portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2024-11-14 01:35:36 +00:00
Code Releases Activity

resilient I/O fix

Browse Source
This commit is contained in:
Antonin Portelli 2018-11-27 20:17:09 +00:00
parent 01e8cf5017
commit 49b934310b
1 changed files with 12 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

18
Grid/parallelIO/BinaryIO.h
View File

@ -371,7 +371,7 @@ PARALLEL_CRITICAL
#endif #endif
} else { } else {
std::cout << GridLogMessage <<"IOobject: C++ read I/O " << file << " : " std::cout << GridLogMessage <<"IOobject: C++ read I/O " << file << " : "
<< iodata.size() * sizeof(fobj) << " bytes" << std::endl; << iodata.size() * sizeof(fobj) << " bytes and offset " << offset << std::endl;
std::ifstream fin; std::ifstream fin;
fin.open(file, std::ios::binary | std::ios::in); fin.open(file, std::ios::binary | std::ios::in);
if (control & BINARYIO_MASTER_APPEND) if (control & BINARYIO_MASTER_APPEND)
@ -583,7 +583,7 @@ PARALLEL_CRITICAL
typedef typename vobj::scalar_object sobj; typedef typename vobj::scalar_object sobj;
typedef typename vobj::Realified::scalar_type word; word w=0; typedef typename vobj::Realified::scalar_type word; word w=0;
GridBase *grid = Umu._grid; GridBase *grid = Umu._grid;
uint64_t lsites = grid->lSites(); uint64_t lsites = grid->lSites(), offsetCopy = offset;
int attemptsLeft = std::max(0, BinaryIO::latticeWriteMaxRetry); int attemptsLeft = std::max(0, BinaryIO::latticeWriteMaxRetry);
bool checkWrite = (BinaryIO::latticeWriteMaxRetry >= 0); bool checkWrite = (BinaryIO::latticeWriteMaxRetry >= 0);
@ -600,7 +600,6 @@ PARALLEL_CRITICAL
grid->Barrier(); grid->Barrier();
timer.Stop(); timer.Stop();
while (attemptsLeft >= 0) while (attemptsLeft >= 0)
{ {
grid->Barrier(); grid->Barrier();
@ -610,14 +609,21 @@ PARALLEL_CRITICAL
{ {
std::vector<fobj> ckiodata(lsites); std::vector<fobj> ckiodata(lsites);
uint32_t cknersc_csum, ckscidac_csuma, ckscidac_csumb; uint32_t cknersc_csum, ckscidac_csuma, ckscidac_csumb;
uint64_t ckoffset = offsetCopy;
std::cout << GridLogMessage << "writeLatticeObject: read back object to check" << std::endl; std::cout << GridLogMessage << "writeLatticeObject: read back object" << std::endl;
grid->Barrier(); grid->Barrier();
IOobject(w,grid,ckiodata,file,offset,format,BINARYIO_READ|BINARYIO_LEXICOGRAPHIC, IOobject(w,grid,ckiodata,file,ckoffset,format,BINARYIO_READ|BINARYIO_LEXICOGRAPHIC,
cknersc_csum,ckscidac_csuma,ckscidac_csumb); cknersc_csum,ckscidac_csuma,ckscidac_csumb);
if ((cknersc_csum != nersc_csum) or (ckscidac_csuma != scidac_csuma) or (ckscidac_csumb != scidac_csumb)) if ((cknersc_csum != nersc_csum) or (ckscidac_csuma != scidac_csuma) or (ckscidac_csumb != scidac_csumb))
{ {
std::cout << GridLogMessage << "writeLatticeObject: checksum failure in test read (" << attemptsLeft << " write attempt(s) remaining)" << std::endl; std::cout << GridLogMessage << "writeLatticeObject: read test checksum failure, re-writing (" << attemptsLeft << " attempt(s) remaining)" << std::endl;
offset = offsetCopy;
}
else
{
std::cout << GridLogMessage << "writeLatticeObject: read test checksum correct" << std::endl;
break;
} }
} }
attemptsLeft--; attemptsLeft--;