmax(A_g_m_k * B0_g_k_n) * B1_g_n_o
|-----------------|
Gemm0
|-------------------------------------|
Gemm1
*/
#define DIM 128 // DIM should be a multiple of 8.
#include <iostream>
#include <numeric>
#include <initializer_list>
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/tensor_specialization.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_batched_mha_infer_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/utility/check_err.hpp"
#include "ck/library/utility/device_memory.hpp"
#include "ck/library/utility/host_tensor.hpp"
#include "ck/library/utility/host_tensor_generator.hpp"
#include "ck/library/utility/literals.hpp"
#include "ck/library/reference_tensor_operation/cpu/reference_batched_gemm.hpp"
#include "ck/library/reference_tensor_operation/cpu/reference_softmax.hpp"
template <ck::index_t... Is>
using S = ck::Sequence<Is...>;
using F16 = ck::half_t;
using F32 = float;
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
using ADataType = F16;
using B0DataType = F16;
using B1DataType = F16;
using AccDataType = F32;
using CShuffleDataType = F32;
using CDataType = F16;
using Acc0BiasDataType = void;
using Acc1BiasDataType = void;
static constexpr ck::index_t NumDimG = 2;
static constexpr ck::index_t NumDimM = 1;
static constexpr ck::index_t NumDimN = 1;
static constexpr ck::index_t NumDimK = 1;
static constexpr ck::index_t NumDimO = 1;
using AElementOp = PassThrough;
using B0ElementOp = PassThrough;
using Acc0ElementOp = ck::tensor_operation::element_wise::Scale;
using B1ElementOp = PassThrough;
using CElementOp = PassThrough;
static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKOPadding;
static constexpr auto MaskingSpec =
ck::tensor_operation::device::MaskingSpecialization::MaskDisabled;
static constexpr auto TensorSpecA = ck::tensor_operation::device::TensorSpecialization::Default;
static constexpr auto TensorSpecB0 = ck::tensor_operation::device::TensorSpecialization::Default;
static constexpr auto TensorSpecB1 = ck::tensor_operation::device::TensorSpecialization::Default;
static constexpr auto TensorSpecC = ck::tensor_operation::device::TensorSpecialization::Default;
using DeviceGemmInstance =
ck::tensor_operation::device::DeviceBatchedMultiheadAttentionInfer_Xdl_CShuffle<
NumDimG,
NumDimM,
NumDimN,
NumDimK,
NumDimO,
ADataType,
B0DataType,
B1DataType,
CDataType,
Acc0BiasDataType,
Acc1BiasDataType,
AccDataType,
CShuffleDataType,
AElementOp,
B0ElementOp,
Acc0ElementOp,
B1ElementOp,
CElementOp,
GemmSpec,
TensorSpecA,
TensorSpecB0,
TensorSpecB1,
TensorSpecC,
1,
256,
128, // MPerBlock
128, // NPerBlock
32, // KPerBlock
DIM, // Gemm1NPerBlock
32, // Gemm1KPerBlock
8, // AK1
8, // BK1
2, // B1K1
32, // MPerXDL
32, // NPerXDL
1, // MXdlPerWave
4, // NXdlPerWave
DIM / 32, // Gemm1NXdlPerWave
S<4, 64, 1>, // ABlockTransfer
S<1, 0, 2>,
S<1, 0, 2>,
2,
8,
8,
true,
S<4, 64, 1>, // BBlockTransfer
S<1, 0, 2>,
S<1, 0, 2>,
2,
8,
8,
true,
4,
S<16, 16, 1>, // B1BlockTransfer
S<0, 2, 1>,
S<0, 2, 1>,
1,
4,
2,
false,
1, // CShuffleMXdlPerWavePerShuffle
2, // CShuffleNXdlPerWavePerShuffle
S<1, 32, 1, 8>, // CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
8, // CShuffleBlockTransferScalarPerVector_NPerBlock
MaskingSpec>; // MaskingSpecialization
// Ref Gemm0: fp16 in, fp32 out
using ReferenceGemm0Instance = ck::tensor_operation::host::ReferenceBatchedGemm<ADataType,
B0DataType,
AccDataType,
AccDataType,
AElementOp,
B0ElementOp,
Acc0ElementOp>;
// Ref Softmax: fp32 in, fp16 out
using ReferenceSoftmaxInstance =
ck::tensor_operation::host::ReferenceSoftmax<AccDataType, ADataType, AccDataType>;
// Ref Gemm1: fp16 in, fp16 out
using ReferenceGemm1Instance = ck::tensor_operation::host::ReferenceBatchedGemm<ADataType,
B1DataType,
CDataType,
AccDataType,
AElementOp,
B1ElementOp,
CElementOp>;
#include "run_batched_multihead_attention_infer.inc"
int main(int argc, char* argv[]) { return run(argc, argv); }
example/52_flash_atten_bias/batched_multihead_attention_bias_forward_v2.cpp
View file @ b23b3d71
...@@ -327,6 +327,6 @@ using ReferenceGemm1Instance = ck::tensor_operation::host::ReferenceBatchedGemm< ...@@ -327,6 +327,6 @@ using ReferenceGemm1Instance = ck::tensor_operation::host::ReferenceBatchedGemm<
using ReferenceDropoutInstance = using ReferenceDropoutInstance =
ck::tensor_operation::host::ReferenceDropout<ZDataType, ADataType, ADataType>; ck::tensor_operation::host::ReferenceDropout<ZDataType, ADataType, ADataType>;
#include "run_batched_multihead_attention_bias_forward.inc" #include "run_batched_multihead_attention_bias_forward_v2.inc"
int main(int argc, char* argv[]) { return run(argc, argv); } int main(int argc, char* argv[]) { return run(argc, argv); }
example/52_flash_atten_bias/grouped_multihead_attention_bias_forward_v2.cpp
View file @ b23b3d71
...@@ -327,6 +327,6 @@ using ReferenceGemm1Instance = ck::tensor_operation::host::ReferenceBatchedGemm< ...@@ -327,6 +327,6 @@ using ReferenceGemm1Instance = ck::tensor_operation::host::ReferenceBatchedGemm<
using ReferenceDropoutInstance = using ReferenceDropoutInstance =
ck::tensor_operation::host::ReferenceDropout<ZDataType, ADataType, ADataType>; ck::tensor_operation::host::ReferenceDropout<ZDataType, ADataType, ADataType>;
#include "run_grouped_multihead_attention_bias_forward.inc" #include "run_grouped_multihead_attention_bias_forward_v2.inc"
int main(int argc, char* argv[]) { return run(argc, argv); } int main(int argc, char* argv[]) { return run(argc, argv); }
example/52_flash_atten_bias/grouped_mutihead_attention_bias_infer.cpp 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
/*
Gemm + Softmax + Gemm fused operation. Computes C_g_m_o = Softmax(A_g_m_k * B0_g_k_n) * B1_g_n_o
|-----------------|
Gemm0
|-------------------------------------|
Gemm1
*/
#include <iostream>
#include <numeric>
#include <initializer_list>
#include <cstdlib>
#include "ck/ck.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/tensor_specialization.hpp"
#include "ck/tensor_operation/gpu/device/impl/device_grouped_mha_infer_xdl_cshuffle.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/library/utility/check_err.hpp"
#include "ck/library/utility/device_memory.hpp"
#include "ck/library/utility/host_tensor.hpp"
#include "ck/library/utility/host_tensor_generator.hpp"
#include "ck/library/utility/literals.hpp"
#include "ck/library/reference_tensor_operation/cpu/reference_batched_gemm.hpp"
#include "ck/library/reference_tensor_operation/cpu/reference_softmax.hpp"
template <ck::index_t... Is>
using S = ck::Sequence<Is...>;
using F16 = ck::half_t;
using F32 = float;
using PassThrough = ck::tensor_operation::element_wise::PassThrough;
using ADataType = F16;
using B0DataType = F16;
using B1DataType = F16;
using AccDataType = F32;
using CShuffleDataType = F32;
using CDataType = F16;
using Acc0BiasDataType = F16;
using Acc1BiasDataType = void;
static constexpr ck::index_t NumDimG = 2;
static constexpr ck::index_t NumDimM = 1;
static constexpr ck::index_t NumDimN = 1;
static constexpr ck::index_t NumDimK = 1;
static constexpr ck::index_t NumDimO = 1;
using AElementOp = PassThrough;
using B0ElementOp = PassThrough;
using Acc0ElementOp = ck::tensor_operation::element_wise::Scale;
using B1ElementOp = PassThrough;
using CElementOp = PassThrough;
static constexpr auto GemmSpec = ck::tensor_operation::device::GemmSpecialization::MNKOPadding;
static constexpr auto MaskingSpec =
ck::tensor_operation::device::MaskingSpecialization::MaskDisabled;
static constexpr auto TensorSpecA = ck::tensor_operation::device::TensorSpecialization::Default;
static constexpr auto TensorSpecB0 = ck::tensor_operation::device::TensorSpecialization::Default;
static constexpr auto TensorSpecB1 = ck::tensor_operation::device::TensorSpecialization::Default;
static constexpr auto TensorSpecC = ck::tensor_operation::device::TensorSpecialization::Default;
using DeviceGemmInstance =
ck::tensor_operation::device::DeviceGroupedMultiheadAttentionInfer_Xdl_CShuffle<
NumDimG,
NumDimM,
NumDimN,
NumDimK,
NumDimO,
ADataType,
B0DataType,
B1DataType,
CDataType,
Acc0BiasDataType,
Acc1BiasDataType,
AccDataType,
CShuffleDataType,
AElementOp,
B0ElementOp,
Acc0ElementOp,
B1ElementOp,
CElementOp,
GemmSpec,
TensorSpecA,
TensorSpecB0,
TensorSpecB1,
TensorSpecC,
1,
256,
128, // MPerBlock
128, // NPerBlock
32, // KPerBlock
64, // Gemm1NPerBlock
32, // Gemm1KPerBlock
8, // AK1
8, // BK1
2, // B1K1
32, // MPerXDL
32, // NPerXDL
1, // MXdlPerWave
4, // NXdlPerWave
2, // Gemm1NXdlPerWave
S<4, 64, 1>, // ABlockTransfer
S<1, 0, 2>,
S<1, 0, 2>,
2,
8,
8,
true,
S<4, 64, 1>, // BBlockTransfer
S<1, 0, 2>,
S<1, 0, 2>,
2,
8,
8,
true,
1,
S<16, 16, 1>, // B1BlockTransfer
S<0, 2, 1>,
S<0, 2, 1>,
1,
4,
2,
false,
1, // CShuffleMXdlPerWavePerShuffle
2, // CShuffleNXdlPerWavePerShuffle
S<1, 32, 1, 8>, // CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock
8, // CShuffleBlockTransferScalarPerVector_NPerBlock
MaskingSpec>; // MaskingSpecialization
// Ref Gemm0: fp16 in, fp32 out
using ReferenceGemm0Instance = ck::tensor_operation::host::ReferenceBatchedGemm<ADataType,
B0DataType,
AccDataType,
AccDataType,
AElementOp,
B0ElementOp,
Acc0ElementOp>;
// Ref Softmax: fp32 in, fp16 out
using ReferenceSoftmaxInstance =
ck::tensor_operation::host::ReferenceSoftmax<AccDataType, ADataType, AccDataType>;
// Ref Gemm1: fp16 in, fp16 out
using ReferenceGemm1Instance = ck::tensor_operation::host::ReferenceBatchedGemm<ADataType,
B1DataType,
CDataType,
AccDataType,
AElementOp,
B1ElementOp,
CElementOp>;
#include "run_grouped_multihead_attention_bias_infer.inc"
int main(int argc, char* argv[]) { return run(argc, argv); }
example/52_flash_atten_bias/run_batched_multihead_attention_bias_infer.inc 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
int run(int argc, char* argv[])
{
bool do_verification = true;
int init_method = 1;
bool time_kernel = false;
// GEMM shape for A/B0/B1/C
// C_g_m_o = A_g_m_k * B0_g_k_n * B1_g_n_o
ck::index_t M = 1024;
ck::index_t N = 1024;
ck::index_t K = DIM;
ck::index_t O = DIM;
// Output shape C[G0, M, G1, O]. Batch dim, outer dim, inner dim must match GEMM shape
// C_g0_g1_m_o = reshape(C_g_m_o, [g0, g1, m, o])
// C_g0_m_g1_o = permute(C_g0_g1_m_o, [0, 2, 1, 3])
ck::index_t G0 = 7;
ck::index_t G1 = 13;
float alpha = 1;
bool input_permute = false;
bool output_permute = true;
if(argc == 1)
{
// use default case
}
else if(argc == 4)
{
do_verification = std::stoi(argv[1]);
init_method = std::stoi(argv[2]);
time_kernel = std::stoi(argv[3]);
}
else if(argc == 13)
{
do_verification = std::stoi(argv[1]);
init_method = std::stoi(argv[2]);
time_kernel = std::stoi(argv[3]);
M = std::stoi(argv[4]);
N = std::stoi(argv[5]);
K = std::stoi(argv[6]);
O = std::stoi(argv[7]);
G0 = std::stoi(argv[8]);
G1 = std::stoi(argv[9]);
alpha = std::stof(argv[10]);
input_permute = std::stoi(argv[11]);
output_permute = std::stoi(argv[12]);
}
else
{
printf("arg1: verification (0=no, 1=yes)\n");
printf("arg2: initialization (0=no init, 1=integer value, 2=decimal value)\n");
printf("arg3: time kernel (0=no, 1=yes)\n");
printf("arg4 to 11: M, N, K, O, G0, G1\n");
printf("arg10: scale (alpha)\n");
printf("arg11 to 12: input / output permute\n");
exit(0);
}
std::vector<ck::index_t> a_gs_ms_ks_lengths{G0, G1, M, K};
std::vector<ck::index_t> a_gs_ms_ks_strides =
input_permute
? std::vector<ck::index_t>{M * G1 * K, K, G1 * K, 1} // A layout [G0, M, G1, K]
: std::vector<ck::index_t>{G1 * M * K, M * K, K, 1}; // A layout [G0, G1, M, K]
std::vector<ck::index_t> b0_gs_ns_ks_lengths{G0, G1, N, K};
std::vector<ck::index_t> b0_gs_ns_ks_strides =
input_permute
? std::vector<ck::index_t>{N * G1 * K, K, G1 * K, 1} // B0 layout [G0, N, G1, K]
: std::vector<ck::index_t>{G1 * N * K, N * K, K, 1}; // B0 layout [G0, G1, N, K]
std::vector<ck::index_t> b1_gs_os_ns_lengths{G0, G1, O, N};
std::vector<ck::index_t> b1_gs_os_ns_strides =
input_permute
? std::vector<ck::index_t>{N * G1 * O, O, 1, G1 * O} // B1 layout [G0, N, G1, O]
: std::vector<ck::index_t>{G1 * N * O, N * O, 1, O}; // B1 layout [G0, G1, N, O]
std::vector<ck::index_t> c_gs_ms_os_lengths{G0, G1, M, O};
std::vector<ck::index_t> c_gs_ms_os_strides =
output_permute
? std::vector<ck::index_t>{M * G1 * O, O, G1 * O, 1} // C layout [G0, M, G1, O]
: std::vector<ck::index_t>{G1 * M * O, M * O, O, 1}; // C layout [G0, G1, M, O]
std::vector<ck::index_t> d0_gs_ms_ns_lengths{G0, G1, M, N};
std::vector<ck::index_t> d0_gs_ms_ns_strides =
input_permute
? std::vector<ck::index_t>{M * G1 * N, N, G1 * N, 1} // D0 layout [G0, M, G1, N]
: std::vector<ck::index_t>{G1 * M * N, M * N, N, 1}; // D0 layout [G0, G1, M, N]
Tensor<ADataType> a_gs_ms_ks(a_gs_ms_ks_lengths, a_gs_ms_ks_strides);
Tensor<B0DataType> b0_gs_ns_ks(b0_gs_ns_ks_lengths, b0_gs_ns_ks_strides);
Tensor<Acc0BiasDataType> d0_gs_ms_ns(d0_gs_ms_ns_lengths, d0_gs_ms_ns_strides);
Tensor<B1DataType> b1_gs_os_ns(b1_gs_os_ns_lengths, b1_gs_os_ns_strides);
Tensor<CDataType> c_gs_ms_os_host_result(c_gs_ms_os_lengths, c_gs_ms_os_strides);
Tensor<CDataType> c_gs_ms_os_device_result(c_gs_ms_os_lengths, c_gs_ms_os_strides);
std::cout << "a_gs_ms_ks: " << a_gs_ms_ks.mDesc << std::endl;
std::cout << "b0_gs_ns_ks: " << b0_gs_ns_ks.mDesc << std::endl;
std::cout << "b1_gs_os_ns: " << b1_gs_os_ns.mDesc << std::endl;
std::cout << "c_gs_ms_os: " << c_gs_ms_os_host_result.mDesc << std::endl;
switch(init_method)
{
case 0: break;
case 1:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_2<ADataType>{-2, 2});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_2<B0DataType>{-2, 2});
d0_gs_ms_ns.GenerateTensorValue(GeneratorTensor_2<Acc0BiasDataType>{-2, 2});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_2<B1DataType>{-2, 2});
break;
case 2:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_3<B0DataType>{0.0, 1.0});
d0_gs_ms_ns.GenerateTensorValue(GeneratorTensor_3<Acc0BiasDataType>{-0.5, 0.5});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_3<B1DataType>{-0.5, 0.5});
break;
case 3:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_2<ADataType>{-2, 2});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_Diagonal<B0DataType>{});
d0_gs_ms_ns.GenerateTensorValue(GeneratorTensor_1<Acc0BiasDataType>{1});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_Diagonal<B1DataType>{});
break;
default:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_Sequential<2>{});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_Diagonal<B0DataType>{});
d0_gs_ms_ns.GenerateTensorValue(GeneratorTensor_1<Acc0BiasDataType>{1});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_Diagonal<B1DataType>{});
}
DeviceMem a_device_buf(sizeof(ADataType) * a_gs_ms_ks.mDesc.GetElementSpaceSize());
DeviceMem b0_device_buf(sizeof(B0DataType) * b0_gs_ns_ks.mDesc.GetElementSpaceSize());
DeviceMem d0_device_buf(sizeof(Acc0BiasDataType) * d0_gs_ms_ns.mDesc.GetElementSpaceSize());
DeviceMem b1_device_buf(sizeof(B1DataType) * b1_gs_os_ns.mDesc.GetElementSpaceSize());
DeviceMem c_device_buf(sizeof(CDataType) *
c_gs_ms_os_device_result.mDesc.GetElementSpaceSize());
a_device_buf.ToDevice(a_gs_ms_ks.mData.data());
b0_device_buf.ToDevice(b0_gs_ns_ks.mData.data());
d0_device_buf.ToDevice(d0_gs_ms_ns.mData.data());
b1_device_buf.ToDevice(b1_gs_os_ns.mData.data());
auto a_element_op = AElementOp{};
auto b0_element_op = B0ElementOp{};
auto acc0_element_op = Acc0ElementOp{alpha};
auto b1_element_op = B1ElementOp{};
auto c_element_op = CElementOp{};
// do GEMM
// TODO ANT: replace array with vector?
