Mercurial > cgi-bin > hgwebdir.cgi > PR > Applications > SSR > SSR__Blocked_Matrix_Mult__Bench
changeset 29:8972c00c00dd
Merge with perf_tuning_paper branch, which grabs fixes and best performing version
| author | Sean Halle <seanhalle@yahoo.com> |
|---|---|
| date | Sun, 15 Jul 2012 01:27:39 -0700 |
| parents | 233fe8a5208f c35cb1f48f89 |
| children | efc1ee85f1ee |
| files | __brch__default |
| diffstat | 4 files changed, 5 insertions(+), 4 deletions(-) [+] |
line diff
1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000 1.2 +++ b/.hgtags Sun Jul 15 01:27:39 2012 -0700 1.3 @@ -0,0 +1,1 @@ 1.4 +b7d8cf6680a9b615e41df8305cce48195d03c67e portable version
2.1 --- a/SSR_Matrix_Mult/Divide_Pr.c Mon Apr 16 18:27:12 2012 +0200 2.2 +++ b/SSR_Matrix_Mult/Divide_Pr.c Sun Jul 15 01:27:39 2012 -0700 2.3 @@ -261,7 +261,7 @@ 2.4 idealNumWorkUnits = SSR__giveIdealNumWorkUnits(); 2.5 2.6 idealSizeOfSide2 = leftMatrix->numRows / rint(cbrt( idealNumWorkUnits )); 2.7 - idealSizeOfSide2 *= 0.4; //finer granularity to help load balance 2.8 + idealSizeOfSide2 *= 0.5; //finer granularity to help load balance 2.9 2.10 if( idealSizeOfSide1 > idealSizeOfSide2 ) 2.11 idealSizeOfSide = idealSizeOfSide1; 2.12 @@ -367,7 +367,7 @@ 2.13 2.14 numCores = SSR__give_number_of_cores_to_schedule_onto(); 2.15 2.16 - numToPutOntoEachCore = numRowIdxs*numColIdxs/(numCores-1); 2.17 + numToPutOntoEachCore = numRowIdxs*numColIdxs/numCores; 2.18 leftOverFraction = 0; 2.19 numVecOnCurrCore = 0; 2.20 coreToAssignOnto = 1; 2.21 @@ -412,7 +412,7 @@ 2.22 2.23 //Move to next core, max core-value to incr to is numCores -1 2.24 coreToAssignOnto += 1; 2.25 - if( coreToAssignOnto >= numCores ) coreToAssignOnto = 1; 2.26 + if( coreToAssignOnto >= numCores ) coreToAssignOnto = 0; 2.27 } //if 2.28 } //for( vecIdx 2.29 } //for( resColIdx
