# HG changeset patch
# User Nina Engelhardt <nengel@mailbox.tu-berlin.de>
# Date 1334683137 -7200
# Node ID cfca88034000138013f425ad5e2866de7272e414
# Parent  bf6a4474f0e027a9bf40db1766db229dc6ee5891
Performance debug series: original code (with divider bug)

diff -r bf6a4474f0e0 -r cfca88034000 SSR_Matrix_Mult/Divide_Pr.c
--- a/SSR_Matrix_Mult/Divide_Pr.c	Tue Apr 17 19:32:08 2012 +0200
+++ b/SSR_Matrix_Mult/Divide_Pr.c	Tue Apr 17 19:18:57 2012 +0200
@@ -196,7 +196,7 @@
 
             DEBUG__printf(dbgAppFlow,"**create result Pr**")
       resultPr =
-         SSR__create_procr_with_affinity( &gatherResults, resultsParams, animPr,0);
+         SSR__create_procr_with( &gatherResults, resultsParams, animPr);
 
          //Make the sub-matrices, and pair them up, and make processor to
          // calc product of each pair.
@@ -261,7 +261,7 @@
    idealNumWorkUnits = SSR__giveIdealNumWorkUnits();
    
    idealSizeOfSide2 = leftMatrix->numRows / rint(cbrt( idealNumWorkUnits ));
-   idealSizeOfSide2 *= 0.4; //finer granularity to help load balance
+   idealSizeOfSide2 *= 0.6; //finer granularity to help load balance
 
    if( idealSizeOfSide1 > idealSizeOfSide2 )
       idealSizeOfSide = idealSizeOfSide1;
@@ -367,10 +367,10 @@
 
    numCores = SSR__give_number_of_cores_to_schedule_onto();
 
-   numToPutOntoEachCore = numRowIdxs*numColIdxs/(numCores-1);
+   numToPutOntoEachCore = numRowIdxs*numColIdxs/numCores;
    leftOverFraction = 0;
    numVecOnCurrCore = 0;
-   coreToAssignOnto = 1;
+   coreToAssignOnto = 0;
 
    for( resRowIdx = 0; resRowIdx < numRowIdxs; resRowIdx++ )
     {
@@ -402,7 +402,7 @@
                // that each core gets the same number of vectors, with a max
                // imbalance of 1 vector more on some cores than others
             numVecOnCurrCore += 1;                 //incr before checking, so
-            if( numVecOnCurrCore > numToPutOntoEachCore ) //actual num 1 less
+         }   if( numVecOnCurrCore > numToPutOntoEachCore ) //actual num 1 less
              {
                   //deal with fractional part, to ensure that imbalance is 1 max
                   // IE, core with most has only 1 more than core with least
@@ -412,9 +412,9 @@
 
                   //Move to next core, max core-value to incr to is numCores -1
                coreToAssignOnto += 1;
-               if( coreToAssignOnto >= numCores ) coreToAssignOnto = 1;
+               if( coreToAssignOnto >= numCores ) coreToAssignOnto = 0;
              } //if
-          } //for( vecIdx
+           //for( vecIdx
        } //for( resColIdx
     } //for( resRowIdx