auto gemm = DeviceGemmInstance{};
auto invoker = gemm.MakeInvoker();
auto argument = gemm.MakeArgument(
static_cast<ADataType*>(a_device_buf.GetDeviceBuffer()),
static_cast<B0DataType*>(b0_device_buf.GetDeviceBuffer()),
static_cast<B1DataType*>(b1_device_buf.GetDeviceBuffer()),
static_cast<CDataType*>(c_device_buf.GetDeviceBuffer()),
static_cast<Acc0BiasDataType*>(d0_device_buf.GetDeviceBuffer()), // p_acc0_bias;
nullptr, // p_acc1_bias;
a_gs_ms_ks_lengths,
a_gs_ms_ks_strides,
b0_gs_ns_ks_lengths,
b0_gs_ns_ks_strides,
b1_gs_os_ns_lengths,
b1_gs_os_ns_strides,
c_gs_ms_os_lengths,
c_gs_ms_os_strides,
d0_gs_ms_ns_lengths, // acc0_bias_gs_ms_ns_lengths
d0_gs_ms_ns_strides, // acc0_bias_gs_ms_ns_strides
{}, // std::vector<ck::index_t>{acc1_biases_gs_ms_os_lengths},
{}, // std::vector<ck::index_t>{acc1_biases_gs_ms_os_strides},
a_element_op,
b0_element_op,
acc0_element_op,
b1_element_op,
c_element_op);
if(!gemm.IsSupportedArgument(argument))
{
std::cout << gemm.GetTypeString() << " does not support this problem" << std::endl;
return 0;
}
ck::index_t BatchCount = G0 * G1;
float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
std::size_t flop = (size_t(M) * N * K * 2 + size_t(M) * N * O * 2 + size_t(M) * N) * BatchCount;
std::size_t num_btype = (sizeof(ADataType) * M * K + sizeof(B0DataType) * K * N +
sizeof(B1DataType) * N * O + sizeof(CDataType) * M * O + sizeof(Acc0BiasDataType) * M * N) *
BatchCount;
float tflops = static_cast<float>(flop) / 1.E9 / ave_time;
float gb_per_sec = num_btype / 1.E6 / ave_time;
std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
<< gemm.GetTypeString() << std::endl;
if(do_verification)
{
c_device_buf.FromDevice(c_gs_ms_os_device_result.mData.data());
Tensor<ADataType> a_g_m_k({BatchCount, M, K});
Tensor<B0DataType> b0_g_k_n({BatchCount, K, N});
Tensor<B1DataType> b1_g_n_o({BatchCount, N, O});
Tensor<AccDataType> acc0_g_m_n({BatchCount, M, N}); // scratch object after gemm0
Tensor<Acc0BiasDataType> d0_g_m_n({BatchCount, M, N});
Tensor<ADataType> a1_g_m_n({BatchCount, M, N}); // scratch object after softmax
Tensor<CDataType> c_g_m_o_host_result({BatchCount, M, O}); // scratch object after gemm1
// permute
a_gs_ms_ks.ForEach([&](auto& self, auto idx) {
a_g_m_k(idx[0] * G1 + idx[1], idx[2], idx[3]) = self(idx);
});
b0_gs_ns_ks.ForEach([&](auto& self, auto idx) {
b0_g_k_n(idx[0] * G1 + idx[1], idx[3], idx[2]) = self(idx);
});
d0_gs_ms_ns.ForEach([&](auto& self, auto idx) {
d0_g_m_n(idx[0] * G1 + idx[1], idx[2], idx[3]) = self(idx);
});
b1_gs_os_ns.ForEach([&](auto& self, auto idx) {
b1_g_n_o(idx[0] * G1 + idx[1], idx[3], idx[2]) = self(idx);
});
// gemm 0
auto ref_gemm0 = ReferenceGemm0Instance{};
auto ref_gemm0_invoker = ref_gemm0.MakeInvoker();
auto ref_gemm0_argument = ref_gemm0.MakeArgument(
a_g_m_k, b0_g_k_n, acc0_g_m_n, a_element_op, b0_element_op, acc0_element_op);
ref_gemm0_invoker.Run(ref_gemm0_argument);
// bias
acc0_g_m_n.ForEach([&](auto& self, auto idx) {
self(idx) += ck::type_convert<AccDataType>(d0_g_m_n(idx));
});
// masking
const auto mask = DeviceGemmInstance::C0MatrixMask(M, N);
acc0_g_m_n.ForEach([&](auto& self, auto idx) {
if(mask.IsMaskedElement(idx[1], idx[2]))
self(idx) = -ck::NumericLimits<float>::Infinity();
});
// softmax
auto ref_softmax = ReferenceSoftmaxInstance{};
auto ref_softmax_invoker = ref_softmax.MakeInvoker();
auto ref_softmax_argument = ref_softmax.MakeArgument(acc0_g_m_n, a1_g_m_n, 1, 0, {2});
ref_softmax_invoker.Run(ref_softmax_argument);
// gemm1
auto ref_gemm1 = ReferenceGemm1Instance{};
auto ref_gemm1_invoker = ref_gemm1.MakeInvoker();
auto ref_gemm1_argument = ref_gemm1.MakeArgument(
a1_g_m_n, b1_g_n_o, c_g_m_o_host_result, PassThrough{}, b1_element_op, c_element_op);
ref_gemm1_invoker.Run(ref_gemm1_argument);
// permute
c_gs_ms_os_host_result.ForEach([&](auto& self, auto idx) {
const size_t& g0 = idx[0];
const size_t& g1 = idx[1];
const size_t g = g0 * G1 + g1;
self(idx) = c_g_m_o_host_result(g, idx[2], idx[3]);
});
// default absolute error and relative error is 0.001
double rtol = 1e-3;
double atol = 1e-3;
// when BF16 is taken, set absolute error and relative error to 0.01
if(std::is_same_v<ADataType, ck::bhalf_t> && std::is_same_v<B0DataType, ck::bhalf_t> &&
std::is_same_v<B1DataType, ck::bhalf_t> && std::is_same_v<CDataType, ck::bhalf_t>)
{
rtol = 1e-2;
atol = 1e-2;
}
return ck::utils::check_err(c_gs_ms_os_device_result.mData,
c_gs_ms_os_host_result.mData,
"Error: Incorrect results!",
rtol,
atol)
? 0
: 1;
}
return 0;
}
example/52_flash_atten_bias/run_batched_multihead_attention_infer.inc 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
int run(int argc, char* argv[])
{
bool do_verification = true;
int init_method = 1;
bool time_kernel = false;
// GEMM shape for A/B0/B1/C
// C_g_m_o = A_g_m_k * B0_g_k_n * B1_g_n_o
ck::index_t M = 1024;
ck::index_t N = 1024;
ck::index_t K = DIM;
ck::index_t O = DIM;
// Output shape C[G0, M, G1, O]. Batch dim, outer dim, inner dim must match GEMM shape
// C_g0_g1_m_o = reshape(C_g_m_o, [g0, g1, m, o])
// C_g0_m_g1_o = permute(C_g0_g1_m_o, [0, 2, 1, 3])
ck::index_t G0 = 7;
ck::index_t G1 = 13;
float alpha = 1;
bool input_permute = false;
bool output_permute = true;
if(argc == 1)
{
// use default case
}
else if(argc == 4)
{
do_verification = std::stoi(argv[1]);
init_method = std::stoi(argv[2]);
time_kernel = std::stoi(argv[3]);
}
else if(argc == 13)
{
do_verification = std::stoi(argv[1]);
init_method = std::stoi(argv[2]);
time_kernel = std::stoi(argv[3]);
M = std::stoi(argv[4]);
N = std::stoi(argv[5]);
K = std::stoi(argv[6]);
O = std::stoi(argv[7]);
G0 = std::stoi(argv[8]);
G1 = std::stoi(argv[9]);
alpha = std::stof(argv[10]);
input_permute = std::stoi(argv[11]);
output_permute = std::stoi(argv[12]);
}
else
{
printf("arg1: verification (0=no, 1=yes)\n");
printf("arg2: initialization (0=no init, 1=integer value, 2=decimal value)\n");
printf("arg3: time kernel (0=no, 1=yes)\n");
printf("arg4 to 11: M, N, K, O, G0, G1\n");
printf("arg10: scale (alpha)\n");
printf("arg11 to 12: input / output permute\n");
exit(0);
}
std::vector<ck::index_t> a_gs_ms_ks_lengths{G0, G1, M, K};
std::vector<ck::index_t> a_gs_ms_ks_strides =
input_permute
? std::vector<ck::index_t>{M * G1 * K, K, G1 * K, 1} // A layout [G0, M, G1, K]
: std::vector<ck::index_t>{G1 * M * K, M * K, K, 1}; // A layout [G0, G1, M, K]
std::vector<ck::index_t> b0_gs_ns_ks_lengths{G0, G1, N, K};
std::vector<ck::index_t> b0_gs_ns_ks_strides =
input_permute
? std::vector<ck::index_t>{N * G1 * K, K, G1 * K, 1} // B0 layout [G0, N, G1, K]
: std::vector<ck::index_t>{G1 * N * K, N * K, K, 1}; // B0 layout [G0, G1, N, K]
std::vector<ck::index_t> b1_gs_os_ns_lengths{G0, G1, O, N};
std::vector<ck::index_t> b1_gs_os_ns_strides =
input_permute
? std::vector<ck::index_t>{N * G1 * O, O, 1, G1 * O} // B1 layout [G0, N, G1, O]
: std::vector<ck::index_t>{G1 * N * O, N * O, 1, O}; // B1 layout [G0, G1, N, O]
std::vector<ck::index_t> c_gs_ms_os_lengths{G0, G1, M, O};
std::vector<ck::index_t> c_gs_ms_os_strides =
output_permute
? std::vector<ck::index_t>{M * G1 * O, O, G1 * O, 1} // C layout [G0, M, G1, O]
: std::vector<ck::index_t>{G1 * M * O, M * O, O, 1}; // C layout [G0, G1, M, O]
Tensor<ADataType> a_gs_ms_ks(a_gs_ms_ks_lengths, a_gs_ms_ks_strides);
Tensor<B0DataType> b0_gs_ns_ks(b0_gs_ns_ks_lengths, b0_gs_ns_ks_strides);
Tensor<B1DataType> b1_gs_os_ns(b1_gs_os_ns_lengths, b1_gs_os_ns_strides);
Tensor<CDataType> c_gs_ms_os_host_result(c_gs_ms_os_lengths, c_gs_ms_os_strides);
Tensor<CDataType> c_gs_ms_os_device_result(c_gs_ms_os_lengths, c_gs_ms_os_strides);
std::cout << "a_gs_ms_ks: " << a_gs_ms_ks.mDesc << std::endl;
std::cout << "b0_gs_ns_ks: " << b0_gs_ns_ks.mDesc << std::endl;
std::cout << "b1_gs_os_ns: " << b1_gs_os_ns.mDesc << std::endl;
std::cout << "c_gs_ms_os: " << c_gs_ms_os_host_result.mDesc << std::endl;
switch(init_method)
{
case 0: break;
case 1:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_2<ADataType>{-2, 2});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_2<B0DataType>{-2, 2});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_2<B1DataType>{-2, 2});
break;
case 2:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_3<B0DataType>{0.0, 1.0});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_3<B1DataType>{-0.5, 0.5});
break;
case 3:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_2<ADataType>{-2, 2});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_Diagonal<B0DataType>{});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_Diagonal<B1DataType>{});
break;
default:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_Sequential<2>{});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_Diagonal<B0DataType>{});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_Diagonal<B1DataType>{});
}
DeviceMem a_device_buf(sizeof(ADataType) * a_gs_ms_ks.mDesc.GetElementSpaceSize());
DeviceMem b0_device_buf(sizeof(B0DataType) * b0_gs_ns_ks.mDesc.GetElementSpaceSize());
DeviceMem b1_device_buf(sizeof(B1DataType) * b1_gs_os_ns.mDesc.GetElementSpaceSize());
DeviceMem c_device_buf(sizeof(CDataType) *
c_gs_ms_os_device_result.mDesc.GetElementSpaceSize());
a_device_buf.ToDevice(a_gs_ms_ks.mData.data());
b0_device_buf.ToDevice(b0_gs_ns_ks.mData.data());
b1_device_buf.ToDevice(b1_gs_os_ns.mData.data());
auto a_element_op = AElementOp{};
auto b0_element_op = B0ElementOp{};
auto acc0_element_op = Acc0ElementOp{alpha};
auto b1_element_op = B1ElementOp{};
auto c_element_op = CElementOp{};
// do GEMM
// TODO ANT: replace array with vector?
auto gemm = DeviceGemmInstance{};
auto invoker = gemm.MakeInvoker();
auto argument = gemm.MakeArgument(
static_cast<ADataType*>(a_device_buf.GetDeviceBuffer()),
static_cast<B0DataType*>(b0_device_buf.GetDeviceBuffer()),
static_cast<B1DataType*>(b1_device_buf.GetDeviceBuffer()),
static_cast<CDataType*>(c_device_buf.GetDeviceBuffer()),
nullptr, // p_acc0_bias;
nullptr, // p_acc1_bias;
a_gs_ms_ks_lengths,
a_gs_ms_ks_strides,
b0_gs_ns_ks_lengths,
b0_gs_ns_ks_strides,
b1_gs_os_ns_lengths,
b1_gs_os_ns_strides,
c_gs_ms_os_lengths,
c_gs_ms_os_strides,
{}, // acc0_bias_gs_ms_ns_lengths
{}, // acc0_bias_gs_ms_ns_strides
{}, // std::vector<ck::index_t>{acc1_bias_gs_ms_os_lengths},
{}, // std::vector<ck::index_t>{acc1_bias_gs_ms_os_strides},
a_element_op,
b0_element_op,
acc0_element_op,
b1_element_op,
c_element_op);
if(!gemm.IsSupportedArgument(argument))
{
std::cout << gemm.GetTypeString() << " does not support this problem" << std::endl;
return 0;
}
ck::index_t BatchCount = G0 * G1;
float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
std::size_t flop = (size_t(M) * N * K * 2 + size_t(M) * N * O * 2) * BatchCount;
std::size_t num_btype = (sizeof(ADataType) * M * K + sizeof(B0DataType) * K * N +
sizeof(B1DataType) * N * O + sizeof(CDataType) * M * O) *
BatchCount;
float tflops = static_cast<float>(flop) / 1.E9 / ave_time;
float gb_per_sec = num_btype / 1.E6 / ave_time;
std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
<< gemm.GetTypeString() << std::endl;
if(do_verification)
{
c_device_buf.FromDevice(c_gs_ms_os_device_result.mData.data());
Tensor<ADataType> a_g_m_k({BatchCount, M, K});
Tensor<B0DataType> b0_g_k_n({BatchCount, K, N});
Tensor<B1DataType> b1_g_n_o({BatchCount, N, O});
Tensor<AccDataType> acc0_g_m_n({BatchCount, M, N}); // scratch object after gemm0
Tensor<ADataType> a1_g_m_n({BatchCount, M, N}); // scratch object after softmax
Tensor<CDataType> c_g_m_o_host_result({BatchCount, M, O}); // scratch object after gemm1
// permute
a_gs_ms_ks.ForEach([&](auto& self, auto idx) {
a_g_m_k(idx[0] * G1 + idx[1], idx[2], idx[3]) = self(idx);
});
b0_gs_ns_ks.ForEach([&](auto& self, auto idx) {
b0_g_k_n(idx[0] * G1 + idx[1], idx[3], idx[2]) = self(idx);
});
b1_gs_os_ns.ForEach([&](auto& self, auto idx) {
b1_g_n_o(idx[0] * G1 + idx[1], idx[3], idx[2]) = self(idx);
});
// gemm 0
auto ref_gemm0 = ReferenceGemm0Instance{};
auto ref_gemm0_invoker = ref_gemm0.MakeInvoker();
auto ref_gemm0_argument = ref_gemm0.MakeArgument(
a_g_m_k, b0_g_k_n, acc0_g_m_n, a_element_op, b0_element_op, acc0_element_op);
ref_gemm0_invoker.Run(ref_gemm0_argument);
// masking
const auto mask = DeviceGemmInstance::C0MatrixMask(M, N);
acc0_g_m_n.ForEach([&](auto& self, auto idx) {
if(mask.IsMaskedElement(idx[1], idx[2]))
self(idx) = -ck::NumericLimits<float>::Infinity();
});
// softmax
auto ref_softmax = ReferenceSoftmaxInstance{};
auto ref_softmax_invoker = ref_softmax.MakeInvoker();
auto ref_softmax_argument = ref_softmax.MakeArgument(acc0_g_m_n, a1_g_m_n, 1, 0, {2});
ref_softmax_invoker.Run(ref_softmax_argument);
// gemm1
auto ref_gemm1 = ReferenceGemm1Instance{};
auto ref_gemm1_invoker = ref_gemm1.MakeInvoker();
auto ref_gemm1_argument = ref_gemm1.MakeArgument(
a1_g_m_n, b1_g_n_o, c_g_m_o_host_result, PassThrough{}, b1_element_op, c_element_op);
ref_gemm1_invoker.Run(ref_gemm1_argument);
// permute
c_gs_ms_os_host_result.ForEach([&](auto& self, auto idx) {
const size_t& g0 = idx[0];
const size_t& g1 = idx[1];
const size_t g = g0 * G1 + g1;
self(idx) = c_g_m_o_host_result(g, idx[2], idx[3]);
});
// default absolute error and relative error is 0.001
double rtol = 1e-3;
double atol = 1e-3;
// when BF16 is taken, set absolute error and relative error to 0.01
if(std::is_same_v<ADataType, ck::bhalf_t> && std::is_same_v<B0DataType, ck::bhalf_t> &&
std::is_same_v<B1DataType, ck::bhalf_t> && std::is_same_v<CDataType, ck::bhalf_t>)
{
rtol = 1e-2;
atol = 1e-2;
}
return ck::utils::check_err(c_gs_ms_os_device_result.mData,
c_gs_ms_os_host_result.mData,
"Error: Incorrect results!",
rtol,
atol)
? 0
: 1;
}
return 0;
}
example/52_flash_atten_bias/run_grouped_multihead_attention_bias_infer.inc 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
int run(int argc, char* argv[])
{
bool do_verification = true;
int init_method = 1;
bool time_kernel = false;
bool input_permute = false;
bool output_permute = true;
if(argc == 1)
{
// use default case
}
else if(argc == 4)
{
do_verification = std::stoi(argv[1]);
init_method = std::stoi(argv[2]);
time_kernel = std::stoi(argv[3]);
}
else if(argc == 6)
{
do_verification = std::stoi(argv[1]);
init_method = std::stoi(argv[2]);
time_kernel = std::stoi(argv[3]);
input_permute = std::stoi(argv[4]);
output_permute = std::stoi(argv[5]);
}
else
{
printf("arg1: verification (0=no, 1=yes)\n");
printf("arg2: initialization (0=no init, 1=integer value, 2=decimal value)\n");
printf("arg3: time kernel (0=no, 1=yes)\n");
printf("arg4 to 5: input / output permute\n");
exit(0);
}
float alpha = 1; // scaling after 1st gemm
std::size_t group_count = 7;
// Problem descs
std::vector<DeviceGemmInstance::ProblemDesc> problem_descs;
std::vector<const void*> p_a;
std::vector<const void*> p_b0;
std::vector<const void*> p_d0;
std::vector<const void*> p_b1;
std::vector<void*> p_c;
std::vector<std::vector<int>> g0_g1_m_n_k_o;
std::vector<Tensor<ADataType>> a_tensors;
std::vector<Tensor<B0DataType>> b0_tensors;
std::vector<Tensor<Acc0BiasDataType>> d0_tensors;
std::vector<Tensor<B1DataType>> b1_tensors;
std::vector<Tensor<CDataType>> c_tensors;
using DeviceMemPtr = std::unique_ptr<DeviceMem>;
std::vector<DeviceMemPtr> a_tensors_device;
std::vector<DeviceMemPtr> b0_tensors_device;
std::vector<DeviceMemPtr> d0_tensors_device;
std::vector<DeviceMemPtr> b1_tensors_device;
std::vector<DeviceMemPtr> c_tensors_device;
std::size_t flop = 0, num_byte = 0;
std::cout << "group count " << group_count << ". printing first 4 groups\n";
for(std::size_t i = 0; i < group_count; i++)
{
int M = 128 * (rand() % 8 + 1);
int N = 128 * (rand() % 8 + 1);
int K = 40;
int O = 40 * (rand() % 2 + 1);
int G0 = rand() % 3 + 1;
int G1 = rand() % 5 + 1;
g0_g1_m_n_k_o.push_back({G0, G1, M, N, K, O});
std::vector<ck::index_t> a_gs_ms_ks_lengths{G0, G1, M, K};
std::vector<ck::index_t> a_gs_ms_ks_strides =
input_permute
? std::vector<ck::index_t>{M * G1 * K, K, G1 * K, 1} // A layout [G0, M, G1, K]
: std::vector<ck::index_t>{G1 * M * K, M * K, K, 1}; // A layout [G0, G1, M, K]
std::vector<ck::index_t> b0_gs_ns_ks_lengths{G0, G1, N, K};
std::vector<ck::index_t> b0_gs_ns_ks_strides =
input_permute
? std::vector<ck::index_t>{N * G1 * K, K, G1 * K, 1} // B0 layout [G0, N, G1, K]
: std::vector<ck::index_t>{G1 * N * K, N * K, K, 1}; // B0 layout [G0, G1, N, K]
std::vector<ck::index_t> b1_gs_os_ns_lengths{G0, G1, O, N};
std::vector<ck::index_t> b1_gs_os_ns_strides =
input_permute
? std::vector<ck::index_t>{N * G1 * O, O, 1, G1 * O} // B1 layout [G0, N, G1, O]
: std::vector<ck::index_t>{G1 * N * O, N * O, 1, O}; // B1 layout [G0, G1, N, O]
std::vector<ck::index_t> c_gs_ms_os_lengths{G0, G1, M, O};
std::vector<ck::index_t> c_gs_ms_os_strides =
output_permute
? std::vector<ck::index_t>{M * G1 * O, O, G1 * O, 1} // C layout [G0, M, G1, O]
: std::vector<ck::index_t>{G1 * M * O, M * O, O, 1}; // C layout [G0, G1, M, O]
std::vector<ck::index_t> d0_gs_ms_ns_lengths{G0, G1, M, N};
std::vector<ck::index_t> d0_gs_ms_ns_strides =
input_permute
? std::vector<ck::index_t>{M * G1 * N, N, G1 * N, 1} // d0 layout [G0, M, G1, N]
: std::vector<ck::index_t>{G1 * M * N, M * N, N, 1}; // d0 layout [G0, G1, M, N]
problem_descs.push_back({a_gs_ms_ks_lengths,
a_gs_ms_ks_strides,
b0_gs_ns_ks_lengths,
b0_gs_ns_ks_strides,
b1_gs_os_ns_lengths,
b1_gs_os_ns_strides,
c_gs_ms_os_lengths,
c_gs_ms_os_strides,
d0_gs_ms_ns_lengths, // acc0_bias_gs_ms_ns_lengths
d0_gs_ms_ns_strides, // acc0_bias_gs_ms_ns_strides
{}, // acc1_bias_gs_ms_os_lengths
{}}); // acc1_bias_gs_ms_os_strides
// C_m_o = (A_m_k * B0_k_n + bias) * B1_n_o
Tensor<ADataType> a_gs_ms_ks(a_gs_ms_ks_lengths, a_gs_ms_ks_strides);
Tensor<B0DataType> b0_gs_ns_ks(b0_gs_ns_ks_lengths, b0_gs_ns_ks_strides);
Tensor<Acc0BiasDataType> d0_gs_ms_ns(d0_gs_ms_ns_lengths, d0_gs_ms_ns_strides);
Tensor<B1DataType> b1_gs_os_ns(b1_gs_os_ns_lengths, b1_gs_os_ns_strides);
Tensor<CDataType> c_gs_ms_os_device_result(c_gs_ms_os_lengths, c_gs_ms_os_strides);
int Batch = G0 * G1;
flop += (size_t(M) * N * K * 2 + size_t(M) * N * O * 2 + size_t(M) * N) * Batch;
num_byte +=
(sizeof(ADataType) * M * K + sizeof(B0DataType) * K * N + sizeof(B1DataType) * N * O +
sizeof(CDataType) * M * O + sizeof(Acc0BiasDataType) * M * N) *
Batch;
if(i < 4)
{
std::cout << "a_gs_ms_ks[" << i << "]: " << a_gs_ms_ks.mDesc << ", "
<< "b0_gs_ns_ks[" << i << "]: " << b0_gs_ns_ks.mDesc << ", "
<< "b1_gs_os_ns[" << i << "]: " << b1_gs_os_ns.mDesc << ", "
<< "c_gs_ms_os[" << i << "]: " << c_gs_ms_os_device_result.mDesc << std::endl;
}
switch(init_method)
{
case 0: break;
case 1:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_2<ADataType>{-2, 2});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_2<B0DataType>{-2, 2});
d0_gs_ms_ns.GenerateTensorValue(GeneratorTensor_2<Acc0BiasDataType>{-2, 2});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_2<B1DataType>{-2, 2});
break;
case 2:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_3<ADataType>{0.0, 1.0});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_3<B0DataType>{0.0, 1.0});
d0_gs_ms_ns.GenerateTensorValue(GeneratorTensor_3<Acc0BiasDataType>{0.0, 1.0});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_3<B1DataType>{-0.5, 0.5});
break;
case 3:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_2<ADataType>{-2, 2});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_Diagonal<B0DataType>{});
d0_gs_ms_ns.GenerateTensorValue(GeneratorTensor_Diagonal<Acc0BiasDataType>{});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_Diagonal<B1DataType>{});
break;
default:
a_gs_ms_ks.GenerateTensorValue(GeneratorTensor_1<ADataType>{1});
b0_gs_ns_ks.GenerateTensorValue(GeneratorTensor_Sequential<1>{});
d0_gs_ms_ns.GenerateTensorValue(GeneratorTensor_Sequential<1>{});
b1_gs_os_ns.GenerateTensorValue(GeneratorTensor_Diagonal<B1DataType>{});
}
a_tensors.push_back(a_gs_ms_ks);
b0_tensors.push_back(b0_gs_ns_ks);
d0_tensors.push_back(d0_gs_ms_ns);
b1_tensors.push_back(b1_gs_os_ns);
c_tensors.push_back(c_gs_ms_os_device_result);
a_tensors_device.emplace_back(std::make_unique<DeviceMem>(
sizeof(ADataType) * a_gs_ms_ks.mDesc.GetElementSpaceSize()));
b0_tensors_device.emplace_back(std::make_unique<DeviceMem>(
