mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-04-20 06:49:15 +00:00
[rocm-libraries] ROCm/rocm-libraries#4340 (commit 70a312f)
Implement device_grouped_gemm_fixed_nk_bias for RDNA4 ## Proposed changes Summary: - Modified implementation for grouped_gemm_fixed_nk_bias - FP16 WMMA examples - WMMA instances - Profiler for grouped_gemm_fixed_nk_bias - Add WMMA instances to existing tests **This PR depends on PR https://github.com/ROCm/rocm-libraries/pull/4299 and should be merged after it. Only the last 6 commits are in the scope of this PR.** ## Checklist Please put an `x` into the boxes that apply. You can also fill these out after creating the PR. If you're not sure, please don't hesitate to ask. - [x] I have added tests relevant to the introduced functionality, and the unit tests are passing locally - [x] I have added the test to REGRESSION_TESTS list defined at the top of CMakeLists.txt in tests/CMakeLists.txt, **IF** the test takes more than 30 seconds to run. - [x] I have added inline documentation which enables the maintainers with understanding the motivation - [x] I have removed the stale documentation which is no longer relevant after this pull request - [ ] (If this change is user-facing) I have added release notes which provide the end users with a brief summary of the improvement from this pull request - [x] I have run `clang-format` on all changed files - [ ] Any dependent changes have been merged ## Discussion If this is a relatively large or complex change, feel free to start a discussion by explaining why you chose the solution you did and what alternatives you considered ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.
This commit is contained in:
Yung-sheng Tu
committed by
assistant-librarian[bot]
assistant-librarian[bot]
parent
9a32f0ea19
commit
75aea70c2c
@@ -0,0 +1,411 @@
|
||||
// Copyright (c) Advanced Micro Devices, Inc., or its affiliates.
|
||||
// SPDX-License-Identifier: MIT
|
||||
|
||||
#pragma once
|
||||
|
||||
#include "ck/ck.hpp"
|
||||
#include "ck/library/reference_tensor_operation/cpu/reference_gemm.hpp"
|
||||
#include "ck/library/tensor_operation_instance/gpu/grouped_gemm_bias.hpp"
|
||||
#include "ck/library/utility/check_err.hpp"
|
||||
#include "ck/library/utility/device_memory.hpp"
|
||||
#include "ck/library/utility/fill.hpp"
|
||||
#include "ck/library/utility/host_tensor.hpp"
|
||||
#include "ck/library/utility/literals.hpp"
|
||||
#include "ck/stream_config.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/device_grouped_gemm.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/device_grouped_gemm_fixed_nk.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/impl/device_grouped_gemm_wmma_fixed_nk.hpp"
|
||||
#include "ck/tensor_operation/gpu/device/tensor_layout.hpp"
|
||||
#include "ck/tensor_operation/gpu/element/binary_element_wise_operation.hpp"
|
||||
#include "ck/tensor_operation/gpu/element/unary_element_wise_operation.hpp"
|
||||
#include "ck/utility/env.hpp"
|
||||
#include "ck/utility/tuple.hpp"
|
||||
#include "ck/utility/type.hpp"
|
||||
|
||||
#include <array>
|
||||
#include <cstddef>
|
||||
#include <iomanip>
|
||||
#include <iostream>
|
||||
#include <memory>
|
||||
#include <stdexcept>
|
||||
#include <string>
|
||||
#include <vector>
|
||||
|
||||
namespace ck {
|
||||
namespace profiler {
|
||||
|
||||
