1.1 --- a/Defines/VMS_defs__HW_constants.h	Fri Mar 16 09:39:53 2012 -0700
     1.2 +++ b/Defines/VMS_defs__HW_constants.h	Fri Mar 16 10:43:15 2012 -0700
     1.3 @@ -18,7 +18,7 @@
     1.4  
     1.5     // tradeoff amortizing master fixed overhead vs imbalance potential
     1.6     // when work-stealing, can make bigger, at risk of losing cache affinity
     1.7 -#define NUM_SCHED_SLOTS  3
     1.8 +#define NUM_SCHED_SLOTS  1
     1.9  
    1.10     //These are for backoff inside core-loop, which reduces lock contention
    1.11  #define NUM_REPS_W_NO_WORK_BEFORE_YIELD      10

     2.1 --- a/Defines/VMS_defs__turn_on_and_off.h	Fri Mar 16 09:39:53 2012 -0700
     2.2 +++ b/Defines/VMS_defs__turn_on_and_off.h	Fri Mar 16 10:43:15 2012 -0700
     2.3 @@ -21,7 +21,7 @@
     2.4  /*turns on the probe-instrumentation in the application -- when not
     2.5   * defined, the calls to the probe functions turn into comments
     2.6   */
     2.7 -#define DEBUG__TURN_ON_DEBUG_PRINT
     2.8 +//#define DEBUG__TURN_ON_DEBUG_PRINT
     2.9  
    2.10  /*These defines turn types of bug messages on and off
    2.11   */
    2.12 @@ -60,6 +60,9 @@
    2.13     /*turn on/off subtraction of create measurements from plugin meas*/
    2.14  //#define MEAS__TURN_ON_EXCLUDE_CREATION_TIME 
    2.15  
    2.16 +#define HOLISTIC__TURN_ON_PERF_COUNTERS
    2.17 +#define HOLISTIC__TURN_ON_OBSERVE_UCC
    2.18 +#define DETECT_LOOP_GRAPH
    2.19  
    2.20  //===================  Turn on or off system options  =======================
    2.21  //

     3.1 --- a/Hardware_Dependent/VMS__HW_measurement.h	Fri Mar 16 09:39:53 2012 -0700
     3.2 +++ b/Hardware_Dependent/VMS__HW_measurement.h	Fri Mar 16 10:43:15 2012 -0700
     3.3 @@ -57,6 +57,6 @@
     3.4     // different cores.
     3.5  //#define NUM_TSC_ROUND_TRIPS 10
     3.6  
     3.7 -
     3.8 +void setup_perf_counters();
     3.9  #endif	/* */
    3.10  

     4.1 --- a/Hardware_Dependent/VMS__primitives_asm.s	Fri Mar 16 09:39:53 2012 -0700
     4.2 +++ b/Hardware_Dependent/VMS__primitives_asm.s	Fri Mar 16 10:43:15 2012 -0700
     4.3 @@ -100,7 +100,7 @@
     4.4      movq    $_VMSMasterEnv, %rcx
     4.5      movq        (%rcx), %rcx         #_VMSMasterEnv is pointer to struct
     4.6      movq    0x00(%rcx), %rax         #get CoreCtlr return pt
     4.7 -    movl    $0x0      , 0x08(%rcx)   #release lock
     4.8 +    movl    $0x0      , 0x100(%rcx)   #release lock
     4.9      jmp     *%rax                    #jmp to CoreCtlr
    4.10  MasterReturn:
    4.11      ret

