1.1 --- a/DESIGN_NOTES.txt	Tue Mar 13 10:04:14 2012 -0700
     1.2 +++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
     1.3 @@ -1,212 +0,0 @@
     1.4 -
     1.5 -From e-mail to Albert, on design of app-virt-procr to core-loop animation
     1.6 -switch and back.
     1.7 -
     1.8 -====================
     1.9 -General warnings about this code:
    1.10 -It only compiles in GCC 4.x  (label addr and computed goto)
    1.11 -Has assembly for x86  32bit
    1.12 -
    1.13 -
    1.14 -====================
    1.15 -AVProcr data-struc has: stack-ptr, jump-ptr, data-ptr, slotNum, coreloop-ptr
    1.16 - and semantic-custom-ptr
    1.17 -
    1.18 -The VMS Creator: takes ptr to function and ptr to initial data
    1.19 --- creates a new AVProcr struc
    1.20 --- sets the jmp-ptr field to the ptr-to-function passed in
    1.21 --- sets the data-ptr to ptr to initial data passed in
    1.22 --- if this is for a suspendable virt  processor, then create a stack and set
    1.23 -   the stack-ptr
    1.24 -
    1.25 -VMS_int__create_slaveVP( AVProcrFnPtr fnPtr, void *initialData )
    1.26 -{
    1.27 -AVProcr   newSlv = malloc( sizeof(AVProcr) );
    1.28 -newSlv->jmpPtr = fnPtr;
    1.29 -newSlv->coreCtlrDonePt = &CoreCtlrDonePt; //label is in coreCtlr
    1.30 -newSlv->data = initialData;
    1.31 -newSlv->stackPtr = createNewStack();
    1.32 -return newSlv;
    1.33 -}
    1.34 -
    1.35 -The semantic layer can then add its own state in the cusom-ptr field
    1.36 -
    1.37 -The Scheduler plug-in:
    1.38 --- Sets slave-ptr in AVProcr, and points the slave to AVProcr
    1.39 --- if non-suspendable, sets the AVProcr's stack-ptr to the slave's stack-ptr
    1.40 -
    1.41 -MasterLoop:
    1.42 --- puts AVProcr structures onto the workQ
    1.43 -
    1.44 -CoreCtlr:
    1.45 --- gets stack-ptr out of AVProcr and sets the core's stack-ptr to that
    1.46 --- gets data-ptr out of AVProcr and puts it into reg GCC uses for that param
    1.47 --- puts AVProcr's addr into reg GCC uses for the AVProcr-pointer param
    1.48 --- jumps to the addr in AVProcr's jmp-ptr field
    1.49 -CoreCtlr()
    1.50 -{ while( FOREVER )
    1.51 - { nextSlv = readQ( workQ );  //workQ is static (global) var declared volatile
    1.52 -   <dataPtr-param-register>       = nextSlv->data;
    1.53 -   <AVProcrPtr-param-register> = nextSlv;
    1.54 -   <stack-pointer register>          = nextSlv->stackPtr;
    1.55 -   jmp nextSlv->jmpPtr;
    1.56 -CoreCtlrDonePt:   //label's addr put into AVProcr when create new one
    1.57 - }
    1.58 -}
    1.59 -(Note, for suspendable processors coming back from suspension, there is no
    1.60 - need to fill the parameter registers -- they will be discarded)
    1.61 -
    1.62 -Suspend an application-level virtual processor:
    1.63 -VMS_int__AVPSuspend( AVProcr *pr )
    1.64 -{
    1.65 -pr->jmpPtr = &ResumePt;  //label defined a few lines below
    1.66 -pr->slave->doneFlag = TRUE;
    1.67 -pr->stackPtr = <current SP reg value>;
    1.68 -jmp pr->coreCtlrDonePt;
    1.69 -ResumePt: return;
    1.70 -}
    1.71 -
    1.72 -This works because the core controller will have switched back to this stack
    1.73 - before jumping to ResumePt..    also, the core controller never modifies the
    1.74 - stack pointer, it simply switches to whatever stack pointer is in the
    1.75 - next AVProcr it gets off the workQ.
    1.76 -
    1.77 -
    1.78 -
    1.79 -=============================================================================
    1.80 -As it is now, there's only one major unknown about GCC (first thing below
    1.81 -  the line),  and there are a few restrictions, the most intrusive being
    1.82 -  that the functions the application gives to the semantic layer have a
    1.83 -  pre-defined prototype -- return nothing, take a pointer to initial data
    1.84 -  and a pointer to an AVProcr struc, which they're not allowed to modify
    1.85 -  -- only pass it to semantic-lib calls.
    1.86 -
    1.87 -So, here are the assumptions, restrictions, and so forth:
    1.88 -===========================
    1.89 -Major assumption:  that GCC will do the following the same way every time:
    1.90 -  say the application defines a function that fits this typedef:
    1.91 -typedef void (*AVProcrFnPtr)  ( void *, AVProcr * );
    1.92 -
    1.93 -and let's say somewhere in the code they do this:
    1.94 -AVProcrFnPtr   fnPtr = &someFunc;
    1.95 -
    1.96 -then they do this:
    1.97 -(*fnPtr)( dataPtr, animatingSlaveVPPtr );
    1.98 -
    1.99 -Can the registers that GCC uses to pass the two pointers be predicted?
   1.100 - Will they always be the same registers, in every program that has the
   1.101 - same typedef?
   1.102 -If that typedef fixes, guaranteed, the registers (on x86) that GCC will use
   1.103 - to send the two pointers, then the rest of this solution works.
   1.104 -
   1.105 -Change in model: Instead of a virtual processor whose execution trace is
   1.106 - divided into work-units, replacing that with the pattern that a virtual
   1.107 - processor is suspended.  Which means, no more "work unit" data structure
   1.108 - -- instead, it's now an "Application Virtual Processor" structure
   1.109 - -- AVProcr -- which is given directly to the application function!
   1.110 -
   1.111 -   -- You were right, don't need slaves to be virtual processors, only need
   1.112 -      "scheduling buckets" -- just a way to keep track of things..
   1.113 -
   1.114 -Restrictions:
   1.115 --- the  "virtual entities"  created by the semantic layer must be virtual
   1.116 -   processors, created with a function-to-execute and initial data -- the
   1.117 -   function is restricted to return nothing and only take a pointer to the
   1.118 -   initial data plus a pointer to an AVProcr structure, which represents
   1.119 -   "self", the virtual processor created.  (This is the interface I showed
   1.120 -   you for "Hello World" semantic layer).
   1.121 -What this means for synchronous dataflow, is that the nodes in the graph
   1.122 -  are virtual processors that in turn spawn a new virtual processor for
   1.123 -  every "firing" of the node.  This should be fine because the function
   1.124 -  that the node itself is created with is a "canned" function that is part
   1.125 -  of the semantic layer -- the function that is spawned is the user-provided
   1.126 -  function.  The restriction only means that the values from the inputs to
   1.127 -  the node are packaged as the "initial data" given to the spawned virtual
   1.128 -  processor -- so the user-function has to cast a void * to the
   1.129 -  semantic-layer-defined structure by which it gets the inputs to the node.
   1.130 -
   1.131 --- Second restriction is that the semantic layer has to use VMS supplied
   1.132 -   stuff -- for example, the data structure that represents the
   1.133 -   application-level virtual processor is defined in VMS, and the semantic
   1.134 -   layer has to call a VMS function in order to suspend a virtual processor.