template <typename ADataType,
|
||||
typename BDataType,
|
||||
typename DsDataType,
|
||||
typename EDataType,
|
||||
typename AccDataType,
|
||||
typename ALayout,
|
||||
typename BLayout,
|
||||
typename DsLayout,
|
||||
typename ELayout>
|
||||
bool profile_grouped_gemm_fixed_nk_bias_impl(int do_verification,
|
||||
int init_method,
|
||||
bool do_log,
|
||||
bool time_kernel,
|
||||
const std::vector<int>& Ms,
|
||||
const std::vector<int>& Ns,
|
||||
const std::vector<int>& Ks,
|
||||
const std::vector<int>& StrideAs,
|
||||
const std::vector<int>& StrideBs,
|
||||
const std::vector<int>& StrideDs,
|
||||
const std::vector<int>& StrideEs,
|
||||
const std::vector<int>& kbatches = {1},
|
||||
int n_warmup = 1,
|
||||
int n_iter = 10)
|
||||
{
|
||||
bool pass = true;
|
||||
using ComputeDataType = ADataType;
|
||||
|
||||
auto f_host_tensor_descriptor =
|
||||
[](std::size_t row, std::size_t col, std::size_t stride, auto layout) {
|
||||
using namespace ck::literals;
|
||||
|
||||
if(is_same<decltype(layout), tensor_layout::gemm::RowMajor>::value)
|
||||
{
|
||||
return HostTensorDescriptor({row, col}, {stride, 1_uz}, layout);
|
||||
}
|
||||
else
|
||||
{
|
||||
return HostTensorDescriptor({row, col}, {1_uz, stride}, layout);
|
||||
}
|
||||
};
|
||||
|
||||
std::size_t group_count = Ms.size();
|
||||
|
||||
if(!(group_count == Ns.size() && group_count == Ks.size() && group_count == StrideAs.size() &&
|
||||
group_count == StrideBs.size() && group_count == StrideDs.size() &&
|
||||
group_count == StrideEs.size()))
|
||||
{
|
||||
throw std::runtime_error("wrong! inconsistent M/N/Ks, StrideA/B/Cs size\n");
|
||||
}
|
||||
|
||||
using D0DataType = remove_cvref_t<ck::tuple_element_t<Number<0>{}, DsDataType>>;
|
||||
|
||||
std::vector<Tensor<ADataType>> a_tensors;
|
||||
std::vector<Tensor<BDataType>> b_tensors;
|
||||
std::vector<Tensor<D0DataType>> d0_tensors;
|
||||
std::vector<Tensor<EDataType>> e_host_tensors;
|
||||
std::vector<Tensor<EDataType>> e_device_tensors;
|
||||
|
||||
a_tensors.reserve(group_count);
|
||||
b_tensors.reserve(group_count);
|
||||
d0_tensors.reserve(group_count);
|
||||
e_host_tensors.reserve(group_count);
|
||||
e_device_tensors.reserve(group_count);
|
||||
|
||||
double max_abs_in_val = 0.f;
|
||||
int sum_of_m = 0;
|
||||
|
||||
using D0Layout = remove_cvref_t<ck::tuple_element_t<Number<0>{}, DsLayout>>;
|
||||
|
||||
for(std::size_t i = 0; i < group_count; ++i)
|
||||
{
|
||||
sum_of_m += Ms[i];
|
||||
a_tensors.push_back(
|
||||
Tensor<ADataType>(f_host_tensor_descriptor(Ms[i], Ks[i], StrideAs[i], ALayout{})));
|
||||
b_tensors.push_back(
|
||||
Tensor<BDataType>(f_host_tensor_descriptor(Ks[i], Ns[i], StrideBs[i], BLayout{})));
|
||||
d0_tensors.push_back(
|
||||
Tensor<D0DataType>(f_host_tensor_descriptor(Ms[i], Ns[i], StrideDs[i], D0Layout{})));
|
||||
e_host_tensors.push_back(
|
||||
Tensor<EDataType>(f_host_tensor_descriptor(Ms[i], Ns[i], StrideEs[i], ELayout{})));
|
||||
e_device_tensors.push_back(
|
||||
Tensor<EDataType>(f_host_tensor_descriptor(Ms[i], Ns[i], StrideEs[i], ELayout{})));
|
||||
|
||||
if(ck::EnvIsEnabled(CK_ENV(CK_LOGGING)))
|
||||
{
|
||||
std::cout << "group: " << i << " a_m_k[" << i << "]:" << a_tensors[i].mDesc
|
||||
<< ", b_k_n[" << i << "]:" << b_tensors[i].mDesc << ", d_m_n[" << i
|
||||
