portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-06-17 15:27:06 +01:00
Code Releases Activity

Corrected some compilation errors (zolotarev.h) and SSE4 vsplat and conj to make cshift test pass.

Browse Source
This commit is contained in:
neo
2015-05-18 16:48:14 +09:00
parent 1887c77498
commit 6d2accba7b
26 changed files with 482 additions and 73 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
lib/simd/Grid_vComplexF.h
View File

@ -219,7 +219,7 @@ namespace Grid {
ret.v = _mm256_set_ps(b,a,b,a,b,a,b,a);
#endif
#ifdef SSE4
ret.v = _mm_set_ps(a,b,a,b);
ret.v = _mm_set_ps(b,a,b,a);
#endif
#ifdef AVX512
ret.v = _mm512_set_ps(b,a,b,a,b,a,b,a,b,a,b,a,b,a,b,a);
@ -354,9 +354,7 @@ namespace Grid {
#endif
#ifdef SSE4
cvec tmp;
tmp = _mm_addsub_ps(ret.v,_mm_shuffle_ps(in.v,in.v,_MM_SHUFFLE(2,3,0,1))); // ymm1 <- br,bi
ret.v=_mm_shuffle_ps(tmp,tmp,_MM_SHUFFLE(2,3,0,1));
ret.v = _mm_xor_ps(_mm_addsub_ps(ret.v,in.v), _mm_set1_ps(-0.f));
#endif
#ifdef AVX512
ret.v = _mm512_mask_sub_ps(in.v,0xaaaa,ret.v,in.v); // Zero out 0+real 0-imag