* 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 "unfold_grad.h"
#include "opdev/data_type_utils.h"
#include "opdev/make_op_executor.h"
#include "opdev/op_log.h"
#include "opdev/op_dfx.h"
#include "opdev/shape_utils.h"
#include "opdev/make_op_executor.h"
#include "aclnn_kernels/common/op_error_check.h"
#include "opdev/aicpu/aicpu_task.h"
using namespace op;
namespace l0op {
OP_TYPE_REGISTER(UnfoldGrad);
static inline bool IsAiCoreSupport(
const aclTensor* gradOut, int64_t dim, int64_t size, int64_t step)
{
int64_t dimNum = gradOut->GetViewShape().GetDimNum()-1;
int64_t maxv = size >= step ? size : step;
op::DataType gradOutDtype = gradOut->GetDataType();
if ((dim == dimNum - 1) &&
(((gradOutDtype == DataType::DT_FLOAT) && (maxv <= 49088)) ||
((gradOutDtype == DataType::DT_FLOAT16 || gradOutDtype == DataType::DT_BF16) && (maxv <= 32720)))) {
return true;
} else if ((dim == dimNum - 2) &&
(((gradOutDtype == DataType::DT_FLOAT) && (maxv <= 88)) ||
((gradOutDtype == DataType::DT_FLOAT16 || gradOutDtype == DataType::DT_BF16) && (maxv <= 72)))) {
return true;
} else {
return false;
}
}
static const aclTensor* UnfoldGradAiCpu(
const aclTensor* gradOut, const aclTensor* inputSizes, int64_t dim, int64_t size, int64_t step, const aclTensor* out,
aclOpExecutor* executor)
{
L0_DFX(UnfoldGradAiCpu, gradOut, inputSizes, dim, size, step, out);
static internal::AicpuTaskSpace space("UnfoldGrad");
ADD_TO_LAUNCHER_LIST_AICPU(
UnfoldGrad, OP_ATTR_NAMES({"dim", "size", "step"}), OP_INPUT(gradOut, inputSizes), OP_OUTPUT(out),
OP_ATTR(dim, size, step));
return out;
}
static const aclTensor* UnfoldGradAiCore(
const aclTensor* gradOut, const aclTensor* inputSizes, int64_t dim, int64_t size, int64_t step, const aclTensor* out,
aclOpExecutor* executor)
{
L0_DFX(UnfoldGradAiCore, gradOut, inputSizes, dim, size, step, out);
ADD_TO_LAUNCHER_LIST_AICORE(UnfoldGrad, OP_INPUT(gradOut, inputSizes), OP_OUTPUT(out), OP_ATTR(dim, size, step));
return out;
}
const aclTensor* UnfoldGrad(
const aclTensor* gradOut, const aclTensor* inputSizes, int64_t dim, int64_t size, int64_t step,
aclOpExecutor* executor)
{
L0_DFX(UnfoldGrad, gradOut, inputSizes, dim, size, step);
auto out = executor->AllocTensor(gradOut->GetDataType(), gradOut->GetStorageFormat(), gradOut->GetOriginalFormat());
auto ret = INFER_SHAPE(UnfoldGrad, OP_INPUT(gradOut, inputSizes), OP_OUTPUT(out), OP_ATTR(dim, size, step));
if (ret != ACLNN_SUCCESS) {
OP_LOGE(ACLNN_ERR_PARAM_INVALID, "infershape failed.");
return nullptr;
}
if (IsAiCoreSupport(gradOut, dim, size, step)) {
return UnfoldGradAiCore(gradOut, inputSizes, dim, size, step, out, executor);
} else {
return UnfoldGradAiCpu(gradOut, inputSizes, dim, size, step, out, executor);
}
}
}