sizeof(B0DataType) * b0_gs_ns_ks.mDesc.GetElementSpaceSize()));
d0_tensors_device.emplace_back(std::make_unique<DeviceMem>(
sizeof(Acc0BiasDataType) * d0_gs_ms_ns.mDesc.GetElementSpaceSize()));
b1_tensors_device.emplace_back(std::make_unique<DeviceMem>(
sizeof(B1DataType) * b1_gs_os_ns.mDesc.GetElementSpaceSize()));
c_tensors_device.emplace_back(std::make_unique<DeviceMem>(
sizeof(CDataType) * c_gs_ms_os_device_result.mDesc.GetElementSpaceSize()));
a_tensors_device[i]->ToDevice(a_gs_ms_ks.mData.data());
b0_tensors_device[i]->ToDevice(b0_gs_ns_ks.mData.data());
d0_tensors_device[i]->ToDevice(d0_gs_ms_ns.mData.data());
b1_tensors_device[i]->ToDevice(b1_gs_os_ns.mData.data());
p_a.push_back(a_tensors_device[i]->GetDeviceBuffer());
p_b0.push_back(b0_tensors_device[i]->GetDeviceBuffer());
p_d0.push_back(d0_tensors_device[i]->GetDeviceBuffer());
p_b1.push_back(b1_tensors_device[i]->GetDeviceBuffer());
p_c.push_back(c_tensors_device[i]->GetDeviceBuffer());
}
auto a_element_op = AElementOp{};
auto b0_element_op = B0ElementOp{};
auto acc0_element_op = Acc0ElementOp{alpha};
auto b1_element_op = B1ElementOp{};
auto c_element_op = CElementOp{};
// do GEMM
auto gemm = DeviceGemmInstance{};
auto invoker = gemm.MakeInvoker();
auto argument = gemm.MakeArgument(p_a,
p_b0,
p_b1,
p_c,
p_d0, // p_acc0_bias
{}, // p_acc1_bias
problem_descs,
a_element_op,
b0_element_op,
acc0_element_op,
b1_element_op,
c_element_op);
// specify workspace for problem_desc
DeviceMem problem_desc_workspace(gemm.GetWorkSpaceSize(&argument));
gemm.SetWorkSpacePointer(&argument, problem_desc_workspace.GetDeviceBuffer());
if(!gemm.IsSupportedArgument(argument))
{
std::cout << gemm.GetTypeString() << " does not support this problem" << std::endl;
return 0;
}
float ave_time = invoker.Run(argument, StreamConfig{nullptr, time_kernel});
float tflops = static_cast<float>(flop) / 1.E9 / ave_time;
float gb_per_sec = num_byte / 1.E6 / ave_time;
std::cout << "Perf: " << ave_time << " ms, " << tflops << " TFlops, " << gb_per_sec << " GB/s, "
<< gemm.GetTypeString() << std::endl;
bool pass = true;
if(do_verification)
{
for(std::size_t i = 0; i < group_count; i++)
{
const int& G0 = g0_g1_m_n_k_o[i][0];
const int& G1 = g0_g1_m_n_k_o[i][1];
const int& M = g0_g1_m_n_k_o[i][2];
const int& N = g0_g1_m_n_k_o[i][3];
const int& K = g0_g1_m_n_k_o[i][4];
const int& O = g0_g1_m_n_k_o[i][5];
const auto& c_gs_ms_os_lengths = problem_descs[i].c_gs_ms_os_lengths;
const auto& c_gs_ms_os_strides = problem_descs[i].c_gs_ms_os_strides;
const auto& a_gs_ms_ks = a_tensors[i];
const auto& b0_gs_ns_ks = b0_tensors[i];
const auto& d0_gs_ms_ns = d0_tensors[i];
const auto& b1_gs_os_ns = b1_tensors[i];
auto& c_gs_ms_os_device_result = c_tensors[i];
auto& c_gs_ms_os_device_buf = *c_tensors_device[i];
c_gs_ms_os_device_buf.FromDevice(c_gs_ms_os_device_result.mData.data());
Tensor<ADataType> a_g_m_k({G0 * G1, M, K});
Tensor<B0DataType> b0_g_k_n({G0 * G1, K, N});
Tensor<Acc0BiasDataType> d0_g_m_n({G0 * G1, M, N});
Tensor<B1DataType> b1_g_n_o({G0 * G1, N, O});
Tensor<AccDataType> acc0_g_m_n({G0 * G1, M, N}); // scratch object after gemm0
Tensor<ADataType> a1_g_m_n({G0 * G1, M, N}); // scratch object after softmax
Tensor<CDataType> c_g_m_o_host_result({G0 * G1, M, O}); // scratch object after gemm1
Tensor<CDataType> c_gs_ms_os_host_result(c_gs_ms_os_lengths, c_gs_ms_os_strides);
// permute
a_gs_ms_ks.ForEach([&](auto& self, auto idx) {
a_g_m_k(idx[0] * G1 + idx[1], idx[2], idx[3]) = self(idx);
});
b0_gs_ns_ks.ForEach([&](auto& self, auto idx) {
b0_g_k_n(idx[0] * G1 + idx[1], idx[3], idx[2]) = self(idx);
});
d0_gs_ms_ns.ForEach([&](auto& self, auto idx) {
d0_g_m_n(idx[0] * G1 + idx[1], idx[2], idx[3]) = self(idx);
});
b1_gs_os_ns.ForEach([&](auto& self, auto idx) {
b1_g_n_o(idx[0] * G1 + idx[1], idx[3], idx[2]) = self(idx);
});
// gemm 0
auto ref_gemm0 = ReferenceGemm0Instance{};
auto ref_gemm0_invoker = ref_gemm0.MakeInvoker();
auto ref_gemm0_argument = ref_gemm0.MakeArgument(
a_g_m_k, b0_g_k_n, acc0_g_m_n, a_element_op, b0_element_op, acc0_element_op);
ref_gemm0_invoker.Run(ref_gemm0_argument);
// bias
acc0_g_m_n.ForEach([&](auto& self, auto idx) {
self(idx) += ck::type_convert<AccDataType>(d0_g_m_n(idx));
});
// masking
const auto mask = DeviceGemmInstance::C0MatrixMask(M, N);
acc0_g_m_n.ForEach([&](auto& self, auto idx) {
if(mask.IsMaskedElement(idx[1], idx[2]))
self(idx) = -ck::NumericLimits<float>::Infinity();
});
// softmax
auto ref_softmax = ReferenceSoftmaxInstance{};
auto ref_softmax_invoker = ref_softmax.MakeInvoker();
auto ref_softmax_argument = ref_softmax.MakeArgument(acc0_g_m_n, a1_g_m_n, 1, 0, {2});
ref_softmax_invoker.Run(ref_softmax_argument);
// gemm 1
auto ref_gemm1 = ReferenceGemm1Instance{};
auto ref_gemm1_invoker = ref_gemm1.MakeInvoker();
auto ref_gemm1_argument = ref_gemm1.MakeArgument(a1_g_m_n,
b1_g_n_o,
c_g_m_o_host_result,
PassThrough{},
b1_element_op,
c_element_op);
ref_gemm1_invoker.Run(ref_gemm1_argument);
// permute
c_gs_ms_os_host_result.ForEach([&](auto& self, auto idx) {
const size_t& g0 = idx[0];
const size_t& g1 = idx[1];
const size_t g = g0 * G1 + g1;
self(idx) = c_g_m_o_host_result(g, idx[2], idx[3]);
});
bool pass_ =
ck::utils::check_err(c_gs_ms_os_device_result.mData, c_gs_ms_os_host_result.mData);
pass &= pass_;
}
}
return pass ? 0 : 1;
}
include/ck/tensor_operation/gpu/device/device_batched_mha_infer.hpp 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <vector>
#include <tuple>
#include "device_base.hpp"
#include "ck/tensor_operation/gpu/device/masking_specialization.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <index_t NumDimG,
index_t NumDimM,
index_t NumDimN,
index_t NumDimK,
index_t NumDimO,
typename ADataType,
typename B0DataType,
typename B1DataType,
typename CDataType,
typename Acc0BiasDataType,
typename Acc1BiasDataType,
typename AElementwiseOperation,
typename B0ElementwiseOperation,
typename C0ElementwiseOperation,
typename B1ElementwiseOperation,
typename C1DEElementwiseOperation,
MaskingSpecialization MaskingSpec>
struct DeviceBatchedMultiheadAttentionInfer : public BaseOperator
{
virtual std::unique_ptr<BaseArgument> MakeArgumentPointer(
const void* p_a,
const void* p_b0,
const void* p_b1,
void* p_c,
const void* p_acc0_bias,
const void* p_acc1_bias,
const std::vector<index_t>& a_gs_ms_ks_lengths,
const std::vector<index_t>& a_gs_ms_ks_strides,
const std::vector<index_t>& b_gs_ns_ks_lengths,
const std::vector<index_t>& b_gs_ns_ks_strides,
const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_lengths, // b1_gs_os_ns_lengths
const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_strides, // b1_gs_os_ns_strides
const std::vector<index_t>& c_gs_ms_gemm1ns_lengths, // c_gs_ms_os_lengths
const std::vector<index_t>& c_gs_ms_gemm1ns_strides, // c_gs_ms_os_strides
const std::vector<index_t>& acc0_bias_gs_ms_ns_lengths,
const std::vector<index_t>& acc0_bias_gs_ms_ns_strides,
const std::vector<index_t>& acc1_bias_gs_ms_gemm1ns_lengths, // acc1_biases_gs_ms_os_lengths
const std::vector<index_t>&
acc1_biases_gs_ms_gemm1ns_strides, // acc1_biases_gs_ms_os_strides
AElementwiseOperation a_element_op,
B0ElementwiseOperation b0_element_op,
C0ElementwiseOperation c0_element_op,
B1ElementwiseOperation b1_element_op,
C1DEElementwiseOperation c1de_element_op) = 0;
virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
include/ck/tensor_operation/gpu/device/device_grouped_mha_infer.hpp 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <vector>
#include "device_base.hpp"
#include "ck/tensor_operation/gpu/device/masking_specialization.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <index_t NumDimG,
index_t NumDimM,
index_t NumDimN,
index_t NumDimK,
index_t NumDimO,
typename ADataType,
typename B0DataType,
typename B1DataType,
typename CDataType,
typename Acc0BiasDataType,
typename Acc1BiasDataType,
typename AElementwiseOperation,
typename B0ElementwiseOperation,
typename Acc0ElementwiseOperation,
typename B1ElementwiseOperation,
typename CElementwiseOperation,
MaskingSpecialization MaskingSpec>
struct DeviceGroupedMultiheadAttentionInfer : public BaseOperator
{
struct ProblemDesc
{
std::vector<index_t> a_gs_ms_ks_lengths;
std::vector<index_t> a_gs_ms_ks_strides;
std::vector<index_t> b0_gs_ns_ks_lengths;
std::vector<index_t> b0_gs_ns_ks_strides;
std::vector<index_t> b1_gs_os_ns_lengths;
std::vector<index_t> b1_gs_os_ns_strides;
std::vector<index_t> c_gs_ms_os_lengths;
std::vector<index_t> c_gs_ms_os_strides;
std::vector<index_t> acc0_bias_gs_ms_ns_lengths;
std::vector<index_t> acc0_bias_gs_ms_ns_strides;
std::vector<index_t> acc1_bias_gs_ms_os_lengths;
std::vector<index_t> acc1_bias_gs_ms_os_strides;
};
virtual std::unique_ptr<BaseArgument>
MakeArgumentPointer(std::vector<const void*> p_a_vec,
std::vector<const void*> p_b0_vec,
std::vector<const void*> p_b1_vec,
std::vector<void*> p_c_vec,
std::vector<const void*> p_acc0_bias_vec,
std::vector<const void*> p_acc1_bias_vec,
std::vector<ProblemDesc> problem_desc_vec,
AElementwiseOperation a_element_op,
B0ElementwiseOperation b0_element_op,
Acc0ElementwiseOperation acc0_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op) = 0;
virtual std::unique_ptr<BaseInvoker> MakeInvokerPointer() = 0;
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
include/ck/tensor_operation/gpu/device/impl/device_batched_mha_fwd_xdl_cshuffle_v2.hpp
View file @ b23b3d71
...@@ -1026,12 +1026,12 @@ struct DeviceBatchedMultiheadAttentionForward_Xdl_CShuffle_V2 ...@@ -1026,12 +1026,12 @@ struct DeviceBatchedMultiheadAttentionForward_Xdl_CShuffle_V2
if constexpr(!is_same<D0DataType, void>::value) if constexpr(!is_same<D0DataType, void>::value)
{ {
if(arg.d0_n_length_stride_[1] == 1 && if(arg.d0_n_length_stride_[1] == 1)
arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector != 0)
{ {
if(arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector != 0)
return false; return false;
} }
if(arg.d0_n_length_stride_[1] != 1 && Acc0BiasTransferSrcScalarPerVector != 1) else if(Acc0BiasTransferSrcScalarPerVector != 1)
{ {
return false; return false;
} }
......
include/ck/tensor_operation/gpu/device/impl/device_batched_mha_infer_xdl_cshuffle.hpp 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/device_batched_mha_infer.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_batched_mha_infer_xdl_cshuffle.hpp"
#include "ck/tensor_operation/operator_transform/transform_contraction_to_gemm.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename GridwiseGemm,
typename FloatAB,
typename D0DataType,
typename FloatC,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename C0ElementwiseOperation,
typename B1ElementwiseOperation,
typename C1DEElementwiseOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename B1GridDesc_BK0_N_BK1,
typename C1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename D0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5,
typename Block2CTileMap,
typename ComputeBasePtrOfStridedBatch,
typename C0MatrixMask,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_batched_multiple_head_flash_attention_infer(
const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
const D0DataType* p_d0_grid,
const FloatAB* __restrict__ p_b1_grid,
FloatC* __restrict__ p_c_grid,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const C0ElementwiseOperation c0_element_op,
const B1ElementwiseOperation b1_element_op,
const C1DEElementwiseOperation c1de_element_op,
const AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1,
const B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1,
const C1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c1_grid_desc_mblock_mperblock_nblock_nperblock,
const D0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5
d0_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
const Block2CTileMap block_2_ctile_map,
const index_t batch_count,
const ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch,
const C0MatrixMask c0_matrix_mask)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__) || \
defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
const index_t num_blocks_per_batch =
__builtin_amdgcn_readfirstlane(get_grid_size() / batch_count);
const index_t g_idx = __builtin_amdgcn_readfirstlane(get_block_1d_id() / num_blocks_per_batch);
const long_index_t a_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetABasePtr(g_idx)));
const long_index_t b_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetBBasePtr(g_idx)));
const long_index_t b1_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetB1BasePtr(g_idx)));
const long_index_t c_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetCBasePtr(g_idx)));
const D0DataType* tmp_p_d0_grid = nullptr;
if constexpr(!is_same<D0DataType, void>::value)
{
const long_index_t d0_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(compute_base_ptr_of_batch.GetD0BasePtr(g_idx)));
if(p_d0_grid != nullptr)
{
tmp_p_d0_grid = p_d0_grid + d0_batch_offset;
}
}
GridwiseGemm::template Run<HasMainKBlockLoop>(p_a_grid + a_batch_offset,
p_b_grid + b_batch_offset,
tmp_p_d0_grid,
p_b1_grid + b1_batch_offset,
p_c_grid + c_batch_offset,
p_shared,
a_element_op,
b_element_op,
c0_element_op,
b1_element_op,
c1de_element_op,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
d0_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
b1_grid_desc_bk0_n_bk1,
c1_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_ctile_map,
c0_matrix_mask);
#else
ignore = p_a_grid;
ignore = p_b_grid;
ignore = p_b1_grid;
ignore = p_c_grid;
ignore = p_d0_grid;
ignore = a_element_op;
ignore = b_element_op;
ignore = c0_element_op;
ignore = b1_element_op;
ignore = c1de_element_op;
ignore = a_grid_desc_ak0_m_ak1;
ignore = b_grid_desc_bk0_n_bk1;
ignore = b1_grid_desc_bk0_n_bk1;
ignore = c1_grid_desc_mblock_mperblock_nblock_nperblock;
ignore = d0_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5;
ignore = block_2_ctile_map;
ignore = batch_count;
ignore = compute_base_ptr_of_batch;
ignore = c0_matrix_mask;
#endif // end of if (defined(__gfx908__) || defined(__gfx90a__))
}
// Computes C = A * B0 * B1
// ^^^^^^ (Acc0)
// ^^^^^^^^^^^ (Acc1)
template <index_t NumDimG,
index_t NumDimM,
index_t NumDimN,
index_t NumDimK,
index_t NumDimO, // NumDimGemm1N
typename ADataType,
typename BDataType,
typename B1DataType,
typename CDataType,
typename Acc0BiasDataType,
typename Acc1BiasDataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename C0ElementwiseOperation,
typename B1ElementwiseOperation,
typename C1DEElementwiseOperation,
GemmSpecialization GemmSpec,
TensorSpecialization ASpec,
TensorSpecialization BSpec,
TensorSpecialization B1Spec,
TensorSpecialization CSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock, // Gemm0NPerBlock
index_t KPerBlock, // Gemm0KPerBlock
index_t Gemm1NPerBlock,
index_t Gemm1KPerBlock,
index_t AK1,
index_t BK1,
index_t B1K1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
index_t Gemm1NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t Acc0BiasTransferSrcScalarPerVector,
typename B1BlockTransferThreadClusterLengths_BK0_N_BK1,
typename B1BlockTransferThreadClusterArrangeOrder,
typename B1BlockTransferSrcAccessOrder,
index_t B1BlockTransferSrcVectorDim,
index_t B1BlockTransferSrcScalarPerVector,
index_t B1BlockTransferDstScalarPerVector_BK1,
bool B1BlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
MaskingSpecialization MaskingSpec,
int D0sTransferSrcScalarPerVector = 4,
LoopScheduler LoopSched = LoopScheduler::Default>
struct DeviceBatchedMultiheadAttentionInfer_Xdl_CShuffle
: public DeviceBatchedMultiheadAttentionInfer<NumDimG,
NumDimM,
NumDimN,
NumDimK,
NumDimO,
ADataType,
BDataType,
B1DataType,
CDataType,
Acc0BiasDataType,
Acc1BiasDataType,
AElementwiseOperation,
BElementwiseOperation,
C0ElementwiseOperation,
B1ElementwiseOperation,
C1DEElementwiseOperation,
MaskingSpec>
{
static_assert(NumDimG > 0 && NumDimM > 0 && NumDimN > 0 && NumDimK > 0 && NumDimO > 0,
"Number of dimension must be greater than 0");
using D0DataType = Acc0BiasDataType;
using D1DataType = Acc1BiasDataType;
// TODO: implement bias combination
static_assert(std::is_void<D1DataType>::value, "Acc1 Bias addition is unimplemented");
#if 0
// TODO ANT: use alias
static constexpr index_t NumDimGemm0M = NumDimM;
static constexpr index_t NumDimGemm0N = NumDimN;
static constexpr index_t NumDimGemm0K = NumDimK;
static constexpr index_t NumDimGemm1M = NumDimM;
static constexpr index_t NumDimGemm1N = NumDimO;
static constexpr index_t NumDimGemm1K = NumDimN;
#endif
using DeviceOp = DeviceBatchedMultiheadAttentionInfer_Xdl_CShuffle;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
using Transform = TransformBatchedContractionContractionToBatchedGemmGemm<
Sequence<NumDimG, NumDimM, NumDimN, NumDimK, NumDimO>,
Sequence<MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock>,
GemmSpec,
ASpec,
BSpec,
B1Spec,
CSpec>;
static auto MakeAGridDescriptor_AK0_M_AK1(const std::vector<index_t>& a_gs_ms_ks_lengths_vec,
const std::vector<index_t>& a_gs_ms_ks_strides_vec)
{
return Transform::MakeAGridDescriptor_AK0_M_AK1(
Transform::MakeAGridDescriptor_M_K(a_gs_ms_ks_lengths_vec, a_gs_ms_ks_strides_vec),
Number<AK1>{});
}
static auto MakeBGridDescriptor_BK0_N_BK1(const std::vector<index_t>& b_gs_ns_ks_lengths_vec,
const std::vector<index_t>& b_gs_ns_ks_strides_vec)
{
return Transform::MakeB0GridDescriptor_BK0_N_BK1(
Transform::MakeB0GridDescriptor_N_K(b_gs_ns_ks_lengths_vec, b_gs_ns_ks_strides_vec),
Number<BK1>{});
}
static auto
MakeB1GridDescriptor_BK0_N_BK1(const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_lengths_vec,
const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_strides_vec)
{
return Transform::MakeB1GridDescriptor_BK0_N_BK1(
Transform::MakeB1GridDescriptor_N_K(b1_gs_gemm1ns_gemm1ks_lengths_vec,
b1_gs_gemm1ns_gemm1ks_strides_vec),
Number<B1K1>{});
}
static auto MakeD0GridDescriptor_M_N(const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_lengths,
const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_strides)
{
return Transform::MakeC0GridDescriptor_M_N(acc0_bias_gs_ms_ns_lengths,
acc0_bias_gs_ms_ns_strides);
}
static auto
MakeD0GridDescriptor_G_M_N(const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_lengths,
const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_strides)
{
return Transform::MakeC0GridDescriptor_G_M_N(acc0_bias_gs_ms_ns_lengths,
acc0_bias_gs_ms_ns_strides);
}
using AGridDesc_AK0_M_AK1 = decltype(MakeAGridDescriptor_AK0_M_AK1({}, {}));
using BGridDesc_BK0_N_BK1 = decltype(MakeBGridDescriptor_BK0_N_BK1({}, {}));
using B1GridDesc_BK0_N_BK1 = decltype(MakeB1GridDescriptor_BK0_N_BK1({}, {}));
using C1GridDesc_M_N = decltype(Transform::MakeCGridDescriptor_M_N({}, {}));
using AGridDesc_G_M_K = decltype(Transform::MakeAGridDescriptor_G_M_K({}, {}));
using BGridDesc_G_N_K = decltype(Transform::MakeB0GridDescriptor_G_N_K({}, {}));
using B1GridDesc_G_N_K = decltype(Transform::MakeB1GridDescriptor_G_N_K({}, {}));
using C1GridDesc_G_M_N = decltype(Transform::MakeCGridDescriptor_G_M_N({}, {}));
using D0GridDesc_M_N = decltype(MakeD0GridDescriptor_M_N({}, {}));
using D0GridDesc_G_M_N = decltype(MakeD0GridDescriptor_G_M_N({}, {}));
constexpr static auto make_MaskOutPredicate()
{
if constexpr(MaskingSpec == MaskingSpecialization::MaskDisabled)
{
return MaskDisabledPredicate{};
}
else if constexpr(MaskingSpec == MaskingSpecialization::MaskUpperTriangleFromTopLeft)
{
return MaskUpperTriangleFromTopLeftPredicate{};
}
else if constexpr(MaskingSpec == MaskingSpecialization::MaskUpperTriangleFromBottomRight)
{
return MaskUpperTriangleFromBottomRightPredicate{};
}
}
using C0MatrixMask = C0MatrixMask_impl<decltype(make_MaskOutPredicate())>;
struct ComputeBasePtrOfStridedBatch
{
ComputeBasePtrOfStridedBatch() {}
ComputeBasePtrOfStridedBatch(const AGridDesc_G_M_K& a_grid_desc_g_m_k,
const BGridDesc_G_N_K& b_grid_desc_g_n_k,
const B1GridDesc_G_N_K& b1_grid_desc_g_n_k,
const C1GridDesc_G_M_N& c1_grid_desc_g_m_n,
const D0GridDesc_G_M_N& d0_grid_desc_g_m_n)
: a_grid_desc_g_m_k_(a_grid_desc_g_m_k),
b_grid_desc_g_n_k_(b_grid_desc_g_n_k),
b1_grid_desc_g_n_k_(b1_grid_desc_g_n_k),
c1_grid_desc_g_m_n_(c1_grid_desc_g_m_n),
d0_grid_desc_g_m_n_(d0_grid_desc_g_m_n)