<< "]:" << d0_tensors[i].mDesc << ", e_m_n_device_results[" << i
|
||||
<< "]:" << e_device_tensors[i].mDesc << std::endl;
|
||||
}
|
||||
switch(init_method)
|
||||
{
|
||||
case 0: break;
|
||||
case 1:
|
||||
ck::utils::FillUniformDistributionIntegerValue<ADataType>{-5.f, 5.f}(a_tensors[i]);
|
||||
ck::utils::FillUniformDistributionIntegerValue<BDataType>{-5.f, 5.f}(b_tensors[i]);
|
||||
max_abs_in_val = 10.f;
|
||||
break;
|
||||
default:
|
||||
ck::utils::FillUniformDistribution<ADataType>{0.0f, 1.0f}(a_tensors[i]);
|
||||
ck::utils::FillUniformDistribution<BDataType>{-0.5f, 0.5f}(b_tensors[i]);
|
||||
max_abs_in_val = 1.0f;
|
||||
}
|
||||
ck::utils::FillUniformDistribution<D0DataType>{-0.5, 0.5}(d0_tensors[i]);
|
||||
}
|
||||
|
||||
using AElementOp = ck::tensor_operation::element_wise::PassThrough;
|
||||
using BElementOp = ck::tensor_operation::element_wise::PassThrough;
|
||||
using CDEElementOp = ck::tensor_operation::element_wise::SplitKAdd;
|
||||
|
||||
constexpr auto a_element_op = AElementOp{};
|
||||
constexpr auto b_element_op = BElementOp{};
|
||||
constexpr auto cde_element_op = CDEElementOp{};
|
||||
|
||||
using DeviceMemPtr = std::unique_ptr<DeviceMem>;
|
||||
std::vector<DeviceMemPtr> a_tensors_device, b_tensors_device, d0_tensors_device,
|
||||
e_tensors_device;
|
||||
|
||||
a_tensors_device.reserve(group_count);
|
||||
b_tensors_device.reserve(group_count);
|
||||
d0_tensors_device.reserve(group_count);
|
||||
e_tensors_device.reserve(group_count);
|
||||
|
||||
std::vector<const void*> p_a, p_b;
|
||||
std::vector<std::array<const void*, 1>> p_ds;
|
||||
std::vector<void*> p_e;
|
||||
|
||||
p_a.reserve(group_count);
|
||||
p_b.reserve(group_count);
|
||||
p_ds.reserve(group_count);
|
||||
p_e.reserve(group_count);
|
||||
|
||||
std::vector<ck::tensor_operation::device::GemmDesc> gemm_descs;
|
||||
gemm_descs.reserve(group_count);
|
||||
|
||||
std::vector<ck::tensor_operation::device::GroupedGemmKernelArgument<1>>
|
||||
grouped_gemm_kernel_args_;
|
||||
grouped_gemm_kernel_args_.reserve(group_count);
|
||||
|
||||
for(std::size_t i = 0; i < group_count; ++i)
|
||||
{
|
||||
a_tensors_device.emplace_back(std::make_unique<DeviceMem>(
|
||||
sizeof(ADataType) * a_tensors[i].mDesc.GetElementSpaceSize()));
|
||||
b_tensors_device.emplace_back(std::make_unique<DeviceMem>(
|
||||
sizeof(BDataType) * b_tensors[i].mDesc.GetElementSpaceSize()));
|
||||
d0_tensors_device.emplace_back(std::make_unique<DeviceMem>(
|
||||
sizeof(D0DataType) * d0_tensors[i].mDesc.GetElementSpaceSize()));
|
||||
e_tensors_device.emplace_back(std::make_unique<DeviceMem>(
|
||||
sizeof(EDataType) * e_device_tensors[i].mDesc.GetElementSpaceSize()));
|
||||
|
||||
a_tensors_device[i]->ToDevice(a_tensors[i].mData.data());
|
||||
b_tensors_device[i]->ToDevice(b_tensors[i].mData.data());
|
||||
d0_tensors_device[i]->ToDevice(d0_tensors[i].mData.data());
|
||||
|
||||
gemm_descs.push_back(
|
||||
{sum_of_m, Ns[i], Ks[i], StrideAs[i], StrideBs[i], StrideEs[i], {StrideDs[i]}});
|
||||
|
||||
p_a.push_back(a_tensors_device[i]->GetDeviceBuffer());
|
||||
p_b.push_back(b_tensors_device[i]->GetDeviceBuffer());
|
||||
p_ds.push_back(std::array<const void*, 1>{d0_tensors_device[i]->GetDeviceBuffer()});
|
||||
p_e.push_back(e_tensors_device[i]->GetDeviceBuffer());
|
||||
|
||||
grouped_gemm_kernel_args_.push_back(
|
||||
