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

axpy_norm_fast -- wasn't using the determinstic MPI sum causing issues

Browse Source
This commit is contained in:
Peter Boyle 2024-10-11 03:22:18 +00:00
parent fd58f0b669
commit 54f1999030
1 changed files with 37 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

38
Grid/lattice/Lattice_reduction.h
View File

@ -375,8 +375,44 @@ axpby_norm_fast(Lattice<vobj> &z,sobj a,sobj b,const Lattice<vobj> &x,const Latt
coalescedWrite(inner_tmp_v[ss],innerProduct(tmp,tmp)); coalescedWrite(inner_tmp_v[ss],innerProduct(tmp,tmp));
coalescedWrite(z_v[ss],tmp); coalescedWrite(z_v[ss],tmp);
}); });
bool ok;
uint64_t csum=0;
uint64_t csum2=0;
#ifdef GRID_SYCL
if ( FlightRecorder::LoggingMode != FlightRecorder::LoggingModeNone)
{
// z_v
{
Integer words = sites*sizeof(vobj)/sizeof(uint64_t);
uint64_t *base= (uint64_t *)&z_v[0];
csum=svm_xor(base,words);
ok = FlightRecorder::CsumLog(csum);
if ( !ok ) {
csum2=svm_xor(base,words);
std::cerr<< " Bad z_v CSUM " << std::hex<< csum << " recomputed as "<<csum2<<std::dec<<std::endl;
}
assert(ok);
}
// inner_v
{
Integer words = sites*sizeof(inner_t)/sizeof(uint64_t);
uint64_t *base= (uint64_t *)&inner_tmp_v[0];
csum=svm_xor(base,words);
ok = FlightRecorder::CsumLog(csum);
if ( !ok ) {
csum2=svm_xor(base,words);
std::cerr<< " Bad inner_tmp_v CSUM " << std::hex<< csum << " recomputed as "<<csum2<<std::dec<<std::endl;
}
assert(ok);
}
}
#endif
nrm = real(TensorRemove(sumD(inner_tmp_v,sites))); nrm = real(TensorRemove(sumD(inner_tmp_v,sites)));
grid->GlobalSum(nrm); ok = FlightRecorder::NormLog(real(nrm));
assert(ok);
RealD local = real(nrm);
grid->GlobalSumP2P(nrm);
FlightRecorder::ReductionLog(local,real(nrm));
return nrm; return nrm;
} }