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

Reworking to keep intel compiler happy

Browse Source
This commit is contained in:
Peter Boyle
2015-05-19 21:29:07 +01:00
parent ee8cf77071
commit efc0d1e0b9
8 changed files with 62 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
lib/simd/Grid_vComplexF.h
View File

@ -234,26 +234,34 @@ namespace Grid {
}
friend inline ComplexF Reduce(const vComplexF & in)
{
vComplexF v1,v2;
union {
cvec v;
float f[sizeof(cvec)/sizeof(float)];
} conv;
#ifdef SSE4
vComplexF v1;
permute(v1,in,0); // sse 128; paired complex single
v1=v1+in;
return ComplexF(v1.v[0],v1.v[1]);
#endif
#if defined(AVX1) || defined (AVX2)
vComplexF v1,v2;
permute(v1,in,0); // sse 128; paired complex single
v1=v1+in;
permute(v2,v1,1); // avx 256; quad complex single
v1=v1+v2;
return ComplexF(v1.v[0],v1.v[1]);
#endif
#ifdef AVX512
return ComplexF(_mm512_mask_reduce_add_ps(0x5555, in.v),_mm512_mask_reduce_add_ps(0xAAAA, in.v));
permute(v1,in,0); // avx512 octo-complex single
v1=v1+in;
permute(v2,v1,1);
v1=v1+v2;
permute(v2,v1,2);
v1=v1+v2;
#endif
#ifdef QPX
#error
#endif
conv.v = v1.v;
return ComplexF(conv.f[0],conv.f[1]);
}
friend inline vComplexF operator * (const ComplexF &a, vComplexF b){