   1.135 -
   1.136 --- Third restriction is that the application code never do anything with
   1.137 -   the AVProcr structure except pass it to semantic-layer lib calls.
   1.138 -
   1.139 --- Fourth restriction is that every virtual processor must call a
   1.140 -   "dissipate" function as its last act -- the user-supplied
   1.141 -   virtual-processor function can't just end -- it has to call
   1.142 -   SemLib__dissipate( AVProcr ) before the closing brace.. and after the
   1.143 -   semantic layer is done cleaning up its own data, it has to in turn call
   1.144 -   VMS_int__disspate( AVProcr ).
   1.145 -
   1.146 --- For performance reasons, I think I want to have two different kinds of
   1.147 -   app-virtual processor -- suspendable ones and non-suspendable -- where
   1.148 -   non-suspendable are not allowed to perform any communication with other
   1.149 -   virtual processors, except at birth and death.  Suspendable ones, of
   1.150 -   course can perform communications, create other processors, and so forth
   1.151 -   -- all of which cause it to suspend.
   1.152 -The performance difference is that I need a separate stack for each
   1.153 -  suspendable, but non-suspendable can re-use a fixed number of stacks
   1.154 -  (one for each slave).
   1.155 -
   1.156 -
   1.157 -==================== May 29
   1.158 -
   1.159 -Qs:
   1.160 ---1 how to safely jump between virt processor's trace and coreloop
   1.161 ---2 how to set up __cdecl style stack + frame for just-born virtual processor
   1.162 ---3 how to switch stack-pointers + frame-pointers
   1.163 -
   1.164 -
   1.165 ---1:
   1.166 -Not sure if GCC's computed goto is safe, because modify the stack pointer
   1.167 -without GCC's knowledge -- although, don't use the stack in the coreloop
   1.168 -segment, so, actually, that should be safe!
   1.169 -
   1.170 -So, GCC has its own special C extensions, one of which gets address of label:
   1.171 -
   1.172 -void *labelAddr;
   1.173 -labelAddr = &&label;
   1.174 -goto *labelAddr;
   1.175 -
   1.176 ---2
   1.177 -In CoreCtlr, will check whether VirtProc just born, or was suspended.
   1.178 -If just born, do bit of code that sets up the virtual processor's stack
   1.179 -and frame according to the __cdecl convention for the standard virt proc
   1.180 -fn typedef -- save the pointer to data and pointer to virt proc struc into
   1.181 -correct places in the frame
   1.182 -   __cdecl says, according to:
   1.183 -http://unixwiz.net/techtips/win32-callconv-asm.html
   1.184 -To do this:
   1.185 -push the parameters onto the stack, right most first, working backwards to
   1.186 - the left.
   1.187 -Then perform call instr, which pushes return addr onto stack.
   1.188 -Then callee first pushes the frame pointer, %EBP followed by placing the
   1.189 -then-current value of stack pointer into %EBP
   1.190 -push ebp
   1.191 -mov  ebp, esp    // ebp « esp
   1.192 -
   1.193 -Once %ebp has been changed, it can now refer directly to the function's
   1.194 - arguments as 8(%ebp), 12(%ebp). Note that 0(%ebp) is the old base pointer
   1.195 - and 4(%ebp) is the old instruction pointer.
   1.196 -
   1.197 -Then callee pushes regs it will use then adds to stack pointer the size of
   1.198 - its local vars.
   1.199 -
   1.200 -Stack in callee looks like this:
   1.201 -16(%ebp)	 - third function parameter
   1.202 -12(%ebp)	 - second function parameter
   1.203 -8(%ebp)	 - first function parameter
   1.204 -4(%ebp)	 - old %EIP (the function's "return address")
   1.205 -----------^^ State seen at first instr of callee ^^-----------
   1.206 -0(%ebp)	- old %EBP (previous function's base pointer)
   1.207 --4(%ebp)	 - save of EAX, the only reg used in function
   1.208 --8(%ebp)	 - first local variable
   1.209 --12(%ebp)	 - second local variable
   1.210 --16(%ebp)	 - third local variable
   1.211 -
   1.212 -
   1.213 ---3
   1.214 -It might be just as simple as two mov instrs, one for %ESP, one for %EBP..
   1.215 - the stack and frame pointer regs

     2.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     2.2 +++ b/SSR.c	Tue Mar 13 18:30:05 2012 -0700
     2.3 @@ -0,0 +1,785 @@
     2.4 +/*
     2.5 + * Copyright 2010  OpenSourceCodeStewardshipFoundation
     2.6 + *
     2.7 + * Licensed under BSD
     2.8 + */
     2.9 +
    2.10 +#include <stdio.h>
    2.11 +#include <stdlib.h>
    2.12 +#include <malloc.h>
    2.13 +
    2.14 +#include "Queue_impl/PrivateQueue.h"
    2.15 +#include "Hash_impl/PrivateHash.h"
    2.16 +
    2.17 +#include "SSR.h"
    2.18 +#include "SSR_Counter_Recording.h"
    2.19 +
    2.20 +//==========================================================================
    2.21 +
    2.22 +void
    2.23 +SSR__init();
    2.24 +
    2.25 +void
    2.26 +SSR__init_Helper();
    2.27 +//==========================================================================
    2.28 +
    2.29 +
    2.30 +/*TODO: Q: dealing with library f()s and DKU vs WT vs FoR
    2.31 + * (still want to do FoR, with time-lines as syntax, could be super cool)
    2.32 + * A: thinking pin the coreCtlrs for all of BLIS -- let Master arbitrate
    2.33 + * among library, DKU, WT, FoR -- all the patterns in terms of virtual
    2.34 + * processors (or equivalently work-units), so Master picks which virt procr
    2.35 + * from which portions of app (DKU, WT, FoR) onto which sched slots
    2.36 + *Might even do hierarchy of masters -- group of sched slots for each core
    2.37 + * has its own master, that keeps generated work local
    2.38 + * single-reader-single-writer sync everywhere -- no atomic primitives
    2.39 + * Might have the different assigners talk to each other, to negotiate
    2.40 + * larger-grain sharing of resources, according to predicted critical
    2.41 + * path, and expansion of work
    2.42 + */
    2.43 +
    2.44 +
    2.45 +
    2.46 +//===========================================================================
    2.47 +
    2.48 +
    2.49 +/*These are the library functions *called in the application*
    2.50 + * 
    2.51 + *There's a pattern for the outside sequential code to interact with the
    2.52 + * VMS_HW code.
    2.53 + *The VMS_HW system is inside a boundary..  every SSR system is in its
    2.54 + * own directory that contains the functions for each of the processor types.
    2.55 + * One of the processor types is the "seed" processor that starts the
    2.56 + * cascade of creating all the processors that do the work.
    2.57 + *So, in the directory is a file called "EntryPoint.c" that contains the
    2.58 + * function, named appropriately to the work performed, that the outside
    2.59 + * sequential code calls.  This function follows a pattern:
    2.60 + *1) it calls SSR__init()
    2.61 + *2) it creates the initial data for the seed processor, which is passed
    2.62 + *    in to the function
    2.63 + *3) it creates the seed SSR processor, with the data to start it with.