     5.1 --- a/Services_Offered_by_VMS/Measurement_and_Stats/MEAS__macros.h	Fri Mar 16 09:39:53 2012 -0700
     5.2 +++ b/Services_Offered_by_VMS/Measurement_and_Stats/MEAS__macros.h	Fri Mar 16 10:43:15 2012 -0700
     5.3 @@ -292,12 +292,57 @@
     5.4     #define MEAS__Print_Hists_for_System_Meas 
     5.5  #endif
     5.6  
     5.7 + #ifdef HOLISTIC__TURN_ON_PERF_COUNTERS
     5.8 +   
     5.9 +#define MEAS__Insert_Counter_Handler \
    5.10 +   typedef void (*CounterHandler) (int,int,int,SlaveVP*,uint64,uint64);
    5.11 + 
    5.12 +   enum eventType {
    5.13 +    DebugEvt = 0,
    5.14 +    AppResponderInvocation_start,
    5.15 +    AppResponder_start,
    5.16 +    AppResponder_end,
    5.17 +    AssignerInvocation_start,
    5.18 +    NextAssigner_start,
    5.19 +    Assigner_start,
    5.20 +    Assigner_end,
    5.21 +    Work_start,
    5.22 +    Work_end,
    5.23 +    HwResponderInvocation_start,
    5.24 +    Timestamp_start,
    5.25 +    Timestamp_end
    5.26 +   };
    5.27 +   
    5.28 +
    5.29 +   #define MEAS__Insert_Counter_Meas_Fields_into_MasterEnv \
    5.30 +     int cycles_counter_fd[NUM_CORES]; \
    5.31 +     int instrs_counter_fd[NUM_CORES]; \
    5.32 +     uint64 start_master_lock[NUM_CORES][2]; \
    5.33 +     CounterHandler counterHandler;
    5.34 +
    5.35 +   #define HOLISTIC__Setup_Perf_Counters void setup_perf_counters();
    5.36 +   
    5.37 +   #define HOLISTIC__Start_Perf_Counters prctl(PR_TASK_PERF_EVENTS_ENABLE);
    5.38 +
    5.39 +   #define HOLISTIC__Record_HwResponderInvocation_start \
    5.40 +        uint64 cycles,instrs; \
    5.41 +        saveCyclesAndInstrs(animatingPr->coreAnimatedBy,cycles, instrs); \
    5.42 +        (*(_VMSMasterEnv->counterHandler))(HwResponderInvocation_start,animatingPr->procrID,animatingPr->numTimesScheduled,animatingPr,cycles,instrs); 
    5.43 +
    5.44 +   
    5.45 +   
    5.46 +#else  
    5.47 +   #define MEAS__Insert_Counter_Handler
    5.48 +   #define MEAS__Insert_Counter_Meas_Fields_into_MasterEnv
    5.49 +   #define HOLISTIC__Setup_Perf_Counters
    5.50 +   #define HOLISTIC__Start_Perf_Counters
    5.51 +#endif
    5.52  
    5.53  //Experiment in two-step macros -- if doesn't work, insert each separately
    5.54  #define MEAS__Insert_Meas_Fields_into_Slave  \
    5.55     MEAS__Insert_Susp_Meas_Fields_into_Slave \
    5.56     MEAS__Insert_Master_Meas_Fields_into_Slave \
    5.57 -   MEAS__Insert_System_Meas_Fields_into_Slave
    5.58 +   MEAS__Insert_System_Meas_Fields_into_Slave 
    5.59  
    5.60  
    5.61  //======================  Histogram Macros -- Create ========================

     6.1 --- a/VMS.h	Fri Mar 16 09:39:53 2012 -0700
     6.2 +++ b/VMS.h	Fri Mar 16 10:43:15 2012 -0700
     6.3 @@ -43,6 +43,9 @@
     6.4  typedef void     (*TopLevelFnPtr)  ( void *, SlaveVP * ); //initData, animSlv
     6.5  typedef void       TopLevelFn      ( void *, SlaveVP * ); //initData, animSlv
     6.6  typedef void     (*ResumeSlvFnPtr) ( SlaveVP *, void * );
     6.7 +      //=========== MEASUREMENT STUFF ==========
     6.8 +        MEAS__Insert_Counter_Handler
     6.9 +      //========================================
    6.10  
    6.11  //============================ HW Dependent Fns ================================
    6.12  
    6.13 @@ -105,6 +108,12 @@
    6.14   };
    6.15  //SchedSlot
    6.16  
    6.17 + enum VPtype {
    6.18 +     Slave = 1, //default
    6.19 +     Master,
    6.20 +     Shutdown
    6.21 + };
    6.22 + 
    6.23  /*This structure embodies the state of a slaveVP.  It is reused for masterVP
    6.24   * and shutdownVPs.
    6.25   */
    6.26 @@ -133,7 +142,10 @@
    6.27         MEAS__Insert_Meas_Fields_into_Slave;
    6.28        //========================================
    6.29     
    6.30 -   float64      createPtInSecs;  //have space but don't use on some configs
    6.31 +   enum VPtype type;
    6.32 +   int         numTimesAssigned;
    6.33 +       
    6.34 +   float64     createPtInSecs;  //have space but don't use on some configs
    6.35   };
    6.36  //SlaveVP
    6.37  
    6.38 @@ -181,6 +193,7 @@
    6.39         MEAS__Insert_Malloc_Meas_Fields_into_MasterEnv;
    6.40         MEAS__Insert_Plugin_Meas_Fields_into_MasterEnv;
    6.41         MEAS__Insert_System_Meas_Fields_into_MasterEnv;
    6.42 +       MEAS__Insert_Counter_Meas_Fields_into_MasterEnv;
    6.43        //==========================================
    6.44   }
    6.45  MasterEnv;
    6.46 @@ -246,6 +259,9 @@
    6.47  void
    6.48  VMS_SS__start_the_work_then_wait_until_done();
    6.49  
    6.50 +SlaveVP* 
    6.51 +VMS_SS__create_shutdown_slave();
    6.52 +
    6.53  void
    6.54  VMS_SS__shutdown();
    6.55  

