* Copyright (c) 2026 Huawei Technologies Co., Ltd.
* This program is free software, you can redistribute it and/or modify it under the terms and conditions of
* CANN Open Software License Agreement Version 2.0 (the "License").
* Please refer to the License for details. You may not use this file except in compliance with the License.
* THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
* INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
* See LICENSE in the root of the software repository for the full text of the License.
*/
#ifndef K_MAX_SHAPE_DIM
#define K_MAX_SHAPE_DIM 0
#endif
#include "catlass/gemm/kernel/matrix_inverse.hpp"
#include "catlass/arch/arch.hpp"
#include "catlass/catlass.hpp"
#include "catlass/gemm/block/block_mmad.hpp"
#include "catlass/gemm/block/block_swizzle.hpp"
#include "catlass/gemm/device/device_gemm.hpp"
#include "catlass/gemm/dispatch_policy.hpp"
#include "catlass/gemm/gemm_type.hpp"
#include "catlass/layout/layout.hpp"
#include "catlass/status.hpp"
#include "tla/layout.hpp"
#include "tla/tensor.hpp"
#include <cmath>
#include <cstdlib>
#include "golden.hpp"
#include "helper.hpp"
using namespace Catlass;
using namespace tla;
static void Run(uint32_t N, int32_t deviceId)
{
aclrtStream stream{nullptr};
ACL_CHECK(aclInit(nullptr));
ACL_CHECK(aclrtSetDevice(deviceId));
ACL_CHECK(aclrtCreateStream(&stream));
using Element = float;
using LayoutTag = layout::RowMajor;
size_t lenA = static_cast<size_t>(N) * N;
size_t lenIpiv = static_cast<size_t>(N);
size_t sizeA = lenA * sizeof(Element);
size_t sizeIpiv = lenIpiv * sizeof(int32_t);
std::vector<float> hostA(lenA);
golden::FillRandomData(hostA, -1.0f, 1.0f);
for (uint32_t i = 0; i < N; ++i) {
hostA[i * N + i] += static_cast<float>(N);
}
std::vector<float> hostOriginal = hostA;
uint8_t* deviceA{nullptr};
ACL_CHECK(aclrtMalloc(reinterpret_cast<void**>(&deviceA), sizeA, ACL_MEM_MALLOC_HUGE_FIRST));
ACL_CHECK(aclrtMemcpy(deviceA, sizeA, hostA.data(), sizeA, ACL_MEMCPY_HOST_TO_DEVICE));
uint8_t* deviceIpiv{nullptr};
ACL_CHECK(aclrtMalloc(reinterpret_cast<void**>(&deviceIpiv), sizeIpiv, ACL_MEM_MALLOC_HUGE_FIRST));
auto aicCoreNum = platform_ascendc::PlatformAscendCManager::GetInstance()->GetCoreNumAic();
using ArchTag = Arch::AtlasA2;
constexpr bool enableUnitFlag = true;
constexpr bool useHF32 = false;
using DispatchPolicy = Gemm::MmadPingpong<ArchTag, enableUnitFlag, useHF32>;
using L1TileShape = Shape<_128, _128, _256>;
using L0TileShape = Shape<_128, _128, _64>;
using TileCopy = Gemm::Tile::PackedTileCopyTla<ArchTag, Element, LayoutTag, Element, LayoutTag, Element, LayoutTag>;
using BlockMmadType =
Gemm::Block::BlockMmadTla<DispatchPolicy, L1TileShape, L0TileShape, Element, Element, Element, void, TileCopy>;
using BlockSchedulerType = Gemm::Block::GemmIdentityBlockSwizzle<>;
using InverterKernel = Gemm::Kernel::MatrixInverse<ArchTag, float, BlockMmadType, BlockSchedulerType>;
using InverterAdapter = Gemm::Device::DeviceGemm<InverterKernel>;
auto layoutA = tla::MakeLayout<Element, LayoutTag>(N, N);
InverterKernel::Arguments arguments{N, deviceA, layoutA, deviceIpiv, nullptr};
InverterAdapter invOp;
invOp.CanImplement(arguments);
size_t sizeWorkspace = invOp.GetWorkspaceSize(arguments);
uint8_t* deviceWorkspace = nullptr;
if (sizeWorkspace > 0) {
ACL_CHECK(aclrtMalloc(reinterpret_cast<void**>(&deviceWorkspace), sizeWorkspace, ACL_MEM_MALLOC_HUGE_FIRST));
arguments = InverterKernel::Arguments{N, deviceA, layoutA, deviceIpiv, deviceWorkspace};
}
invOp.Initialize(arguments, deviceWorkspace);
aclrtEvent startEvent, endEvent;
ACL_CHECK(aclrtCreateEvent(&startEvent));
ACL_CHECK(aclrtCreateEvent(&endEvent));
ACL_CHECK(aclrtRecordEvent(startEvent, stream));
invOp(stream, aicCoreNum);
ACL_CHECK(aclrtRecordEvent(endEvent, stream));
ACL_CHECK(aclrtSynchronizeEvent(endEvent));
float kernelTimeMs = 0.0f;
ACL_CHECK(aclrtEventElapsedTime(&kernelTimeMs, startEvent, endEvent));
aclrtDestroyEvent(startEvent);
aclrtDestroyEvent(endEvent);
std::cout << "Kernel time: " << kernelTimeMs << " ms" << std::endl;
if (sizeWorkspace > 0) {
ACL_CHECK(aclrtFree(deviceWorkspace));
}
std::vector<float> hostResult(lenA);
ACL_CHECK(aclrtMemcpy(hostResult.data(), sizeA, deviceA, sizeA, ACL_MEMCPY_DEVICE_TO_HOST));
std::vector<float> hostGolden = hostOriginal;
int info = golden::ComputeInverseInplace(N, hostGolden);
if (info != 0) {
std::cerr << "Golden reference failed: matrix is singular (info=" << info << ")" << std::endl;
}
std::vector<uint64_t> errorIndices = golden::CompareData(hostResult, hostGolden, N * N);
if (errorIndices.empty()) {
std::cout << "Compare success." << std::endl;
} else {
std::cerr << "Compare failed. Error count: " << errorIndices.size() << " / " << lenA << std::endl;
}
ACL_CHECK(aclrtFree(deviceA));
ACL_CHECK(aclrtFree(deviceIpiv));
ACL_CHECK(aclrtDestroyStream(stream));
ACL_CHECK(aclrtResetDevice(deviceId));
ACL_CHECK(aclFinalize());
}
int main(int argc, const char** argv)
{
uint32_t N = 64;
int32_t deviceId = 0;
if (argc >= 2) {
N = static_cast<uint32_t>(std::atoi(argv[1]));
}
if (argc >= 3) {
deviceId = std::atoi(argv[2]);
}
std::cout << "Matrix Inverse: N=" << N << ", device=" << deviceId << std::endl;
Run(N, deviceId);
return 0;
}