    2.64 + *4) it calls startSSRThenWaitUntilWorkDone
    2.65 + *5) it gets the returnValue from the transfer struc and returns that
    2.66 + *    from the function
    2.67 + *
    2.68 + *For now, a new SSR system has to be created via SSR__init every
    2.69 + * time an entry point function is called -- later, might add letting the
    2.70 + * SSR system be created once, and let all the entry points just reuse
    2.71 + * it -- want to be as simple as possible now, and see by using what makes
    2.72 + * sense for later..
    2.73 + */
    2.74 +
    2.75 +
    2.76 +
    2.77 +//===========================================================================
    2.78 +
    2.79 +/*This is the "border crossing" function -- the thing that crosses from the
    2.80 + * outside world, into the VMS_HW world.  It initializes and starts up the
    2.81 + * VMS system, then creates one processor from the specified function and
    2.82 + * puts it into the readyQ.  From that point, that one function is resp.
    2.83 + * for creating all the other processors, that then create others, and so
    2.84 + * forth.
    2.85 + *When all the processors, including the seed, have dissipated, then this
    2.86 + * function returns.  The results will have been written by side-effect via
    2.87 + * pointers read from, or written into initData.
    2.88 + *
    2.89 + *NOTE: no Threads should exist in the outside program that might touch
    2.90 + * any of the data reachable from initData passed in to here
    2.91 + */
    2.92 +void
    2.93 +SSR__create_seed_procr_and_do_work( TopLevelFnPtr fnPtr, void *initData )
    2.94 + { SSRSemEnv *semEnv;
    2.95 +   SlaveVP *seedPr;
    2.96 +
    2.97 +   SSR__init();      //normal multi-thd
    2.98 +   
    2.99 +   semEnv = _VMSMasterEnv->semanticEnv;
   2.100 +
   2.101 +      //SSR starts with one processor, which is put into initial environ,
   2.102 +      // and which then calls create() to create more, thereby expanding work
   2.103 +   seedPr = SSR__create_procr_helper( fnPtr, initData,
   2.104 +                                      semEnv, semEnv->nextCoreToGetNewPr++ );
   2.105 +
   2.106 +   resume_slaveVP( seedPr, semEnv );
   2.107 +   
   2.108 +   VMS_SS__start_the_work_then_wait_until_done();      //normal multi-thd
   2.109 +
   2.110 +   SSR__cleanup_after_shutdown();
   2.111 + }
   2.112 +
   2.113 +
   2.114 +int32
   2.115 +SSR__giveMinWorkUnitCycles( float32 percentOverhead )
   2.116 + {
   2.117 +   return MIN_WORK_UNIT_CYCLES;
   2.118 + }
   2.119 +
   2.120 +int32
   2.121 +SSR__giveIdealNumWorkUnits()
   2.122 + {
   2.123 +   return NUM_SCHED_SLOTS * NUM_CORES;
   2.124 + }
   2.125 +
   2.126 +int32
   2.127 +SSR__give_number_of_cores_to_schedule_onto()
   2.128 + {
   2.129 +   return NUM_CORES;
   2.130 + }
   2.131 +
   2.132 +/*For now, use TSC -- later, make these two macros with assembly that first
   2.133 + * saves jump point, and second jumps back several times to get reliable time
   2.134 + */
   2.135 +void
   2.136 +SSR__start_primitive()
   2.137 + { saveLowTimeStampCountInto( ((SSRSemEnv *)(_VMSMasterEnv->semanticEnv))->
   2.138 +                              primitiveStartTime );
   2.139 + }
   2.140 +
   2.141 +/*Just quick and dirty for now -- make reliable later
   2.142 + * will want this to jump back several times -- to be sure cache is warm
   2.143 + * because don't want comm time included in calc-time measurement -- and
   2.144 + * also to throw out any "weird" values due to OS interrupt or TSC rollover
   2.145 + */
   2.146 +int32
   2.147 +SSR__end_primitive_and_give_cycles()
   2.148 + { int32 endTime, startTime;
   2.149 +   //TODO: fix by repeating time-measurement
   2.150 +   saveLowTimeStampCountInto( endTime );
   2.151 +   startTime =((SSRSemEnv*)(_VMSMasterEnv->semanticEnv))->primitiveStartTime;
   2.152 +   return (endTime - startTime);
   2.153 + }
   2.154 +
   2.155 +//===========================================================================
   2.156 +
   2.157 +/*Initializes all the data-structures for a SSR system -- but doesn't
   2.158 + * start it running yet!
   2.159 + *
   2.160 + *This runs in the main thread -- before VMS starts up
   2.161 + * 
   2.162 + *This sets up the semantic layer over the VMS system
   2.163 + *
   2.164 + *First, calls VMS_Setup, then creates own environment, making it ready
   2.165 + * for creating the seed processor and then starting the work.
   2.166 + */
   2.167 +void
   2.168 +SSR__init()
   2.169 + {
   2.170 +   VMS_SS__init();
   2.171 +      //masterEnv, a global var, now is partially set up by init_VMS
   2.172 +      // after this, have VMS_int__malloc and VMS_int__free available
   2.173 +
   2.174 +   SSR__init_Helper();
   2.175 + }
   2.176 +
   2.177 +
   2.178 +void idle_fn(void* data, SlaveVP *animatingSlv){
   2.179 +    while(1){
   2.180 +        VMS_int__suspend_slaveVP_and_send_req(animatingSlv);
   2.181 +    }
   2.182 +}
   2.183 +
   2.184 +void
   2.185 +SSR__init_Helper()
   2.186 + { SSRSemEnv       *semanticEnv;
   2.187 +   PrivQueueStruc **readyVPQs;
   2.188 +   int              coreIdx, i, j;
   2.189 + 
   2.190 +      //Hook up the semantic layer's plug-ins to the Master virt procr
   2.191 +   _VMSMasterEnv->requestHandler = &SSR__Request_Handler;
   2.192 +   _VMSMasterEnv->slaveAssigner  = &SSR__assign_slaveVP;
   2.193 +   #ifdef HOLISTIC__TURN_ON_PERF_COUNTERS
   2.194 +   _VMSMasterEnv->counterHandler = &SSR__counter_handler;
   2.195 +   #endif
   2.196 +
   2.197 +      //create the semantic layer's environment (all its data) and add to
   2.198 +      // the master environment
