* 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.
*/
* \file kernel_base_check.h
* \brief
*/
#ifndef ASCENDC_BASE_CHECK_H
#define ASCENDC_BASE_CHECK_H
#include <vector>
#include <string>
#include "kernel_check_util.h"
namespace AscendC {
namespace check {
const uint32_t MASK_ARRAY_LEN = 2;
const uint32_t NAME_MAX_LEN = 128;
enum class ModeType : uint8_t { NONE_MODE = 0, NORM_MODE = 1, COUNTER_MODE = 2 };
struct TensorOverflowParams {
TensorOverflowParams() = default;
TensorOverflowParams(
const uint64_t bufferSizeIn, const uint64_t dtypeSizeIn, const uint64_t repeatTimesIn,
const uint64_t blkStrideIn, const uint64_t repStrideIn, const bool isCounterIn)
: bufferSize(bufferSizeIn),
dtypeSize(dtypeSizeIn),
repeatTimes(repeatTimesIn),
blkStride(blkStrideIn),
repStride(repStrideIn),
isCounter(isCounterIn)
{}
uint64_t bufferSize = 0;
uint64_t dtypeSize = 1;
uint64_t repeatTimes = 1;
uint64_t blkStride = DEFAULT_BLK_STRIDE;
uint64_t repStride = DEFAULT_REPEAT_STRIDE;
bool isCounter = false;
};
* @funcname: GetMaskLength
* @brief: the maximum number of elements calculated this repeat.
* For example mask is 0x30fff + 0xffffffff. The maximum num per repeat is 64 + 18 = 82
* @return: uint64_t
*/
uint64_t GetMaskLength(std::vector<uint64_t>& maskArray, const uint32_t dtypeSize);
* @funcname: CheckTensorSizeOverflow
* @brief: common method used to compare tensor size and minimum needed size, and report error if overflow
* mode 0: none, mode 1: norm mode, mode 2: counter mode
* @return: true/false
*/
bool CheckTensorSizeOverflow(
uint64_t expectedSize, uint64_t tensorSize, const std::string& tensorName, const std::string& apiName,
const ModeType mode = ModeType::NONE_MODE);
void CounterSplitMainTail(
std::vector<uint64_t>& maskArray, const uint32_t dtypeBytes, uint64_t& mainRepeatTimes, uint64_t& tailRepeatTimes,
std::vector<uint64_t>& mainMaskArray, std::vector<uint64_t>& tailMaskArray);
uint64_t CalculateVectorMaxOffset(
const uint64_t repeatTimes, const uint64_t blkStride, const uint64_t repStride, const uint64_t maskLen,
const uint64_t blockLen);
class TikcppBaseCheck {
public:
explicit TikcppBaseCheck(const std::string& name) : apiName(name) {}
virtual ~TikcppBaseCheck() {}
* @funcname: CheckTensorScope
* @brief: check whether the tensor hardware position equal to the expect position.
* @params: tensorPos, tensor physical position
* expectedPos, expected tensor physical position
* tensorInfo, src0 / src1/ dst for print information
* posInfo, expected position info for print information
* @return: true/false
*/
bool CheckTensorScope(
const uint8_t logicPos, const uint8_t expectedPos, const std::string& tensorInfo,
const std::string& posInfo) const;
* @funcname: CheckBufferSizeOverFlow
* @brief: check whether the tensor allocate size equal to the buffer limited size.
* @return: true/false
*/
bool CheckBufferSizeOverFlow(const uint64_t localSize, const uint64_t bufferSize, const std::string& errMsg) const;
* @funcname: CheckMaskArray
* @brief: check the mask in bits mode
* @return: true/false
*/
bool CheckMaskArray(std::vector<uint64_t> maskArray) const;
* @funcname: CheckMaskImm
* @brief: check the mask in continuous mode
* @return: true/false
*/
bool CheckMaskImm(const uint64_t mask) const;
* @funcname: CheckTensorOverflowLow
* @brief: check whether tensor used size over the allocated size in low api level
* @return: true/false
*/
bool CheckTensorOverflowLow(
std::vector<uint64_t>& maskArray, const TensorOverflowParams& params, const std::string& tensorName) const;
* @funcname: CheckTensorOverflowLowGatherMask
* @brief: check whether tensor used size over the allocated size in low api level for GatherMask
* @return: true/false
*/
bool CheckTensorOverflowLowGathermask(
std::vector<uint64_t>& maskArray, const TensorOverflowParams& params, const std::string& tensorName) const;
* @funcname: CheckTensorOverflowLowBrcb
* @brief: check whether tensor used size over the allocated size in low api level for Brcb
* Note that there is no counter mode for brcb + mask is not used
* @return: true/false
*/
bool CheckTensorOverflowLowBrcb(const TensorOverflowParams& params, const std::string& tensorName) const;
* @funcname: CheckTensorOverflowHigh
* @brief: check whether tensor used size over the allocated size in high api level
* @return: true/false
*/
bool CheckTensorOverflowHigh(
const uint32_t dtypeSize, const uint64_t bufferSize, const uint32_t calCount,
const std::string& tensorName) const;
* @funcname: UpdateMaskArrayAndCheck
* @brief: If isSetMask = false, replace maskArray with maskHigh and maskLow value in registers.
* Check the latest maskArray value is valid with given dtype
* @params: maskArray, mask value given by user. Can be len 1 or len 2
* maxByteLen, among all dtypes given by function, the largest value of sizeof(dtype)
* @return: true/false
*/
bool UpdateMaskArrayAndCheck(std::vector<uint64_t>& maskArray, const uint32_t maxByteLen) const;
* @funcname: CheckTensorAddrAlign
* @brief: Check tensor start address is aligned with alignBytes
* @params: tensorAddr, tensor address
* phyPos, tensor physical position, used to calculate real offset of tensor
* alignBytes, 32B aligned / 512B aligned etc
* tensorInfo, src0 / src1/ dst for print information
* @return: true/false
*/
bool CheckTensorAddrAlign(
const uint64_t tensorAddr, const uint8_t phyPos, const uint64_t alignBytes,
const std::string& tensorInfo) const;
protected:
std::string apiName = "";
};
}
}
#endif
