* 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 "aclnn_bitwiseand.h"
#include "math/logical_and/op_api/logical_and.h"
#include "bitwiseand.h"
#include "aclnn_kernels/cast.h"
#include "aclnn_kernels/contiguous.h"
#include "opdev/common_types.h"
#include "opdev/data_type_utils.h"
#include "opdev/format_utils.h"
#include "opdev/op_dfx.h"
#include "opdev/op_executor.h"
#include "opdev/op_log.h"
#include "opdev/shape_utils.h"
#include "opdev/tensor_view_utils.h"
#include "aclnn_kernels/common/op_error_check.h"
constexpr int BITWISE_AND_MAX_TENSOR_DIM = 8;
using namespace op;
#ifdef __cplusplus
extern "C" {
#endif
* self.dtype == bool
* self other
* | |
* \ /
* Contiguous(workspace_0) Contiguous(workspace_2)
* \ /
* Cast(workspace_1) Cast(workspace_3)
* \ /
* LogicalAnd(workspace_4)
* |
* Cast(workspace_5)
* |
* ViewCopy
* |
* result
* self.dtype == INT
* self other
* | |
* \ /
* Contiguous(workspace_0) Contiguous(workspace_2)
* \ /
* Cast(workspace_1) Cast(workspace_3)
* \ /
* BitwiseAnd(workspace_4)
* |
* Cast(workspace_5)
* |
* ViewCopy
* |
* result
*/
static const std::initializer_list<op::DataType> DTYPE_SUPPORT_LIST = {
op::DataType::DT_INT8, op::DataType::DT_INT16, op::DataType::DT_INT32,
op::DataType::DT_INT64, op::DataType::DT_UINT8, op::DataType::DT_UINT16,
op::DataType::DT_UINT32, op::DataType::DT_UINT64, op::DataType::DT_BOOL};
static bool CheckNotNull(const aclTensor* self, const aclTensor* other, const aclTensor* out)
{
OP_CHECK_NULL(self, return false);
OP_CHECK_NULL(other, return false);
OP_CHECK_NULL(out, return false);
return true;
}
static bool CheckDtypeValid(const aclTensor* self, const aclTensor* other)
{
OP_CHECK_DTYPE_NOT_SUPPORT(self, DTYPE_SUPPORT_LIST, return false);
OP_CHECK_DTYPE_NOT_SUPPORT(other, DTYPE_SUPPORT_LIST, return false);
return true;
}
static bool CheckPromoteType(const aclTensor* self, const aclTensor* other, const aclTensor* y)
{
op::DataType promoteType = op::PromoteType(self->GetDataType(), other->GetDataType());
if (promoteType == DataType::DT_UNDEFINED) {
OP_LOGE(ACLNN_ERR_PARAM_INVALID, "Self dtype %s and other dtype %s can not promote dtype.",
op::ToString(self->GetDataType()).GetString(), op::ToString(other->GetDataType()).GetString());
return false;
}
OP_CHECK_RESULT_DTYPE_CAST_FAILED(promoteType, y->GetDataType(), return false);
return true;
}
static bool CheckFormat(const aclTensor* self, const aclTensor* other, const aclTensor* y)
{
if (self->GetStorageFormat() != other->GetStorageFormat() || self->GetStorageFormat() != y->GetStorageFormat()) {
OP_LOGE(ACLNN_ERR_PARAM_INVALID, "Format of input and output should be equal, self [%s], other [%s], out [%s].",
op::ToString(self->GetStorageFormat()).GetString(), op::ToString(other->GetStorageFormat()).GetString(),
op::ToString(y->GetStorageFormat()).GetString());
return false;
}
if (op::IsPrivateFormat(self->GetStorageFormat())) {
OP_LOGE(ACLNN_ERR_PARAM_INVALID, "Format only supports ND, NCHW, NHWC, HWCN, NDHWC, NCDHW.");
return false;
}
return true;
}
static bool CheckShape(const aclTensor* self, const aclTensor* other, const aclTensor* y)
{
OP_CHECK_MAX_DIM(self, BITWISE_AND_MAX_TENSOR_DIM, return false);
OP_CHECK_MAX_DIM(other, BITWISE_AND_MAX_TENSOR_DIM, return false);
op::Shape broadcastShape;
