portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-15 02:05:37 +00:00
Code Releases Activity

Merge pull request #145 from yongchull/fix-skylake-avx512

Browse Source 
patch to compile with AVX512 for SkyLake Xeon processor using GCC7.2.…
This commit is contained in:
Peter Boyle 2018-01-28 11:01:43 +00:00 committed by GitHub
commit ebb1bebf24
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 8 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
configure.ac
View File

@ -249,6 +249,9 @@ case ${ax_cv_cxx_compiler_vendor} in
AVX512)
AC_DEFINE([AVX512],[1],[AVX512 intrinsics])
SIMD_FLAGS='-mavx512f -mavx512pf -mavx512er -mavx512cd';;
SKL)
AC_DEFINE([AVX512],[1],[AVX512 intrinsics for SkyLake Xeon])
SIMD_FLAGS='-march=skylake-avx512';;
KNC)
AC_DEFINE([IMCI],[1],[IMCI intrinsics for Knights Corner])
SIMD_FLAGS='';;

4
lib/qcd/action/fermion/CayleyFermion5Dvec.cc
View File

@ -469,7 +469,7 @@ void CayleyFermion5D<Impl>::MooeeInternalAsm(const FermionField &psi, FermionFie
}
a0 = a0+incr;
a1 = a1+incr;
a2 = a2+sizeof(Simd::scalar_type);
a2 = a2+sizeof(typename Simd::scalar_type);
}}
{
int lexa = s1+LLs*site;
@ -701,7 +701,7 @@ void CayleyFermion5D<Impl>::MooeeInternalZAsm(const FermionField &psi, FermionFi
}
a0 = a0+incr;
a1 = a1+incr;
a2 = a2+sizeof(Simd::scalar_type);
a2 = a2+sizeof(typename Simd::scalar_type);
}}
{
int lexa = s1+LLs*site;

2
lib/qcd/action/fermion/DomainWallEOFAFermionvec.cc
View File

@ -475,7 +475,7 @@ namespace QCD {
}
a0 = a0 + incr;
a1 = a1 + incr;
a2 = a2 + sizeof(Simd::scalar_type);
a2 = a2 + sizeof(typename Simd::scalar_type);
}
}

2
lib/qcd/action/fermion/MobiusEOFAFermionvec.cc
View File

@ -853,7 +853,7 @@ namespace QCD {
a0 = a0 + incr;
a1 = a1 + incr;
a2 = a2 + sizeof(Simd::scalar_type);
a2 = a2 + sizeof(typename Simd::scalar_type);
}
}

2
lib/simd/Grid_avx512.h
View File

@ -556,7 +556,7 @@ namespace Optimization {
v3 = _mm256_add_epi32(v1, v2);
v1 = _mm256_hadd_epi32(v3, v3);
v2 = _mm256_hadd_epi32(v1, v1);
u1 = _mm256_castsi256_si128(v2) // upper half
u1 = _mm256_castsi256_si128(v2); // upper half
u2 = _mm256_extracti128_si256(v2, 1); // lower half
ret = _mm_add_epi32(u1, u2);
return _mm_cvtsi128_si32(ret);