portelli/Grid
1
0
Fork 0
mirror of https://github.com/paboyle/Grid.git synced 2025-04-09 21:50:45 +01:00
Code Releases Activity

binary parallel IO can do read tests and eventually re-write in case of failure

Browse Source
This commit is contained in:
Antonin Portelli 2018-11-27 18:38:24 +00:00
parent a4c55406ed
commit 136d3802cb
2 changed files with 29 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

3
Grid/parallelIO/BinaryIO.cc Normal file
View File

@ -0,0 +1,3 @@
#include <Grid/GridCore.h>
int Grid::BinaryIO::latticeWriteMaxRetry = -1;

28
Grid/parallelIO/BinaryIO.h
View File

@ -81,6 +81,7 @@ inline void removeWhitespace(std::string &key)
/////////////////////////////////////////////////////////////////////////////////////////////////// ///////////////////////////////////////////////////////////////////////////////////////////////////
class BinaryIO { class BinaryIO {
public: public:
static int latticeWriteMaxRetry;
///////////////////////////////////////////////////////////////////////////// /////////////////////////////////////////////////////////////////////////////
// more byte manipulation helpers // more byte manipulation helpers
@ -583,6 +584,8 @@ PARALLEL_CRITICAL
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();
int attemptsLeft = std::max(0, BinaryIO::latticeWriteMaxRetry);
bool checkWrite = (BinaryIO::latticeWriteMaxRetry >= 0);
std::vector<sobj> scalardata(lsites); std::vector<sobj> scalardata(lsites);
std::vector<fobj> iodata(lsites); // Munge, checksum, byte order in here std::vector<fobj> iodata(lsites); // Munge, checksum, byte order in here
@ -598,8 +601,28 @@ PARALLEL_CRITICAL
grid->Barrier(); grid->Barrier();
timer.Stop(); timer.Stop();
IOobject(w,grid,iodata,file,offset,format,BINARYIO_WRITE|BINARYIO_LEXICOGRAPHIC, while (attemptsLeft >= 0)
nersc_csum,scidac_csuma,scidac_csumb); {
grid->Barrier();
IOobject(w,grid,iodata,file,offset,format,BINARYIO_WRITE|BINARYIO_LEXICOGRAPHIC,
nersc_csum,scidac_csuma,scidac_csumb);
if (checkWrite)
{
std::vector<fobj> ckiodata(lsites);
uint32_t cknersc_csum, ckscidac_csuma, ckscidac_csumb;
std::cout << GridLogMessage << "writeLatticeObject: read back object to check" << std::endl;
grid->Barrier();
IOobject(w,grid,ckiodata,file,offset,format,BINARYIO_READ|BINARYIO_LEXICOGRAPHIC,
cknersc_csum,ckscidac_csuma,ckscidac_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;
}
}
attemptsLeft--;
}
std::cout<<GridLogMessage<<"writeLatticeObject: unvectorize overhead "<<timer.Elapsed() <<std::endl; std::cout<<GridLogMessage<<"writeLatticeObject: unvectorize overhead "<<timer.Elapsed() <<std::endl;
} }
@ -725,5 +748,6 @@ PARALLEL_CRITICAL
std::cout << GridLogMessage << "RNG state overhead " << timer.Elapsed() << std::endl; std::cout << GridLogMessage << "RNG state overhead " << timer.Elapsed() << std::endl;
} }
}; };
} }
#endif #endif