asc_loadunalign_postupdate【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Ascend 950PR/Ascend 950DT支持Atlas A3 训练系列产品/Atlas A3 推理系列产品不支持Atlas A2 训练系列产品/Atlas A2 推理系列产品不支持Atlas 200I/500 A2 推理产品不支持Atlas 推理系列产品AI Core不支持Atlas 推理系列产品Vector Core不支持Atlas 训练系列产品不支持功能说明头文件路径c_api/reg_compute/reg_load.h。数据搬运接口适用于从UB非32B对齐地址起始连续搬入。asc_loadunalign_postupdate不带addr_reg入参配合asc_loadunalign_pre不带addr_reg入参接口使用。asc_loadunalign_postupdate带addr_reg入参配合asc_loadunalign_pre带addr_reg入参接口使用。函数原型__simd_callee__ inline void asc_loadunalign_postupdate(vector_int8_t dst, vector_load_unalign src0, __ubuf__ int8_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint8_t dst, vector_load_unalign src0, __ubuf__ uint8_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp4x2_e2m1_t dst, vector_load_unalign src0, __ubuf__ fp4x2_e2m1_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp4x2_e1m2_t dst, vector_load_unalign src0, __ubuf__ fp4x2_e1m2_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e8m0_t dst, vector_load_unalign src0, __ubuf__ fp8_e8m0_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e5m2_t dst, vector_load_unalign src0, __ubuf__ fp8_e5m2_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e4m3fn_t dst, vector_load_unalign src0, __ubuf__ fp8_e4m3fn_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_hifloat8_t dst, vector_load_unalign src0, __ubuf__ hifloat8_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_half dst, vector_load_unalign src0, __ubuf__ half* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int16_t dst, vector_load_unalign src0, __ubuf__ int16_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint16_t dst, vector_load_unalign src0, __ubuf__ uint16_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_bfloat16_t dst, vector_load_unalign src0, __ubuf__ bfloat16_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int32_t dst, vector_load_unalign src0, __ubuf__ int32_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint32_t dst, vector_load_unalign src0, __ubuf__ uint32_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_float dst, vector_load_unalign src0, __ubuf__ float* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int64_t dst, vector_load_unalign src0, __ubuf__ int64_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int4x2_t dst, vector_load_unalign src0, __ubuf__ int4b_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int8_t dst, vector_load_unalign src0, __ubuf__ int8_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint8_t dst, vector_load_unalign src0, __ubuf__ uint8_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp4x2_e2m1_t dst, vector_load_unalign src0, __ubuf__ fp4x2_e2m1_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp4x2_e1m2_t dst, vector_load_unalign src0, __ubuf__ fp4x2_e1m2_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e8m0_t dst, vector_load_unalign src0, __ubuf__ fp8_e8m0_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e5m2_t dst, vector_load_unalign src0, __ubuf__ fp8_e5m2_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e4m3fn_t dst, vector_load_unalign src0, __ubuf__ fp8_e4m3fn_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_hifloat8_t dst, vector_load_unalign src0, __ubuf__ hifloat8_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int16_t dst, vector_load_unalign src0, __ubuf__ int16_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint16_t dst, vector_load_unalign src0, __ubuf__ uint16_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_half dst, vector_load_unalign src0, __ubuf__ half *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_bfloat16_t dst, vector_load_unalign src0, __ubuf__ bfloat16_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int32_t dst, vector_load_unalign src0, __ubuf__ int32_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint32_t dst, vector_load_unalign src0, __ubuf__ uint32_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_float dst, vector_load_unalign src0, __ubuf__ float *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int64_t dst, vector_load_unalign src0, __ubuf__ int64_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int4x2_t dst, vector_load_unalign src0, __ubuf__ int4b_t *src1, addr_reg offset, uint32_t inc)参数说明参数名输入/输出描述dst输出目的操作数矢量。src0输入源操作数矢量数据寄存器。src1输入/输出源操作数矢量的起始地址。当使用不含addr_reg参数的接口时搬运完成后该地址会被自动更新。offset输入/输出存储地址的偏移量。当使用含addr_reg参数的接口时搬运完成后该偏移量会被自动更新。count输入搬运元素个数。inc输入在多次循环中多次调用该接口时每次搬运完成后下次搬运的offset在自增前会更新为offset inc。建议设置为0。矢量数据寄存器和掩码寄存器的详细说明请参见data_type_definition.md。返回值说明无流水类型PIPE_V约束说明接口中的src1不需要32B对齐。调用示例vector_int8_t dst; vector_load_unalign src0; __ubuf__ int8_t* src_ptr (__ubuf__ int8_t*)asc_get_phy_buf_addr(0); uint32_t count 16; asc_loadunalign_postupdate(dst, src0, src_ptr, count);【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