{a_tensors_device[i]->GetDeviceBuffer(),
|
||||
b_tensors_device[i]->GetDeviceBuffer(),
|
||||
std::array<const void*, 1>{d0_tensors_device[i]->GetDeviceBuffer()},
|
||||
e_tensors_device[i]->GetDeviceBuffer(),
|
||||
Ms[i],
|
||||
Ns[i],
|
||||
Ks[i],
|
||||
StrideAs[i],
|
||||
StrideBs[i],
|
||||
std::array<ck::index_t, 1>{StrideDs[i]},
|
||||
StrideEs[i]});
|
||||
}
|
||||
|
||||
using DeviceOp = ck::tensor_operation::device::DeviceGroupedGemmFixedNK<ALayout,
|
||||
BLayout,
|
||||
DsLayout,
|
||||
ELayout,
|
||||
ADataType,
|
||||
BDataType,
|
||||
DsDataType,
|
||||
EDataType,
|
||||
AElementOp,
|
||||
BElementOp,
|
||||
CDEElementOp>;
|
||||
|
||||
const auto op_ptrs = ck::tensor_operation::device::instance::DeviceOperationInstanceFactory<
|
||||
DeviceOp>::GetInstances();
|
||||
|
||||
if(op_ptrs.size() <= 0)
|
||||
{
|
||||
std::cerr << "Skip! no device GEMM instance found" << std::endl;
|
||||
return true;
|
||||
}
|
||||
|
||||
std::string best_gemm_name;
|
||||
float best_ave_time = 0;
|
||||
float best_tflops = 0;
|
||||
float best_gb_per_sec = 0;
|
||||
float best_kbatch = 0;
|
||||
|
||||
if(do_verification)
|
||||
{
|
||||
for(std::size_t i = 0; i < gemm_descs.size(); ++i)
|
||||
{
|
||||
using ReferenceGemmInstance = ck::tensor_operation::host::ReferenceGemm<
|
||||
ADataType,
|
||||
BDataType,
|
||||
EDataType,
|
||||
AccDataType,
|
||||
AElementOp,
|
||||
BElementOp,
|
||||
ck::tensor_operation::element_wise::PassThrough>;
|
||||
|
||||
auto ref_gemm = ReferenceGemmInstance{};
|
||||
auto ref_invoker = ref_gemm.MakeInvoker();
|
||||
|
||||
auto ref_argument =
|
||||
ref_gemm.MakeArgument(a_tensors[i],
|
||||
b_tensors[i],
|
||||
e_host_tensors[i],
|
||||
a_element_op,
|
||||
b_element_op,
|
||||
ck::tensor_operation::element_wise::PassThrough{});
|
||||
|
||||
ref_invoker.Run(ref_argument);
|
||||
|
||||
for(int m = 0; m < Ms[i]; ++m)
|
||||
{
|
||||
for(int n = 0; n < Ns[i]; ++n)
|
||||
{
|
||||
cde_element_op(
|
||||
e_host_tensors[i](m, n), e_host_tensors[i](m, n), d0_tensors[i](m, n));
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// profile device GEMM instances
|
||||
for(auto& gemm_ptr : op_ptrs)
|
||||
{
|
||||
auto argument_ptr = gemm_ptr->MakeArgumentPointer(
|
||||
p_a, p_b, p_ds, p_e, gemm_descs, a_element_op, b_element_op, cde_element_op);
|
||||
|
||||
auto invoker_ptr = gemm_ptr->MakeInvokerPointer();
|
||||
|
||||
DeviceMem gemm_desc_workspace(gemm_ptr->GetWorkSpaceSize(argument_ptr.get()));
|
||||
|
||||
DeviceMem grouped_gemm_kernel_args_dev(
|
||||
gemm_ptr->GetDeviceKernelArgSize(argument_ptr.get()));
|
||||
|
||||
hipGetErrorString(hipMemcpy(grouped_gemm_kernel_args_dev.GetDeviceBuffer(),
|
||||
grouped_gemm_kernel_args_.data(),
|
||||
gemm_ptr->GetDeviceKernelArgSize(argument_ptr.get()),
|
||||
hipMemcpyHostToDevice));
|
||||
|
||||
gemm_ptr->SetWorkSpacePointer(argument_ptr.get(), gemm_desc_workspace.GetDeviceBuffer());
|
||||
|
||||
gemm_ptr->SetDeviceKernelArgs(argument_ptr.get(),
|
||||
grouped_gemm_kernel_args_dev.GetDeviceBuffer());
|
||||
|
||||
std::string gemm_name = gemm_ptr->GetTypeString();
|
||||
|
||||
for(std::size_t j = 0; j < kbatches.size(); ++j)
|
||||
{
|
||||
|
||||
auto kbatch_curr = kbatches[j];
|
||||
|
||||
gemm_ptr->SetKBatch(argument_ptr.get(), kbatch_curr);
|
||||
|
||||
if(gemm_ptr->IsSupportedArgument(argument_ptr.get()))