   2.199 +   semanticEnv = VMS_int__malloc( sizeof( SSRSemEnv ) );
   2.200 +   _VMSMasterEnv->semanticEnv = semanticEnv;
   2.201 +   
   2.202 +   #ifdef HOLISTIC__TURN_ON_PERF_COUNTERS
   2.203 +   SSR__init_counter_data_structs();
   2.204 +   #endif
   2.205 +   for(i=0;i<NUM_CORES;++i){
   2.206 +       for(j=0;j<NUM_SCHED_SLOTS;++j){
   2.207 +           semanticEnv->idlePr[i][j] = VMS_int__create_slaveVP(&idle_fn,NULL);
   2.208 +           semanticEnv->idlePr[i][j]->coreAnimatedBy = i;
   2.209 +       }
   2.210 +   }
   2.211 +
   2.212 +   #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
   2.213 +   semanticEnv->unitList = makeListOfArrays(sizeof(Unit),128);
   2.214 +   semanticEnv->ctlDependenciesList = makeListOfArrays(sizeof(Dependency),128);
   2.215 +   semanticEnv->commDependenciesList = makeListOfArrays(sizeof(Dependency),128);
   2.216 +   semanticEnv->dynDependenciesList = makeListOfArrays(sizeof(Dependency),128);
   2.217 +   semanticEnv->ntonGroupsInfo = makePrivDynArrayOfSize((void***)&(semanticEnv->ntonGroups),8);
   2.218 +   
   2.219 +   semanticEnv->hwArcs = makeListOfArrays(sizeof(Dependency),128);
   2.220 +   memset(semanticEnv->last_in_slot,0,sizeof(NUM_CORES * NUM_SCHED_SLOTS * sizeof(Unit)));
   2.221 +   #endif
   2.222 +
   2.223 +      //create the ready queue, hash tables used for pairing send to receive
   2.224 +      // and so forth
   2.225 +      //TODO: add hash tables for pairing sends with receives, and
   2.226 +      // initialize the data ownership system
   2.227 +   readyVPQs = VMS_int__malloc( NUM_CORES * sizeof(PrivQueueStruc *) );
   2.228 +
   2.229 +   for( coreIdx = 0; coreIdx < NUM_CORES; coreIdx++ )
   2.230 +    {
   2.231 +      readyVPQs[ coreIdx ] = makeVMSQ();
   2.232 +    }
   2.233 +   
   2.234 +   semanticEnv->readyVPQs = readyVPQs;
   2.235 +   
   2.236 +   semanticEnv->nextCoreToGetNewPr = 0;
   2.237 +   semanticEnv->numSlaveVP = 0;
   2.238 +   
   2.239 +   semanticEnv->commHashTbl  = makeHashTable( 1<<16, &VMS_int__free );//start big
   2.240 +
   2.241 +   //TODO: bug -- turn these arrays into dyn arrays to eliminate limit
   2.242 +   //semanticEnv->singletonHasBeenExecutedFlags = makeDynArrayInfo( );
   2.243 +   //semanticEnv->transactionStrucs = makeDynArrayInfo( );
   2.244 +   for( i = 0; i < NUM_STRUCS_IN_SEM_ENV; i++ )
   2.245 +    {
   2.246 +      semanticEnv->fnSingletons[i].endInstrAddr      = NULL;
   2.247 +      semanticEnv->fnSingletons[i].hasBeenStarted    = FALSE;
   2.248 +      semanticEnv->fnSingletons[i].hasFinished       = FALSE;
   2.249 +      semanticEnv->fnSingletons[i].waitQ             = makeVMSQ();
   2.250 +      semanticEnv->transactionStrucs[i].waitingVPQ   = makeVMSQ();
   2.251 +    }
   2.252 + }
   2.253 +
   2.254 +
   2.255 +/*Frees any memory allocated by SSR__init() then calls VMS_int__shutdown
   2.256 + */
   2.257 +void
   2.258 +SSR__cleanup_after_shutdown()
   2.259 + { SSRSemEnv *semanticEnv;
   2.260 +   
   2.261 +   semanticEnv = _VMSMasterEnv->semanticEnv;
   2.262 +
   2.263 +   #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
   2.264 +   //UCC
   2.265 +   FILE* output;
   2.266 +   int n;
   2.267 +   char filename[255];    
   2.268 +    for(n=0;n<255;n++)
   2.269 +    {
   2.270 +        sprintf(filename, "./counters/UCC.%d",n);
   2.271 +        output = fopen(filename,"r");
   2.272 +        if(output)
   2.273 +        {
   2.274 +            fclose(output);
   2.275 +        }else{
   2.276 +            break;
   2.277 +        }
   2.278 +    }
   2.279 +   if(n<255){
   2.280 +    printf("Saving UCC to File: %s ...\n", filename);
   2.281 +    output = fopen(filename,"w+");
   2.282 +    if(output!=NULL){
   2.283 +        set_dependency_file(output);
   2.284 +        //fprintf(output,"digraph Dependencies {\n");
   2.285 +        //set_dot_file(output);
   2.286 +        //FIXME:  first line still depends on counters being enabled, replace w/ unit struct!
   2.287 +        //forAllInDynArrayDo(_VMSMasterEnv->counter_history_array_info, &print_dot_node_info );
   2.288 +        forAllInListOfArraysDo(semanticEnv->unitList, &print_unit_to_file);
   2.289 +        forAllInListOfArraysDo( semanticEnv->commDependenciesList, &print_comm_dependency_to_file );
   2.290 +        forAllInListOfArraysDo( semanticEnv->ctlDependenciesList, &print_ctl_dependency_to_file );
   2.291 +        forAllInDynArrayDo(semanticEnv->ntonGroupsInfo,&print_nton_to_file);
   2.292 +        //fprintf(output,"}\n");
   2.293 +        fflush(output);
   2.294 +
   2.295 +    } else
   2.296 +        printf("Opening UCC file failed. Please check that folder \"counters\" exists in run directory and has write permission.\n");
   2.297 +   } else {
   2.298 +       printf("Could not open UCC file, please clean \"counters\" folder. (Must contain less than 255 files.)\n");
   2.299 +   }
   2.300 +   //Loop Graph
   2.301 +   for(n=0;n<255;n++)
   2.302 +    {
   2.303 +        sprintf(filename, "./counters/LoopGraph.%d",n);
   2.304 +        output = fopen(filename,"r");
   2.305 +        if(output)
   2.306 +        {
   2.307 +            fclose(output);
   2.308 +        }else{
   2.309 +            break;
   2.310 +        }
   2.311 +    }
   2.312 +   if(n<255){
   2.313 +    printf("Saving LoopGraph to File: %s ...\n", filename);
   2.314 +    output = fopen(filename,"w+");
   2.315 +    if(output!=NULL){
   2.316 +        set_dependency_file(output);
   2.317 +        //fprintf(output,"digraph Dependencies {\n");
   2.318 +        //set_dot_file(output);
   2.319 +        //FIXME:  first line still depends on counters being enabled, replace w/ unit struct!
