# HG changeset patch
# User Merten Sach <msach@mailbox.tu-berlin.de>
# Date 1324396221 -3600
# Node ID c3f458403cd6441e3d4b1db1471990b78d9a525c
# Parent  bfaebdf60df3045cb7f9abc4ba9db49a8a349626
Always allocate more than 256 byte, measurements removed in master loop, aligned slots

diff -r bfaebdf60df3 -r c3f458403cd6 MasterLoop.c
--- a/MasterLoop.c	Tue Dec 20 15:39:30 2011 +0100
+++ b/MasterLoop.c	Tue Dec 20 16:50:21 2011 +0100
@@ -95,7 +95,18 @@
        uint64 numCycles;
       //==========================================================
    
-   //masterLoopStartPt:
+   masterEnv        = (MasterEnv*)_VMSMasterEnv;
+   
+   masterVP         = (VirtProcr*)volatilemasterVP;  //just to make sure after jmp
+   thisCoresIdx     = masterVP->coreAnimatedBy;
+   readyToAnimateQ  = masterEnv->readyToAnimateQs[thisCoresIdx];
+   schedSlots       = masterEnv->allSchedSlots[thisCoresIdx];
+
+   requestHandler   = masterEnv->requestHandler;
+   slaveScheduler   = masterEnv->slaveScheduler;
+   semanticEnv      = masterEnv->semanticEnv;
+   
+      //masterLoopStartPt:
    while(1){    //switch to core_loop and back to here is at end of loop
        
        //============================= MEASUREMENT STUFF =======================
@@ -107,22 +118,12 @@
        #endif
        //=======================================================================
 
-   masterEnv        = (MasterEnv*)_VMSMasterEnv;
-   
-      //GCC may optimize so doesn't always re-define from frame-storage
-   masterVP         = (VirtProcr*)volatilemasterVP;  //just to make sure after jmp
-   thisCoresIdx     = masterVP->coreAnimatedBy;
-   readyToAnimateQ  = masterEnv->readyToAnimateQs[thisCoresIdx];
-   schedSlots       = masterEnv->allSchedSlots[thisCoresIdx];
-
-   requestHandler   = masterEnv->requestHandler;
-   slaveScheduler   = masterEnv->slaveScheduler;
-   semanticEnv      = masterEnv->semanticEnv;
-
 
       //Poll each slot's Done flag
    numSlotsFilled = 0;
+   /*
          Meas_startMasterLoop
+    */
    for( slotIdx = 0; slotIdx < NUM_SCHED_SLOTS; slotIdx++)
     {
       currSlot = schedSlots[ slotIdx ];
@@ -133,9 +134,13 @@
          currSlot->needsProcrAssigned = TRUE;
 
             //process requests from slave to master
+         /*
                Meas_startReqHdlr
+          */
          (*requestHandler)( currSlot->procrAssignedToSlot, semanticEnv );
+         /*
                Meas_endReqHdlr
+          */
        }
       if( currSlot->needsProcrAssigned )
        {    //give slot a new virt procr
@@ -152,7 +157,9 @@
           }
        }
     }
+        /*
          Meas_endMasterLoop
+         */
    
    #ifdef USE_WORK_STEALING
       //If no slots filled, means no more work, look for work to steal.
diff -r bfaebdf60df3 -r c3f458403cd6 VMS.c
--- a/VMS.c	Tue Dec 20 15:39:30 2011 +0100
+++ b/VMS.c	Tue Dec 20 16:50:21 2011 +0100
@@ -187,11 +187,13 @@
  { SchedSlot  **schedSlots;
    int i;
 
-   schedSlots  = VMS__malloc( NUM_SCHED_SLOTS * sizeof(SchedSlot *) );
+   //schedSlots  = VMS__malloc( NUM_SCHED_SLOTS * sizeof(SchedSlot *) );
+   posix_memalign(&schedSlots, CACHELINE_SIZE, NUM_SCHED_SLOTS * sizeof(SchedSlot *));
 
    for( i = 0; i < NUM_SCHED_SLOTS; i++ )
     {
-      schedSlots[i] = VMS__malloc( sizeof(SchedSlot) );
+      //schedSlots[i] = VMS__malloc( sizeof(SchedSlot) );
+       posix_memalign(&schedSlots[i], CACHELINE_SIZE, sizeof(SchedSlot) );
 
          //Set state to mean "handling requests done, slot needs filling"
       schedSlots[i]->workIsDone         = FALSE;
@@ -590,8 +592,8 @@
       // itself
       //Note, should not stack-allocate initial data -- no guarantee, in
       // general that creating processor will outlive ones it creates.
-   //VMS__free( animatingPr->startOfStack );
-   //VMS__free( animatingPr );
+   free( animatingPr->startOfStack );
+   free( animatingPr );
  }
 
 
diff -r bfaebdf60df3 -r c3f458403cd6 vmalloc.c
--- a/vmalloc.c	Tue Dec 20 15:39:30 2011 +0100
+++ b/vmalloc.c	Tue Dec 20 16:50:21 2011 +0100
@@ -212,6 +212,8 @@
    #endif
    //========================================================================
    
+   sizeRequested += CACHELINE_SIZE; //Allocate more than cacheline_size to avoid false sharing
+   
    MallocArrays* freeLists = _VMSMasterEnv->freeLists;
    MallocProlog* foundChunk;