* 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.
*/
#include <iostream>
#include <vector>
#include <cmath>
#include <cstring>
#include "acl/acl.h"
#include "aclnnop/aclnn_spherical_harmonics_bwd.h"
#include "test_utils.h"
using namespace std;
namespace {
struct TensorResources {
void* dirsDeviceAddr = nullptr;
void* coeffsDeviceAddr = nullptr;
void* vColorsDeviceAddr = nullptr;
void* vDirsDeviceAddr = nullptr;
void* vCoeffsDeviceAddr = nullptr;
aclTensor* dirsTensor = nullptr;
aclTensor* coeffsTensor = nullptr;
aclTensor* vColorsTensor = nullptr;
aclTensor* vDirsTensor = nullptr;
aclTensor* vCoeffsTensor = nullptr;
};
int InitializeTensors(TensorResources& resources) {
int64_t batchSize = 1;
int64_t gaussNum = 1024;
int64_t degree = 3;
int64_t K = (degree + 1) * (degree + 1);
std::vector<int64_t> dirsShape = {batchSize, 3, gaussNum};
std::vector<int64_t> coeffsShape = {batchSize, K, 3, gaussNum};
std::vector<int64_t> vColorsShape = {batchSize, 3, gaussNum};
std::vector<int64_t> vDirsShape = {batchSize, 3, gaussNum};
std::vector<int64_t> vCoeffsShape = {batchSize, K, 3, gaussNum};
std::vector<float> dirsHostData(GetShapeSize(dirsShape), 0.5f);
std::vector<float> coeffsHostData(GetShapeSize(coeffsShape), 0.1f);
std::vector<float> vColorsHostData(GetShapeSize(vColorsShape), 1.0f);
std::vector<float> vDirsHostData(GetShapeSize(vDirsShape), 0.0f);
std::vector<float> vCoeffsHostData(GetShapeSize(vCoeffsShape), 0.0f);
int ret = CreateAclTensor(dirsHostData, dirsShape, &resources.dirsDeviceAddr,
aclDataType::ACL_FLOAT, &resources.dirsTensor);
if (!CHECK_RET(ret == ACL_SUCCESS)) return ret;
ret = CreateAclTensor(coeffsHostData, coeffsShape, &resources.coeffsDeviceAddr,
aclDataType::ACL_FLOAT, &resources.coeffsTensor);
if (!CHECK_RET(ret == ACL_SUCCESS)) return ret;
ret = CreateAclTensor(vColorsHostData, vColorsShape, &resources.vColorsDeviceAddr,
aclDataType::ACL_FLOAT, &resources.vColorsTensor);
if (!CHECK_RET(ret == ACL_SUCCESS)) return ret;
ret = CreateAclTensor(vDirsHostData, vDirsShape, &resources.vDirsDeviceAddr,
aclDataType::ACL_FLOAT, &resources.vDirsTensor);
if (!CHECK_RET(ret == ACL_SUCCESS)) return ret;
ret = CreateAclTensor(vCoeffsHostData, vCoeffsShape, &resources.vCoeffsDeviceAddr,
aclDataType::ACL_FLOAT, &resources.vCoeffsTensor);
if (!CHECK_RET(ret == ACL_SUCCESS)) return ret;
return ACL_SUCCESS;
}
int ExecuteSphericalHarmonicsBwd(TensorResources& resources, aclrtStream stream,
void** workspaceAddr, uint64_t* workspaceSize) {
int64_t degree = 3;
aclOpExecutor* executor;
int ret = aclnnSphericalHarmonicsBwdGetWorkspaceSize(
resources.dirsTensor, resources.coeffsTensor, resources.vColorsTensor, degree,
resources.vDirsTensor, resources.vCoeffsTensor, workspaceSize, &executor);
if (!CHECK_RET(ret == ACL_SUCCESS)) {
LOG_PRINT("aclnnSphericalHarmonicsBwdGetWorkspaceSize failed. ERROR: %d\n", ret);
return ret;
}
if (*workspaceSize > 0ULL) {
ret = aclrtMalloc(workspaceAddr, *workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST);
if (!CHECK_RET(ret == ACL_SUCCESS)) return ret;
}
ret = aclnnSphericalHarmonicsBwd(*workspaceAddr, *workspaceSize, executor, stream);
if (!CHECK_RET(ret == ACL_SUCCESS)) {
LOG_PRINT("aclnnSphericalHarmonicsBwd failed. ERROR: %d\n", ret);
return ret;
}
return ACL_SUCCESS;
}
void CleanupResources(TensorResources& resources, void* workspaceAddr,
aclrtStream stream, int32_t deviceId) {
if (resources.dirsTensor) aclDestroyTensor(resources.dirsTensor);
if (resources.coeffsTensor) aclDestroyTensor(resources.coeffsTensor);
if (resources.vColorsTensor) aclDestroyTensor(resources.vColorsTensor);
if (resources.vDirsTensor) aclDestroyTensor(resources.vDirsTensor);
if (resources.vCoeffsTensor) aclDestroyTensor(resources.vCoeffsTensor);
if (resources.dirsDeviceAddr) aclrtFree(resources.dirsDeviceAddr);
if (resources.coeffsDeviceAddr) aclrtFree(resources.coeffsDeviceAddr);
if (resources.vColorsDeviceAddr) aclrtFree(resources.vColorsDeviceAddr);
if (resources.vDirsDeviceAddr) aclrtFree(resources.vDirsDeviceAddr);
if (resources.vCoeffsDeviceAddr) aclrtFree(resources.vCoeffsDeviceAddr);
if (workspaceAddr) aclrtFree(workspaceAddr);
if (stream) aclrtDestroyStream(stream);
aclrtResetDevice(deviceId);
aclFinalize();
}
}
int main() {
int32_t deviceId = 0;
aclrtStream stream = nullptr;
TensorResources resources = {};
void* workspaceAddr = nullptr;
uint64_t workspaceSize = 0;
int ret = ACL_SUCCESS;
ret = Init(deviceId, &stream);
if (!CHECK_RET(ret == ACL_SUCCESS)) return ret;
ret = InitializeTensors(resources);
if (!CHECK_RET(ret == ACL_SUCCESS)) {
CleanupResources(resources, workspaceAddr, stream, deviceId);
return ret;
}
ret = ExecuteSphericalHarmonicsBwd(resources, stream, &workspaceAddr, &workspaceSize);
if (!CHECK_RET(ret == ACL_SUCCESS)) {
CleanupResources(resources, workspaceAddr, stream, deviceId);
return ret;
}
ret = aclrtSynchronizeStream(stream);
if (!CHECK_RET(ret == ACL_SUCCESS)) {
LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret);
}
CleanupResources(resources, workspaceAddr, stream, deviceId);
return 0;
}