   2.320 +        //forAllInDynArrayDo(_VMSMasterEnv->counter_history_array_info, &print_dot_node_info );
   2.321 +        forAllInListOfArraysDo( semanticEnv->unitList, &print_unit_to_file );
   2.322 +        forAllInListOfArraysDo( semanticEnv->commDependenciesList, &print_comm_dependency_to_file );
   2.323 +        forAllInListOfArraysDo( semanticEnv->ctlDependenciesList, &print_ctl_dependency_to_file );
   2.324 +        forAllInListOfArraysDo( semanticEnv->dynDependenciesList, &print_dyn_dependency_to_file );
   2.325 +        forAllInListOfArraysDo( semanticEnv->hwArcs, &print_hw_dependency_to_file );
   2.326 +        //fprintf(output,"}\n");
   2.327 +        fflush(output);
   2.328 +
   2.329 +    } else
   2.330 +        printf("Opening LoopGraph file failed. Please check that folder \"counters\" exists in run directory and has write permission.\n");
   2.331 +   } else {
   2.332 +       printf("Could not open LoopGraph file, please clean \"counters\" folder. (Must contain less than 255 files.)\n");
   2.333 +   }
   2.334 +   
   2.335 +   
   2.336 +   freeListOfArrays(semanticEnv->unitList);
   2.337 +   freeListOfArrays(semanticEnv->commDependenciesList);
   2.338 +   freeListOfArrays(semanticEnv->ctlDependenciesList);
   2.339 +   freeListOfArrays(semanticEnv->dynDependenciesList);
   2.340 +   
   2.341 +   #endif
   2.342 +#ifdef HOLISTIC__TURN_ON_PERF_COUNTERS    
   2.343 +    for(n=0;n<255;n++)
   2.344 +    {
   2.345 +        sprintf(filename, "./counters/Counters.%d.csv",n);
   2.346 +        output = fopen(filename,"r");
   2.347 +        if(output)
   2.348 +        {
   2.349 +            fclose(output);
   2.350 +        }else{
   2.351 +            break;
   2.352 +        }
   2.353 +    }
   2.354 +    if(n<255){
   2.355 +    printf("Saving Counter measurements to File: %s ...\n", filename);
   2.356 +    output = fopen(filename,"w+");
   2.357 +    if(output!=NULL){
   2.358 +        set_counter_file(output);
   2.359 +        int i;
   2.360 +        for(i=0;i<NUM_CORES;i++){
   2.361 +            forAllInListOfArraysDo( semanticEnv->counterList[i], &print_counter_events_to_file );
   2.362 +            fflush(output);
   2.363 +        }
   2.364 +
   2.365 +    } else
   2.366 +        printf("Opening UCC file failed. Please check that folder \"counters\" exists in run directory and has write permission.\n");
   2.367 +   } else {
   2.368 +       printf("Could not open UCC file, please clean \"counters\" folder. (Must contain less than 255 files.)\n");
   2.369 +   }
   2.370 +    
   2.371 +#endif
   2.372 +/* It's all allocated inside VMS's big chunk -- that's about to be freed, so
   2.373 + *  nothing to do here
   2.374 +   
   2.375 +
   2.376 +   for( coreIdx = 0; coreIdx < NUM_CORES; coreIdx++ )
   2.377 +    {
   2.378 +      VMS_int__free( semanticEnv->readyVPQs[coreIdx]->startOfData );
   2.379 +      VMS_int__free( semanticEnv->readyVPQs[coreIdx] );
   2.380 +    }
   2.381 +   VMS_int__free( semanticEnv->readyVPQs );
   2.382 +   
   2.383 +   freeHashTable( semanticEnv->commHashTbl );
   2.384 +   VMS_int__free( _VMSMasterEnv->semanticEnv );
   2.385 + */
   2.386 +   VMS_SS__cleanup_at_end_of_shutdown();
   2.387 + }
   2.388 +
   2.389 +
   2.390 +//===========================================================================
   2.391 +
   2.392 +/*
   2.393 + */
   2.394 +  SlaveVP *
   2.395 +SSR__create_procr_with( TopLevelFnPtr fnPtr,   void *initData,
   2.396 +                        SlaveVP *creatingPr )
   2.397 + { SSRSemReq reqData;
   2.398 +
   2.399 +      //the semantic request data is on the stack and disappears when this
   2.400 +      // call returns -- it's guaranteed to remain in the VP's stack for as
   2.401 +      // long as the VP is suspended.
   2.402 +   reqData.reqType            = 0; //know type because in a VMS create req
   2.403 +   reqData.coreToAssignOnto = -1; //means round-robin assign
   2.404 +   reqData.fnPtr              = fnPtr;
   2.405 +   reqData.initData           = initData;
   2.406 +   reqData.sendPr             = creatingPr;
   2.407 +
   2.408 +   VMS_WL__send_create_slaveVP_req( &reqData, creatingPr );
   2.409 +
   2.410 +   return creatingPr->dataRetFromReq;
   2.411 + }
   2.412 +
   2.413 +  SlaveVP *
   2.414 +SSR__create_procr_with_affinity( TopLevelFnPtr fnPtr, void *initData,
   2.415 +                        SlaveVP *creatingPr,  int32  coreToAssignOnto )
   2.416 + { SSRSemReq  reqData;
   2.417 +
   2.418 +      //the semantic request data is on the stack and disappears when this
   2.419 +      // call returns -- it's guaranteed to remain in the VP's stack for as
   2.420 +      // long as the VP is suspended.
   2.421 +   reqData.reqType            = 0; //know type because in a VMS create req
   2.422 +   reqData.coreToAssignOnto = coreToAssignOnto;
   2.423 +   reqData.fnPtr              = fnPtr;
   2.424 +   reqData.initData           = initData;
   2.425 +   reqData.sendPr             = creatingPr;
   2.426 +
   2.427 +   VMS_WL__send_create_slaveVP_req( &reqData, creatingPr );
   2.428 +
   2.429 +   return creatingPr->dataRetFromReq;
   2.430 + }
   2.431 +
   2.432 +
   2.433 +  void
   2.434 +SSR__dissipate_procr( SlaveVP *procrToDissipate )
   2.435 + {
   2.436 +   VMS_WL__send_dissipate_req( procrToDissipate );
   2.437 + }
   2.438 +
   2.439 +
   2.440 +//===========================================================================
   2.441 +
   2.442 +void *
   2.443 +SSR__malloc_to( int32 sizeToMalloc, SlaveVP *owningPr )
   2.444 + { SSRSemReq reqData;
   2.445 +
   2.446 +   reqData.reqType      = malloc_req;
   2.447 +   reqData.sendPr       = owningPr;
   2.448 +   reqData.sizeToMalloc = sizeToMalloc;
   2.449 +
   2.450 +   VMS_WL__send_sem_request( &reqData, owningPr );
   2.451 +
   2.452 +   return owningPr->dataRetFromReq;
   2.453 + }
   2.454 +
   2.455 +
   2.456 +/*Sends request to Master, which does the work of freeing
   2.457 + */
   2.458 +void
   2.459 +SSR__free( void *ptrToFree, SlaveVP *owningPr )
   2.460 + { SSRSemReq reqData;
   2.461 +
   2.462 +   reqData.reqType      = free_req;
   2.463 +   reqData.sendPr       = owningPr;
   2.464 +   reqData.ptrToFree    = ptrToFree;
   2.465 +
   2.466 +   VMS_WL__send_sem_request( &reqData, owningPr );
   2.467 + }
   2.468 +
   2.469 +
   2.470 +void
   2.471 +SSR__transfer_ownership_of_from_to( void *data, SlaveVP *oldOwnerSlv,
   2.472 +                                                  SlaveVP *newOwnerPr )
   2.473 + {
   2.474 +   //TODO: put in the ownership system that automatically frees when no
   2.475 +   // owners of data left -- will need keeper for keeping data around when
   2.476 +   // future created processors might need it but don't exist yet
   2.477 + }
   2.478 +
   2.479 +
   2.480 +void
   2.481 +SSR__add_ownership_by_to( SlaveVP *newOwnerSlv, void *data )
   2.482 + {
   2.483 +
   2.484 + }
   2.485 +
   2.486 +
   2.487 +void
   2.488 +SSR__remove_ownership_by_from( SlaveVP *loserSlv, void *dataLosing )
   2.489 + {
   2.490 +
   2.491 + }
   2.492 +
   2.493 +
   2.494 +/*Causes the SSR system to remove internal ownership, so data won't be
   2.495 + * freed when SSR shuts down, and will persist in the external program.
   2.496 + *
   2.497 + *Must be called from the processor that currently owns the data.
   2.498 + *
   2.499 + *IMPL: Transferring ownership touches two different virtual processor's
   2.500 + * state -- which means it has to be done carefully -- the VMS rules for
   2.501 + * semantic layers say that a work-unit is only allowed to touch the
   2.502 + * virtual processor it is part of, and that only a single work-unit per
   2.503 + * virtual processor be assigned to a slave at a time.  So, this has to
   2.504 + * modify the virtual processor that owns the work-unit that called this
   2.505 + * function, then create a request to have the other processor modified.