asc_loadunalign_postupdate【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Ascend 950PR/Ascend 950DT支持Atlas A3 训练系列产品/Atlas A3 推理系列产品不支持Atlas A2 训练系列产品/Atlas A2 推理系列产品不支持Atlas 200I/500 A2 推理产品不支持Atlas 推理系列产品AI Core不支持Atlas 推理系列产品Vector Core不支持Atlas 训练系列产品不支持功能说明头文件路径c_api/reg_compute/reg_load.h。数据搬运接口适用于从UB非32B对齐地址起始连续搬入。asc_loadunalign_postupdate不带addr_reg入参配合asc_loadunalign_pre不带addr_reg入参接口使用。asc_loadunalign_postupdate带addr_reg入参配合asc_loadunalign_pre带addr_reg入参接口使用。函数原型__simd_callee__ inline void asc_loadunalign_postupdate(vector_int8_t dst, vector_load_unalign src0, __ubuf__ int8_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint8_t dst, vector_load_unalign src0, __ubuf__ uint8_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp4x2_e2m1_t dst, vector_load_unalign src0, __ubuf__ fp4x2_e2m1_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp4x2_e1m2_t dst, vector_load_unalign src0, __ubuf__ fp4x2_e1m2_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e8m0_t dst, vector_load_unalign src0, __ubuf__ fp8_e8m0_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e5m2_t dst, vector_load_unalign src0, __ubuf__ fp8_e5m2_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e4m3fn_t dst, vector_load_unalign src0, __ubuf__ fp8_e4m3fn_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_hifloat8_t dst, vector_load_unalign src0, __ubuf__ hifloat8_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_half dst, vector_load_unalign src0, __ubuf__ half* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int16_t dst, vector_load_unalign src0, __ubuf__ int16_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint16_t dst, vector_load_unalign src0, __ubuf__ uint16_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_bfloat16_t dst, vector_load_unalign src0, __ubuf__ bfloat16_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int32_t dst, vector_load_unalign src0, __ubuf__ int32_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint32_t dst, vector_load_unalign src0, __ubuf__ uint32_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_float dst, vector_load_unalign src0, __ubuf__ float* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int64_t dst, vector_load_unalign src0, __ubuf__ int64_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int4x2_t dst, vector_load_unalign src0, __ubuf__ int4b_t* src1, uint32_t count) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int8_t dst, vector_load_unalign src0, __ubuf__ int8_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint8_t dst, vector_load_unalign src0, __ubuf__ uint8_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp4x2_e2m1_t dst, vector_load_unalign src0, __ubuf__ fp4x2_e2m1_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp4x2_e1m2_t dst, vector_load_unalign src0, __ubuf__ fp4x2_e1m2_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e8m0_t dst, vector_load_unalign src0, __ubuf__ fp8_e8m0_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e5m2_t dst, vector_load_unalign src0, __ubuf__ fp8_e5m2_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e4m3fn_t dst, vector_load_unalign src0, __ubuf__ fp8_e4m3fn_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_hifloat8_t dst, vector_load_unalign src0, __ubuf__ hifloat8_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int16_t dst, vector_load_unalign src0, __ubuf__ int16_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint16_t dst, vector_load_unalign src0, __ubuf__ uint16_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_half dst, vector_load_unalign src0, __ubuf__ half *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_bfloat16_t dst, vector_load_unalign src0, __ubuf__ bfloat16_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int32_t dst, vector_load_unalign src0, __ubuf__ int32_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_uint32_t dst, vector_load_unalign src0, __ubuf__ uint32_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_float dst, vector_load_unalign src0, __ubuf__ float *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int64_t dst, vector_load_unalign src0, __ubuf__ int64_t *src1, addr_reg offset, uint32_t inc) __simd_callee__ inline void asc_loadunalign_postupdate(vector_int4x2_t dst, vector_load_unalign src0, __ubuf__ int4b_t *src1, addr_reg offset, uint32_t inc)参数说明参数名输入/输出描述dst输出目的操作数矢量。src0输入源操作数矢量数据寄存器。src1输入/输出源操作数矢量的起始地址。当使用不含addr_reg参数的接口时搬运完成后该地址会被自动更新。offset输入/输出存储地址的偏移量。当使用含addr_reg参数的接口时搬运完成后该偏移量会被自动更新。count输入搬运元素个数。inc输入在多次循环中多次调用该接口时每次搬运完成后下次搬运的offset在自增前会更新为offset inc。建议设置为0。矢量数据寄存器和掩码寄存器的详细说明请参见data_type_definition.md。返回值说明无流水类型PIPE_V约束说明接口中的src1不需要32B对齐。调用示例vector_int8_t dst; vector_load_unalign src0; __ubuf__ int8_t* src_ptr (__ubuf__ int8_t*)asc_get_phy_buf_addr(0); uint32_t count 16; asc_loadunalign_postupdate(dst, src0, src_ptr, count);【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