{
}
__host__ __device__ constexpr long_index_t GetABasePtr(index_t g_idx) const
{
return a_grid_desc_g_m_k_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
__host__ __device__ constexpr long_index_t GetBBasePtr(index_t g_idx) const
{
return b_grid_desc_g_n_k_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
__host__ __device__ constexpr long_index_t GetD0BasePtr(index_t g_idx) const
{
return d0_grid_desc_g_m_n_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
__host__ __device__ constexpr long_index_t GetB1BasePtr(index_t g_idx) const
{
return b1_grid_desc_g_n_k_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
__host__ __device__ constexpr long_index_t GetCBasePtr(index_t g_idx) const
{
return c1_grid_desc_g_m_n_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
private:
AGridDesc_G_M_K a_grid_desc_g_m_k_;
BGridDesc_G_N_K b_grid_desc_g_n_k_;
B1GridDesc_G_N_K b1_grid_desc_g_n_k_;
C1GridDesc_G_M_N c1_grid_desc_g_m_n_;
D0GridDesc_G_M_N d0_grid_desc_g_m_n_;
};
using GridwiseGemm = GridwiseMultiHeadFlashAttentionInfer_Xdl_CShuffle<
ADataType, // TODO: distinguish A/B datatype
D0DataType,
GemmAccDataType,
CShuffleDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
C0ElementwiseOperation,
B1ElementwiseOperation,
C1DEElementwiseOperation,
InMemoryDataOperationEnum::Set,
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
D0GridDesc_M_N,
B1GridDesc_BK0_N_BK1,
C1GridDesc_M_N,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
Gemm1NPerBlock,
Gemm1KPerBlock,
AK1,
BK1,
B1K1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
Gemm1NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
true,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
true,
BBlockLdsExtraN,
Acc0BiasTransferSrcScalarPerVector,
B1BlockTransferThreadClusterLengths_BK0_N_BK1,
B1BlockTransferThreadClusterArrangeOrder,
B1BlockTransferSrcAccessOrder,
B1BlockTransferSrcVectorDim,
B1BlockTransferSrcScalarPerVector,
B1BlockTransferDstScalarPerVector_BK1,
false,
B1BlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopSched,
Transform::matrix_padder.PadN,
MaskingSpec != MaskingSpecialization::MaskDisabled>;
// Argument
// FIXME: constness
struct Argument : public BaseArgument
{
Argument(const ADataType* p_a_grid,
const BDataType* p_b_grid,
const B1DataType* p_b1_grid,
CDataType* p_c_grid,
const Acc0BiasDataType* p_acc0_bias,
const Acc1BiasDataType* p_acc1_bias,
const std::vector<index_t>& a_gs_ms_ks_lengths,
const std::vector<index_t>& a_gs_ms_ks_strides,
const std::vector<index_t>& b_gs_ns_ks_lengths,
const std::vector<index_t>& b_gs_ns_ks_strides,
const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_lengths, // b1_gs_os_ns_lengths
const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_strides, // b1_gs_os_ns_strides
const std::vector<index_t>& c_gs_ms_gemm1ns_lengths, // c_gs_ms_os_lengths
const std::vector<index_t>& c_gs_ms_gemm1ns_strides, // c_gs_ms_os_strides
const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_lengths,
const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_strides,
const std::vector<ck::index_t>&
acc1_bias_gs_ms_gemm1ns_lengths, // acc1_biases_gs_ms_os_lengths
const std::vector<ck::index_t>&
acc1_bias_gs_ms_gemm1ns_strides, // acc1_biases_gs_ms_os_strides
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
C0ElementwiseOperation c0_element_op,
B1ElementwiseOperation b1_element_op,
C1DEElementwiseOperation c1de_element_op)
: p_a_grid_{p_a_grid},
p_b_grid_{p_b_grid},
p_b1_grid_{p_b1_grid},
p_c_grid_{p_c_grid},
p_d0_grid_{p_acc0_bias},
a_grid_desc_ak0_m_ak1_{
DeviceOp::MakeAGridDescriptor_AK0_M_AK1(a_gs_ms_ks_lengths, a_gs_ms_ks_strides)},
b_grid_desc_bk0_n_bk1_{
DeviceOp::MakeBGridDescriptor_BK0_N_BK1(b_gs_ns_ks_lengths, b_gs_ns_ks_strides)},
b1_grid_desc_bk0_n_bk1_{DeviceOp::MakeB1GridDescriptor_BK0_N_BK1(
b1_gs_gemm1ns_gemm1ks_lengths, b1_gs_gemm1ns_gemm1ks_strides)},
c1_grid_desc_m_n_{Transform::MakeCGridDescriptor_M_N(c_gs_ms_gemm1ns_lengths,
c_gs_ms_gemm1ns_strides)},
a_grid_desc_g_m_k_{
Transform::MakeAGridDescriptor_G_M_K(a_gs_ms_ks_lengths, a_gs_ms_ks_strides)},
b_grid_desc_g_n_k_{
Transform::MakeB0GridDescriptor_G_N_K(b_gs_ns_ks_lengths, b_gs_ns_ks_strides)},
b1_grid_desc_g_n_k_{Transform::MakeB1GridDescriptor_G_N_K(
b1_gs_gemm1ns_gemm1ks_lengths, b1_gs_gemm1ns_gemm1ks_strides)},
c1_grid_desc_g_m_n_{Transform::MakeCGridDescriptor_G_M_N(c_gs_ms_gemm1ns_lengths,
c_gs_ms_gemm1ns_strides)},
c1_grid_desc_mblock_mperblock_nblock_nperblock_{},
// d0_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_{},
block_2_ctile_map_{GridwiseGemm::MakeDefaultBlock2CTileMap(c1_grid_desc_m_n_)},
a_element_op_{a_element_op},
b_element_op_{b_element_op},
c0_element_op_{c0_element_op},
b1_element_op_{b1_element_op},
c1de_element_op_{c1de_element_op},
c0_matrix_mask_{a_grid_desc_g_m_k_.GetLength(I1), b_grid_desc_g_n_k_.GetLength(I1)},
raw_lengths_mz_nz_kz_gemm1nz_{a_gs_ms_ks_lengths[NumDimG + NumDimM - 1],
b_gs_ns_ks_lengths[NumDimG + NumDimN - 1],
b_gs_ns_ks_lengths[NumDimG + NumDimN + NumDimK - 1],
b1_gs_gemm1ns_gemm1ks_lengths[NumDimG + NumDimO - 1]},
a_mz_kz_strides_{a_gs_ms_ks_strides[NumDimG + NumDimM - 1],
a_gs_ms_ks_strides[NumDimG + NumDimM + NumDimK - 1]},
b_nz_kz_strides_{b_gs_ns_ks_strides[NumDimG + NumDimN - 1],
b_gs_ns_ks_strides[NumDimG + NumDimN + NumDimK - 1]},
b1_nz_kz_strides_{b1_gs_gemm1ns_gemm1ks_strides[NumDimG + NumDimO - 1],
b1_gs_gemm1ns_gemm1ks_strides[NumDimG + NumDimO + NumDimN - 1]},
c_mz_gemm1nz_strides_{c_gs_ms_gemm1ns_strides[NumDimG + NumDimM - 1],
c_gs_ms_gemm1ns_strides[NumDimG + NumDimM + NumDimO - 1]},
batch_count_{c1_grid_desc_g_m_n_.GetLength(I0)}
{
// TODO ANT: implement bias addition
ignore = p_acc1_bias;
ignore = acc1_bias_gs_ms_gemm1ns_lengths;
ignore = acc1_bias_gs_ms_gemm1ns_strides;
if(GridwiseGemm::CheckValidity(a_grid_desc_ak0_m_ak1_,
b_grid_desc_bk0_n_bk1_,
b1_grid_desc_bk0_n_bk1_,
c1_grid_desc_m_n_,
block_2_ctile_map_))
{
c1_grid_desc_mblock_mperblock_nblock_nperblock_ =
GridwiseGemm::MakeC1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c1_grid_desc_m_n_);
if constexpr(!is_same<D0DataType, void>::value)
{
D0GridDesc_M_N d0_grid_desc_m_n_ = MakeD0GridDescriptor_M_N(
acc0_bias_gs_ms_ns_lengths, acc0_bias_gs_ms_ns_strides);
d0_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_ =
GridwiseGemm::MakeD0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5(
d0_grid_desc_m_n_);
d0_grid_desc_g_m_n_ = MakeD0GridDescriptor_G_M_N(acc0_bias_gs_ms_ns_lengths,
acc0_bias_gs_ms_ns_strides);
d0_n_length_stride_.push_back(acc0_bias_gs_ms_ns_lengths[NumDimG + NumDimM]);
d0_n_length_stride_.push_back(acc0_bias_gs_ms_ns_strides[NumDimG + NumDimM]);
}
compute_base_ptr_of_batch_ = ComputeBasePtrOfStridedBatch(a_grid_desc_g_m_k_,
b_grid_desc_g_n_k_,
b1_grid_desc_g_n_k_,
c1_grid_desc_g_m_n_,
d0_grid_desc_g_m_n_);
}
}
void Print() const
{
std::cout << "a_grid_desc_g_m_k_: " << a_grid_desc_g_m_k_.GetLength(I0) << ", "
<< a_grid_desc_g_m_k_.GetLength(I1) << ", "
<< a_grid_desc_g_m_k_.GetLength(I2) << '\n';
std::cout << "b_grid_desc_g_n_k_: " << b_grid_desc_g_n_k_.GetLength(I0) << ", "
<< b_grid_desc_g_n_k_.GetLength(I1) << ", "
<< b_grid_desc_g_n_k_.GetLength(I2) << '\n';
std::cout << "b1_grid_desc_g_n_k_: " << b1_grid_desc_g_n_k_.GetLength(I0) << ", "
<< b1_grid_desc_g_n_k_.GetLength(I1) << ", "
<< b1_grid_desc_g_n_k_.GetLength(I2) << '\n';
std::cout << "c1_grid_desc_g_m_n_: " << c1_grid_desc_g_m_n_.GetLength(I0) << ", "
<< c1_grid_desc_g_m_n_.GetLength(I1) << ", "
<< c1_grid_desc_g_m_n_.GetLength(I2) << '\n';
}
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
const B1DataType* p_b1_grid_;
CDataType* p_c_grid_;
const D0DataType* p_d0_grid_;
// tensor descriptor
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1_;
C1GridDesc_M_N c1_grid_desc_m_n_;
AGridDesc_G_M_K a_grid_desc_g_m_k_;
BGridDesc_G_N_K b_grid_desc_g_n_k_;
B1GridDesc_G_N_K b1_grid_desc_g_n_k_;
C1GridDesc_G_M_N c1_grid_desc_g_m_n_;
D0GridDesc_G_M_N d0_grid_desc_g_m_n_;
typename GridwiseGemm::C1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c1_grid_desc_mblock_mperblock_nblock_nperblock_;
typename GridwiseGemm::D0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5
d0_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_;
// block-to-c-tile map
typename GridwiseGemm::DefaultBlock2CTileMap block_2_ctile_map_;
// element-wise op
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
C0ElementwiseOperation c0_element_op_;
B1ElementwiseOperation b1_element_op_;
C1DEElementwiseOperation c1de_element_op_;
// check C0 masking and padding
C0MatrixMask c0_matrix_mask_;
// For robust IsSupportedArgument() check
std::vector<index_t> raw_lengths_mz_nz_kz_gemm1nz_;
std::vector<index_t> a_mz_kz_strides_;
std::vector<index_t> b_nz_kz_strides_;
std::vector<index_t> b1_nz_kz_strides_;
std::vector<index_t> c_mz_gemm1nz_strides_;
std::vector<ck::index_t> d0s_nl_ns_lengths_strides_;
index_t batch_count_;
ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch_;
// raw data
std::vector<ck::index_t> d0_n_length_stride_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!DeviceOp::IsSupportedArgument(arg))
{
throw std::runtime_error("wrong! unsupported argument");
}
const index_t grid_size =
arg.block_2_ctile_map_.CalculateGridSize(arg.c1_grid_desc_m_n_) * arg.batch_count_;
// Gemm0_K
const auto K =
arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) * arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
float ave_time = 0;
auto launch_kernel = [&](auto has_main_k_block_loop_) {
const auto kernel = kernel_batched_multiple_head_flash_attention_infer<
GridwiseGemm,
ADataType, // TODO: distiguish A/B datatype
D0DataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
C0ElementwiseOperation,
B1ElementwiseOperation,
C1DEElementwiseOperation,
DeviceOp::AGridDesc_AK0_M_AK1,
DeviceOp::BGridDesc_BK0_N_BK1,
DeviceOp::B1GridDesc_BK0_N_BK1,
typename GridwiseGemm::C1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock,
typename GridwiseGemm::D0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5,
typename GridwiseGemm::DefaultBlock2CTileMap,
ComputeBasePtrOfStridedBatch,
C0MatrixMask,
has_main_k_block_loop_>;
return launch_and_time_kernel(stream_config,
kernel,
dim3(grid_size),
dim3(BlockSize),
0,
arg.p_a_grid_,
arg.p_b_grid_,
arg.p_d0_grid_,
arg.p_b1_grid_,
arg.p_c_grid_,
arg.a_element_op_,
arg.b_element_op_,
arg.c0_element_op_,
arg.b1_element_op_,
arg.c1de_element_op_,
arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.c1_grid_desc_mblock_mperblock_nblock_nperblock_,
arg.d0_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_,
arg.block_2_ctile_map_,
arg.batch_count_,
arg.compute_base_ptr_of_batch_,
arg.c0_matrix_mask_);
};
// Gemm1_K is split into Gemm1_K0/K1 where K1 is known at compile time, so we only need
// to concern Gemm0's loop
if(GridwiseGemm::CalculateHasMainKBlockLoop(K))
{
ave_time = launch_kernel(integral_constant<bool, true>{});
}
else
{
ave_time = launch_kernel(integral_constant<bool, false>{});
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
#if DEBUG_LOG
arg.Print();
#endif
if(!ck::is_xdl_supported())
{
return false;
}
// TODO ANT: Check if tensor specialization & strides mismatch
// Check if C permute dimension matches GEMM + GEMM shape
const index_t c_g = arg.c1_grid_desc_g_m_n_.GetLength(I0); // unpadded
const index_t c_m = arg.c1_grid_desc_m_n_.GetLength(I0);
const index_t c_gemm1n = arg.c1_grid_desc_m_n_.GetLength(I1);
const index_t a_m = arg.a_grid_desc_ak0_m_ak1_.GetLength(I1);
const index_t b1_gemm1n = arg.b1_grid_desc_bk0_n_bk1_.GetLength(I1);
if(!(c_g == arg.batch_count_ && c_m == a_m && c_gemm1n == b1_gemm1n))
{
return false;
}
if constexpr(!is_same<D0DataType, void>::value)
{
if(arg.d0_n_length_stride_[1] == 1)
{
if(arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector != 0)
return false;
}
else if(Acc0BiasTransferSrcScalarPerVector != 1)
{
return false;
}
}
// Note: we need raw lengths since threadwise copy can not handle vector load when part of
// vector is out of bounds
// Note: need lowest dim in Ms/Ns/Ks/Os, not merged M/N/K/O
const auto MzRaw = arg.raw_lengths_mz_nz_kz_gemm1nz_[0];
const auto NzRaw = arg.raw_lengths_mz_nz_kz_gemm1nz_[1];
const auto KzRaw = arg.raw_lengths_mz_nz_kz_gemm1nz_[2];
const auto Gemm1NzRaw = arg.raw_lengths_mz_nz_kz_gemm1nz_[3];
// Check scalar per vector requirement
const auto a_extent_lowest = ABlockTransferSrcVectorDim == 2 ? KzRaw : MzRaw;
const auto b_extent_lowest = BBlockTransferSrcVectorDim == 2 ? KzRaw : NzRaw;
const auto b1_extent_lowest = B1BlockTransferSrcVectorDim == 2 ? NzRaw : Gemm1NzRaw;
const auto c_extent_lowest = Gemm1NzRaw;
if(!(a_extent_lowest % ABlockTransferSrcScalarPerVector == 0 &&
b_extent_lowest % BBlockTransferSrcScalarPerVector == 0 &&
b1_extent_lowest % B1BlockTransferSrcScalarPerVector == 0 &&
c_extent_lowest % CShuffleBlockTransferScalarPerVector_NPerBlock == 0))
{
return false;
}
// Check vector load/store requirement
const auto a_stride_lowest =
ABlockTransferSrcVectorDim == 2 ? arg.a_mz_kz_strides_[1] : arg.a_mz_kz_strides_[0];
const auto b_stride_lowest =
BBlockTransferSrcVectorDim == 2 ? arg.b_nz_kz_strides_[1] : arg.b_nz_kz_strides_[0];
const auto b1_stride_lowest =
B1BlockTransferSrcVectorDim == 2 ? arg.b1_nz_kz_strides_[1] : arg.b1_nz_kz_strides_[0];
const auto c_stride_lowest =
arg.c_mz_gemm1nz_strides_[1]; // cshuffle assumes lowest dim in Gemm1Ns to be contiguous
if(!(a_stride_lowest == 1 || b_stride_lowest == 1 || b1_stride_lowest == 1 ||
c_stride_lowest == 1))
{
return false;
}
return GridwiseGemm::CheckValidity(arg.a_grid_desc_ak0_m_ak1_,
arg.b_grid_desc_bk0_n_bk1_,
arg.b1_grid_desc_bk0_n_bk1_,
arg.c1_grid_desc_m_n_,
arg.block_2_ctile_map_);
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto
MakeArgument(const ADataType* p_a,
const BDataType* p_b,
const B1DataType* p_b1,
CDataType* p_c,
const Acc0BiasDataType* p_acc0_bias,
const Acc1BiasDataType* p_acc1_bias,
const std::vector<index_t>& a_gs_ms_ks_lengths,
const std::vector<index_t>& a_gs_ms_ks_strides,
const std::vector<index_t>& b_gs_ns_ks_lengths,
const std::vector<index_t>& b_gs_ns_ks_strides,
const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_lengths, // b1_gs_os_ns_lengths
const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_strides, // b1_gs_os_ns_strides
const std::vector<index_t>& c_gs_ms_gemm1ns_lengths, // c_gs_ms_os_lengths
const std::vector<index_t>& c_gs_ms_gemm1ns_strides, // c_gs_ms_os_strides
const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_lengths,
const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_strides,
const std::vector<ck::index_t>&
acc1_bias_gs_ms_gemm1ns_lengths, // acc1_biases_gs_ms_os_lengths
const std::vector<ck::index_t>&
acc1_bias_gs_ms_gemm1ns_strides, // acc1_biases_gs_ms_os_strides
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
C0ElementwiseOperation c0_element_op,
B1ElementwiseOperation b1_element_op,
C1DEElementwiseOperation c1de_element_op)
{
return Argument{p_a,
p_b,
p_b1,
p_c,
p_acc0_bias,
p_acc1_bias,
a_gs_ms_ks_lengths,
a_gs_ms_ks_strides,
b_gs_ns_ks_lengths,
b_gs_ns_ks_strides,
b1_gs_gemm1ns_gemm1ks_lengths, // b1_gs_os_ns_lengths
b1_gs_gemm1ns_gemm1ks_strides, // b1_gs_os_ns_strides
c_gs_ms_gemm1ns_lengths, // c_gs_ms_os_lengths
c_gs_ms_gemm1ns_strides, // c_gs_ms_os_strides
acc0_bias_gs_ms_ns_lengths,
acc0_bias_gs_ms_ns_strides,
acc1_bias_gs_ms_gemm1ns_lengths, // acc1_biases_gs_ms_os_lengths
acc1_bias_gs_ms_gemm1ns_strides, // acc1_biases_gs_ms_os_strides
a_element_op,
b_element_op,
c0_element_op,
b1_element_op,
c1de_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
// FIXME: constness
std::unique_ptr<BaseArgument> MakeArgumentPointer(
const void* p_a,
const void* p_b,
const void* p_b1,
void* p_c,
const void* p_acc0_bias,
const void* p_acc1_bias,
const std::vector<index_t>& a_gs_ms_ks_lengths,
const std::vector<index_t>& a_gs_ms_ks_strides,
const std::vector<index_t>& b_gs_ns_ks_lengths,
const std::vector<index_t>& b_gs_ns_ks_strides,
const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_lengths, // b1_gs_os_ns_lengths
const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_strides, // b1_gs_os_ns_strides
const std::vector<index_t>& c_gs_ms_gemm1ns_lengths, // c_gs_ms_os_lengths
const std::vector<index_t>& c_gs_ms_gemm1ns_strides, // c_gs_ms_os_strides
const std::vector<index_t>& acc0_bias_gs_ms_ns_lengths,
const std::vector<index_t>& acc0_bias_gs_ms_ns_strides,
const std::vector<ck::index_t>&
acc1_bias_gs_ms_gemm1ns_lengths, // acc1_biases_gs_ms_os_lengths
const std::vector<ck::index_t>&
acc1_bias_gs_ms_gemm1ns_strides, // acc1_biases_gs_ms_os_strides
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
C0ElementwiseOperation c0_element_op,
B1ElementwiseOperation b1_element_op,
C1DEElementwiseOperation c1de_element_op) override
{
return std::make_unique<Argument>(
static_cast<const ADataType*>(p_a),
static_cast<const BDataType*>(p_b),
static_cast<const B1DataType*>(p_b1),
static_cast<CDataType*>(p_c),
static_cast<const Acc0BiasDataType*>(p_acc0_bias), // cast in struct Argument
static_cast<const Acc1BiasDataType*>(p_acc1_bias), // cast in struct Argument
a_gs_ms_ks_lengths,
a_gs_ms_ks_strides,
b_gs_ns_ks_lengths,
b_gs_ns_ks_strides,
b1_gs_gemm1ns_gemm1ks_lengths, // b1_gs_os_ns_lengths
b1_gs_gemm1ns_gemm1ks_strides, // b1_gs_os_ns_strides
c_gs_ms_gemm1ns_lengths, // c_gs_ms_os_lengths
c_gs_ms_gemm1ns_strides, // c_gs_ms_os_strides
acc0_bias_gs_ms_ns_lengths,
acc0_bias_gs_ms_ns_strides,
acc1_bias_gs_ms_gemm1ns_lengths,
acc1_bias_gs_ms_gemm1ns_strides,
a_element_op,
b_element_op,
c0_element_op,
b1_element_op,
c1de_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceBatchedMultiheadAttentionInfer_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< MPerBlock << ", "
<< Gemm1NPerBlock << ", "
<< Gemm1KPerBlock << ", "
<< B1K1 << ", "
<< getGemmSpecializationString(GemmSpec) << ", "
<< "ASpec" << getTensorSpecializationString(ASpec) << ", "
<< "B0Spec" << getTensorSpecializationString(BSpec) << ", "
<< "B1Spec" << getTensorSpecializationString(B1Spec) << ", "
<< "CSpec" << getTensorSpecializationString(CSpec) << ", "
<< getMaskingSpecializationString(MaskingSpec) << ">";
// clang-format on
return str.str();
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
include/ck/tensor_operation/gpu/device/impl/device_grouped_mha_fwd_xdl_cshuffle_v2.hpp
View file @ b23b3d71
...@@ -416,7 +416,7 @@ struct DeviceGroupedMultiheadAttentionForward_Xdl_CShuffle_V2 ...@@ -416,7 +416,7 @@ struct DeviceGroupedMultiheadAttentionForward_Xdl_CShuffle_V2
const std::vector<ck::index_t>& acc0_biases_gs_ms_ns_strides) const std::vector<ck::index_t>& acc0_biases_gs_ms_ns_strides)
{ {
return Transform::MakeCGridDescriptor_M_N(acc0_biases_gs_ms_ns_lengths, return Transform::MakeC0GridDescriptor_M_N(acc0_biases_gs_ms_ns_lengths,
acc0_biases_gs_ms_ns_strides); acc0_biases_gs_ms_ns_strides);
} }
...@@ -425,7 +425,7 @@ struct DeviceGroupedMultiheadAttentionForward_Xdl_CShuffle_V2 ...@@ -425,7 +425,7 @@ struct DeviceGroupedMultiheadAttentionForward_Xdl_CShuffle_V2
const std::vector<ck::index_t>& acc0_biases_gs_ms_ns_strides) const std::vector<ck::index_t>& acc0_biases_gs_ms_ns_strides)
{ {
return Transform::MakeCGridDescriptor_G_M_N(acc0_biases_gs_ms_ns_lengths, return Transform::MakeC0GridDescriptor_G_M_N(acc0_biases_gs_ms_ns_lengths,
acc0_biases_gs_ms_ns_strides); acc0_biases_gs_ms_ns_strides);
} }
...@@ -1102,13 +1102,12 @@ struct DeviceGroupedMultiheadAttentionForward_Xdl_CShuffle_V2 ...@@ -1102,13 +1102,12 @@ struct DeviceGroupedMultiheadAttentionForward_Xdl_CShuffle_V2
if constexpr(!is_same<D0DataType, void>::value) if constexpr(!is_same<D0DataType, void>::value)
{ {
if(device_arg.d0_n_length_stride_[1] == 1 && if(device_arg.d0_n_length_stride_[1] == 1)
device_arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector != 0)
{ {
if(device_arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector != 0)
return false; return false;
} }
if(device_arg.d0_n_length_stride_[1] != 1 && else if(Acc0BiasTransferSrcScalarPerVector != 1)
Acc0BiasTransferSrcScalarPerVector != 1)
{ {
return false; return false;
} }
......