   2.506 + *However, in this case, the TO processor is the outside, and transfers
   2.507 + * are only allowed to be called by the giver-upper, so can mark caller of
   2.508 + * this function as no longer owner, and return -- done.
   2.509 + */
   2.510 +void
   2.511 +SSR__transfer_ownership_to_outside( void *data )
   2.512 + {
   2.513 +   //TODO: removeAllOwnersFrom( data );
   2.514 + }
   2.515 +
   2.516 +
   2.517 +//===========================================================================
   2.518 +
   2.519 +void
   2.520 +SSR__send_of_type_to( SlaveVP *sendPr, void *msg, const int type,
   2.521 +                        SlaveVP *receivePr)
   2.522 + { SSRSemReq  reqData;
   2.523 +
   2.524 +   reqData.receivePr = receivePr;
   2.525 +   reqData.sendPr    = sendPr;
   2.526 +   reqData.reqType   = send_type;
   2.527 +   reqData.msgType   = type;
   2.528 +   reqData.msg       = msg;
   2.529 +   reqData.nextReqInHashEntry = NULL;
   2.530 +
   2.531 +      //On ownership -- remove inside the send and let ownership sit in limbo
   2.532 +      // as a potential in an entry in the hash table, when this receive msg
   2.533 +      // gets paired to a send, the ownership gets added to the receivePr --
   2.534 +      // the next work-unit in the receivePr's trace will have ownership.
   2.535 +   VMS_WL__send_sem_request( &reqData, sendPr );
   2.536 +
   2.537 +      //When come back from suspend, no longer own data reachable from msg
   2.538 +      //TODO: release ownership here
   2.539 + }
   2.540 +
   2.541 +void
   2.542 +SSR__send_from_to( void *msg, SlaveVP *sendPr, SlaveVP *receivePr )
   2.543 + { SSRSemReq  reqData;
   2.544 +
   2.545 +      //hash on the receiver, 'cause always know it, but sometimes want to
   2.546 +      // receive from anonymous sender
   2.547 +
   2.548 +   reqData.receivePr = receivePr;
   2.549 +   reqData.sendPr    = sendPr;
   2.550 +   reqData.reqType   = send_from_to;
   2.551 +   reqData.msg       = msg;
   2.552 +   reqData.nextReqInHashEntry = NULL;
   2.553 +
   2.554 +   VMS_WL__send_sem_request( &reqData, sendPr );
   2.555 + }
   2.556 +
   2.557 +
   2.558 +//===========================================================================
   2.559 +
   2.560 +void *
   2.561 +SSR__receive_any_to( SlaveVP *receivePr )
   2.562 + {
   2.563 +
   2.564 + }
   2.565 +
   2.566 +void *
   2.567 +SSR__receive_type_to( const int type, SlaveVP *receivePr )
   2.568 + { 
   2.569 +   SSRSemReq  reqData;
   2.570 +
   2.571 +   reqData.receivePr = receivePr;
   2.572 +   reqData.reqType   = receive_type;
   2.573 +   reqData.msgType   = type;
   2.574 +   reqData.nextReqInHashEntry = NULL;
   2.575 +
   2.576 +   VMS_WL__send_sem_request( &reqData, receivePr );
   2.577 +   
   2.578 +   return receivePr->dataRetFromReq;
   2.579 + }
   2.580 +
   2.581 +
   2.582 +
   2.583 +/*Call this at point receiving virt pr wants in-coming data.
   2.584 + * 
   2.585 + *The reason receivePr must call this is that it modifies the receivPr
   2.586 + * loc structure directly -- and the VMS rules state a virtual processor
   2.587 + * loc structure can only be modified by itself.
   2.588 + */
   2.589 +void *
   2.590 +SSR__receive_from_to( SlaveVP *sendPr, SlaveVP *receivePr )
   2.591 + { SSRSemReq  reqData;
   2.592 +
   2.593 +      //hash on the receiver, 'cause always know it, but sometimes want to
   2.594 +      // receive from anonymous sender
   2.595 +
   2.596 +   reqData.receivePr = receivePr;
   2.597 +   reqData.sendPr    = sendPr;
   2.598 +   reqData.reqType   = receive_from_to;
   2.599 +   reqData.nextReqInHashEntry = NULL;
   2.600 +
   2.601 +   VMS_WL__send_sem_request( &reqData, receivePr );
   2.602 +
   2.603 +   return receivePr->dataRetFromReq;
   2.604 + }
   2.605 +
   2.606 +
   2.607 +//===========================================================================
   2.608 +//
   2.609 +/*A function singleton is a function whose body executes exactly once, on a
   2.610 + * single core, no matter how many times the fuction is called and no
   2.611 + * matter how many cores or the timing of cores calling it.
   2.612 + *
   2.613 + *A data singleton is a ticket attached to data.  That ticket can be used
   2.614 + * to get the data through the function exactly once, no matter how many
   2.615 + * times the data is given to the function, and no matter the timing of
   2.616 + * trying to get the data through from different cores.
   2.617 + */
   2.618 +
   2.619 +/*asm function declarations*/
   2.620 +void asm_save_ret_to_singleton(SSRSingleton *singletonPtrAddr);
   2.621 +void asm_write_ret_from_singleton(SSRSingleton *singletonPtrAddr);
   2.622 +
   2.623 +/*Fn singleton uses ID as index into array of singleton structs held in the
   2.624 + * semantic environment.
   2.625 + */
   2.626 +void
   2.627 +SSR__start_fn_singleton( int32 singletonID,   SlaveVP *animPr )
   2.628 + {
   2.629 +   SSRSemReq  reqData;
   2.630 +
   2.631 +      //
   2.632 +   reqData.reqType     = singleton_fn_start;
   2.633 +   reqData.singletonID = singletonID;
   2.634 +
   2.635 +   VMS_WL__send_sem_request( &reqData, animPr );
   2.636 +   if( animPr->dataRetFromReq ) //will be 0 or addr of label in end singleton
   2.637 +    {
   2.638 +       SSRSemEnv *semEnv = VMS_int__give_sem_env_for( animPr );
   2.639 +       asm_write_ret_from_singleton(&(semEnv->fnSingletons[ singletonID]));
   2.640 +    }
   2.641 + }
   2.642 +
   2.643 +/*Data singleton hands addr of loc holding a pointer to a singleton struct.
   2.644 + * The start_data_singleton makes the structure and puts its addr into the
   2.645 + * location.
   2.646 + */
   2.647 +void
   2.648 +SSR__start_data_singleton( SSRSingleton **singletonAddr,  SlaveVP *animPr )
   2.649 + {
   2.650 +   SSRSemReq  reqData;
   2.651 +
   2.652 +   if( *singletonAddr && (*singletonAddr)->hasFinished )
   2.653 +       goto JmpToEndSingleton;
   2.654 +   
   2.655 +   reqData.reqType          = singleton_data_start;
   2.656 +   reqData.singletonPtrAddr = singletonAddr;
   2.657 +
   2.658 +   VMS_WL__send_sem_request( &reqData, animPr );
   2.659 +   if( animPr->dataRetFromReq ) //either 0 or end singleton's return addr
   2.660 +    {    //Assembly code changes the return addr on the stack to the one
   2.661 +         // saved into the singleton by the end-singleton-fn
   2.662 +         //The return addr is at 0x4(%%ebp)
   2.663 +        JmpToEndSingleton:
   2.664 +          asm_write_ret_from_singleton(*singletonAddr);
   2.665 +    }
   2.666 +   //now, simply return
   2.667 +   //will exit either from the start singleton call or the end-singleton call
   2.668 + }
   2.669 +
   2.670 +/*Uses ID as index into array of flags.  If flag already set, resumes from
   2.671 + * end-label.  Else, sets flag and resumes normally.
