文档HOT
EN
注册
我要评分
文档获取效率
文档正确性
内容完整性
文档易理解
提交
在线提单
论坛求助

Convert

此类函数转换源向量中的每个元素的数据类型，结果保存在目的向量中。

带_S后缀的函数根据scale值对结果值进行缩放。如果转换后的结果超出输出数据范围，则将达到饱和。

对float16_t型数据进行转换时函数不支持HMPP_RND_FINANCIAL舍入模式。

函数接口声明如下：

  • 整型转整型的操作：

    HmppResult HMPPS_Convert_24u32u(const uint8_t *src, uint32_t *dst, int32_t len);

    HmppResult HMPPS_Convert_8s8u(const int8_t *src, uint8_t *dst, int32_t len);

    HmppResult HMPPS_Convert_8s16s(const int8_t *src, int16_t *dst, int32_t len);

    HmppResult HMPPS_Convert_16s32s(const int16_t *src, int32_t *dst, int32_t len);

    HmppResult HMPPS_Convert_24s32s(const uint8_t *src, int32_t *dst, int32_t len);

    HmppResult HMPPS_Convert_32s16s(const int32_t *src, int16_t *dst, int32_t len);

  • 整型转浮点的操作：

    HmppResult HMPPS_Convert_24u32f(const uint8_t *src, float *dst, int32_t len);

    HmppResult HMPPS_Convert_16s16f(const int16_t *src, float16_t *dst, int32_t len, HmppRoundMode roundMode);

    HmppResult HMPPS_Convert_8u32f(const uint8_t *src, float *dst, int32_t len);

    HmppResult HMPPS_Convert_8s32f(const int8_t *src, float *dst, int32_t len);

    HmppResult HMPPS_Convert_16u32f(const uint16_t *src, float *dst, int32_t len);

    HmppResult HMPPS_Convert_16s32f(const int16_t *src, float *dst, int32_t len);

    HmppResult HMPPS_Convert_24s32f(const uint8_t *src, float *dst, int32_t len);

    HmppResult HMPPS_Convert_32s32f(const int32_t *src, float *dst, int32_t len);

    HmppResult HMPPS_Convert_32s64f(const int32_t *src, double *dst, int32_t len);

    HmppResult HMPPS_Convert_64s64f(const int64_t *src, double *dst, int32_t len);

  • 浮点转浮点的操作：

    HmppResult HMPPS_Convert_32f16f(const float *src, float16_t *dst, int32_t len, HmppRoundMode roundMode);

    HmppResult HMPPS_Convert_32f64f(const float *src, double *dst, int32_t len);

    HmppResult HMPPS_Convert_16f32f(const float16_t *src, float *dst, int32_t len);

    HmppResult HMPPS_Convert_64f32f(const double *src, float *dst, int32_t len);

  • 有缩放的整型转整型操作：

    HmppResult HMPPS_Convert_8u8s_S(const uint8_t *src, int8_t *dst, int32_t len, HmppRoundMode rndMode, double scale)

    HmppResult HMPPS_Convert_16s8s_S(const int16_t *src, int8_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_32u24u_S(const uint32_t *src, uint8_t *dst, int32_t len, double scale);

    HmppResult HMPPS_Convert_32s16s_S(const int32_t *src, int16_t *dst, int32_t len, double scale);

    HmppResult HMPPS_Convert_32s24s_S(const int32_t *src, uint8_t *dst, int32_t len, double scale);

    HmppResult HMPPS_Convert_64s32s_S(const int64_t *src, int32_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

  • 有缩放的整型转浮点操作：

    HmppResult HMPPS_Convert_16s32f_S(const int16_t *src, float *dst, int32_t len, double scale);

    HmppResult HMPPS_Convert_16s64f_S(const int16_t *src, double *dst, int32_t len, double scale);

    HmppResult HMPPS_Convert_32s32f_S(const int32_t *src, float *dst, int32_t len, double scale);

    HmppResult HMPPS_Convert_32s64f_S(const int32_t *src, double *dst, int32_t len, double scale);

  • 有缩放的浮点转整型操作：

    HmppResult HMPPS_Convert_32f8u_S(const float *src, uint8_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_32f8s_S(const float *src, int8_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_32f16u_S(const float *src, uint16_t *dst, int32_t len,HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_32f16s_S(const float *src, int16_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_32f32s_S(const float *src, int32_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_64f8u_S(const double *src, uint8_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_64f8s_S(const double *src, int8_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_64f16u_S(const double *src, uint16_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_64f16s_S(const double *src, int16_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_64f32s_S(const double *src, int32_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_64f64s_S(const double *src, int64_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_16f16s_S(const float16_t *src, int16_t *dst, int32_t len, HmppRoundMode rndMode, double scale);

    HmppResult HMPPS_Convert_32f24u_S(const float *src, uint8_t *dst, int32_t len, double scale);

    HmppResult HMPPS_Convert_32f24s_S(const float *src, uint8_t *dst, int32_t len, double scale);

参数

参数名

描述

取值范围

输入/输出

src

指向源向量的指针。

非空

输入

dst

指向目标向量的指针。

非空

输出

len

源向量长度。

(0, INT_MAX]或[3, INT_MAX]

输入

scale

缩放因数。

(0,INF)且输入为2^n

输入

rndMode

舍入模式。

HmppRndZero、HmppRndNear、HmppRndFinancial

输入

返回值

  • 成功：返回HMPP_STS_NO_ERR。
  • 失败：返回错误码。

错误码

错误码

描述

HMPP_STS_NULL_PTR_ERR

src、dst这两个入参中存在空指针。

HMPP_STS_SIZE_ERR

len小于或等于0或len小于3。

HMPP_STS_NOT_SUPPORT

当前数据类型转换不支持参数传入的round mode。

HMPP_STS_SCALE_ERR

scale不在(0,INF)范围内或输入为nan。

示例

#define BUFFER_SIZE_S 9

void ConvertExample()
{
    const int8_t src[BUFFER_SIZE_S] = { 123, 32, 0, 254, 3, 255, 32, 129, 254 };
    uint8_t dst[BUFFER_SIZE_S] = {0};
    int32_t i;
    HmppResult result = HMPPS_Convert_8s8u(src, dst, BUFFER_SIZE_S);
    printf("result = %d \ndst =", result);
    for(i = 0; i < BUFFER_SIZE_S; i++){
        printf(" %u ", dst[i]);
    }
}

运行结果：

result = 0 
dst = 123  32  0  0  3  0  32  0  0
父主题： 基础与通用运算