Line data    Source code

       1              : /**
       2              :  * Copyright (c) 2025 Huawei Technologies Co., Ltd.
       3              :  * This program is free software, you can redistribute it and/or modify it under the terms and conditions of
       4              :  * CANN Open Software License Agreement Version 2.0 (the "License").
       5              :  * Please refer to the License for details. You may not use this file except in compliance with the License.
       6              :  * THIS SOFTWARE IS PROVIDED ON AN "AS IS" BASIS, WITHOUT WARRANTIES OF ANY KIND, EITHER EXPRESS OR IMPLIED,
       7              :  * INCLUDING BUT NOT LIMITED TO NON-INFRINGEMENT, MERCHANTABILITY, OR FITNESS FOR A PARTICULAR PURPOSE.
       8              :  * See LICENSE in the root of the software repository for the full text of the License.
       9              :  */
      10              : 
      11              : /*!
      12              :  * \file kernel_vec_binary_scalar_check.cpp
      13              :  * \brief
      14              :  */
      15              : 
      16              : #include "kernel_check_params.h"
      17              : #include "kernel_vec_binary_scalar_check.h"
      18              : 
      19              : namespace AscendC {
      20              : namespace check {
      21              : 
      22           32 : bool TikcppVecBinaryScalarCheck::CommonCheck()
      23              : {
      24           32 :     const std::string supportPos = "VECIN/VECOUT/VECCALC";
      25           96 :     ASCENDC_CHECK(CheckTensorScope(param_.dstLogicPos, static_cast<uint8_t>(HardWareIndex::UB), "dst", supportPos));
      26           72 :     ASCENDC_CHECK(CheckTensorScope(param_.src0LogicPos, static_cast<uint8_t>(HardWareIndex::UB), "src", supportPos));
      27           72 :     ASCENDC_CHECK(CheckTensorAddrAlign(param_.dstAddr, param_.dstPos, ONE_BLK_SIZE, "dst"));
      28           72 :     ASCENDC_CHECK(CheckTensorAddrAlign(param_.src0Addr, param_.src0Pos, ONE_BLK_SIZE, "src"));
      29           72 :     ASCENDC_CHECK(CheckBufferSizeOverFlow(
      30              :         param_.dstSize, GlobalParams::Instance().bufferSizeMap.at(param_.dstPos),
      31              :         "check dst tensor buffersize failed"));
      32           72 :     ASCENDC_CHECK(CheckBufferSizeOverFlow(
      33              :         param_.src0Size, GlobalParams::Instance().bufferSizeMap.at(param_.src0Pos),
      34              :         "check src tensor buffersize failed"));
      35           24 :     return true;
      36           32 : }
      37              : 
      38           16 : bool TikcppVecBinaryScalarCheck::CheckAllLowLevel(std::vector<uint64_t> maskArray)
      39              : {
      40           16 :     uint32_t maxByteLen = std::max(param_.dstDtypeBytes, param_.src0DtypeBytes);
      41           16 :     ASCENDC_CHECK(UpdateMaskArrayAndCheck(maskArray, maxByteLen));
      42           16 :     ASCENDC_CHECK(CommonCheck());
      43              : 
      44              :     // check tensor overflow
      45              :     TensorOverflowParams params = {
      46           12 :         param_.dstSize,
      47           12 :         param_.dstDtypeBytes,
      48           12 :         static_cast<uint64_t>(param_.repeatTimes),
      49           12 :         static_cast<uint64_t>(param_.dstBlockStride),
      50           12 :         static_cast<uint64_t>(param_.dstRepeatStride),
      51           12 :         false};
      52              :     // check dst src0 tensor overflow
      53           12 :     if (Int4Setter::Instance().GetDstInt4()) {
      54            0 :         Int4Setter::Instance().SetInt4();
      55              :     }
      56           36 :     ASCENDC_CHECK(CheckTensorOverflowLow(maskArray, params, "dstLocal"));
      57            8 :     params = {
      58            8 :         param_.src0Size,
      59            8 :         param_.src0DtypeBytes,
      60            8 :         static_cast<uint64_t>(param_.repeatTimes),
      61            8 :         static_cast<uint64_t>(param_.src0BlockStride),
      62            8 :         static_cast<uint64_t>(param_.src0RepeatStride),
      63            8 :         false};
      64            8 :     if (Int4Setter::Instance().GetSrcInt4()) {
      65            0 :         Int4Setter::Instance().SetInt4();
      66              :     }
      67              : #if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3510) || (__NPU_ARCH__ == 5102))
      68              :     if (param_.enableFlexibleScalar != 0) {
      69              :         if (param_.scalarPos == 1) {
      70              :             ASCENDC_CHECK(CheckTensorOverflowLow(maskArray, params, "src0"));
      71              :         } else {
      72              :             ASCENDC_CHECK(CheckTensorOverflowLow(maskArray, params, "src1"));
      73              :         }
      74              :     } else {
      75              :         ASCENDC_CHECK(CheckTensorOverflowLow(maskArray, params, "srcLocal"));
      76              :     }
      77              : #else
      78           24 :     ASCENDC_CHECK(CheckTensorOverflowLow(maskArray, params, "srcLocal"));
      79              : #endif
      80            8 :     return true;
      81              : }
      82              : 
      83           16 : bool TikcppVecBinaryScalarCheck::CheckAllHighLevel()
      84              : {
      85           16 :     ASCENDC_CHECK(CommonCheck());
      86           12 :     if (Int4Setter::Instance().GetDstInt4()) {
      87            0 :         Int4Setter::Instance().SetInt4();
      88              :     }
      89           36 :     ASCENDC_CHECK(CheckTensorOverflowHigh(param_.dstDtypeBytes, param_.dstSize, param_.calCount, "dstLocal"));
      90            8 :     if (Int4Setter::Instance().GetSrcInt4()) {
      91            0 :         Int4Setter::Instance().SetInt4();
      92              :     }
      93              : 
      94              : #if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3510) || (__NPU_ARCH__ == 5102))
      95              :     if (param_.enableFlexibleScalar != 0) {
      96              :         if (param_.scalarPos == 1) {
      97              :             ASCENDC_CHECK(CheckTensorOverflowHigh(param_.src0DtypeBytes, param_.src0Size, param_.calCount, "src0"));
      98              :         } else {
      99              :             ASCENDC_CHECK(CheckTensorOverflowHigh(param_.src0DtypeBytes, param_.src0Size, param_.calCount, "src1"));
     100              :         }
     101              :     } else {
     102              :         ASCENDC_CHECK(CheckTensorOverflowHigh(param_.src0DtypeBytes, param_.src0Size, param_.calCount, "srcLocal"));
     103              :     }
     104              : #else
     105           24 :     ASCENDC_CHECK(CheckTensorOverflowHigh(param_.src0DtypeBytes, param_.src0Size, param_.calCount, "srcLocal"));
     106              : #endif
     107            8 :     return true;
     108              : }
     109              : 
     110              : } // namespace check
     111              : } // namespace AscendC