include/ck/tensor_operation/gpu/device/impl/device_grouped_mha_infer_xdl_cshuffle.hpp 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include <iostream>
#include <sstream>
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
#include "ck/tensor_operation/gpu/device/device_grouped_mha_infer.hpp"
#include "ck/tensor_operation/gpu/device/gemm_specialization.hpp"
#include "ck/tensor_operation/gpu/device/matrix_padder.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_batched_mha_infer_xdl_cshuffle.hpp"
#include "ck/tensor_operation/operator_transform/transform_contraction_to_gemm.hpp"
#include "ck/host_utility/device_prop.hpp"
#include "ck/host_utility/kernel_launch.hpp"
namespace ck {
namespace tensor_operation {
namespace device {
template <typename GridwiseGemm,
typename D0DataType,
typename GroupKernelArg,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename AccElementwiseOperation,
typename B1ElementwiseOperation,
typename CElementwiseOperation,
bool HasMainKBlockLoop>
__global__ void
#if CK_USE_LAUNCH_BOUNDS
__launch_bounds__(CK_MAX_THREAD_PER_BLOCK, CK_MIN_BLOCK_PER_CU)
#endif
kernel_grouped_multiple_head_flash_attention_infer(
const void CK_CONSTANT_ADDRESS_SPACE* group_kernel_args,
const index_t group_count,
const AElementwiseOperation a_element_op,
const BElementwiseOperation b_element_op,
const AccElementwiseOperation acc_element_op,
const B1ElementwiseOperation b1_element_op,
const CElementwiseOperation c_element_op)
{
#if(!defined(__HIP_DEVICE_COMPILE__) || defined(__gfx908__) || defined(__gfx90a__) || \
defined(__gfx940__) || defined(__gfx941__) || defined(__gfx942__))
__shared__ char p_shared[GridwiseGemm::GetSharedMemoryNumberOfByte()];
const index_t block_id = get_block_1d_id();
const auto arg_ptr = reinterpret_cast<const GroupKernelArg*>(
cast_pointer_to_generic_address_space(group_kernel_args));
index_t left = 0;
index_t right = group_count;
index_t group_id = index_t((left + right) / 2);
while(
(!(block_id >= arg_ptr[group_id].block_start_ && block_id < arg_ptr[group_id].block_end_)))
{
if(block_id < arg_ptr[group_id].block_start_)
{
right = group_id;
}
else
{
left = group_id;
}
group_id = index_t((left + right) / 2);
}
// per-group batch offset
const index_t num_blocks_per_batch = arg_ptr[group_id].num_blocks_per_batch_;
const index_t g_idx = __builtin_amdgcn_readfirstlane(
(block_id - arg_ptr[group_id].block_start_) / num_blocks_per_batch);
const long_index_t a_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(arg_ptr[group_id].compute_base_ptr_of_batch_.GetABasePtr(g_idx)));
const long_index_t b_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(arg_ptr[group_id].compute_base_ptr_of_batch_.GetBBasePtr(g_idx)));
const long_index_t b1_batch_offset = __builtin_amdgcn_readfirstlane(static_cast<long_index_t>(
arg_ptr[group_id].compute_base_ptr_of_batch_.GetB1BasePtr(g_idx)));
const long_index_t c_batch_offset = __builtin_amdgcn_readfirstlane(
static_cast<long_index_t>(arg_ptr[group_id].compute_base_ptr_of_batch_.GetCBasePtr(g_idx)));
const D0DataType* tmp_p_d0_grid = nullptr;
if constexpr(!is_same<D0DataType, void>::value)
{
const long_index_t d0_batch_offset =
__builtin_amdgcn_readfirstlane(static_cast<long_index_t>(
arg_ptr[group_id].compute_base_ptr_of_batch_.GetD0BasePtr(g_idx)));
tmp_p_d0_grid = arg_ptr[group_id].p_d0_grid_ + d0_batch_offset;
}
GridwiseGemm::template Run<HasMainKBlockLoop>(
arg_ptr[group_id].p_a_grid_ + a_batch_offset,
arg_ptr[group_id].p_b_grid_ + b_batch_offset,
tmp_p_d0_grid,
arg_ptr[group_id].p_b1_grid_ + b1_batch_offset,
arg_ptr[group_id].p_c_grid_ + c_batch_offset,
p_shared,
a_element_op,
b_element_op,
acc_element_op,
b1_element_op,
c_element_op,
arg_ptr[group_id].a_grid_desc_ak0_m_ak1_,
arg_ptr[group_id].b_grid_desc_bk0_n_bk1_,
arg_ptr[group_id].d0_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_,
arg_ptr[group_id].b1_grid_desc_bk0_n_bk1_,
arg_ptr[group_id].c_grid_desc_mblock_mperblock_nblock_nperblock_,
arg_ptr[group_id].block_2_ctile_map_,
arg_ptr[group_id].c0_matrix_mask_);
#else
ignore = group_kernel_args;
ignore = group_count;
ignore = a_element_op;
ignore = b_element_op;
ignore = acc_element_op;
ignore = b1_element_op;
ignore = c_element_op;
#endif // end of if (defined(__gfx908__) || defined(__gfx90a__))
}
// Computes C = A * B0 * B1
// ^^^^^^ (Acc0)
// ^^^^^^^^^^^ (Acc1)
template <index_t NumDimG,
index_t NumDimM,
index_t NumDimN,
index_t NumDimK,
index_t NumDimO, // NumDimGemm1N
typename ADataType,
typename BDataType,
typename B1DataType,
typename CDataType,
typename Acc0BiasDataType,
typename Acc1BiasDataType,
typename GemmAccDataType,
typename CShuffleDataType,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename AccElementwiseOperation,
typename B1ElementwiseOperation,
typename CElementwiseOperation,
GemmSpecialization GemmSpec,
TensorSpecialization ASpec,
TensorSpecialization BSpec,
TensorSpecialization B1Spec,
TensorSpecialization CSpec,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock, // Gemm0NPerBlock
index_t KPerBlock, // Gemm0KPerBlock
index_t Gemm1NPerBlock,
index_t Gemm1KPerBlock,
index_t AK1,
index_t BK1,
index_t B1K1,
index_t MPerXDL,
index_t NPerXDL,
index_t MXdlPerWave,
index_t NXdlPerWave,
index_t Gemm1NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BBlockLdsExtraN,
index_t Acc0BiasTransferSrcScalarPerVector,
typename B1BlockTransferThreadClusterLengths_BK0_N_BK1,
typename B1BlockTransferThreadClusterArrangeOrder,
typename B1BlockTransferSrcAccessOrder,
index_t B1BlockTransferSrcVectorDim,
index_t B1BlockTransferSrcScalarPerVector,
index_t B1BlockTransferDstScalarPerVector_BK1,
bool B1BlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
MaskingSpecialization MaskingSpec,
LoopScheduler LoopSched = LoopScheduler::Default>
struct DeviceGroupedMultiheadAttentionInfer_Xdl_CShuffle
: public DeviceGroupedMultiheadAttentionInfer<NumDimG,
NumDimM,
NumDimN,
NumDimK,
NumDimO,
ADataType,
BDataType,
B1DataType,
CDataType,
Acc0BiasDataType,
Acc1BiasDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation,
MaskingSpec>
{
static_assert(NumDimG > 0 && NumDimM > 0 && NumDimN > 0 && NumDimK > 0 && NumDimO > 0,
"Number of dimension must be greater than 0");
using D0DataType = Acc0BiasDataType;
using D1DataType = Acc1BiasDataType;
// TODO ANT: implement bias combination
static_assert(std::is_void<Acc1BiasDataType>::value, "Acc1 Bias addition is unimplemented");
#if 0
// TODO ANT: use alias
static constexpr index_t NumDimGemm0M = NumDimM;
static constexpr index_t NumDimGemm0N = NumDimN;
static constexpr index_t NumDimGemm0K = NumDimK;
static constexpr index_t NumDimGemm1M = NumDimM;
static constexpr index_t NumDimGemm1N = NumDimO;
static constexpr index_t NumDimGemm1K = NumDimN;
#endif
using DeviceOp = DeviceGroupedMultiheadAttentionInfer_Xdl_CShuffle;
using ProblemDesc = typename DeviceGroupedMultiheadAttentionInfer<NumDimG,
NumDimM,
NumDimN,
NumDimK,
NumDimO,
ADataType,
BDataType,
B1DataType,
CDataType,
Acc0BiasDataType,
Acc1BiasDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation,
MaskingSpec>::ProblemDesc;
static constexpr auto I0 = Number<0>{};
static constexpr auto I1 = Number<1>{};
static constexpr auto I2 = Number<2>{};
using Transform = TransformBatchedContractionContractionToBatchedGemmGemm<
Sequence<NumDimG, NumDimM, NumDimN, NumDimK, NumDimO>,
Sequence<MPerBlock, NPerBlock, KPerBlock, Gemm1NPerBlock>,
GemmSpec,
ASpec,
BSpec,
B1Spec,
CSpec>;
static auto MakeAGridDescriptor_AK0_M_AK1(const std::vector<index_t>& a_gs_ms_ks_lengths_vec,
const std::vector<index_t>& a_gs_ms_ks_strides_vec)
{
return Transform::MakeAGridDescriptor_AK0_M_AK1(
Transform::MakeAGridDescriptor_M_K(a_gs_ms_ks_lengths_vec, a_gs_ms_ks_strides_vec),
Number<AK1>{});
}
static auto MakeBGridDescriptor_BK0_N_BK1(const std::vector<index_t>& b_gs_ns_ks_lengths_vec,
const std::vector<index_t>& b_gs_ns_ks_strides_vec)
{
return Transform::MakeB0GridDescriptor_BK0_N_BK1(
Transform::MakeB0GridDescriptor_N_K(b_gs_ns_ks_lengths_vec, b_gs_ns_ks_strides_vec),
Number<BK1>{});
}
static auto
MakeB1GridDescriptor_BK0_N_BK1(const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_lengths_vec,
const std::vector<index_t>& b1_gs_gemm1ns_gemm1ks_strides_vec)
{
return Transform::MakeB1GridDescriptor_BK0_N_BK1(
Transform::MakeB1GridDescriptor_N_K(b1_gs_gemm1ns_gemm1ks_lengths_vec,
b1_gs_gemm1ns_gemm1ks_strides_vec),
Number<B1K1>{});
}
static auto MakeD0GridDescriptor_M_N(const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_lengths,
const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_strides)
{
return Transform::MakeC0GridDescriptor_M_N(acc0_bias_gs_ms_ns_lengths,
acc0_bias_gs_ms_ns_strides);
}
static auto
MakeD0GridDescriptor_G_M_N(const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_lengths,
const std::vector<ck::index_t>& acc0_bias_gs_ms_ns_strides)
{
return Transform::MakeC0GridDescriptor_G_M_N(acc0_bias_gs_ms_ns_lengths,
acc0_bias_gs_ms_ns_strides);
}
using AGridDesc_AK0_M_AK1 = decltype(MakeAGridDescriptor_AK0_M_AK1({}, {}));
using BGridDesc_BK0_N_BK1 = decltype(MakeBGridDescriptor_BK0_N_BK1({}, {}));
using D0GridDesc_M_N = decltype(MakeD0GridDescriptor_M_N({}, {}));
using B1GridDesc_BK0_N_BK1 = decltype(MakeB1GridDescriptor_BK0_N_BK1({}, {}));
using C1GridDesc_M_N = decltype(Transform::MakeCGridDescriptor_M_N({}, {}));
using AGridDesc_G_M_K = decltype(Transform::MakeAGridDescriptor_G_M_K({}, {}));
using BGridDesc_G_N_K = decltype(Transform::MakeB0GridDescriptor_G_N_K({}, {}));
using D0GridDesc_G_M_N = decltype(MakeD0GridDescriptor_G_M_N({}, {}));
using B1GridDesc_G_N_K = decltype(Transform::MakeB1GridDescriptor_G_N_K({}, {}));
using C1GridDesc_G_M_N = decltype(Transform::MakeCGridDescriptor_G_M_N({}, {}));
constexpr static auto make_MaskOutPredicate()
{
if constexpr(MaskingSpec == MaskingSpecialization::MaskDisabled)
{
return MaskDisabledPredicate{};
}
else if constexpr(MaskingSpec == MaskingSpecialization::MaskUpperTriangleFromTopLeft)
{
return MaskUpperTriangleFromTopLeftPredicate{};
}
else if constexpr(MaskingSpec == MaskingSpecialization::MaskUpperTriangleFromBottomRight)
{
return MaskUpperTriangleFromBottomRightPredicate{};
}
}
using C0MatrixMask = C0MatrixMask_impl<decltype(make_MaskOutPredicate())>;
struct ComputeBasePtrOfStridedBatch
{
ComputeBasePtrOfStridedBatch(const AGridDesc_G_M_K& a_grid_desc_g_m_k,
const BGridDesc_G_N_K& b_grid_desc_g_n_k,
const D0GridDesc_G_M_N& d0_grid_desc_g_m_n,
const B1GridDesc_G_N_K& b1_grid_desc_g_n_k,
const C1GridDesc_G_M_N& c1_grid_desc_g_m_n)
: a_grid_desc_g_m_k_(a_grid_desc_g_m_k),
b_grid_desc_g_n_k_(b_grid_desc_g_n_k),
d0_grid_desc_g_m_n_(d0_grid_desc_g_m_n),
b1_grid_desc_g_n_k_(b1_grid_desc_g_n_k),
c1_grid_desc_g_m_n_(c1_grid_desc_g_m_n)
{
}
__host__ __device__ constexpr long_index_t GetABasePtr(index_t g_idx) const
{
return a_grid_desc_g_m_k_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
__host__ __device__ constexpr long_index_t GetBBasePtr(index_t g_idx) const
{
return b_grid_desc_g_n_k_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
__host__ __device__ constexpr long_index_t GetD0BasePtr(index_t g_idx) const
{
return d0_grid_desc_g_m_n_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
__host__ __device__ constexpr long_index_t GetB1BasePtr(index_t g_idx) const
{
return b1_grid_desc_g_n_k_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
__host__ __device__ constexpr long_index_t GetCBasePtr(index_t g_idx) const
{
return c1_grid_desc_g_m_n_.CalculateOffset(make_multi_index(g_idx, 0, 0));
}
private:
AGridDesc_G_M_K a_grid_desc_g_m_k_;
BGridDesc_G_N_K b_grid_desc_g_n_k_;
D0GridDesc_G_M_N d0_grid_desc_g_m_n_;
B1GridDesc_G_N_K b1_grid_desc_g_n_k_;
C1GridDesc_G_M_N c1_grid_desc_g_m_n_;
};
// GridwiseGemm
using GridwiseGemm = GridwiseMultiHeadFlashAttentionInfer_Xdl_CShuffle<
ADataType, // TODO: distinguish A/B datatype
Acc0BiasDataType,
GemmAccDataType,
CShuffleDataType,
CDataType,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation,
InMemoryDataOperationEnum::Set,
AGridDesc_AK0_M_AK1,
BGridDesc_BK0_N_BK1,
D0GridDesc_M_N,
B1GridDesc_BK0_N_BK1,
C1GridDesc_M_N,
NumGemmKPrefetchStage,
BlockSize,
MPerBlock,
NPerBlock,
KPerBlock,
Gemm1NPerBlock,
Gemm1KPerBlock,
AK1,
BK1,
B1K1,
MPerXDL,
NPerXDL,
MXdlPerWave,
NXdlPerWave,
Gemm1NXdlPerWave,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
ABlockTransferSrcAccessOrder,
ABlockTransferSrcVectorDim,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
true,
ABlockLdsExtraM,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
BBlockTransferSrcAccessOrder,
BBlockTransferSrcVectorDim,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
true,
BBlockLdsExtraN,
Acc0BiasTransferSrcScalarPerVector,
B1BlockTransferThreadClusterLengths_BK0_N_BK1,
B1BlockTransferThreadClusterArrangeOrder,
B1BlockTransferSrcAccessOrder,
B1BlockTransferSrcVectorDim,
B1BlockTransferSrcScalarPerVector,
B1BlockTransferDstScalarPerVector_BK1,
false,
B1BlockLdsExtraN,
CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopSched,
Transform::matrix_padder.PadN,
MaskingSpec != MaskingSpecialization::MaskDisabled>;
using Block2CTileMap = OffsettedBlockToCTileMap<typename GridwiseGemm::DefaultBlock2CTileMap>;
struct GroupKernelArg
{
// pointers
const ADataType* p_a_grid_;
const BDataType* p_b_grid_;
const D0DataType* p_d0_grid_;
const B1DataType* p_b1_grid_;
CDataType* p_c_grid_;
// tensor descriptors for block/thread-wise copy
AGridDesc_AK0_M_AK1 a_grid_desc_ak0_m_ak1_;
BGridDesc_BK0_N_BK1 b_grid_desc_bk0_n_bk1_;
typename GridwiseGemm::D0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5
d0_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5_;
B1GridDesc_BK0_N_BK1 b1_grid_desc_bk0_n_bk1_;
typename GridwiseGemm::C1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock
c_grid_desc_mblock_mperblock_nblock_nperblock_;
// batch & stride
index_t num_blocks_per_batch_;
ComputeBasePtrOfStridedBatch compute_base_ptr_of_batch_;
// check C0 masking and padding
C0MatrixMask c0_matrix_mask_;
// block-to-c-tile map
Block2CTileMap block_2_ctile_map_;
index_t block_start_, block_end_;
};
struct GroupDeviceArg
{
// lengths for the last dimensions of overall problem for sanity check of vector load/store
std::vector<index_t> raw_lengths_mz_nz_kz_gemm1nz_;
// strides for the last dimensions of each tensor for sanity check of vector load/store
std::vector<index_t> a_mz_kz_strides_;
std::vector<index_t> b_nz_kz_strides_;
std::vector<index_t> b1_nz_kz_strides_;
std::vector<index_t> c_mz_gemm1nz_strides_;
// for gridwise gemm check
C1GridDesc_M_N c1_grid_desc_m_n_;
// raw data
std::vector<ck::index_t> d0_n_length_stride_;
};
// Argument
// FIXME: constness
struct Argument : public BaseArgument
{
Argument(std::vector<const void*> p_a_vec,
std::vector<const void*> p_b_vec,
std::vector<const void*> p_b1_vec,
std::vector<void*> p_c_vec,
std::vector<const void*> p_acc0_bias_vec,
std::vector<const void*> p_acc1_bias_vec,
std::vector<ProblemDesc> problem_desc_vec,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op)
: a_element_op_{a_element_op},
b_element_op_{b_element_op},
acc_element_op_{acc_element_op},
b1_element_op_{b1_element_op},
c_element_op_{c_element_op}
{
ignore = p_acc0_bias_vec;
ignore = p_acc1_bias_vec;
// TODO ANT: implement bias addition
group_count_ = problem_desc_vec.size();
if(!(group_count_ == p_a_vec.size() && group_count_ == p_b_vec.size() &&
group_count_ == p_b1_vec.size() && group_count_ == p_c_vec.size() &&
(group_count_ == p_acc0_bias_vec.size() || p_acc0_bias_vec.size() == 0)))
{
throw std::runtime_error("wrong! group_count_ != a/b/b1/c_vec.size");
}
grid_size_ = 0;
for(std::size_t i = 0; i < group_count_; i++)
{
const auto p_a_grid = static_cast<const ADataType*>(p_a_vec[i]);
const auto p_b_grid = static_cast<const BDataType*>(p_b_vec[i]);
const auto p_d0_grid = (p_acc0_bias_vec.size() == group_count_)
? static_cast<const D0DataType*>(p_acc0_bias_vec[i])
: nullptr;
const auto p_b1_grid = static_cast<const B1DataType*>(p_b1_vec[i]);
const auto p_c_grid = static_cast<CDataType*>(p_c_vec[i]);
const auto& problem_desc = problem_desc_vec[i];
const auto a_grid_desc_ak0_m_ak1 = MakeAGridDescriptor_AK0_M_AK1(
problem_desc.a_gs_ms_ks_lengths, problem_desc.a_gs_ms_ks_strides);
const auto b_grid_desc_bk0_n_bk1 = MakeBGridDescriptor_BK0_N_BK1(
problem_desc.b0_gs_ns_ks_lengths, problem_desc.b0_gs_ns_ks_strides);
std::vector<index_t> tmp_d0_gs_ms_ns_lengths;
std::vector<index_t> tmp_d0_gs_ms_ns_strides;
if constexpr(!is_same<D0DataType, void>::value)
{
tmp_d0_gs_ms_ns_lengths = problem_desc.acc0_bias_gs_ms_ns_lengths;
tmp_d0_gs_ms_ns_strides = problem_desc.acc0_bias_gs_ms_ns_strides;
}
else
{
tmp_d0_gs_ms_ns_lengths = {1, 1, 1, 1};
tmp_d0_gs_ms_ns_strides = {0, 0, 0, 0};
}
const D0GridDesc_M_N d0_grid_desc_m_n{DeviceOp::MakeD0GridDescriptor_M_N(
tmp_d0_gs_ms_ns_lengths, tmp_d0_gs_ms_ns_strides)};
const auto d0_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5 =
GridwiseGemm::MakeD0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5(
d0_grid_desc_m_n);
const auto b1_grid_desc_bk0_n_bk1 = MakeB1GridDescriptor_BK0_N_BK1(
problem_desc.b1_gs_os_ns_lengths, problem_desc.b1_gs_os_ns_strides);
const auto c_grid_desc_m_n = Transform::MakeCGridDescriptor_M_N(
problem_desc.c_gs_ms_os_lengths, problem_desc.c_gs_ms_os_strides);
const auto a_grid_desc_g_m_k = Transform::MakeAGridDescriptor_G_M_K(
problem_desc.a_gs_ms_ks_lengths, problem_desc.a_gs_ms_ks_strides);
const auto b_grid_desc_g_n_k = Transform::MakeB0GridDescriptor_G_N_K(
problem_desc.b0_gs_ns_ks_lengths, problem_desc.b0_gs_ns_ks_strides);
const auto d0_grid_desc_g_m_n = DeviceOp::MakeD0GridDescriptor_G_M_N(
tmp_d0_gs_ms_ns_lengths, tmp_d0_gs_ms_ns_strides);
const auto b1_grid_desc_g_n_k = Transform::MakeB1GridDescriptor_G_N_K(
problem_desc.b1_gs_os_ns_lengths, problem_desc.b1_gs_os_ns_strides);
const auto c1_grid_desc_g_m_n = Transform::MakeCGridDescriptor_G_M_N(
problem_desc.c_gs_ms_os_lengths, problem_desc.c_gs_ms_os_strides);
const auto c_grid_desc_mblock_mperblock_nblock_nperblock =
GridwiseGemm::MakeC1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(
c_grid_desc_m_n);
const index_t BlockStart = grid_size_;
const auto block_2_ctile_map = Block2CTileMap(c_grid_desc_m_n, BlockStart);
const index_t batch_count = c1_grid_desc_g_m_n.GetLength(I0);
