asc_loadalign_brc_elem【免费下载链接】asc-devkit本项目是CANN 推出的昇腾AI处理器专用的算子程序开发语言原生支持C和C标准规范主要由类库和语言扩展层构成提供多层级API满足多维场景算子开发诉求。项目地址: https://gitcode.com/cann/asc-devkit产品支持情况产品是否支持Ascend 950PR/Ascend 950DT√功能说明对齐数据搬运接口从UB连续对齐搬入目的操作数实现BRC搬入模式搬运一个b8/b16/b32类型的数据并Broadcast到所有元素位置。支持三种偏移方式偏移固定传入0由用户自行更新源操作数的地址。通过int32_t传入偏移用户可以选择更新偏移或者更新源操作数的地址。通过iter_reg地址寄存器传入偏移用户可以选择更新偏移或者更新源操作数的地址。函数原型偏移固定传入0由用户自行更新源操作数的地址__simd_callee__ inline void asc_loadalign_brc_elem(vector_int8_t dst, __ubuf__ int8_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_uint8_t dst, __ubuf__ uint8_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp4x2_e2m1_t dst, __ubuf__ fp4x2_e2m1_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp4x2_e1m2_t dst, __ubuf__ fp4x2_e1m2_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp8_e8m0_t dst, __ubuf__ fp8_e8m0_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp8_e5m2_t dst, __ubuf__ fp8_e5m2_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp8_e4m3fn_t dst, __ubuf__ fp8_e4m3fn_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_hifloat8_t dst, __ubuf__ hifloat8_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_int16_t dst, __ubuf__ int16_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_uint16_t dst, __ubuf__ uint16_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_half dst, __ubuf__ half* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_bfloat16_t dst, __ubuf__ bfloat16_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_int32_t dst, __ubuf__ int32_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_uint32_t dst, __ubuf__ uint32_t* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_float dst, __ubuf__ float* src) __simd_callee__ inline void asc_loadalign_brc_elem(vector_int4x2_t dst, __ubuf__ int4b_t* src)通过int32_t传入偏移__simd_callee__ inline void asc_loadalign_brc_elem(vector_int8_t dst, __ubuf__ int8_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_uint8_t dst, __ubuf__ uint8_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp4x2_e2m1_t dst, __ubuf__ fp4x2_e2m1_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp4x2_e1m2_t dst, __ubuf__ fp4x2_e1m2_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp8_e8m0_t dst, __ubuf__ fp8_e8m0_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp8_e5m2_t dst, __ubuf__ fp8_e5m2_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp8_e4m3fn_t dst, __ubuf__ fp8_e4m3fn_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_hifloat8_t dst, __ubuf__ hifloat8_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_int16_t dst, __ubuf__ int16_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_uint16_t dst, __ubuf__ uint16_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_half dst, __ubuf__ half* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_bfloat16_t dst, __ubuf__ bfloat16_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_int32_t dst, __ubuf__ int32_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_uint32_t dst, __ubuf__ uint32_t* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_float dst, __ubuf__ float* src, int32_t offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_int4x2_t dst, __ubuf__ int4b_t* src, int32_t offset)通过iter_reg地址寄存器传入偏移__simd_callee__ inline void asc_loadalign_brc_elem(vector_fp4x2_e1m2_t dst, __ubuf__ fp4x2_e1m2_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp4x2_e2m1_t dst, __ubuf__ fp4x2_e2m1_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_int4x2_t dst, __ubuf__ int4b_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_uint8_t dst, __ubuf__ uint8_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_int8_t dst, __ubuf__ int8_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp8_e4m3fn_t dst, __ubuf__ fp8_e4m3fn_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_hifloat8_t dst, __ubuf__ hifloat8_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp8_e5m2_t dst, __ubuf__ fp8_e5m2_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_fp8_e8m0_t dst, __ubuf__ fp8_e8m0_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_uint16_t dst, __ubuf__ uint16_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_int16_t dst, __ubuf__ int16_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_bfloat16_t dst, __ubuf__ bfloat16_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_half dst, __ubuf__ half* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_uint32_t dst, __ubuf__ uint32_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_int32_t dst, __ubuf__ int32_t* src, iter_reg offset) __simd_callee__ inline void asc_loadalign_brc_elem(vector_float dst, __ubuf__ float* src, iter_reg offset)参数说明参数名输入/输出描述dst输出目的操作数矢量数据寄存器。src输入源操作数矢量的起始地址。offset输入偏移量。矢量数据寄存器的详细说明请参见reg数据类型定义.md。返回值说明无流水类型PIPE_V约束说明无调用示例vector_half dst; __ubuf__ half* src; asc_loadalign_brc_elem(dst, src);【免费下载链接】asc-devkit本项目是CANN 推出的昇腾AI处理器专用的算子程序开发语言原生支持C和C标准规范主要由类库和语言扩展层构成提供多层级API满足多维场景算子开发诉求。项目地址: https://gitcode.com/cann/asc-devkit创作声明：本文部分内容由AI辅助生成（AIGC），仅供参考