   2.672 + *
   2.673 + *Note, this call cannot be inlined because the instr addr at the label
   2.674 + * inside is shared by all invocations of a given singleton ID.
   2.675 + */
   2.676 +void
   2.677 +SSR__end_fn_singleton( int32 singletonID, SlaveVP *animPr )
   2.678 + {
   2.679 +   SSRSemReq  reqData;
   2.680 +
   2.681 +      //don't need this addr until after at least one singleton has reached
   2.682 +      // this function
   2.683 +   SSRSemEnv *semEnv = VMS_int__give_sem_env_for( animPr );
   2.684 +   asm_write_ret_from_singleton(&(semEnv->fnSingletons[ singletonID]));
   2.685 +
   2.686 +   reqData.reqType     = singleton_fn_end;
   2.687 +   reqData.singletonID = singletonID;
   2.688 +
   2.689 +   VMS_WL__send_sem_request( &reqData, animPr );
   2.690 +
   2.691 +EndSingletonInstrAddr:
   2.692 +   return;
   2.693 + }
   2.694 +
   2.695 +void
   2.696 +SSR__end_data_singleton(  SSRSingleton **singletonPtrAddr, SlaveVP *animPr )
   2.697 + {
   2.698 +   SSRSemReq  reqData;
   2.699 +
   2.700 +      //don't need this addr until after singleton struct has reached
   2.701 +      // this function for first time
   2.702 +      //do assembly that saves the return addr of this fn call into the
   2.703 +      // data singleton -- that data-singleton can only be given to exactly
   2.704 +      // one instance in the code of this function.  However, can use this
   2.705 +      // function in different places for different data-singletons.
   2.706 +//   (*(singletonAddr))->endInstrAddr =  &&EndDataSingletonInstrAddr;
   2.707 +
   2.708 +
   2.709 +   asm_save_ret_to_singleton(*singletonPtrAddr);
   2.710 +
   2.711 +   reqData.reqType          = singleton_data_end;
   2.712 +   reqData.singletonPtrAddr = singletonPtrAddr;
   2.713 +
   2.714 +   VMS_WL__send_sem_request( &reqData, animPr );
   2.715 + }
   2.716 +
   2.717 +/*This executes the function in the masterVP, so it executes in isolation
   2.718 + * from any other copies -- only one copy of the function can ever execute
   2.719 + * at a time.
   2.720 + *
   2.721 + *It suspends to the master, and the request handler takes the function
   2.722 + * pointer out of the request and calls it, then resumes the VP.
   2.723 + *Only very short functions should be called this way -- for longer-running
   2.724 + * isolation, use transaction-start and transaction-end, which run the code
   2.725 + * between as work-code.
   2.726 + */
   2.727 +void
   2.728 +SSR__animate_short_fn_in_isolation( PtrToAtomicFn ptrToFnToExecInMaster,
   2.729 +                                    void *data, SlaveVP *animPr )
   2.730 + {
   2.731 +   SSRSemReq  reqData;
   2.732 +
   2.733 +      //
   2.734 +   reqData.reqType          = atomic;
   2.735 +   reqData.fnToExecInMaster = ptrToFnToExecInMaster;
   2.736 +   reqData.dataForFn        = data;
   2.737 +
   2.738 +   VMS_WL__send_sem_request( &reqData, animPr );
   2.739 + }
   2.740 +
   2.741 +
   2.742 +/*This suspends to the master.
   2.743 + *First, it looks at the VP's data, to see the highest transactionID that VP
   2.744 + * already has entered.  If the current ID is not larger, it throws an
   2.745 + * exception stating a bug in the code.  Otherwise it puts the current ID
   2.746 + * there, and adds the ID to a linked list of IDs entered -- the list is
   2.747 + * used to check that exits are properly ordered.
   2.748 + *Next it is uses transactionID as index into an array of transaction
   2.749 + * structures.
   2.750 + *If the "VP_currently_executing" field is non-null, then put requesting VP
   2.751 + * into queue in the struct.  (At some point a holder will request
   2.752 + * end-transaction, which will take this VP from the queue and resume it.)
   2.753 + *If NULL, then write requesting into the field and resume.
   2.754 + */
   2.755 +void
   2.756 +SSR__start_transaction( int32 transactionID, SlaveVP *animPr )
   2.757 + {
   2.758 +   SSRSemReq  reqData;
   2.759 +
   2.760 +      //
   2.761 +   reqData.sendPr      = animPr;
   2.762 +   reqData.reqType     = trans_start;
   2.763 +   reqData.transID     = transactionID;
   2.764 +
   2.765 +   VMS_WL__send_sem_request( &reqData, animPr );
   2.766 + }
   2.767 +
   2.768 +/*This suspends to the master, then uses transactionID as index into an
   2.769 + * array of transaction structures.
   2.770 + *It looks at VP_currently_executing to be sure it's same as requesting VP.
   2.771 + * If different, throws an exception, stating there's a bug in the code.
   2.772 + *Next it looks at the queue in the structure.
   2.773 + *If it's empty, it sets VP_currently_executing field to NULL and resumes.
   2.774 + *If something in, gets it, sets VP_currently_executing to that VP, then
   2.775 + * resumes both.