const index_t grid_size_grp =
block_2_ctile_map.CalculateGridSize(c_grid_desc_m_n) * batch_count;
const index_t BlockEnd = grid_size_ + grid_size_grp;
// batch stride
const auto compute_base_ptr_of_batch =
ComputeBasePtrOfStridedBatch(a_grid_desc_g_m_k,
b_grid_desc_g_n_k,
d0_grid_desc_g_m_n,
b1_grid_desc_g_n_k,
c1_grid_desc_g_m_n);
// C0 mask
const auto c0_matrix_mask =
C0MatrixMask(a_grid_desc_g_m_k.GetLength(I1), b_grid_desc_g_n_k.GetLength(I1));
grid_size_ += grid_size_grp;
group_kernel_args_.push_back({p_a_grid,
p_b_grid,
p_d0_grid,
p_b1_grid,
p_c_grid,
a_grid_desc_ak0_m_ak1,
b_grid_desc_bk0_n_bk1,
d0_grid_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
b1_grid_desc_bk0_n_bk1,
c_grid_desc_mblock_mperblock_nblock_nperblock,
block_2_ctile_map.CalculateGridSize(c_grid_desc_m_n),
compute_base_ptr_of_batch,
c0_matrix_mask,
block_2_ctile_map,
BlockStart,
BlockEnd});
std::vector<ck::index_t> d0_n_length_stride;
d0_n_length_stride.push_back(tmp_d0_gs_ms_ns_lengths[NumDimG + NumDimM]);
d0_n_length_stride.push_back(tmp_d0_gs_ms_ns_strides[NumDimG + NumDimM]);
group_device_args_.push_back(
{{problem_desc.a_gs_ms_ks_lengths[NumDimG + NumDimM - 1],
problem_desc.b0_gs_ns_ks_lengths[NumDimG + NumDimN - 1],
problem_desc.b0_gs_ns_ks_lengths[NumDimG + NumDimN + NumDimK - 1],
problem_desc.b1_gs_os_ns_lengths[NumDimG + NumDimO - 1]},
{problem_desc.a_gs_ms_ks_strides[NumDimG + NumDimM - 1],
problem_desc.a_gs_ms_ks_strides[NumDimG + NumDimM + NumDimK - 1]},
{problem_desc.b0_gs_ns_ks_strides[NumDimG + NumDimN - 1],
problem_desc.b0_gs_ns_ks_strides[NumDimG + NumDimN + NumDimK - 1]},
{problem_desc.b1_gs_os_ns_strides[NumDimG + NumDimO - 1],
problem_desc.b1_gs_os_ns_strides[NumDimG + NumDimO + NumDimN - 1]},
{problem_desc.c_gs_ms_os_strides[NumDimG + NumDimM - 1],
problem_desc.c_gs_ms_os_strides[NumDimG + NumDimM + NumDimO - 1]},
c_grid_desc_m_n,
d0_n_length_stride});
}
}
std::vector<GroupKernelArg> group_kernel_args_;
std::vector<GroupDeviceArg> group_device_args_;
std::size_t group_count_;
index_t grid_size_;
AElementwiseOperation a_element_op_;
BElementwiseOperation b_element_op_;
AccElementwiseOperation acc_element_op_;
B1ElementwiseOperation b1_element_op_;
CElementwiseOperation c_element_op_;
};
// Invoker
struct Invoker : public BaseInvoker
{
using Argument = DeviceOp::Argument;
float Run(const Argument& arg, const StreamConfig& stream_config = StreamConfig{})
{
if(!DeviceOp::IsSupportedArgument(arg))
{
throw std::runtime_error("wrong! unsupported argument");
}
bool all_has_main_k_block_loop = true;
bool some_has_main_k_block_loop = false;
for(std::size_t i = 0; i < arg.group_count_; i++)
{
const auto K = arg.group_kernel_args_[i].a_grid_desc_ak0_m_ak1_.GetLength(I0) *
arg.group_kernel_args_[i].a_grid_desc_ak0_m_ak1_.GetLength(I2);
const bool y = GridwiseGemm::CalculateHasMainKBlockLoop(K);
all_has_main_k_block_loop &= y;
some_has_main_k_block_loop |= y;
}
hipGetErrorString(
hipMemcpyWithStream(arg.p_workspace_,
arg.group_kernel_args_.data(),
arg.group_kernel_args_.size() * sizeof(GroupKernelArg),
hipMemcpyHostToDevice,
stream_config.stream_id_));
float ave_time = 0;
auto launch_kernel = [&](auto has_main_k_block_loop_) {
const auto kernel =
kernel_grouped_multiple_head_flash_attention_infer<GridwiseGemm,
D0DataType,
GroupKernelArg,
AElementwiseOperation,
BElementwiseOperation,
AccElementwiseOperation,
B1ElementwiseOperation,
CElementwiseOperation,
has_main_k_block_loop_>;
return launch_and_time_kernel(
stream_config,
kernel,
dim3(arg.grid_size_),
dim3(BlockSize),
0,
cast_pointer_to_constant_address_space(arg.p_workspace_),
arg.group_count_,
arg.a_element_op_,
arg.b_element_op_,
arg.acc_element_op_,
arg.b1_element_op_,
arg.c_element_op_);
};
// Gemm1_K is split into Gemm1_K0/K1 where K1 is known at compile time, so we only need
// to concern Gemm0's loop
if(all_has_main_k_block_loop)
{
ave_time = launch_kernel(integral_constant<bool, true>{});
}
else if(!some_has_main_k_block_loop)
{
ave_time = launch_kernel(integral_constant<bool, false>{});
}
else
{
throw std::runtime_error("wrong! all gemm problems have to simultaneously meet "
"has_main_k_block_loop or no_main_k_block_loop");
}
return ave_time;
}
// polymorphic
float Run(const BaseArgument* p_arg,
const StreamConfig& stream_config = StreamConfig{}) override
{
return Run(*dynamic_cast<const Argument*>(p_arg), stream_config);
}
};
static constexpr bool IsValidCompilationParameter()
{
// TODO: properly implement this check
return true;
}
static bool IsSupportedArgument(const Argument& arg)
{
if(!ck::is_xdl_supported())
{
return false;
}
// TODO ANT: Check if tensor specialization & strides mismatch
bool all_has_main_k_block_loop = true;
bool some_has_main_k_block_loop = false;
for(std::size_t i = 0; i < arg.group_count_; i++)
{
const auto& kernel_arg = arg.group_kernel_args_[i];
const auto& device_arg = arg.group_device_args_[i];
// Check if C permute dimension matches GEMM + GEMM shape
const index_t c_m = device_arg.c1_grid_desc_m_n_.GetLength(I0);
const index_t c_gemm1n = device_arg.c1_grid_desc_m_n_.GetLength(I1);
const index_t a_m = kernel_arg.a_grid_desc_ak0_m_ak1_.GetLength(I1);
const index_t b1_gemm1n = kernel_arg.b1_grid_desc_bk0_n_bk1_.GetLength(I1);
if(!(c_m == a_m && c_gemm1n == b1_gemm1n))
{
return false;
}
if constexpr(!is_same<D0DataType, void>::value)
{
if(device_arg.d0_n_length_stride_[1] == 1)
{
if(device_arg.d0_n_length_stride_[0] % Acc0BiasTransferSrcScalarPerVector != 0)
{
return false;
}
}
else if(Acc0BiasTransferSrcScalarPerVector != 1)
{
return false;
}
}
// Check if having main loop
const auto K = kernel_arg.a_grid_desc_ak0_m_ak1_.GetLength(I0) *
kernel_arg.a_grid_desc_ak0_m_ak1_.GetLength(I2);
const bool y = GridwiseGemm::CalculateHasMainKBlockLoop(K);
all_has_main_k_block_loop &= y;
some_has_main_k_block_loop |= y;
// Note: we need raw lengths since threadwise copy can not handle vector load when
// part of vector is out of bounds
const auto MzRaw = device_arg.raw_lengths_mz_nz_kz_gemm1nz_[0];
const auto NzRaw = device_arg.raw_lengths_mz_nz_kz_gemm1nz_[1];
const auto KzRaw = device_arg.raw_lengths_mz_nz_kz_gemm1nz_[2];
const auto Gemm1NzRaw = device_arg.raw_lengths_mz_nz_kz_gemm1nz_[3];
// Check scalar per vector requirement
const auto a_extent_lowest = ABlockTransferSrcVectorDim == 2 ? KzRaw : MzRaw;
const auto b_extent_lowest = BBlockTransferSrcVectorDim == 2 ? KzRaw : NzRaw;
const auto b1_extent_lowest = B1BlockTransferSrcVectorDim == 2 ? NzRaw : Gemm1NzRaw;
const auto c_extent_lowest = Gemm1NzRaw;
if(!(a_extent_lowest % ABlockTransferSrcScalarPerVector == 0 &&
b_extent_lowest % BBlockTransferSrcScalarPerVector == 0 &&
b1_extent_lowest % B1BlockTransferSrcScalarPerVector == 0 &&
c_extent_lowest % CShuffleBlockTransferScalarPerVector_NPerBlock == 0))
{
return false;
}
// Check vector load/store requirement
const auto a_stride_lowest = ABlockTransferSrcVectorDim == 2
? device_arg.a_mz_kz_strides_[1]
: device_arg.a_mz_kz_strides_[0];
const auto b_stride_lowest = BBlockTransferSrcVectorDim == 2
? device_arg.b_nz_kz_strides_[1]
: device_arg.b_nz_kz_strides_[0];
const auto b1_stride_lowest = B1BlockTransferSrcVectorDim == 2
? device_arg.b1_nz_kz_strides_[1]
: device_arg.b1_nz_kz_strides_[0];
const auto c_stride_lowest =
device_arg.c_mz_gemm1nz_strides_[1]; // cshuffle assumes lowest dim in Gemm1Ns to be
// contiguous
if(!(a_stride_lowest == 1 || b_stride_lowest == 1 || b1_stride_lowest == 1 ||
c_stride_lowest == 1))
{
return false;
}
if(!GridwiseGemm::CheckValidity(kernel_arg.a_grid_desc_ak0_m_ak1_,
kernel_arg.b_grid_desc_bk0_n_bk1_,
kernel_arg.b1_grid_desc_bk0_n_bk1_,
device_arg.c1_grid_desc_m_n_,
kernel_arg.block_2_ctile_map_))
{
return false;
}
}
// all gemm problems have to simultaneously meet has_main_k_block_loop or
// no_main_k_block_loop
if(!(all_has_main_k_block_loop || !some_has_main_k_block_loop))
{
return false;
}
return true;
}
// polymorphic
bool IsSupportedArgument(const BaseArgument* p_arg) override
{
return IsSupportedArgument(*dynamic_cast<const Argument*>(p_arg));
}
static auto MakeArgument(std::vector<const void*> p_a_vec,
std::vector<const void*> p_b_vec,
std::vector<const void*> p_b1_vec,
std::vector<void*> p_c_vec,
std::vector<const void*> p_acc0_bias_vec,
std::vector<const void*> p_acc1_bias_vec,
std::vector<ProblemDesc> problem_desc_vec,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op)
{
return Argument{p_a_vec,
p_b_vec,
p_b1_vec,
p_c_vec,
p_acc0_bias_vec,
p_acc1_bias_vec,
problem_desc_vec,
a_element_op,
b_element_op,
acc_element_op,
b1_element_op,
c_element_op};
}
static auto MakeInvoker() { return Invoker{}; }
// polymorphic
std::unique_ptr<BaseArgument> MakeArgumentPointer(std::vector<const void*> p_a_vec,
std::vector<const void*> p_b_vec,
std::vector<const void*> p_b1_vec,
std::vector<void*> p_c_vec,
std::vector<const void*> p_acc0_bias_vec,
std::vector<const void*> p_acc1_bias_vec,
std::vector<ProblemDesc> problem_desc_vec,
AElementwiseOperation a_element_op,
BElementwiseOperation b_element_op,
AccElementwiseOperation acc_element_op,
B1ElementwiseOperation b1_element_op,
CElementwiseOperation c_element_op) override
{
return std::make_unique<Argument>(p_a_vec,
p_b_vec,
p_b1_vec,
p_c_vec,
p_acc0_bias_vec,
p_acc1_bias_vec,
problem_desc_vec,
a_element_op,
b_element_op,
acc_element_op,
b1_element_op,
c_element_op);
}
// polymorphic
std::unique_ptr<BaseInvoker> MakeInvokerPointer() override
{
return std::make_unique<Invoker>(Invoker{});
}
// polymorphic
std::string GetTypeString() const override
{
auto str = std::stringstream();
// clang-format off
str << "DeviceGroupedMultiheadAttentionInfer_Xdl_CShuffle"
<< "<"
<< BlockSize << ", "
<< MPerBlock << ", "
<< NPerBlock << ", "
<< KPerBlock << ", "
<< AK1 << ", "
<< BK1 << ", "
<< MPerBlock << ", "
<< Gemm1NPerBlock << ", "
<< Gemm1KPerBlock << ", "
<< B1K1 << ", "
<< getGemmSpecializationString(GemmSpec) << ", "
<< "ASpec" << getTensorSpecializationString(ASpec) << ", "
<< "B0Spec" << getTensorSpecializationString(BSpec) << ", "
<< "B1Spec" << getTensorSpecializationString(B1Spec) << ", "
<< "CSpec" << getTensorSpecializationString(CSpec) << ", "
<< getMaskingSpecializationString(MaskingSpec) << ", "
<< MPerXDL << ", "
<< NPerXDL << ", "
<< MXdlPerWave << ", "
<< NXdlPerWave << ", "
<< ABlockTransferSrcScalarPerVector << ", "
<< BBlockTransferSrcScalarPerVector << ", "
<< CShuffleMXdlPerWavePerShuffle << ", "
<< CShuffleNXdlPerWavePerShuffle
<< ">";
// clang-format on
return str.str();
}
size_t GetWorkSpaceSize(const BaseArgument* p_arg) const override
{
return dynamic_cast<const Argument*>(p_arg)->group_count_ * sizeof(GroupKernelArg);
}
};
} // namespace device
} // namespace tensor_operation
} // namespace ck
include/ck/tensor_operation/gpu/grid/gridwise_batched_mha_infer_xdl_cshuffle.hpp 0 → 100644
View file @ b23b3d71
// SPDX-License-Identifier: MIT
// Copyright (c) 2018-2023, Advanced Micro Devices, Inc. All rights reserved.
#pragma once
#include "ck/utility/common_header.hpp"
#include "ck/tensor_description/multi_index_transform_helper.hpp"
#include "ck/tensor_description/tensor_descriptor.hpp"
#include "ck/tensor_description/tensor_descriptor_helper.hpp"
#include "ck/tensor_operation/gpu/grid/block_to_ctile_map.hpp"
#include "ck/tensor_operation/gpu/grid/gridwise_gemm_pipeline_selector.hpp"
#include "ck/tensor_operation/gpu/block/blockwise_gemm_xdlops.hpp"
#include "ck/tensor_operation/gpu/block/thread_group_tensor_slice_transfer_v4r1.hpp"
#include "ck/tensor_operation/gpu/block/thread_group_tensor_slice_transfer_v6r1.hpp"
#include "ck/tensor_operation/gpu/thread/threadwise_tensor_slice_transfer.hpp"
#include "ck/tensor_operation/gpu/element/element_wise_operation.hpp"
#include "ck/tensor_operation/gpu/block/blockwise_softmax.hpp"
namespace ck {
/**
* @brief Gridwise gemm + softmax + gemm fusion
*
*/
template <typename FloatAB,
typename D0DataType,
typename FloatGemmAcc,
typename FloatCShuffle,
typename FloatC,
typename AElementwiseOperation,
typename BElementwiseOperation,
typename AccElementwiseOperation,
typename B1ElementwiseOperation,
typename CElementwiseOperation,
InMemoryDataOperationEnum CGlobalMemoryDataOperation,
typename AGridDesc_AK0_M_AK1,
typename BGridDesc_BK0_N_BK1,
typename D0GridDesc_M_N,
typename B1GridDesc_BK0_N_BK1,
typename C1GridDesc_M_N,
index_t NumGemmKPrefetchStage,
index_t BlockSize,
index_t MPerBlock,
index_t NPerBlock,
index_t KPerBlock,
index_t Gemm1NPerBlock,
index_t Gemm1KPerBlock,
index_t AK1Value,
index_t BK1Value,
index_t B1K1Value,
index_t MPerXdl,
index_t NPerXdl,
index_t MXdlPerWave,
index_t NXdlPerWave,
index_t Gemm1NXdlPerWave,
typename ABlockTransferThreadClusterLengths_AK0_M_AK1,
typename ABlockTransferThreadClusterArrangeOrder,
typename ABlockTransferSrcAccessOrder,
index_t ABlockTransferSrcVectorDim,
index_t ABlockTransferSrcScalarPerVector,
index_t ABlockTransferDstScalarPerVector_AK1,
bool AThreadTransferSrcResetCoordinateAfterRun, // ignored
index_t ABlockLdsExtraM,
typename BBlockTransferThreadClusterLengths_BK0_N_BK1,
typename BBlockTransferThreadClusterArrangeOrder,
typename BBlockTransferSrcAccessOrder,
index_t BBlockTransferSrcVectorDim,
index_t BBlockTransferSrcScalarPerVector,
index_t BBlockTransferDstScalarPerVector_BK1,
bool BThreadTransferSrcResetCoordinateAfterRun, // ignored
index_t BBlockLdsExtraN,
index_t D0BlockTransferSrcScalarPerVector,
typename B1BlockTransferThreadClusterLengths_BK0_N_BK1,
typename B1BlockTransferThreadClusterArrangeOrder,
typename B1BlockTransferSrcAccessOrder,
index_t B1BlockTransferSrcVectorDim,
index_t B1BlockTransferSrcScalarPerVector,
index_t B1BlockTransferDstScalarPerVector_BK1,
bool B1ThreadTransferSrcResetCoordinateAfterRun,
index_t B1BlockLdsExtraN,
index_t CShuffleMXdlPerWavePerShuffle,
index_t CShuffleNXdlPerWavePerShuffle,
typename CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
index_t CShuffleBlockTransferScalarPerVector_NPerBlock,
LoopScheduler LoopSched,
bool PadN,
bool MaskOutUpperTriangle,
PipelineVersion PipelineVer = PipelineVersion::v1>
struct GridwiseMultiHeadFlashAttentionInfer_Xdl_CShuffle
{
static_assert(D0BlockTransferSrcScalarPerVector == 1 ||
D0BlockTransferSrcScalarPerVector == 2 ||
D0BlockTransferSrcScalarPerVector == 4,
"D0BlockTransferSrcScalarPerVector must be 1 or 2 or 4");
static_assert(LoopSched == LoopScheduler::Default,
"Non-default loop scheduler is currently not supported");
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>{};
static constexpr auto WaveSize = 64;
// K1 should be Number<...>
// Gemm0
static constexpr auto AK0 = Number<KPerBlock / AK1Value>{};
static constexpr auto BK0 = Number<KPerBlock / BK1Value>{};
static constexpr auto AK1 = Number<AK1Value>{};
static constexpr auto BK1 = Number<BK1Value>{};
static constexpr auto Gemm0MWaves = MPerBlock / (MPerXdl * MXdlPerWave);
static constexpr auto Gemm0NWaves = NPerBlock / (NPerXdl * NXdlPerWave);
// Gemm1
static constexpr auto B1K0 = Number<Gemm1KPerBlock / B1K1Value>{};
static constexpr auto B1K1 = Number<B1K1Value>{};
static constexpr auto mfma = MfmaSelector<FloatAB, MPerXdl, NPerXdl>::selected_mfma;
using ThisThreadBlock = ThisThreadBlock<BlockSize>;
using GridwiseGemmPipe = remove_cvref_t<decltype(
GridwiseGemmPipeline_Selector<PipelineVer, NumGemmKPrefetchStage>())>;
__device__ static auto GetGemm0WaveIdx()
{
const index_t thread_id = get_thread_local_1d_id();
constexpr auto threadid_to_wave_idx_adaptor = make_single_stage_tensor_adaptor(
make_tuple(make_merge_transform(make_tuple(Gemm0MWaves, Gemm0NWaves, WaveSize))),
make_tuple(Sequence<0, 1, 2>{}),
make_tuple(Sequence<0>{}));
return threadid_to_wave_idx_adaptor.CalculateBottomIndex(make_multi_index(thread_id));
}
__device__ static auto GetGemm0WaveMNIdx(const index_t thread_id)
{
constexpr auto wave_threadid_to_mn_idx_adaptor = make_single_stage_tensor_adaptor(
make_tuple(make_merge_transform(make_tuple(WaveSize / MPerXdl, MPerXdl))),
make_tuple(Sequence<0, 1>{}),
make_tuple(Sequence<0>{}));
return wave_threadid_to_mn_idx_adaptor.CalculateBottomIndex(make_multi_index(thread_id));
}
template <typename ABlockDesc_AK0_M_AK1>
__host__ __device__ static constexpr auto
MakeGemm0AMmaTileDescriptor_M0_M1_M2_K(const ABlockDesc_AK0_M_AK1&)
{
constexpr index_t MWaves = MPerBlock / (MXdlPerWave * MPerXdl);
return MakeGemmMmaTileDescriptor_MN0_MN1_MN2_K<MXdlPerWave, MWaves, MPerXdl>(
ABlockDesc_AK0_M_AK1{});
}
template <typename BBlockDesc_BK0_N_BK1>
__host__ __device__ static constexpr auto
MakeGemm0BMmaTileDescriptor_N0_N1_N2_K(const BBlockDesc_BK0_N_BK1&)
{
constexpr index_t NWaves = NPerBlock / (NXdlPerWave * NPerXdl);
return MakeGemmMmaTileDescriptor_MN0_MN1_MN2_K<NXdlPerWave, NWaves, NPerXdl>(
BBlockDesc_BK0_N_BK1{});
}
template <typename ABlockDesc_AK0_M_AK1>
__host__ __device__ static constexpr auto
MakeGemm1AMmaTileDescriptor_M0_M1_M2_K(const ABlockDesc_AK0_M_AK1&)
{
return MakeGemmMmaTileDescriptor_MN0_MN1_MN2_K<MXdlPerWave, 1, 1>(ABlockDesc_AK0_M_AK1{});
}
template <typename BBlockDesc_BK0_N_BK1>
__host__ __device__ static constexpr auto
MakeGemm1BMmaTileDescriptor_N0_N1_N2_K(const BBlockDesc_BK0_N_BK1&)
{
constexpr index_t Gemm1NWaves = Gemm1NPerBlock / (Gemm1NXdlPerWave * NPerXdl);
return MakeGemmMmaTileDescriptor_MN0_MN1_MN2_K<Gemm1NXdlPerWave, Gemm1NWaves, NPerXdl>(
BBlockDesc_BK0_N_BK1{});
}
__host__ __device__ static constexpr auto GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1()
{
// A matrix in LDS memory, dst of blockwise copy
return make_naive_tensor_descriptor(
make_tuple(AK0, Number<MPerBlock>{}, AK1),
make_tuple(Number<MPerBlock + ABlockLdsExtraM>{} * AK1, AK1, I1));
}
__host__ __device__ static constexpr auto GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1()
{
// B matrix in LDS memory, dst of blockwise copy
return make_naive_tensor_descriptor(
make_tuple(BK0, Number<NPerBlock>{}, BK1),
make_tuple(Number<NPerBlock + BBlockLdsExtraN>{} * BK1, BK1, I1));
}
__host__ __device__ static constexpr auto GetB1BlockDescriptor_BK0PerBlock_NPerBlock_BK1()
{
// B1 matrix in LDS memory, dst of blockwise copy
return make_naive_tensor_descriptor(
make_tuple(B1K0, Number<Gemm1NPerBlock>{}, B1K1),
make_tuple(Number<Gemm1NPerBlock + B1BlockLdsExtraN>{} * B1K1, B1K1, I1));
}
__host__ __device__ static constexpr auto
GetCShuffleBlockDescriptor_MBlock_MPerBlock_NBlock_NPerBlock()