|
||||
{
|
||||
for(std::size_t i = 0; i < gemm_descs.size(); ++i)
|
||||
{
|
||||
e_tensors_device[i]->SetZero();
|
||||
}
|
||||
|
||||
invoker_ptr->Run(argument_ptr.get(),
|
||||
StreamConfig{nullptr, false, 0, n_warmup, n_iter});
|
||||
|
||||
if(do_verification)
|
||||
{
|
||||
bool instance_pass = true;
|
||||
for(std::size_t i = 0; i < gemm_descs.size(); ++i)
|
||||
{
|
||||
e_tensors_device[i]->FromDevice(e_device_tensors[i].mData.data());
|
||||
auto atol = ck::utils::get_absolute_threshold<ComputeDataType, EDataType>(
|
||||
max_abs_in_val, gemm_descs[i].K_);
|
||||
auto rtol = ck::utils::get_relative_threshold<ComputeDataType, EDataType>(
|
||||
gemm_descs[i].K_);
|
||||
|
||||
instance_pass =
|
||||
instance_pass && ck::utils::check_err(e_device_tensors[i],
|
||||
e_host_tensors[i],
|
||||
"Error: Incorrect results!",
|
||||
rtol,
|
||||
atol);
|
||||
|
||||
if(do_log)
|
||||
{
|
||||
LogRangeAsType<float>(std::cout << "a : ", a_tensors[i].mData, ",")
|
||||
<< std::endl;
|
||||
LogRangeAsType<float>(std::cout << "b: ", b_tensors[i].mData, ",")
|
||||
<< std::endl;
|
||||
LogRangeAsType<float>(std::cout << "d0: ", d0_tensors[i].mData, ",")
|
||||
<< std::endl;
|
||||
LogRangeAsType<float>(
|
||||
std::cout << "e_device: ", e_device_tensors[i].mData, ",")
|
||||
<< std::endl;
|
||||
LogRangeAsType<float>(
|
||||
std::cout << "e_host : ", e_host_tensors[i].mData, ",")
|
||||
<< std::endl;
|
||||
}
|
||||
}
|
||||
|
||||
std::cout << "Instance: " << gemm_name << "; KBatch: " << kbatch_curr << " "
|
||||
<< (instance_pass ? "SUCCEED" : "FAILED") << std::endl;
|
||||
|
||||
pass = pass && instance_pass;
|
||||
}
|
||||
|
||||
float ave_time = invoker_ptr->Run(
|
||||
argument_ptr.get(), StreamConfig{nullptr, time_kernel, 0, n_warmup, n_iter});
|
||||
|
||||
if(time_kernel)
|
||||
{
|
||||
std::size_t flop = 0, num_btype = 0;
|
||||
for(std::size_t i = 0; i < gemm_descs.size(); i++)
|
||||
{
|
||||
flop += std::size_t(2) * Ms[i] * Ns[i] * Ks[i];
|
||||
|
||||
num_btype +=
|
||||
sizeof(ADataType) * Ms[i] * Ks[i] + sizeof(BDataType) * Ks[i] * Ns[i] +
|
||||
sizeof(D0DataType) * Ms[i] * Ns[i] + sizeof(EDataType) * Ms[i] * Ns[i];
|
||||
}
|
||||
|
||||
float tflops = static_cast<float>(flop) / 1.E9 / ave_time;
|
||||
|
||||
float gb_per_sec = num_btype / 1.E6 / ave_time;
|
||||
std::cout << "Perf: " << std::setw(10) << ave_time << " ms, " << tflops
|
||||
<< " TFlops, " << gb_per_sec << " GB/s, " << gemm_name << ", KBatch "
|
||||
<< kbatch_curr << std::endl;
|
||||
|
||||
if(tflops > best_tflops)
|
||||
{
|
||||
best_gemm_name = gemm_name;
|
||||
best_tflops = tflops;
|
||||
best_ave_time = ave_time;
|
||||
best_gb_per_sec = gb_per_sec;
|
||||
best_kbatch = kbatch_curr;
|
||||
}
|
||||
}
|
||||
}
|
||||
else
|
||||
{
|
||||
std::cout << "Instance: " << gemm_name
|
||||
<< ", does not support this GEMM problem (KBatch: " << kbatch_curr << ")"
|
||||
<< std::endl;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if(time_kernel)
|
||||
{
|
||||
std::cout << "Best Perf: " << best_ave_time << " ms, " << best_tflops << " TFlops, "
|
||||
<< best_gb_per_sec << " GB/s, " << best_gemm_name << ", KBatch = " << best_kbatch
|
||||
<< std::endl;
|
||||
}
|
||||
return pass;
|
||||
}
|
||||
|
||||
} // namespace profiler
|
||||
} // namespace ck
|
||||
Reference in New Issue
Block a user