     7.1 --- a/VMS__int.c	Fri Mar 16 09:39:53 2012 -0700
     7.2 +++ b/VMS__int.c	Fri Mar 16 10:43:15 2012 -0700
     7.3 @@ -175,6 +175,8 @@
     7.4     newSlv->slaveID      = _VMSMasterEnv->numSlavesCreated++;
     7.5     newSlv->requests     = NULL;
     7.6     newSlv->schedSlotAssignedTo    = NULL;
     7.7 +   newSlv->type         = Slave;
     7.8 +   newSlv->numTimesAssigned = 0;
     7.9  
    7.10     VMS_int__point_slaveVP_to_Fn( newSlv, fnPtr, dataParam );
    7.11             

     8.1 --- a/VMS__startup_and_shutdown.c	Fri Mar 16 09:39:53 2012 -0700
     8.2 +++ b/VMS__startup_and_shutdown.c	Fri Mar 16 10:43:15 2012 -0700
     8.3 @@ -82,6 +82,7 @@
     8.4        printf( "\n\n Running in SEQUENTIAL mode \n\n" );
     8.5     #else
     8.6        create_masterEnv();
     8.7 +      printf("Offset of lock in masterEnv:%d\n",offsetof(MasterEnv,masterLock));
     8.8        create_the_coreCtlr_OS_threads();
     8.9     #endif
    8.10   }
    8.11 @@ -305,6 +306,7 @@
    8.12           //Q: should give masterVP core-specific info as its init data?
    8.13        masterVPs[ coreIdx ] = VMS_int__create_slaveVP( (TopLevelFnPtr)&animationMaster, (void*)masterEnv );
    8.14        masterVPs[ coreIdx ]->coreAnimatedBy = coreIdx;
    8.15 +      masterVPs[ coreIdx ]->type = Master;
    8.16        allSchedSlots[ coreIdx ] = create_sched_slots(); //makes for one core
    8.17      }
    8.18     _VMSMasterEnv->masterVPs        = masterVPs;
    8.19 @@ -323,6 +325,8 @@
    8.20  
    8.21           PROBES__Create_Probe_Bookkeeping_Vars;
    8.22           
    8.23 +         HOLISTIC__Setup_Perf_Counters;
    8.24 +         
    8.25     //========================================================================
    8.26   }
    8.27  
    8.28 @@ -448,6 +452,15 @@
    8.29   }
    8.30  
    8.31  
    8.32 +SlaveVP* VMS_SS__create_shutdown_slave(){
    8.33 +    SlaveVP* shutdownVP;
    8.34 +    
    8.35 +    shutdownVP = VMS_int__create_slaveVP( &endOSThreadFn, NULL );
    8.36 +    shutdownVP->type = Shutdown;
    8.37 +    
    8.38 +    return shutdownVP;
    8.39 +}
    8.40 +
    8.41  //TODO: look at architecting cleanest separation between request handler
    8.42  // and animation master, for dissipate, create, shutdown, and other non-semantic
    8.43  // requests.  Issue is chain: one removes requests from AppSlv, one dispatches
    8.44 @@ -489,7 +502,7 @@
    8.45        // directly into the Q -- each core will die when gets one
    8.46     for( coreIdx = 0; coreIdx < NUM_CORES; coreIdx++ )
    8.47      {    //Note, this is running in the master
    8.48 -      shutDownSlv = VMS_int__create_slaveVP( &endOSThreadFn, NULL );
    8.49 +      shutDownSlv = VMS_SS__create_shutdown_slave();
    8.50           //last slave has dissipated, so no more in slots, so write
    8.51           // shut down slave into first schedulng slot.
    8.52        schedSlots = _VMSMasterEnv->allSchedSlots[ coreIdx ];