{
constexpr index_t MWave = MPerBlock / (MXdlPerWave * MPerXdl);
constexpr index_t NWave = Gemm1NPerBlock / (Gemm1NXdlPerWave * NPerXdl);
constexpr auto c_shuffle_block_desc_mblock_mperblock_nblock_nperblock =
make_naive_tensor_descriptor_packed(
make_tuple(I1,
Number<CShuffleMXdlPerWavePerShuffle * MWave * MPerXdl>{},
I1,
Number<CShuffleNXdlPerWavePerShuffle * NWave * NPerXdl>{}));
return c_shuffle_block_desc_mblock_mperblock_nblock_nperblock;
}
__host__ __device__ static constexpr index_t GetSharedMemoryNumberOfByte()
{
const index_t gemm0_bytes_end = (SharedMemTrait::a_block_space_size_aligned +
SharedMemTrait::b_block_space_size_aligned) *
sizeof(FloatAB);
const index_t gemm1_bytes_end =
(SharedMemTrait::b1_block_space_offset + SharedMemTrait::b1_block_space_size_aligned) *
sizeof(FloatAB);
const index_t softmax_bytes_end = (SharedMemTrait::reduction_space_offset +
SharedMemTrait::reduction_space_size_aligned) *
sizeof(FloatGemmAcc);
const index_t c_block_bytes_end =
SharedMemTrait::c_block_space_size * sizeof(FloatCShuffle);
return math::max(gemm0_bytes_end, gemm1_bytes_end, softmax_bytes_end, c_block_bytes_end);
}
// block_id to matrix tile idx (m0, n0) mapping are controlled by {M01, N01}
template <typename Block2CTileMap>
__host__ __device__ static constexpr bool
CheckValidity(const AGridDesc_AK0_M_AK1& a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1& b_grid_desc_bk0_n_bk1,
const B1GridDesc_BK0_N_BK1& b1_grid_desc_bk0_n_bk1,
const C1GridDesc_M_N& c_grid_desc_m_n,
const Block2CTileMap& block_2_ctile_map)
{
static_assert((MPerBlock % (MPerXdl * MXdlPerWave) == 0) &&
(NPerBlock % (NXdlPerWave * NPerXdl)) == 0,
"Invalid tuning param!");
const auto M = a_grid_desc_ak0_m_ak1.GetLength(I1);
const auto N = b_grid_desc_bk0_n_bk1.GetLength(I1);
const auto K = a_grid_desc_ak0_m_ak1.GetLength(I0) * a_grid_desc_ak0_m_ak1.GetLength(I2);
const auto Gemm1N = b1_grid_desc_bk0_n_bk1.GetLength(I1);
if(!(M == c_grid_desc_m_n.GetLength(I0) && Gemm1N == c_grid_desc_m_n.GetLength(I1)))
{
return false;
}
if(!(M % MPerBlock == 0 && N % NPerBlock == 0 && K % KPerBlock == 0 &&
Gemm1N % Gemm1NPerBlock == 0))
{
return false;
}
// check gemm0 gridwise gemm pipeline
const auto num_gemm0_k_loop = K / KPerBlock;
if(!GridwiseGemmPipe::IsSupported(num_gemm0_k_loop))
{
return false;
}
// check gemm1 gridwise gemm pipeline
if(!(NPerBlock % Gemm1KPerBlock == 0))
{
return false;
}
const auto num_gemm1_k_inner_loop = NPerBlock / Gemm1KPerBlock;
if(!GridwiseGemmPipe::IsSupported(num_gemm1_k_inner_loop))
{
return false;
}
if(!block_2_ctile_map.CheckValidity(c_grid_desc_m_n))
{
return false;
}
// TODO: also check validity of all components (blockwise-copy, threadwise-copy, etc)
return true;
}
__host__ __device__ static constexpr bool CalculateHasMainKBlockLoop(index_t K)
{
const index_t num_loop = K / KPerBlock;
return GridwiseGemmPipe::CalculateHasMainLoop(num_loop);
}
__host__ __device__ static constexpr auto
MakeC1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(const C1GridDesc_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 auto MBlock = M / MPerBlock;
const auto NBlock = N / Gemm1NPerBlock;
const auto c1_grid_desc_mblock_mperblock_nblock_nperblock = transform_tensor_descriptor(
c_grid_desc_m_n,
make_tuple(make_unmerge_transform(make_tuple(MBlock, Number<MPerBlock>{})),
make_unmerge_transform(make_tuple(NBlock, Number<Gemm1NPerBlock>{}))),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 1>{}, Sequence<2, 3>{}));
return c1_grid_desc_mblock_mperblock_nblock_nperblock;
}
// return block_id to C matrix tile idx (m0, n0) mapping
__host__ __device__ static constexpr auto
MakeDefaultBlock2CTileMap(const C1GridDesc_M_N& c_grid_desc_m_n)
{
return BlockToCTileMap_M00_N0_M01Adapt<MPerBlock, Gemm1NPerBlock, C1GridDesc_M_N>(
c_grid_desc_m_n);
}
// D0 desc for source in blockwise copy
__host__ __device__ static constexpr auto
MakeD0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5(const D0GridDesc_M_N& d0_grid_desc_m_n)
{
const auto M = d0_grid_desc_m_n.GetLength(I0);
const auto N = d0_grid_desc_m_n.GetLength(I1);
constexpr auto N3 = mfma.num_groups_per_blk;
constexpr auto N4 = mfma.num_input_blks;
constexpr auto N5 = mfma.group_size;
return transform_tensor_descriptor(
d0_grid_desc_m_n,
make_tuple(make_unmerge_transform(
make_tuple(M / MPerBlock, MXdlPerWave, Gemm0MWaves, MPerXdl)),
make_unmerge_transform(
make_tuple(N / NPerBlock, NXdlPerWave, Gemm0NWaves, N3, N4, N5))),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2, 4, 6>{}, Sequence<1, 3, 5, 7, 8, 9>{}));
}
using D0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5 = remove_cvref_t<decltype(
MakeD0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5(D0GridDesc_M_N{}))>;
using C1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock = remove_cvref_t<decltype(
MakeC1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock(C1GridDesc_M_N{}))>;
using DefaultBlock2CTileMap =
remove_cvref_t<decltype(MakeDefaultBlock2CTileMap(C1GridDesc_M_N{}))>;
struct SharedMemTrait
{
// LDS allocation for A and B: be careful of alignment
static constexpr auto a_block_desc_ak0_m_ak1 =
GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1();
static constexpr auto b_block_desc_bk0_n_bk1 =
GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1();
static constexpr auto b1_block_desc_bk0_n_bk1 =
GetB1BlockDescriptor_BK0PerBlock_NPerBlock_BK1();
static constexpr auto max_lds_align = math::lcm(math::lcm(AK1, BK1), B1K1);
static constexpr auto a_block_space_size_aligned = math::integer_least_multiple(
a_block_desc_ak0_m_ak1.GetElementSpaceSize(), max_lds_align);
static constexpr auto b_block_space_size_aligned = math::integer_least_multiple(
b_block_desc_bk0_n_bk1.GetElementSpaceSize(), max_lds_align);
static constexpr auto b1_block_space_size_aligned = math::integer_least_multiple(
b1_block_desc_bk0_n_bk1.GetElementSpaceSize(), max_lds_align);
static constexpr auto a_block_space_offset = 0;
static constexpr auto b_block_space_offset = a_block_space_size_aligned.value;
static constexpr auto b1_block_space_offset = 0;
// LDS allocation for reduction
static constexpr index_t reduction_space_size_aligned =
math::integer_least_multiple(BlockSize, max_lds_align);
static constexpr auto reduction_space_offset = 0;
// LDS allocation for C shuffle in LDS
static constexpr auto c_shuffle_block_desc_mblock_mperblock_nblock_nperblock =
GetCShuffleBlockDescriptor_MBlock_MPerBlock_NBlock_NPerBlock();
static constexpr auto c_block_space_size =
c_shuffle_block_desc_mblock_mperblock_nblock_nperblock.GetElementSpaceSize();
};
template <bool HasMainKBlockLoop, typename Block2CTileMap, typename C0MatrixMask>
__device__ static void Run(const FloatAB* __restrict__ p_a_grid,
const FloatAB* __restrict__ p_b_grid,
const D0DataType* __restrict__ p_d0_grid,
const FloatAB* __restrict__ p_b1_grid,
FloatC* __restrict__ p_c_grid,
void* __restrict__ p_shared,
const AElementwiseOperation& a_element_op,
const BElementwiseOperation& b_element_op,
const AccElementwiseOperation& acc_element_op,
const B1ElementwiseOperation& b1_element_op,
const CElementwiseOperation& c_element_op,
const AGridDesc_AK0_M_AK1& a_grid_desc_ak0_m_ak1,
const BGridDesc_BK0_N_BK1& b_grid_desc_bk0_n_bk1,
const D0GridDescriptor_M0_N0_M1_N1_M2_N2_M3_N3_N4_N5&
d0_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
const B1GridDesc_BK0_N_BK1& b1_grid_desc_bk0_n_bk1,
const C1GridDescriptor_MBlock_MPerBlock_NBlock_NPerBlock&
c1_grid_desc_mblock_mperblock_nblock_nperblock,
const Block2CTileMap& block_2_ctile_map,
const C0MatrixMask& c0_matrix_mask)
{
const auto a_grid_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_a_grid, a_grid_desc_ak0_m_ak1.GetElementSpaceSize());
const auto b_grid_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_b_grid, b_grid_desc_bk0_n_bk1.GetElementSpaceSize());
const auto b1_grid_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_b1_grid, b1_grid_desc_bk0_n_bk1.GetElementSpaceSize());
auto c_grid_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_c_grid, c1_grid_desc_mblock_mperblock_nblock_nperblock.GetElementSpaceSize());
// divide block work by [M, N]
const auto block_work_idx =
block_2_ctile_map.CalculateBottomIndex(make_multi_index(get_block_1d_id()));
if(!block_2_ctile_map.ValidCTileIndex(
block_work_idx,
make_tuple(c1_grid_desc_mblock_mperblock_nblock_nperblock.GetLength(I0),
c1_grid_desc_mblock_mperblock_nblock_nperblock.GetLength(I2))))
{
return;
}
// HACK: this force m/gemm1_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 gemm1_n_block_data_idx_on_grid =
__builtin_amdgcn_readfirstlane(block_work_idx[I1] * Gemm1NPerBlock);
// A matrix in LDS memory, dst of blockwise copy
constexpr auto a_block_desc_ak0_m_ak1 = GetABlockDescriptor_AK0PerBlock_MPerBlock_AK1();
// B matrix in LDS memory, dst of blockwise copy
constexpr auto b_block_desc_bk0_n_bk1 = GetBBlockDescriptor_BK0PerBlock_NPerBlock_BK1();
//
// set up Gemm0
//
// A matrix blockwise copy
auto a_blockwise_copy =
ThreadGroupTensorSliceTransfer_v4r1<ThisThreadBlock,
AElementwiseOperation,
tensor_operation::element_wise::PassThrough,
InMemoryDataOperationEnum::Set,
Sequence<AK0, MPerBlock, AK1>,
ABlockTransferThreadClusterLengths_AK0_M_AK1,
ABlockTransferThreadClusterArrangeOrder,
FloatAB,
FloatAB,
decltype(a_grid_desc_ak0_m_ak1),
decltype(a_block_desc_ak0_m_ak1),
ABlockTransferSrcAccessOrder,
Sequence<1, 0, 2>,
ABlockTransferSrcVectorDim,
2,
ABlockTransferSrcScalarPerVector,
ABlockTransferDstScalarPerVector_AK1,
1,
1,
true, // SrcResetCoord
true, // DstResetCoord
NumGemmKPrefetchStage>(
a_grid_desc_ak0_m_ak1,
make_multi_index(0, m_block_data_idx_on_grid, 0),
a_element_op,
a_block_desc_ak0_m_ak1,
make_multi_index(0, 0, 0),
tensor_operation::element_wise::PassThrough{});
// B matrix blockwise copy
auto b_blockwise_copy =
ThreadGroupTensorSliceTransfer_v4r1<ThisThreadBlock,
BElementwiseOperation,
tensor_operation::element_wise::PassThrough,
InMemoryDataOperationEnum::Set,
Sequence<BK0, NPerBlock, BK1>,
BBlockTransferThreadClusterLengths_BK0_N_BK1,
BBlockTransferThreadClusterArrangeOrder,
FloatAB,
FloatAB,
decltype(b_grid_desc_bk0_n_bk1),
decltype(b_block_desc_bk0_n_bk1),
BBlockTransferSrcAccessOrder,
Sequence<1, 0, 2>,
BBlockTransferSrcVectorDim,
2,
BBlockTransferSrcScalarPerVector,
BBlockTransferDstScalarPerVector_BK1,
1,
1,
true, // SrcResetCoord
true, // DstResetCoord
NumGemmKPrefetchStage>(
b_grid_desc_bk0_n_bk1,
make_multi_index(0, 0, 0), // will loop over GemmN dimension
b_element_op,
b_block_desc_bk0_n_bk1,
make_multi_index(0, 0, 0),
tensor_operation::element_wise::PassThrough{});
// Fused Gemm+Gemm pipeline
// for n in N0:
// for k in K0:
// acc[m][n] += A[m][k] * B0[k][n]
// acc1[m][o] += acc[m][n] * B1[n][o]
// sanity check
constexpr index_t KPack = math::max(
math::lcm(AK1, BK1), MfmaSelector<FloatAB, MPerXdl, NPerXdl>::selected_mfma.k_per_blk);
auto blockwise_gemm = BlockwiseGemmXdlops_v2<
BlockSize,
FloatAB,
FloatGemmAcc,
decltype(a_block_desc_ak0_m_ak1),
decltype(b_block_desc_bk0_n_bk1),
decltype(MakeGemm0AMmaTileDescriptor_M0_M1_M2_K(a_block_desc_ak0_m_ak1)),
decltype(MakeGemm0BMmaTileDescriptor_N0_N1_N2_K(b_block_desc_bk0_n_bk1)),
MPerBlock,
NPerBlock,
KPerBlock,
MPerXdl,
NPerXdl,
MXdlPerWave,
NXdlPerWave,
KPack,
true>{}; // TransposeC
auto acc_thread_buf = blockwise_gemm.GetCThreadBuffer();
// LDS allocation for A and B: be careful of alignment
auto a_block_buf = make_dynamic_buffer<AddressSpaceEnum::Lds>(
static_cast<FloatAB*>(p_shared) + SharedMemTrait::a_block_space_offset,
a_block_desc_ak0_m_ak1.GetElementSpaceSize());
auto b_block_buf = make_dynamic_buffer<AddressSpaceEnum::Lds>(
static_cast<FloatAB*>(p_shared) + SharedMemTrait::b_block_space_offset,
b_block_desc_bk0_n_bk1.GetElementSpaceSize());
constexpr auto a_block_slice_copy_step = make_multi_index(KPerBlock / AK1, 0, 0);
constexpr auto b_block_slice_copy_step = make_multi_index(KPerBlock / BK1, 0, 0);
const auto a_block_reset_copy_step =
make_multi_index(-a_grid_desc_ak0_m_ak1.GetLength(I0), 0, 0);
const auto b_block_reset_copy_step =
make_multi_index(-b_grid_desc_bk0_n_bk1.GetLength(I0), NPerBlock, 0);
// gridwise GEMM pipeline
// Only supports LoopScheduler::Default
const auto gridwise_gemm_pipeline = GridwiseGemmPipeline_Selector<PipelineVer,
NumGemmKPrefetchStage,
LoopScheduler::Default>();
const index_t num_k_block_main_loop = __builtin_amdgcn_readfirstlane(
(a_grid_desc_ak0_m_ak1.GetLength(I0) * a_grid_desc_ak0_m_ak1.GetLength(I2)) /
KPerBlock);
//
// set up Gemm1
//
// Acc matrix threadwise copy: AccVGPR to VGPR and downcast to XDL input data type
constexpr auto acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4 =
blockwise_gemm.GetCThreadDescriptor_M0_N0_M1_N1_M2_N2_N3_N4();
constexpr auto m0 = acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I0);
constexpr auto n0 = acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I1);
constexpr auto m1 = acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I2);
constexpr auto n1 = acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I3);
constexpr auto m2 = acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I4);
constexpr auto n2 = acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I5);
constexpr auto n3 = acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I6);
constexpr auto n4 = acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I7);
constexpr auto b1_block_slice_copy_step = make_multi_index(Gemm1KPerBlock / B1K1, 0, 0);
// acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4 to acc_thread_desc_k0_m_k1
// n0_n1_n2_n3 -> k0
// m0_m1_m2 -> m
// n4 -> k1
// NOTE: had to use merge_v3 or will spit out compilation errors
constexpr auto acc_thread_desc_k0_m_k1 = transform_tensor_descriptor(
acc_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4,
make_tuple(make_merge_transform_v3_division_mod(make_tuple(n0, n1, n2, n3)),
make_merge_transform_v3_division_mod(make_tuple(m0, m1, m2)),
make_pass_through_transform(n4)),
make_tuple(Sequence<1, 3, 5, 6>{}, Sequence<0, 2, 4>{}, Sequence<7>{}),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}));
// A1 matrix in AccVGPR
// N2 num_groups_per_blk, N3 num_input_blks, N4 group_size
constexpr auto AccN3 =
blockwise_gemm.GetCBlockDescriptor_M0_N0_M1_N1_M2_N2_N3_N4().GetLength(I6);
constexpr auto A1ThreadSlice_K0_M_K1 =
make_tuple(Number<Gemm1KPerBlock / n4 / AccN3>{}, Number<m0 * m1 * m2>{}, Number<n4>{});
constexpr auto A1ThreadSliceK0 = A1ThreadSlice_K0_M_K1[I0];
constexpr auto A1ThreadSliceM = A1ThreadSlice_K0_M_K1[I1];
constexpr auto A1ThreadSliceK1 = A1ThreadSlice_K0_M_K1[I2];
constexpr auto a1_thread_desc_k0_m_k1 = make_naive_tensor_descriptor(
A1ThreadSlice_K0_M_K1,
make_tuple(A1ThreadSliceM * A1ThreadSliceK1, A1ThreadSliceK1, I1));
// B1 matrix in LDS memory, dst of blockwise copy
constexpr auto b1_block_desc_bk0_n_bk1 = GetB1BlockDescriptor_BK0PerBlock_NPerBlock_BK1();
// A1 matrix blockwise copy
auto a1_blockwise_copy = ThreadwiseTensorSliceTransfer_StaticToStatic<
FloatGemmAcc,
FloatAB,
decltype(acc_thread_desc_k0_m_k1),
decltype(a1_thread_desc_k0_m_k1),
tensor_operation::element_wise::PassThrough,
Sequence<A1ThreadSliceK0, A1ThreadSliceM, A1ThreadSliceK1>,
Sequence<1, 0, 2>,
2,
n4>{tensor_operation::element_wise::PassThrough{}};
// B1 matrix blockwise copy
auto b1_blockwise_copy =
ThreadGroupTensorSliceTransfer_v4r1<ThisThreadBlock,
BElementwiseOperation,
tensor_operation::element_wise::PassThrough,
InMemoryDataOperationEnum::Set,
Sequence<B1K0, Gemm1NPerBlock, B1K1>,
B1BlockTransferThreadClusterLengths_BK0_N_BK1,
B1BlockTransferThreadClusterArrangeOrder,
FloatAB,
FloatAB,
decltype(b1_grid_desc_bk0_n_bk1),
decltype(b1_block_desc_bk0_n_bk1),
B1BlockTransferSrcAccessOrder,
Sequence<1, 0, 2>,
B1BlockTransferSrcVectorDim,
2,
B1BlockTransferSrcScalarPerVector,
B1BlockTransferDstScalarPerVector_BK1,
1,
1,
B1ThreadTransferSrcResetCoordinateAfterRun,
true, // DstResetCoord
NumGemmKPrefetchStage>(
b1_grid_desc_bk0_n_bk1,
make_multi_index(0, gemm1_n_block_data_idx_on_grid, 0),
b1_element_op,
b1_block_desc_bk0_n_bk1,
make_multi_index(0, 0, 0),
tensor_operation::element_wise::PassThrough{});
auto a1_thread_buf = make_static_buffer<AddressSpaceEnum::Vgpr, FloatAB>(
a1_thread_desc_k0_m_k1.GetElementSpaceSize());
// reuse LDS space for gemm0's b_block_buf
auto b1_block_buf = make_dynamic_buffer<AddressSpaceEnum::Lds>(
static_cast<FloatAB*>(p_shared) + SharedMemTrait::b1_block_space_offset,
b1_block_desc_bk0_n_bk1.GetElementSpaceSize());
const auto wave_id = GetGemm0WaveIdx();
const auto wave_m_n_id = GetGemm0WaveMNIdx(wave_id[I2]); // I2: 0~63
// bias (d0 matrix)
constexpr auto d0_thread_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5 =
make_naive_tensor_descriptor_packed(make_tuple(I1, // MBlockId
I1, // NBlockId
m0, // MRepeat
n0, // NRepeat
m1, // MWaveId
n1, // NWaveId
m2, // MPerXdl
n2, // NGroupNum
n3, // NInputNum
n4)); // RegisterNum
auto d0_threadwise_copy =
ThreadwiseTensorSliceTransfer_v2<D0DataType,
D0DataType,
decltype(d0_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5),
decltype(d0_thread_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5),
Sequence<I1, // MBlockId
I1, // NBlockID
m0, // MRepeat
n0, // NRepeat
m1, // MWaveId
n1, // NWaveId
m2, // MPerXdl
n2, // NGroupNum
n3, // NInputNum
n4>,
Sequence<0, 1, 2, 3, 4, 5, 6, 7, 8, 9>,
9,
D0BlockTransferSrcScalarPerVector,
1,
false>(d0_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
make_multi_index(block_work_idx[I0], // MBlockId
0, // NBlockId
0, // mrepeat
0, // nrepeat
wave_id[I0], // MWaveId
wave_id[I1], // NWaveId
wave_m_n_id[I1], // MPerXdl
0, // group
wave_m_n_id[I0], // NInputIndex
0)); // register number
// selected_mfma.group_size or B1K1 <= Gemm1KPack <= selected_mfma.group_size
// selected_mfma.k_per_blk <= Gemm1KPack
//
// Following similar rationale behind Gemm0KPack, let Gemm1KPack be the lowest common
// multiples of A1K1 (predetermined by selected_mfma.group_size) and B1K1. But in this case
// Gemm1KPack can't be higher than A1K1 itself because A1 matrix is distributed in VGPRs
// with 'group_size' amount of contiguous elements. Having Gemm1KPack greater than A1K1 will
// cause mismatch in summation index for example c[0:7] = a1[[0:3, 8:11]] * b1[0:7].
