__lasx_xvsrarn_w_d

LASX intrinsic · group: Shift

Synopsis

__m256i __lasx_xvsrarn_w_d (__m256i a, __m256i b)
#include <lasxintrin.h>
Instruction: xvsrarn.w.d xr, xr, xr
CPU Flags: LASX

Description

Arithmetic right shift with rounding each signed 64-bit element in a by the corresponding element of b modulo 64, write the lower 32 bits of the shifted result into the lower half of each 128-bit lane of dst, and zero the upper half.

Operation

for (int i = 0; i < 4; i++) {
  if (i < 2) {
    u8 shift = (b.dword[i] & 63);
    if (shift == 0) {
      dst.word[i] = (s32)(s64)a.dword[i];
    } else {
      dst.word[i] = (s32)(((s64)a.dword[i] >> shift) +
                          (((s64)a.dword[i] >> (shift - 1)) & 0x1));
    }
  } else {
    dst.word[i] = 0;
  }
}
for (int i = 4; i < 8; i++) {
  if (i < 6) {
    u8 shift = (b.dword[i - 2] & 63);
    if (shift == 0) {
      dst.word[i] = (s32)(s64)a.dword[i - 2];
    } else {
      dst.word[i] = (s32)(((s64)a.dword[i - 2] >> shift) +
                          (((s64)a.dword[i - 2] >> (shift - 1)) & 0x1));
    }
  } else {
    dst.word[i] = 0;
  }
}

Tested on real machine.

Latency and Throughput

CPU	µarch	Latency	Throughput (IPC)
3C5000	LA464	4	1
3A6000	LA664	4	2
3C6000	LA664	4	2

---

All intrinsics in alphabetical order