asc_loadunalign_postupdate

产 品 支 持 情 况

产 品	是 否 支 持
Ascend 950PR/Ascend 950DT	√

功 能 说 明

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iter_reg入 参）：配 合asc_loadunalign（不 带iter_reg入 参）接 口 使 用。
• asc_loadunalign_postupdate（带iter_reg入 参）：配 合asc_loadunalign（带iter_reg入 参）接 口 使 用。

函 数 原 型

C++

__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, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_uint8_t& dst, vector_load_unalign& src0, __ubuf__ uint8_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_fp4x2_e2m1_t& dst, vector_load_unalign& src0, __ubuf__ fp4x2_e2m1_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_fp4x2_e1m2_t& dst, vector_load_unalign& src0, __ubuf__ fp4x2_e1m2_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e8m0_t& dst, vector_load_unalign& src0, __ubuf__ fp8_e8m0_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e5m2_t& dst, vector_load_unalign& src0, __ubuf__ fp8_e5m2_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_fp8_e4m3fn_t& dst, vector_load_unalign& src0, __ubuf__ fp8_e4m3fn_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_hifloat8_t& dst, vector_load_unalign& src0, __ubuf__ hifloat8_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_int16_t& dst, vector_load_unalign& src0, __ubuf__ int16_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_uint16_t& dst, vector_load_unalign& src0, __ubuf__ uint16_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_half& dst, vector_load_unalign& src0, __ubuf__ half *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_bfloat16_t& dst, vector_load_unalign& src0, __ubuf__ bfloat16_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_int32_t& dst, vector_load_unalign& src0, __ubuf__ int32_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_uint32_t& dst, vector_load_unalign& src0, __ubuf__ uint32_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_float& dst, vector_load_unalign& src0, __ubuf__ float *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_int64_t& dst, vector_load_unalign& src0, __ubuf__ int64_t *src1, iter_reg offset, uint32_t count)
__simd_callee__ inline void asc_loadunalign_postupdate(vector_int4x2_t& dst, vector_load_unalign& src0, __ubuf__ int4b_t *src1, iter_reg offset, uint32_t count)

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	源 操 作 数（矢 量）的 起 始 地 址。
offset	输 入	存 储 地 址 的 偏 移 量。实 际 搬 运UB起 始 地 址 为src1+offset
count	输 入	搬 运 完 成 后offset在 自 增 前 被 更 新 为offset+count

矢 量 数 据 寄 存 器 和 掩 码 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

• 接 口 中 的src1不 需 要32B对 齐。

调 用 示 例

C++

vector_int8_t dst;
vector_load_unalign src0;
__ubuf__ int8_t src1[256];
uint32_t count = 16;
asc_loadunalign_postupdate(dst, src0, src1, count);