* Copyright (c) 2025 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.
*/
#include <gtest/gtest.h>
#include "kernel_operator.h"
#include "include/adv_api/matmul/tiling.h"
#include "impl/adv_api/detail/matmul/utils/matmul_param.h"
#include "impl/adv_api/detail/matmul/policy/matmul_policy.h"
#include "impl/adv_api/detail/matmul/policy/matmul_private_modules.h"
#define private public
#define protected public
#include "impl/adv_api/detail/matmul/param/matmul_tensor_info.h"
#include "impl/adv_api/detail/matmul/param/matmul_shape_tiling.h"
using namespace std;
using namespace AscendC;
namespace {
template <class A_TYPE, class B_TYPE, class C_TYPE, class BIAS_TYPE, const MatmulConfig& MM_CFG, class MM_CB,
MATMUL_POLICY_DEFAULT_OF(MatmulPolicy)>
class MatmulImpl
: MATMUL_IMPORT_MODULE_PRIVATE(MatmulShapeInfo)
, MATMUL_IMPORT_MODULE_PRIVATE(MatmulTensorInfoB)
, MATMUL_IMPORT_MODULE_PRIVATE(MatmulShapeTiling)
, MATMUL_IMPORT_MODULE_PRIVATE(MLoop)
, MATMUL_IMPORT_MODULE_PRIVATE(KLoop)
, MATMUL_IMPORT_MODULE_PRIVATE(NLoop)
, MATMUL_IMPORT_MODULE_PRIVATE(MatmulUserDefineInfo)
{
MATMUL_ALLOW_USING_PRIVATE(MatmulShapeInfo);
MATMUL_ALLOW_USING_PRIVATE(MatmulTensorInfoB);
MATMUL_ALLOW_USING_PRIVATE(MatmulShapeTiling);
MATMUL_ALLOW_USING_PRIVATE(MLoop);
MATMUL_ALLOW_USING_PRIVATE(KLoop);
MATMUL_ALLOW_USING_PRIVATE(NLoop);
MATMUL_ALLOW_USING_PRIVATE(MatmulUserDefineInfo);
using SrcT = typename A_TYPE::T;
public:
using VAR_PARAMS =
typename Impl::Detail::MatmulParams<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG, GetMatmulMode(MM_CFG)>::PARAMS;
using IMPL = MatmulImpl<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, MM_CFG, MM_CB, MATMUL_POLICY>;
MATMUL_USE_MODULE(MLoop);
MATMUL_USE_MODULE(KLoop);
MATMUL_USE_MODULE(NLoop);
MATMUL_USE_MODULE(MatmulShapeTiling);
MATMUL_USE_MODULE(MatmulShapeInfo);
MATMUL_USE_MODULE(MatmulUserDefineInfo);
MatmulImpl() {
InitVar();
}
VAR_PARAMS& GetVar() {
return var;
}
void InitVar() {
tiling.depthA1 = 2;
tiling.depthB1 = 2;
tiling.iterateOrder = 0;
tiling.baseM = 16;
tiling.baseN = 16;
tiling.baseK = 16;
MATMUL_MODULE(MatmulShapeTiling)->SetTiling(&tiling);
var.tpipe_ = &pipe;
}
void SetSingleCoreParams(int32_t singleM, int32_t singleN, int32_t singleK, bool isIntraBlock = false)
{
if (isIntraBlock) {
intraBlockMatmul.singleCoreM = singleM;
intraBlockMatmul.singleCoreN = singleN;
intraBlockMatmul.singleCoreK = singleK;
intraBlockMatmul.M = singleM;
intraBlockMatmul.N = singleN;
intraBlockMatmul.Ka = singleK;
intraBlockMatmul.Kb = singleK;
} else {
MATMUL_MODULE(MatmulShapeInfo)->SetSingleShape(singleM, singleN, singleK);
MATMUL_MODULE(MatmulShapeInfo)->SetOrgShape(singleM, singleN, singleK, singleK, singleN);
}
MATMUL_MODULE(MLoop)->Init(MATMUL_MODULE(MatmulShapeInfo)->GetSingleCoreM());
MATMUL_MODULE(KLoop)->Init(MATMUL_MODULE(MatmulShapeInfo)->GetSingleCoreK());
MATMUL_MODULE(NLoop)->Init(MATMUL_MODULE(MatmulShapeInfo)->GetSingleCoreN());
}
void SetInitAParams(int32_t stepM, int32_t stepKa, int32_t baseM, int32_t baseK, int32_t depth,
bool isIntraBlock = false) {
tiling.stepM = stepM;
tiling.stepKa = stepKa;
tiling.baseM = baseM;
tiling.baseK = baseK;
tiling.depthA1 = depth;
var.baseUseK_ = baseK;
var.baseUseStepKa_ = baseK * stepKa;
if (isIntraBlock) {
intraBlockMatmul.baseUseM = baseM;