OP_CHECK_BROADCAST_AND_INFER_SHAPE(self, other, broadcastShape, return false);
OP_CHECK_SHAPE_NOT_EQUAL_WITH_EXPECTED_SIZE(y, broadcastShape, return false);
return true;
}
static aclnnStatus CheckParams(const aclTensor* self, const aclTensor* other, const aclTensor* y)
{
CHECK_RET(CheckNotNull(self, other, y), ACLNN_ERR_PARAM_NULLPTR);
CHECK_RET(CheckDtypeValid(self, other), ACLNN_ERR_PARAM_INVALID);
CHECK_RET(CheckPromoteType(self, other, y), ACLNN_ERR_PARAM_INVALID);
CHECK_RET(CheckFormat(self, other, y), ACLNN_ERR_PARAM_INVALID);
CHECK_RET(CheckShape(self, other, y), ACLNN_ERR_PARAM_INVALID);
return ACLNN_SUCCESS;
}
aclnnStatus aclnnBitwiseAndTensorOutGetWorkspaceSize(const aclTensor* self, const aclTensor* other, aclTensor* out,
uint64_t* workspaceSize, aclOpExecutor** executor)
{
L2_DFX_PHASE_1(aclnnBitwiseAndTensorOut, DFX_IN(self, other), DFX_OUT(out));
auto uniqueExecutor = CREATE_EXECUTOR();
CHECK_RET(uniqueExecutor.get() != nullptr, ACLNN_ERR_INNER_CREATE_EXECUTOR);
auto ret = CheckParams(self, other, out);
CHECK_RET(ret == ACLNN_SUCCESS, ret);
if (self->IsEmpty() || other->IsEmpty()) {
*workspaceSize = 0;
uniqueExecutor.ReleaseTo(executor);
return ACLNN_SUCCESS;
}
auto promoteType = op::PromoteType(self->GetDataType(), other->GetDataType());
auto selfContiguous = l0op::Contiguous(self, uniqueExecutor.get());
CHECK_RET(selfContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);
auto selfCasted = l0op::Cast(selfContiguous, promoteType, uniqueExecutor.get());
CHECK_RET(selfCasted != nullptr, ACLNN_ERR_INNER_NULLPTR);
auto otherContiguous = l0op::Contiguous(other, uniqueExecutor.get());
CHECK_RET(otherContiguous != nullptr, ACLNN_ERR_INNER_NULLPTR);
auto otherCasted = l0op::Cast(otherContiguous, promoteType, uniqueExecutor.get());
CHECK_RET(otherCasted != nullptr, ACLNN_ERR_INNER_NULLPTR);
const aclTensor* andOpOut = nullptr;
if (promoteType == op::DataType::DT_BOOL) {
andOpOut = l0op::LogicalAnd(selfCasted, otherCasted, uniqueExecutor.get());
} else {
andOpOut = l0op::BitwiseAnd(selfCasted, otherCasted, uniqueExecutor.get());
}
CHECK_RET(andOpOut != nullptr, ACLNN_ERR_INNER_NULLPTR);
auto castOut = l0op::Cast(andOpOut, out->GetDataType(), uniqueExecutor.get());
CHECK_RET(castOut != nullptr, ACLNN_ERR_INNER_NULLPTR);
auto viewCopyResult = l0op::ViewCopy(castOut, out, uniqueExecutor.get());
CHECK_RET(viewCopyResult != nullptr, ACLNN_ERR_INNER_NULLPTR);
*workspaceSize = uniqueExecutor->GetWorkspaceSize();
uniqueExecutor.ReleaseTo(executor);
return ACLNN_SUCCESS;
}
aclnnStatus aclnnInplaceBitwiseAndTensorOutGetWorkspaceSize(const aclTensor* selfRef, const aclTensor* other,
uint64_t* workspaceSize, aclOpExecutor** executor)
{
auto out = const_cast<aclTensor*>(selfRef);
return aclnnBitwiseAndTensorOutGetWorkspaceSize(selfRef, other, out, workspaceSize, executor);
}
aclnnStatus aclnnBitwiseAndTensorOut(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor,
aclrtStream stream)
{
L2_DFX_PHASE_2(aclnnBitwiseAndTensorOut);
return CommonOpExecutorRun(workspace, workspaceSize, executor, stream);
}
aclnnStatus aclnnInplaceBitwiseAndTensorOut(void* workspace, uint64_t workspaceSize, aclOpExecutor* executor,
aclrtStream stream)
{
L2_DFX_PHASE_2(aclnnInplaceBitwiseAndTensorOut);
return CommonOpExecutorRun(workspace, workspaceSize, executor, stream);
}
#ifdef __cplusplus
}
#endif