__lsx_vhaddw_d_w
LSX intrinsic · group: Integer Computation
Synopsis
__m128i __lsx_vhaddw_d_w (__m128i a, __m128i b)
#include <lsxintrin.h>
Instruction: vhaddw.d.w vr, vr, vr
CPU Flags: LSX
Description
Add odd-positioned signed 32-bit elements in a to even-positioned signed 32-bit elements in b, sign-extending both to 64-bit, to get the 64-bit result.
Operation
for (int i = 0; i < 2; i++) {
dst.dword[i] = (s64)(s32)a.word[2 * i + 1] + (s64)(s32)b.word[2 * i];
}
// Expands to:
if (0) {
dst.dword[0] = ((s64)((s32)a.word[1])) + ((s64)((s32)b.word[0]));
dst.dword[1] = ((s64)((s32)a.word[3])) + ((s64)((s32)b.word[2]));
}
Tested on real machine.
Latency and Throughput
| CPU | µarch | Latency | Throughput (IPC) |
|---|---|---|---|
| 3C5000 | LA464 | 2 | 2 |
| 3A6000 | LA664 | 2 | 2 |
| 3C6000 | LA664 | 2 | 2 |
| 2K1000LA | LA264 | 2 | 1 |
| 2K3000 | LA364E | 2 | 1 |