Skip to content

Commit

Permalink
Change the target of prefetching
Browse files Browse the repository at this point in the history
  • Loading branch information
0140454 committed Oct 21, 2016
1 parent c3f9238 commit df7d9e8
Showing 1 changed file with 4 additions and 4 deletions.
8 changes: 4 additions & 4 deletions impl.c
Expand Up @@ -169,10 +169,10 @@ void sse_prefetch_multiply(int *src1, int *src2, int *dst, int src1_w,

for (int k = 0; k < src2_w; k += 4) {
#define SSE_PFDIST 8
_mm_prefetch(src1 + (k + SSE_PFDIST + 0) * src1_w + y, _MM_HINT_T1);
_mm_prefetch(src1 + (k + SSE_PFDIST + 1) * src1_w + k, _MM_HINT_T1);
_mm_prefetch(src1 + (k + SSE_PFDIST + 2) * src1_w + k, _MM_HINT_T1);
_mm_prefetch(src1 + (k + SSE_PFDIST + 3) * src1_w + k, _MM_HINT_T1);
_mm_prefetch(src2 + (k + SSE_PFDIST + 0) * src2_w + y, _MM_HINT_T1);
_mm_prefetch(src2 + (k + SSE_PFDIST + 1) * src2_w + y, _MM_HINT_T1);
_mm_prefetch(src2 + (k + SSE_PFDIST + 2) * src2_w + y, _MM_HINT_T1);
_mm_prefetch(src2 + (k + SSE_PFDIST + 3) * src2_w + y, _MM_HINT_T1);

__m128i I0 = _mm_loadu_si128((__m128i *)(src1 + (x + 0) * src1_w + k));
__m128i I1 = _mm_loadu_si128((__m128i *)(src1 + (x + 1) * src1_w + k));
Expand Down

0 comments on commit df7d9e8

Please sign in to comment.