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

'typename' is added to compile with AVX512 using GCC7.2.0; a semicolon was missing in Grid_avx512.h and the bug is fixed. Option SKL is added to configure script for skylake processor specific AVX512 operations. Code can be compiled with --enable-simd=SKL using GCC 7.2.0, but Test_simd fails. AVX512 support for complex double type with non-intel compilers makes this error; it needs a review.

Browse Source
This commit is contained in:
Yong-Chull Jang 2017-12-23 14:54:07 -05:00
parent 89c4e9b168
commit 3cb8cb7282
5 changed files with 10 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lib/algorithms/iterative/ImplicitlyRestartedLanczos.h
View File

@ -267,7 +267,7 @@ until convergence
Qt = Eigen::MatrixXd::Identity(Nm,Nm); Qt = Eigen::MatrixXd::Identity(Nm,Nm);
for(int ip=k2; ip<Nm; ++ip){ for(int ip=k2; ip<Nm; ++ip){
// Eigen replacement for qr_decomp ??? // Eigen replacement for qr_decomp ???
clog << "ckpt B1: shift[" << ip << "] = " << eval2[ip] << endl; clog << "ckpt B1: shift[" << ip << "] = " << eval2[ip] << std::endl;
qr_decomp(eval,lme,Nm,Nm,Qt,eval2[ip],k1,Nm); qr_decomp(eval,lme,Nm,Nm,Qt,eval2[ip],k1,Nm);
} }

6
lib/qcd/action/fermion/CayleyFermion5Dvec.cc
View File

@ -469,7 +469,8 @@ void CayleyFermion5D<Impl>::MooeeInternalAsm(const FermionField &psi, FermionFie
} }
a0 = a0+incr; a0 = a0+incr;
a1 = a1+incr; a1 = a1+incr;
a2 = a2+sizeof(Simd::scalar_type); //a2 = a2+sizeof(Simd::scalar_type);
a2 = a2+sizeof(typename Simd::scalar_type); //ypj [debug]
}} }}
{ {
int lexa = s1+LLs*site; int lexa = s1+LLs*site;
@ -701,7 +702,8 @@ void CayleyFermion5D<Impl>::MooeeInternalZAsm(const FermionField &psi, FermionFi
} }
a0 = a0+incr; a0 = a0+incr;
a1 = a1+incr; a1 = a1+incr;
a2 = a2+sizeof(Simd::scalar_type); //a2 = a2+sizeof(Simd::scalar_type);
a2 = a2+sizeof(typename Simd::scalar_type); // ypj [debug]
}} }}
{ {
int lexa = s1+LLs*site; int lexa = s1+LLs*site;

3
lib/qcd/action/fermion/DomainWallEOFAFermionvec.cc
View File

@ -475,7 +475,8 @@ namespace QCD {
} }
a0 = a0 + incr; a0 = a0 + incr;
a1 = a1 + incr; a1 = a1 + incr;
a2 = a2 + sizeof(Simd::scalar_type); //a2 = a2 + sizeof(Simd::scalar_type);
a2 = a2 + sizeof(typename Simd::scalar_type); // ypj [debug]
} }
} }

3
lib/qcd/action/fermion/MobiusEOFAFermionvec.cc
View File

@ -853,7 +853,8 @@ namespace QCD {
a0 = a0 + incr; a0 = a0 + incr;
a1 = a1 + incr; a1 = a1 + incr;
a2 = a2 + sizeof(Simd::scalar_type); //a2 = a2 + sizeof(Simd::scalar_type);
a2 = a2 + sizeof(typename Simd::scalar_type); // ypj [debug]
} }
} }

2
lib/simd/Grid_avx512.h
View File

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