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_utils.h
      13              :  * \brief
      14              :  */
      15              : #ifndef ASCENDC_MODULE_UTILS_H
      16              : #define ASCENDC_MODULE_UTILS_H
      17              : #include "utils/kernel_utils_ceil_oom_que.h"
      18              : #include "utils/kernel_utils_constants.h"
      19              : #include "utils/kernel_utils_struct_norm_sort.h"
      20              : #include "utils/kernel_utils_mode_cpu.h"
      21              : #include "utils/kernel_utils_ceil_oom_que.h"
      22              : #include "utils/kernel_utils_constants.h"
      23              : #include "utils/kernel_utils_struct_norm_sort.h"
      24              : 
      25              : namespace AscendC {
      26              : #if defined(__NPU_ARCH__) && \
      27              :     ((__NPU_ARCH__ == 2201) || (__NPU_ARCH__ == 3002) || (__NPU_ARCH__ == 3510) || (__NPU_ARCH__ == 5102))
      28              : 
      29              : #if defined(__NPU_ARCH__) && ((__NPU_ARCH__ == 3510) || (__NPU_ARCH__ == 5102))
      30              : 
      31              : namespace FPTranslation {
      32              : // HiFloat8 -> Fp32
      33              : #define HIF8_SIGN_INDEX (7)
      34              : #define HIF8_BIT6_INDEX (6)
      35              : #define HIF8_BIT5_INDEX (5)
      36              : #define HIF8_BIT4_INDEX (4)
      37              : #define HIF8_BIT3_INDEX (3)
      38              : #define HIF8_EXTRACT_SIGN(x) (((x) >> HIF8_SIGN_INDEX) & 0x1)
      39              : #define HIF8_EXTRACT_BIT6(x) (((x) >> HIF8_BIT6_INDEX) & 0x1)
      40              : #define HIF8_EXTRACT_BIT5(x) (((x) >> HIF8_BIT5_INDEX) & 0x1)
      41              : #define HIF8_EXTRACT_BIT4(x) (((x) >> HIF8_BIT4_INDEX) & 0x1)
      42              : #define HIF8_EXTRACT_BIT3(x) (((x) >> HIF8_BIT3_INDEX) & 0x1)
      43              : 
      44              : constexpr int8_t HIF8_NAN = 0x80;
      45              : constexpr int8_t HIF8_POS_INF = 0x6F;
      46              : constexpr int8_t HIF8_NEG_INF = 0xEF;
      47              : constexpr int8_t HIF8_BIT_LEN = 8;
      48              : constexpr uint32_t FP32_NAN = 0x7FFFFFFF;
      49              : constexpr uint32_t FP32_EXP_BIAS = 127;
      50              : 
      51              : #define FP32_MAX_MAN (0x7FFFFF)
      52              : #define FP32_POS_INF (0x7F800000)
      53              : #define FP32_NEG_INF (0xff800000)
      54              : constexpr uint32_t FP32_SIGN_INDEX = 31;
      55              : constexpr uint32_t FP32_MAN_LEN = 23;
      56              : 
      57              : // FP8 (E5M2) -> Fp32
      58              : #define FP8_SIGN_INDEX (7)
      59              : #define FP8_T_NAN (0x7F)
      60              : #define FP8_MAX_MAN (0x7)
      61              : 
      62              : constexpr int16_t FP8E5M2_EXP_MASK = 0x7C;
      63              : constexpr int16_t FP8E5M2_MAN_MASK = 0x3;
      64              : constexpr uint32_t FP8E5M2_MAN_LEN = 2;
      65              : constexpr uint32_t FP8E5M2_EXP_BIAS = 15;
      66              : 
      67              : #define FP8E5M2_MAN_HIDE_BIT (0x4)
      68              : #define FP8E5M2_MAX_EXP (0x1F)
      69              : #define FP8E5M2_MAX_MAN (0x3)
      70              : #define FP8E5M2_INF (0X7C)
      71              : 
      72              : // FP8 (E4M3) -> Fp32
      73              : #define FP8_SIGN_INDEX (7)
      74              : #define FP8_T_NAN (0x7F)
      75              : constexpr uint32_t FP8E4M3_EXP_BIAS = 7;
      76              : constexpr uint32_t FP8E4M3_MAN_LEN = 3;
      77              : #define FP8_MAX_EXP (0xF)
      78              : #define FP8_MAX_MAN (0x7)
      79              : #define FP8_MAN_HIDE_BIT (0x8)
      80              : 
      81              : } // namespace FPTranslation
      82              : #endif
      83              : #endif
      84              : } // namespace AscendC
      85              : 
      86              : namespace AscendC {
      87              : struct SliceInfo {
      88          832 :     __aicore__ SliceInfo() {}
      89              : 
      90          208 :     __aicore__ SliceInfo(
      91              :         const uint32_t startIndexIn, const uint32_t endIndexIn, const uint32_t strideIn, const uint32_t burstLenIn,
      92              :         const uint32_t shapeValueIn = 0)
      93          208 :         : startIndex(startIndexIn),
      94          208 :           endIndex(endIndexIn),
      95          208 :           stride(strideIn),
      96          208 :           burstLen(burstLenIn),
      97          208 :           shapeValue(shapeValueIn)
      98          208 :     {}
      99              : 
     100              :     uint32_t startIndex = 0;
     101              :     uint32_t endIndex = ONE_BLK_SIZE - 1;
     102              :     uint32_t stride = 0;
     103              :     uint32_t burstLen = ONE_BLK_SIZE;
     104              :     uint32_t shapeValue = 0;
     105              : };
     106              : 
     107              : class AscendCUtils {
     108              : public:
     109              :     __aicore__ static inline int32_t GetC0Size() { return DEFAULT_C0_SIZE; }
     110              : 
     111              :     __aicore__ static inline int32_t GetC0Count(const int32_t dtypeSize)
     112              :     {
     113              :         ASCENDC_ASSERT((dtypeSize != 0), { KERNEL_LOG(KERNEL_ERROR, "dtypeSize can not be 0"); });
     114              :         return GetC0Size() / dtypeSize;
     115              :     }
     116              : };
     117              : 
     118              : #ifdef ASCENDC_CPU_DEBUG
     119              : enum AtomicType { SUM, MAX, MIN };
     120              : extern bool g_isAtomic;
     121              : extern AtomicType g_atomicType;
     122              : 
     123              : #endif // ASCENDC_CPU_DEBUG
     124              : 
     125              : } // namespace AscendC
     126              : #endif // ASCENDC_MODULE_UTILS_H