* Copyright (c) 2026 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.
*/
#pragma once
#include <algorithm>
#include <cmath>
#include <cstdint>
#include <cstring>
#include "acl/acl.h"
#include "cann_ops_blas.h"
#include "cblas_compat.h"
#include "scalex_param.h"
inline uint16_t floatToFp16(float v)
{
uint32_t x = *reinterpret_cast<const uint32_t*>(&v);
uint32_t sign = (x >> 16) & 0x8000u;
int32_t exp = static_cast<int32_t>((x >> 23) & 0xFFu) - 127 + 15;
uint32_t mant = x & 0x7FFFFFu;
if (((x >> 23) & 0xFFu) == 0xFFu)
return static_cast<uint16_t>(sign | 0x7C00u | (mant ? 0x200u : 0));
if (exp >= 0x1F)
return static_cast<uint16_t>(sign | 0x7C00u);
if (exp <= 0) {
if (exp < -10) return static_cast<uint16_t>(sign);
mant |= 0x800000u;
int shift = 14 - exp;
uint32_t halfMant = mant >> shift;
uint32_t rem = mant & ((1u << shift) - 1u);
uint32_t halfWay = 1u << (shift - 1u);
if (rem > halfWay || (rem == halfWay && (halfMant & 1u))) halfMant++;
return static_cast<uint16_t>(sign | halfMant);
}
uint32_t halfMant = mant >> 13;
uint32_t rem = mant & 0x1FFFu;
if (rem > 0x1000u || (rem == 0x1000u && (halfMant & 1u))) {
halfMant++;
if (halfMant == 0x400u) { halfMant = 0; exp++; }
}
if (exp >= 0x1F) return static_cast<uint16_t>(sign | 0x7C00u);
return static_cast<uint16_t>(sign | (static_cast<uint32_t>(exp) << 10) | halfMant);
}
inline float fp16ToFloat(uint16_t h)
{
uint32_t hsign = (static_cast<uint32_t>(h) & 0x8000u) << 16;
uint32_t hexp = (h >> 10) & 0x1Fu;
uint32_t hmant = h & 0x3FFu;
uint32_t f;
if (hexp == 0) {
if (hmant == 0) {
f = hsign;
} else {
int e = -1;
do { hmant <<= 1; e++; } while ((hmant & 0x400u) == 0);
hmant &= 0x3FFu;
f = hsign | (static_cast<uint32_t>(127 - 15 - e) << 23) | (hmant << 13);
}
} else if (hexp == 0x1Fu) {
f = hsign | 0x7F800000u | (hmant << 13);
} else {
f = hsign | (static_cast<uint32_t>(hexp - 15 + 127) << 23) | (hmant << 13);
}
float out = *reinterpret_cast<float*>(&f);
return out;
}
inline uint16_t floatToBf16(float v)
{
uint32_t bits = *reinterpret_cast<const uint32_t*>(&v);
uint32_t lsb = (bits >> 16) & 1u;
bits += 0x7FFFu + lsb;
return static_cast<uint16_t>(bits >> 16);
}
inline float bf16ToFloat(uint16_t h)
{
uint32_t bits = static_cast<uint32_t>(h) << 16;
float out = *reinterpret_cast<float*>(&bits);
return out;
}
inline float castToDtype(float v, int32_t dtype)
{
if (dtype == static_cast<int32_t>(ACL_FLOAT16)) return fp16ToFloat(floatToFp16(v));
else if (dtype == static_cast<int32_t>(ACL_BF16)) return bf16ToFloat(floatToBf16(v));
else return v;
}
inline aclblasStatus_t aclblasScalex_cpu(
aclblasHandle_t handle,
int n,
const void* alpha,
aclDataType alphaType,
void* x,
aclDataType xType,
int incx,
aclDataType executionType)
{
if (handle == nullptr) return ACLBLAS_STATUS_HANDLE_IS_NULLPTR;
if (n < 0) return ACLBLAS_STATUS_INVALID_VALUE;
if (n == 0) return ACLBLAS_STATUS_SUCCESS;
if (alpha == nullptr) return ACLBLAS_STATUS_INVALID_VALUE;
if (x == nullptr) return ACLBLAS_STATUS_INVALID_VALUE;
if (incx == 0) return ACLBLAS_STATUS_INVALID_VALUE;
if (alphaType != ACL_FLOAT) return ACLBLAS_STATUS_NOT_SUPPORTED;
if (xType != ACL_FLOAT && xType != ACL_FLOAT16 && xType != ACL_BF16)
return ACLBLAS_STATUS_NOT_SUPPORTED;
if (executionType != ACL_FLOAT) return ACLBLAS_STATUS_NOT_SUPPORTED;
float alphaVal = *static_cast<const float*>(alpha);
float* xFloat = static_cast<float*>(x);
if (xType == ACL_FLOAT) {
cblas_sscal(n, alphaVal, xFloat, incx);
} else {
int absIncx = std::abs(incx);
for (int i = 0; i < n; i++) {
int idx = (incx > 0) ? (i * incx) : ((n - 1 - i) * absIncx);
float val = xFloat[idx];
float result = alphaVal * val;
xFloat[idx] = castToDtype(result, static_cast<int32_t>(xType));
}
}
return ACLBLAS_STATUS_SUCCESS;
}