// therefore we may just as well assign Gemm1KPack = group_size
constexpr index_t Gemm1KPack =
MfmaSelector<FloatAB, MPerXdl, NPerXdl>::selected_mfma.group_size;
auto gemm1_blockwise_gemm = BlockwiseGemmXdlops_v2<
BlockSize,
FloatAB,
FloatGemmAcc,
decltype(a1_thread_desc_k0_m_k1),
decltype(b1_block_desc_bk0_n_bk1),
decltype(MakeGemm1AMmaTileDescriptor_M0_M1_M2_K(a1_thread_desc_k0_m_k1)),
decltype(MakeGemm1BMmaTileDescriptor_N0_N1_N2_K(b1_block_desc_bk0_n_bk1)),
MPerBlock,
Gemm1NPerBlock,
Gemm1KPerBlock,
MPerXdl,
NPerXdl,
MXdlPerWave,
Gemm1NXdlPerWave,
Gemm1KPack,
true, // TransposeC
Gemm1KPack, // AMmaKStride
Gemm1KPack * XdlopsGemm<FloatAB, MPerXdl, NPerXdl, Gemm1KPack, false>{}.K0PerXdlops>{
// BMmaKStride
make_tuple(0, 0, 0, 0)}; // A_origin
auto acc1_thread_buf = gemm1_blockwise_gemm.GetCThreadBuffer();
//
// Blockwise softmax
//
auto workspace_buf = make_dynamic_buffer<AddressSpaceEnum::Lds>(
static_cast<FloatGemmAcc*>(p_shared) + SharedMemTrait::reduction_space_offset,
SharedMemTrait::reduction_space_size_aligned);
// get acc0 8D thread cluster
constexpr auto thread_cluster_m0_n0_m1_n1_m2_n2_n3_n4 =
blockwise_gemm.GetCBlockDescriptor_M0_N0_M1_N1_M2_N2_N3_N4().GetLengths() /
blockwise_gemm.GetCThreadDescriptor_M0_N0_M1_N1_M2_N2_N3_N4().GetLengths();
constexpr auto tm0 = thread_cluster_m0_n0_m1_n1_m2_n2_n3_n4.At(I0);
constexpr auto tn0 = thread_cluster_m0_n0_m1_n1_m2_n2_n3_n4.At(I1);
constexpr auto tm1 = thread_cluster_m0_n0_m1_n1_m2_n2_n3_n4.At(I2);
constexpr auto tn1 = thread_cluster_m0_n0_m1_n1_m2_n2_n3_n4.At(I3);
constexpr auto tm2 = thread_cluster_m0_n0_m1_n1_m2_n2_n3_n4.At(I4);
constexpr auto tn2 = thread_cluster_m0_n0_m1_n1_m2_n2_n3_n4.At(I5);
constexpr auto tn3 = thread_cluster_m0_n0_m1_n1_m2_n2_n3_n4.At(I6);
constexpr auto tn4 = thread_cluster_m0_n0_m1_n1_m2_n2_n3_n4.At(I7);
// get acc0 thread map
constexpr auto m0_n_m1_to_m_n_adaptor = make_single_stage_tensor_adaptor(
make_tuple(make_unmerge_transform(make_tuple(tm0 * tm1, tm2)),
make_pass_through_transform(I1)),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2>{}, Sequence<1>{}));
constexpr auto threadid_to_m0_n_m1_adaptor = make_single_stage_tensor_adaptor(
make_tuple(
make_merge_transform(make_tuple(tm0 * tm1, tn0 * tn1 * tn2 * tn3 * tn4, tm2))),
make_tuple(Sequence<0, 1, 2>{}),
make_tuple(Sequence<0>{}));
const auto threadid_to_m_n_thread_cluster_adaptor =
chain_tensor_adaptors(m0_n_m1_to_m_n_adaptor, threadid_to_m0_n_m1_adaptor);
// get acc0 2D thread cluster & 2D thread slice
constexpr auto thread_cluster_desc_m_n = make_naive_tensor_descriptor_packed(
make_tuple(tm0 * tm1 * tm2, tn0 * tn1 * tn2 * tn3 * tn4));
constexpr auto thread_slice_desc_m_n =
make_naive_tensor_descriptor_packed(make_tuple(m0 * m1 * m2, n0 * n1 * n2 * n3 * n4));
auto blockwise_softmax = BlockwiseSoftmax<BlockSize,
FloatGemmAcc,
decltype(threadid_to_m_n_thread_cluster_adaptor),
decltype(thread_cluster_desc_m_n),
decltype(thread_slice_desc_m_n)>{};
const index_t num_gemm1_k_block_outer_loop =
b_grid_desc_bk0_n_bk1.GetLength(I1) / NPerBlock;
constexpr index_t num_gemm1_k_block_inner_loop = NPerBlock / Gemm1KPerBlock;
// Initialize C
StaticBuffer<AddressSpaceEnum::Vgpr, FloatGemmAcc, acc1_thread_buf.Size(), true>
c_thread_buf;
c_thread_buf.Clear();
// Initialize running sum and max of exponentiating row vectors
using SoftmaxBuf = typename decltype(blockwise_softmax)::BufferType;
SoftmaxBuf running_sum, running_sum_new, running_max, running_max_new;
running_sum = 0;
running_sum_new = 0;
running_max = NumericLimits<FloatGemmAcc>::Lowest();
running_max_new = NumericLimits<FloatGemmAcc>::Lowest();
// gemm1 K loop
index_t gemm1_k_block_outer_index = 0;
do
{
auto n_block_data_idx_on_grid =
__builtin_amdgcn_readfirstlane(gemm1_k_block_outer_index * NPerBlock);
if(c0_matrix_mask.IsTileSkippable(
m_block_data_idx_on_grid, n_block_data_idx_on_grid, MPerBlock, NPerBlock))
{
continue;
}
// gemm0
gridwise_gemm_pipeline.template Run<HasMainKBlockLoop>(a_grid_desc_ak0_m_ak1,
a_block_desc_ak0_m_ak1,
a_blockwise_copy,
a_grid_buf,
a_block_buf,
a_block_slice_copy_step,
b_grid_desc_bk0_n_bk1,
b_block_desc_bk0_n_bk1,
b_blockwise_copy,
b_grid_buf,
b_block_buf,
b_block_slice_copy_step,
blockwise_gemm,
acc_thread_buf,
num_k_block_main_loop);
// do MNK padding or upper triangular masking
if constexpr(MaskOutUpperTriangle || PadN)
{
// 8d thread_desc in thread scope
constexpr auto c_thread_lengths =
blockwise_gemm.GetCThreadDescriptor_M0_N0_M1_N1_M2_N2_N3_N4().GetLengths();
// 8d block_desc in block scope
constexpr auto c_block_lengths =
blockwise_gemm.GetCBlockDescriptor_M0_N0_M1_N1_M2_N2_N3_N4().GetLengths();
constexpr auto M0 = c_block_lengths[I0];
constexpr auto N0 = c_block_lengths[I1];
constexpr auto M1 = c_block_lengths[I2];
constexpr auto N1 = c_block_lengths[I3];
constexpr auto M2 = c_block_lengths[I4];
constexpr auto N2 = c_block_lengths[I5];
constexpr auto N3 = c_block_lengths[I6];
constexpr auto N4 = c_block_lengths[I7];
// works like multi-dimension static_for (static_ford), but provides both the linear
// index as well as n-d index
using Acc0TileIterator = SpaceFillingCurve<
decltype(c_thread_lengths),
typename arithmetic_sequence_gen<0, c_thread_lengths.Size(), 1>::type,
typename uniform_sequence_gen<c_thread_lengths.Size(), 1>::type,
false>; // SnakeCurved
auto acc0_thread_origin = blockwise_gemm.CalculateCThreadOriginDataIndex8D(
Number<0>{}, Number<0>{}, Number<0>{}, Number<0>{});
constexpr auto block_idx_to_m_n_adaptor = make_single_stage_tensor_adaptor(
make_tuple(make_unmerge_transform(make_tuple(M0, M1, M2)),
make_unmerge_transform(make_tuple(N0, N1, N2, N3, N4))),
make_tuple(Sequence<0>{}, Sequence<1>{}),
make_tuple(Sequence<0, 2, 4>{}, Sequence<1, 3, 5, 6, 7>{}));
static_for<0, Acc0TileIterator::GetNumOfAccess(), 1>{}([&](auto i) {
auto acc0_thread_idx = Acc0TileIterator::GetIndex(i) + acc0_thread_origin;
auto m_local =
block_idx_to_m_n_adaptor.CalculateBottomIndex(acc0_thread_idx)[I0];
auto n_local =
block_idx_to_m_n_adaptor.CalculateBottomIndex(acc0_thread_idx)[I1];
auto m_global = m_local + m_block_data_idx_on_grid;
auto n_global = n_local + n_block_data_idx_on_grid;
if(c0_matrix_mask.IsMaskedElement(m_global, n_global))
{
acc_thread_buf(i) = -ck::NumericLimits<float>::Infinity();
}
else
{
acc_element_op(acc_thread_buf(i), acc_thread_buf[i]);
}
});
}
else
{
static_for<0, acc_thread_buf.Size(), 1>{}(
[&](auto i) { acc_element_op(acc_thread_buf(i), acc_thread_buf[i]); });
}
block_sync_lds(); // wait for lds read in gemm0 blockwise gemm
// add bias
if constexpr(!is_same<D0DataType, void>::value)
{
const auto d0_grid_buf = make_dynamic_buffer<AddressSpaceEnum::Global>(
p_d0_grid, d0_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5.GetElementSpaceSize());
// get register
StaticBuffer<AddressSpaceEnum::Vgpr,
D0DataType,
d0_thread_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5.GetElementSpaceSize(),
true>
d0_thread_buf;
// load data from global
d0_threadwise_copy.Run(d0_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
d0_grid_buf,
d0_thread_desc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
make_tuple(I0, I0, I0, I0, I0, I0, I0, I0, I0, I0),
d0_thread_buf);
// acc add bias
static_for<0, m0 * n0 * n2 * n4, 1>{}([&](auto i) {
acc_thread_buf(i) += ck::type_convert<FloatGemmAcc>(d0_thread_buf[i]);
});
d0_threadwise_copy.MoveSrcSliceWindow(
d0_griddesc_m0_n0_m1_n1_m2_n2_m3_n3_n4_n5,
make_multi_index(0, 1, 0, 0, 0, 0, 0, 0, 0, 0));
}
// softmax
SoftmaxBuf& max = blockwise_softmax.max_value_buf;
SoftmaxBuf& sum = blockwise_softmax.sum_value_buf;
blockwise_softmax.Run(acc_thread_buf, workspace_buf);
// TODO: may convert to log domain
running_max_new = mathext::max(max, running_max);
running_sum_new = mathext::exp(running_max - running_max_new) * running_sum +
mathext::exp(max - running_max_new) * sum;
// gemm1
{
// TODO: explore using dynamic buffer for a1 thread buffer
// For a1_blockwise_copy, the goal is to satisfy pipeline requirements RunRead(),
// RunWrite(), and MoveSliceWindow(). But it is impossible to implement given that
// the A1 source buffer is static buffer holding the output of first GEMM and
// requires constexpr offset by design. Therefore, we pass tensor coordinate offset
// explicitly in Run() below.
// Initialize acc1
acc1_thread_buf.Clear();
// preload data into LDS
b1_blockwise_copy.RunRead(b1_grid_desc_bk0_n_bk1, b1_grid_buf);
b1_blockwise_copy.MoveSrcSliceWindow(b1_grid_desc_bk0_n_bk1,
b1_block_slice_copy_step);
block_sync_lds(); // wait for reduction LDS read
b1_blockwise_copy.RunWrite(b1_block_desc_bk0_n_bk1, b1_block_buf);
// main body
if constexpr(num_gemm1_k_block_inner_loop > 1)
{
static_for<0, num_gemm1_k_block_inner_loop - 1, 1>{}([&](auto i) {
a1_blockwise_copy.Run(acc_thread_desc_k0_m_k1,
make_tuple(Number<i * A1ThreadSliceK0>{}, I0, I0),
acc_thread_buf,
a1_thread_desc_k0_m_k1,
make_tuple(I0, I0, I0),
a1_thread_buf);
b1_blockwise_copy.RunRead(b1_grid_desc_bk0_n_bk1, b1_grid_buf);
block_sync_lds();
gemm1_blockwise_gemm.Run(a1_thread_buf, b1_block_buf, acc1_thread_buf);
block_sync_lds();
b1_blockwise_copy.MoveSrcSliceWindow(b1_grid_desc_bk0_n_bk1,
b1_block_slice_copy_step);
b1_blockwise_copy.RunWrite(b1_block_desc_bk0_n_bk1, b1_block_buf);
});
}
// tail
{
a1_blockwise_copy.Run(
acc_thread_desc_k0_m_k1,
make_tuple(
Number<(num_gemm1_k_block_inner_loop - 1) * A1ThreadSliceK0>{}, I0, I0),
acc_thread_buf,
a1_thread_desc_k0_m_k1,
make_tuple(I0, I0, I0),
a1_thread_buf);
block_sync_lds();
gemm1_blockwise_gemm.Run(a1_thread_buf, b1_block_buf, acc1_thread_buf);
}
} // end gemm1
constexpr auto c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4 =
gemm1_blockwise_gemm.GetCThreadDescriptor_M0_N0_M1_N1_M2_N2_N3_N4();
constexpr auto cm0 = c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I0);
constexpr auto cn0 = c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I1);
constexpr auto cm1 = c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I2);
constexpr auto cn1 = c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I3);
constexpr auto cm2 = c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I4);
constexpr auto cn2 = c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I5);
constexpr auto cn3 = c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I6);
constexpr auto cn4 = c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4.GetLength(I7);
constexpr auto c_thread_slice_desc_m_n = make_naive_tensor_descriptor_packed(
make_tuple(cm0 * cm1 * cm2, cn0 * cn1 * cn2 * cn3 * cn4));
constexpr auto c_thread_buf_slice_m = c_thread_slice_desc_m_n.GetLength(I0);
constexpr auto c_thread_buf_slice_n = c_thread_slice_desc_m_n.GetLength(I1);
static_for<0, c_thread_buf_slice_m, 1>{}([&](auto iM) {
static_for<0, c_thread_buf_slice_n, 1>{}([&](auto iN) {
auto I = Number<c_thread_slice_desc_m_n.CalculateOffset(make_tuple(iM, iN))>{};
FloatGemmAcc acc1 = acc1_thread_buf[I]; // P*V
FloatGemmAcc c = c_thread_buf[I]; // O
FloatGemmAcc c_new =
(running_sum[iM] * math::exp(running_max[iM] - running_max_new[iM]) * c +
math::exp(max[iM] - running_max_new[iM]) * acc1) /
running_sum_new[iM]; // Formula by Dao et al.,
// https://arxiv.org/pdf/2205.14135v2.pdf section 3.1
c_thread_buf(I) = c_new; // O_new
});
});
a_blockwise_copy.MoveSrcSliceWindow(a_grid_desc_ak0_m_ak1,
a_block_reset_copy_step); // rewind K
b_blockwise_copy.MoveSrcSliceWindow(b_grid_desc_bk0_n_bk1,
b_block_reset_copy_step); // rewind K and step N
// update before next j iteration
running_max = running_max_new;
running_sum = running_sum_new;
block_sync_lds(); // wait for gemm1 LDS read
} while(++gemm1_k_block_outer_index < num_gemm1_k_block_outer_loop); // end j loop
// shuffle C and write out
{
static_assert(MXdlPerWave % CShuffleMXdlPerWavePerShuffle == 0 &&
Gemm1NXdlPerWave % CShuffleNXdlPerWavePerShuffle == 0,
"wrong!");
constexpr index_t MWave = MPerBlock / (MXdlPerWave * MPerXdl);
constexpr index_t NWave = Gemm1NPerBlock / (Gemm1NXdlPerWave * NPerXdl);
// TODO: hacky, fix it!
constexpr auto c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4 =
gemm1_blockwise_gemm.GetCThreadDescriptor_M0_N0_M1_N1_M2_N2_N3_N4();
// TODO: hacky, fix it!
// c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4_tmp is only used to get lengths
constexpr auto c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4_tmp =
gemm1_blockwise_gemm.GetCBlockDescriptor_M0_N0_M1_N1_M2_N2_N3_N4();
constexpr auto M0 = c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4_tmp.GetLength(I0);
constexpr auto N0 = c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4_tmp.GetLength(I1);
constexpr auto M1 = c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4_tmp.GetLength(I2);
constexpr auto N1 = c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4_tmp.GetLength(I3);
constexpr auto M2 = c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4_tmp.GetLength(I4);
constexpr auto N2 = c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4_tmp.GetLength(I5);
constexpr auto N3 = c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4_tmp.GetLength(I6);
constexpr auto N4 = c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4_tmp.GetLength(I7);
constexpr auto c_shuffle_block_desc_mblock_mperblock_nblock_nperblock =
GetCShuffleBlockDescriptor_MBlock_MPerBlock_NBlock_NPerBlock();
auto c_shuffle_block_buf = make_dynamic_buffer<AddressSpaceEnum::Lds>(
static_cast<FloatCShuffle*>(p_shared),
c_shuffle_block_desc_mblock_mperblock_nblock_nperblock.GetElementSpaceSize());
constexpr auto c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4 = transform_tensor_descriptor(
c_shuffle_block_desc_mblock_mperblock_nblock_nperblock,
make_tuple(
make_freeze_transform(I0),
make_unmerge_transform(make_tuple(
Number<CShuffleMXdlPerWavePerShuffle>{}, // M0 (MXdlPerWave) per shuffle
M1, // M1 = MWave
M2)), // M2 = MPerXdl
make_freeze_transform(I0),
make_unmerge_transform(make_tuple(
Number<CShuffleNXdlPerWavePerShuffle>{}, // N0 (NXdlPerWave) per shuffle
N1, // N1 = NWave
N2, // N2 * N3 * N4 = NPerXdl
N3,
N4))),
make_tuple(Sequence<0>{}, Sequence<1>{}, Sequence<2>{}, Sequence<3>{}),
make_tuple(
Sequence<>{}, Sequence<0, 2, 4>{}, Sequence<>{}, Sequence<1, 3, 5, 6, 7>{}));
// calculate origin of thread output tensor on global memory
// blockwise GEMM c matrix starting index
const auto c_thread_mtx_on_block =
gemm1_blockwise_gemm.CalculateCThreadOriginDataIndex(I0, I0, I0, I0);
const index_t m_thread_data_on_block = c_thread_mtx_on_block[I0];
const index_t n_thread_data_on_block = c_thread_mtx_on_block[I1];
const auto m_thread_data_on_block_to_m0_m1_m2_adaptor =
make_single_stage_tensor_adaptor(
make_tuple(make_merge_transform(make_tuple(M0, M1, M2))),
make_tuple(Sequence<0, 1, 2>{}),
make_tuple(Sequence<0>{}));
const auto m_thread_data_on_block_idx =
m_thread_data_on_block_to_m0_m1_m2_adaptor.CalculateBottomIndex(
make_multi_index(m_thread_data_on_block));
const auto n_thread_data_on_block_to_n0_n1_n2_n3_n4_adaptor =
make_single_stage_tensor_adaptor(
make_tuple(make_merge_transform(make_tuple(N0, N1, N2, N3, N4))),
make_tuple(Sequence<0, 1, 2, 3, 4>{}),
make_tuple(Sequence<0>{}));
const auto n_thread_data_on_block_idx =
n_thread_data_on_block_to_n0_n1_n2_n3_n4_adaptor.CalculateBottomIndex(
make_multi_index(n_thread_data_on_block));
// shuffle: threadwise copy C from VGPR to LDS
auto c_thread_copy_vgpr_to_lds =
ThreadwiseTensorSliceTransfer_v1r3<FloatGemmAcc,
FloatCShuffle,
decltype(c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4),
decltype(c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4),
tensor_operation::element_wise::PassThrough,
Sequence<CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
I1,
I1,
I1,
N2,
I1,
N4>,
Sequence<0, 1, 2, 3, 4, 5, 6, 7>,
7,
1,
InMemoryDataOperationEnum::Set,
1,
true>{
c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4,
make_multi_index(0,
0,
m_thread_data_on_block_idx[I1],
n_thread_data_on_block_idx[I1],
m_thread_data_on_block_idx[I2],
n_thread_data_on_block_idx[I2],
n_thread_data_on_block_idx[I3],
n_thread_data_on_block_idx[I4]),
tensor_operation::element_wise::PassThrough{}};
// shuffle: blockwise copy C from LDS to global
auto c_shuffle_block_copy_lds_to_global = ThreadGroupTensorSliceTransfer_v6r1<
ThisThreadBlock, // ThreadGroup
CElementwiseOperation, // ElementwiseOperation,
CGlobalMemoryDataOperation, // DstInMemOp,
Sequence<1,
CShuffleMXdlPerWavePerShuffle * MWave * MPerXdl,
1,
CShuffleNXdlPerWavePerShuffle * NWave * NPerXdl>, // BlockSliceLengths,
CShuffleBlockTransferClusterLengths_MBlock_MPerBlock_NBlock_NPerBlock,
Sequence<0, 1, 2, 3>, // typename ThreadClusterArrangeOrder,
FloatCShuffle, // typename SrcData,
FloatC, // typename DstData,
decltype(c_shuffle_block_desc_mblock_mperblock_nblock_nperblock),
decltype(c1_grid_desc_mblock_mperblock_nblock_nperblock),
Sequence<0, 1, 2, 3>, // typename DimAccessOrder,
3, // index_t VectorDim,
CShuffleBlockTransferScalarPerVector_NPerBlock, // index_t ScalarPerVector,
true, // bool ThreadTransferSrcResetCoordinateAfterRun,
false> // bool ThreadTransferDstResetCoordinateAfterRun>
{c_shuffle_block_desc_mblock_mperblock_nblock_nperblock,
make_multi_index(0, 0, 0, 0),
c1_grid_desc_mblock_mperblock_nblock_nperblock,
make_multi_index(block_work_idx[I0], 0, block_work_idx[I1], 0),
c_element_op};
// space filling curve for threadwise C in VGPR
constexpr auto sfc_c_vgpr =
SpaceFillingCurve<Sequence<MXdlPerWave, Gemm1NXdlPerWave, 1, 1, 1, N2, 1, N4>,
Sequence<0, 1, 2, 3, 4, 5, 6, 7>,
Sequence<CShuffleMXdlPerWavePerShuffle,
CShuffleNXdlPerWavePerShuffle,
1,
1,
1,
N2,
1,
N4>>{};
// space filling curve for shuffled blockwise C in global mem
constexpr auto sfc_c_global =
SpaceFillingCurve<Sequence<1, MPerBlock, 1, Gemm1NPerBlock>,
Sequence<0, 2, 1, 3>,
Sequence<1,
CShuffleMXdlPerWavePerShuffle * MWave * MPerXdl,
1,
CShuffleNXdlPerWavePerShuffle * NWave * NPerXdl>>{};
constexpr index_t num_access = sfc_c_vgpr.GetNumOfAccess();
static_assert(num_access == sfc_c_global.GetNumOfAccess(), "wrong!");
static_for<0, num_access, 1>{}([&](auto access_id) {
// make sure it's safe to write to LDS
block_sync_lds();
// each thread write its data from VGPR to LDS
c_thread_copy_vgpr_to_lds.Run(c_thread_desc_m0_n0_m1_n1_m2_n2_n3_n4,
sfc_c_vgpr.GetIndexTupleOfNumber(access_id),
c_thread_buf,
c_block_desc_m0_n0_m1_n1_m2_n2_n3_n4,
c_shuffle_block_buf);
// make sure it's safe to read from LDS
block_sync_lds();
// each block copy its data from LDS to global
c_shuffle_block_copy_lds_to_global.Run(
c_shuffle_block_desc_mblock_mperblock_nblock_nperblock,
c_shuffle_block_buf,
c1_grid_desc_mblock_mperblock_nblock_nperblock,
c_grid_buf);
if constexpr(access_id < num_access - 1)
{
constexpr auto c_global_step = sfc_c_global.GetForwardStep(access_id);
// move on C
c_shuffle_block_copy_lds_to_global.MoveDstSliceWindow(
c1_grid_desc_mblock_mperblock_nblock_nperblock, c_global_step);
}
});
}
}
};
} // namespace ck
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