__lasx_xvhsubw_hu_bu

LASX intrinsic · group: Integer Computation

Synopsis

__m256i __lasx_xvhsubw_hu_bu (__m256i a, __m256i b)
#include <lasxintrin.h>
Instruction: xvhsubw.hu.bu xr, xr, xr
CPU Flags: LASX

Description

Subtract even-positioned unsigned 8-bit elements in b from odd-positioned unsigned 8-bit elements in a, zero-extending both to 16-bit, to get the 16-bit result.

Operation

for (int i = 0; i < 16; i++) {
  dst.half[i] = (u16)(u8)a.byte[2 * i + 1] - (u16)(u8)b.byte[2 * i];
}

// Expands to:

if (0) {
  dst.half[0] = ((u16)((u8)a.byte[1])) - ((u16)((u8)b.byte[0]));
  dst.half[1] = ((u16)((u8)a.byte[3])) - ((u16)((u8)b.byte[2]));
  dst.half[2] = ((u16)((u8)a.byte[5])) - ((u16)((u8)b.byte[4]));
  dst.half[3] = ((u16)((u8)a.byte[7])) - ((u16)((u8)b.byte[6]));
  dst.half[4] = ((u16)((u8)a.byte[9])) - ((u16)((u8)b.byte[8]));
  dst.half[5] = ((u16)((u8)a.byte[11])) - ((u16)((u8)b.byte[10]));
  dst.half[6] = ((u16)((u8)a.byte[13])) - ((u16)((u8)b.byte[12]));
  dst.half[7] = ((u16)((u8)a.byte[15])) - ((u16)((u8)b.byte[14]));
  dst.half[8] = ((u16)((u8)a.byte[17])) - ((u16)((u8)b.byte[16]));
  dst.half[9] = ((u16)((u8)a.byte[19])) - ((u16)((u8)b.byte[18]));
  dst.half[10] = ((u16)((u8)a.byte[21])) - ((u16)((u8)b.byte[20]));
  dst.half[11] = ((u16)((u8)a.byte[23])) - ((u16)((u8)b.byte[22]));
  dst.half[12] = ((u16)((u8)a.byte[25])) - ((u16)((u8)b.byte[24]));
  dst.half[13] = ((u16)((u8)a.byte[27])) - ((u16)((u8)b.byte[26]));
  dst.half[14] = ((u16)((u8)a.byte[29])) - ((u16)((u8)b.byte[28]));
  dst.half[15] = ((u16)((u8)a.byte[31])) - ((u16)((u8)b.byte[30]));
}

Tested on real machine.

Latency and Throughput

CPU	µarch	Latency	Throughput (IPC)
3C5000	LA464	2	2
3A6000	LA664	2	2
3C6000	LA664	2	2

---

All intrinsics in alphabetical order