昇腾社区首页
中文
注册
开发者
下载
支持
积分兑换 NEW

vcmp

功能说明

逐元素比较向量 src0 与向量 src1 的每个元素，如果比较后的结果为真，则输出结果的对应比特位为1，否则为0。

元素类型为f16时，向量中元素个数为128，因此结果是一个连续的128bit，写入CMPMASK。

元素类型为f32时，向量中元素个数为64，因此结果是一个连续的64bit，写入CMPMASK的低64位。

读取CMPMASK可以使用get_cmpmask()接口。

支持多种比较接口：

  • EQ：src0等于src1（equal to）
  • NE：src0不等于src1（not equal to）
  • LT：src0小于src1（less than）
  • GT：src0大于src1（greater than）
  • LE：src0小于或等于src1（less than or equal to）
  • GE：src0大于或等于src1（greater than or equal to）

该接口支持MASK配置。如果MASK中的一位为零，则CMPMASK中的相应位不变。

接口原型

// 相同接口的不同原型区别在于源地址和目的地址的数据类型不同
// vcmp_eq
void vcmp_eq(__ubuf__ half *src0, __ubuf__ half *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_eq(__ubuf__ float *src0, __ubuf__ float *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_eq(__ubuf__ uint16_t *src0, __ubuf__ uint16_t *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

// vcmp_ne
void vcmp_ne(__ubuf__ half *src0, __ubuf__ half *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_ne(__ubuf__ float *src0, __ubuf__ float *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_ne(__ubuf__ uint16_t *src0, __ubuf__ uint16_t *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

// vcmp_lt
void vcmp_lt(__ubuf__ half *src0, __ubuf__ half *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_lt(__ubuf__ float *src0, __ubuf__ float *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_lt(__ubuf__ uint16_t *src0, __ubuf__ uint16_t *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

// vcmp_gt
void vcmp_gt(__ubuf__ half *src0, __ubuf__ half *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_gt(__ubuf__ float *src0, __ubuf__ float *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_gt(__ubuf__ uint16_t *src0, __ubuf__ uint16_t *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

// vcmp_le
void vcmp_le(__ubuf__ half *src0, __ubuf__ half *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_le(__ubuf__ float *src0, __ubuf__ float *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_le(__ubuf__ uint16_t *src0, __ubuf__ uint16_t *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

// vcmp_ge
void vcmp_ge(__ubuf__ half *src0, __ubuf__ half *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_ge(__ubuf__ float *src0, __ubuf__ float *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

void vcmp_ge(__ubuf__ uint16_t *src0, __ubuf__ uint16_t *src1, uint8_t repeat, uint8_t dstBlockStride, uint8_t src0BlockStride, uint8_t src1BlockStride, uint8_t dstRepeatStride, uint8_t src0RepeatStride, uint8_t src1RepeatStride);

参数说明

参数含义见 表2 双目运算参数说明。

流水类型

PIPE_V
父主题： 比较运算