intraBlockMatmul.mIter = intraBlockMatmul.singleCoreM / baseM;
intraBlockMatmul.kIter = intraBlockMatmul.singleCoreK / baseK;
} else {
var.baseUseM_ = baseM;
}
}
void SetALayoutInfo(int32_t b, int32_t s, int32_t n, int32_t g, int32_t d, int32_t batch)
{
tiling.ALayoutInfoB = b;
tiling.ALayoutInfoS = s;
tiling.ALayoutInfoN = n;
tiling.ALayoutInfoG = g;
tiling.ALayoutInfoD = d;
batchA_ = batch;
}
void SetInitBParams(int32_t stepN, int32_t stepKb, int32_t baseN, int32_t baseK, int32_t depth, bool isIntraBlock = false) {
tiling.stepN = stepN;
tiling.stepKb = stepKb;
tiling.baseN = baseN;
tiling.baseK = baseK;
tiling.depthB1 = depth;
var.baseUseK_ = baseK;
var.baseUseStepKb_ = baseK * stepKb;
if (isIntraBlock) {
intraBlockMatmul.baseUseN = baseN;
intraBlockMatmul.nIter = intraBlockMatmul.singleCoreN / baseN;
intraBlockMatmul.kIter = intraBlockMatmul.singleCoreK / baseK;
} else {
var.baseUseN_ = baseN;
}
}
void SetBLayoutInfo(int32_t b, int32_t s, int32_t n, int32_t g, int32_t d, int32_t batch)
{
tiling.BLayoutInfoB = b;
tiling.BLayoutInfoS = s;
tiling.BLayoutInfoN = n;
tiling.BLayoutInfoG = g;
tiling.BLayoutInfoD = d;
batchB_ = batch;
}
void SetTranspose(bool isATrans, bool isBTrans, bool isIntraBlock = false)
{
if (isIntraBlock) {
intraBlockMatmul.isTransposeA = isATrans;
intraBlockMatmul.isTransposeB = isBTrans;
} else {
MATMUL_MODULE(MatmulShapeInfo)->SetTransposeA(isATrans);
MATMUL_MODULE(MatmulShapeInfo)->SetTransposeB(isBTrans);
}
}
void SetSelfDefineData(uint64_t dataPtr) {
MATMUL_MODULE(MatmulUserDefineInfo)->SetSelfDefineData(dataPtr);
}
void SetUserDefineInfo(uint64_t tilingPtr) {
MATMUL_MODULE(MatmulUserDefineInfo)->SetUserDefineInfo(tilingPtr);
}
uint32_t GetMIter() {
return MATMUL_MODULE(MLoop)->GetTotalIter();
}
uint32_t GetNIter() {
return MATMUL_MODULE(NLoop)->GetTotalIter();
}
uint32_t GetKIter() {
return MATMUL_MODULE(KLoop)->GetTotalIter();
}
private:
TCubeTiling tiling;
TPipe pipe;
VAR_PARAMS var;
int32_t batchA_;
int32_t batchB_;
struct IntraBlock {
__aicore__ inline IntraBlock(){};
__gm__ SrcT* aGlobal;
__gm__ SrcT* bGlobal;
int M;
int N;
int Ka;
int Kb;
int Kc;
int singleCoreM;
int singleCoreN;
int singleCoreK;
int mIter;
int nIter;
int kIter;
int baseUseM;
int baseUseN;
int blockUseM;
int blockUseN;
int tailM, tailK, tailN;
int cacheProcA = 0;
bool enableBias = false;
bool isTransposeA;
bool isTransposeB;
bool fakeMsg = false;
};
IntraBlock intraBlockMatmul;
};
}
class TestMatmulShapeInfo : public testing::Test {
protected:
void SetUp() {}
void TearDown() {}
private:
using A_TYPE = MatmulType<AscendC::TPosition::GM, CubeFormat::ND, half, false>;
using B_TYPE = MatmulType<AscendC::TPosition::GM, CubeFormat::ND, half, false>;
using C_TYPE = MatmulType<AscendC::TPosition::GM, CubeFormat::ND, float>;
using BIAS_TYPE = MatmulType<AscendC::TPosition::GM, CubeFormat::ND, float>;
MatmulImpl<A_TYPE, B_TYPE, C_TYPE, BIAS_TYPE, CFG_NORM, void> mm;
};
TEST_F(TestMatmulShapeInfo, test_get_singleCore_params)
{
mm.SetSingleCoreParams(32, 32, 16);
EXPECT_EQ(mm.GetSingleCoreM(), 32);
EXPECT_EQ(mm.GetSingleCoreN(), 32);
EXPECT_EQ(mm.GetKIter(), 1);
EXPECT_EQ(mm.GetMIter(), 2);
EXPECT_EQ(mm.GetNIter(), 2);
EXPECT_EQ(mm.GetOrgKa(), 16);
EXPECT_EQ(mm.GetOrgKa(), 16);
EXPECT_EQ(mm.GetOrgKc(), 32);
EXPECT_EQ(mm.GetOrgN(), 32);
EXPECT_EQ(mm.GetOrgM(), 32);
}