   2.776 + */
   2.777 +void
   2.778 +SSR__end_transaction( int32 transactionID, SlaveVP *animPr )
   2.779 + {
   2.780 +   SSRSemReq  reqData;
   2.781 +
   2.782 +      //
   2.783 +   reqData.sendPr      = animPr;
   2.784 +   reqData.reqType     = trans_end;
   2.785 +   reqData.transID     = transactionID;
   2.786 +
   2.787 +   VMS_WL__send_sem_request( &reqData, animPr );
   2.788 + }

     3.1 --- a/SSR.h	Tue Mar 13 10:04:14 2012 -0700
     3.2 +++ b/SSR.h	Tue Mar 13 18:30:05 2012 -0700
     3.3 @@ -84,7 +84,7 @@
     3.4  
     3.5     void              *initData;
     3.6     TopLevelFnPtr     fnPtr;
     3.7 -   int32              coreToScheduleOnto;
     3.8 +   int32              coreToAssignOnto;
     3.9  
    3.10     int32              sizeToMalloc;
    3.11     void              *ptrToFree;
    3.12 @@ -182,7 +182,7 @@
    3.13  
    3.14    SlaveVP *
    3.15  SSR__create_procr_with_affinity( TopLevelFnPtr fnPtr,    void *initData,
    3.16 -                            SlaveVP *creatingPr, int32 coreToScheduleOnto);
    3.17 +                            SlaveVP *creatingPr, int32 coreToAssignOnto);
    3.18  
    3.19  void
    3.20  SSR__dissipate_procr( SlaveVP *procrToDissipate );
    3.21 @@ -253,11 +253,15 @@
    3.22  SSR__Request_Handler( SlaveVP *requestingPr, void *_semEnv );
    3.23  
    3.24  SlaveVP *
    3.25 -SSR__schedule_slaveVP( void *_semEnv, int coreNum, int slotNum );
    3.26 +SSR__assign_slaveVP( void *_semEnv, int coreNum, SchedSlot *slot );
    3.27  
    3.28  SlaveVP*
    3.29  SSR__create_procr_helper( TopLevelFnPtr fnPtr, void *initData,
    3.30 -                          SSRSemEnv *semEnv,    int32 coreToScheduleOnto );
    3.31 +                          SSRSemEnv *semEnv,    int32 coreToAssignOnto );
    3.32  
    3.33 +//=====================  Measurement of Lang Overheads  =====================
    3.34 +#include "SSR_Measurement.h"
    3.35 +
    3.36 +//===========================================================================
    3.37  #endif	/* _SSR_H */
    3.38  

     4.1 --- a/SSR.s	Tue Mar 13 10:04:14 2012 -0700
     4.2 +++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
     4.3 @@ -1,21 +0,0 @@
     4.4 -
     4.5 -//Assembly code takes the return addr off the stack and saves
     4.6 -// into the singleton.  The first field in the singleton is the
     4.7 -// "endInstrAddr" field, and the return addr is at 0x4(%ebp)
     4.8 -.globl asm_save_ret_to_singleton
     4.9 -asm_save_ret_to_singleton:
    4.10 -    movq 0x8(%rbp),     %rax   #get ret address, ebp is the same as in the calling function
    4.11 -    movq     %rax,     (%rdi) #write ret addr to endInstrAddr field
    4.12 -    ret
    4.13 -
    4.14 -
    4.15 -//Assembly code changes the return addr on the stack to the one
    4.16 -// saved into the singleton by the end-singleton-fn
    4.17 -//The stack's return addr is at 0x4(%%ebp)
    4.18 -.globl asm_write_ret_from_singleton
    4.19 -asm_write_ret_from_singleton:
    4.20 -    movq    (%rdi),    %rax  #get endInstrAddr field
    4.21 -    movq      %rax,    0x8(%rbp) #write return addr to the stack of the caller
    4.22 -    ret
    4.23 -
    4.24 -

     5.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     5.2 +++ b/SSR_Measurement.h	Tue Mar 13 18:30:05 2012 -0700
     5.3 @@ -0,0 +1,87 @@
     5.4 +/*
     5.5 + *  Copyright 2009 OpenSourceStewardshipFoundation.org
     5.6 + *  Licensed under GNU General Public License version 2
     5.7 + *
     5.8 + * Author: seanhalle@yahoo.com
     5.9 + *
    5.10 + */
    5.11 +
    5.12 +#ifndef _SSR_MEAS_H
    5.13 +#define	_SSR_MEAS_H
    5.14 +
    5.15 +
    5.16 +#ifdef MEAS__TURN_ON_LANG_MEAS
    5.17 +
    5.18 +   #ifdef MEAS__Make_Meas_Hists_for_Language
    5.19 +   #undef MEAS__Make_Meas_Hists_for_Language
    5.20 +   #endif
    5.21 +
    5.22 +
    5.23 +//===================  Language-specific Measurement Stuff ===================
    5.24 +//
    5.25 +//
    5.26 +   #define SendFromToHistIdx      1 //note: starts at 1
    5.27 +   #define SendOfTypeHistIdx      2
    5.28 +   #define ReceiveFromToHistIdx   3
    5.29 +   #define ReceiveOfTypeHistIdx   4
    5.30 +
    5.31 +   #define MEAS__Make_Meas_Hists_for_Language \
    5.32 +      _VMSMasterEnv->measHistsInfo = \
    5.33 +                 makePrivDynArrayOfSize( (void***)&(_VMSMasterEnv->measHists), 200); \
    5.34 +       makeAMeasHist( SendFromToHistIdx,   "SendFromTo",    50, 0, 100 ) \
    5.35 +       makeAMeasHist( SendOfTypeHistIdx,   "SendOfType",    50, 0, 100 ) \
    5.36 +       makeAMeasHist( ReceiveFromToHistIdx,"ReceiveFromTo", 50, 0, 100 ) \
    5.37 +       makeAMeasHist( ReceiveOfTypeHistIdx,"ReceiveOfType", 50, 0, 100 )
    5.38 +
    5.39 +   #define Meas_startSendFromTo \
    5.40 +       int32 startStamp, endStamp; \
    5.41 +       saveLowTimeStampCountInto( startStamp ); \
    5.42 +
    5.43 +   #define Meas_endSendFromTo \
    5.44 +       saveLowTimeStampCountInto( endStamp ); \
    5.45 +       addIntervalToHist( startStamp, endStamp, \
    5.46 +                                _VMSMasterEnv->measHists[ SendFromToHistIdx ] );
    5.47 +
    5.48 +   #define Meas_startSendOfType \
    5.49 +       int32 startStamp, endStamp; \
    5.50 +       saveLowTimeStampCountInto( startStamp ); \
    5.51 +
    5.52 +   #define Meas_endSendOfType \
    5.53 +       saveLowTimeStampCountInto( endStamp ); \
    5.54 +       addIntervalToHist( startStamp, endStamp, \
    5.55 +                                _VMSMasterEnv->measHists[ SendOfTypeHistIdx ] );
    5.56 +
    5.57 +   #define Meas_startReceiveFromTo \
    5.58 +       int32 startStamp, endStamp; \
    5.59 +       saveLowTimeStampCountInto( startStamp ); \
    5.60 +
    5.61 +   #define Meas_endReceiveFromTo \
    5.62 +       saveLowTimeStampCountInto( endStamp ); \
    5.63 +       addIntervalToHist( startStamp, endStamp, \
    5.64 +                                _VMSMasterEnv->measHists[ ReceiveFromToHistIdx ] );
    5.65 +
    5.66 +   #define Meas_startReceiveOfType \
    5.67 +       int32 startStamp, endStamp; \
    5.68 +       saveLowTimeStampCountInto( startStamp ); \
    5.69 +
    5.70 +   #define Meas_endReceiveOfType \
    5.71 +       saveLowTimeStampCountInto( endStamp ); \
    5.72 +       addIntervalToHist( startStamp, endStamp, \
    5.73 +                                _VMSMasterEnv->measHists[ReceiveOfTypeHistIdx ] );
    5.74 +
    5.75 +#else //===================== turned off ==========================
    5.76 +
    5.77 +   #define MEAS__Make_Meas_Hists_for_Language 
    5.78 +   #define Meas_startSendFromTo
    5.79 +   #define Meas_endSendFromTo
    5.80 +   #define Meas_startSendOfType
    5.81 +   #define Meas_endSendOfType
    5.82 +   #define Meas_startReceiveFromTo
    5.83 +   #define Meas_endReceiveFromTo
    5.84 +   #define Meas_startReceiveOfType
    5.85 +   #define Meas_endReceiveOfType
    5.86 +
    5.87 +#endif  /* MEAS__TURN_ON_LANG_MEAS */
    5.88 +
    5.89 +#endif	/*  */
    5.90 +

     6.1 --- a/SSR_PluginFns.c	Tue Mar 13 10:04:14 2012 -0700
     6.2 +++ b/SSR_PluginFns.c	Tue Mar 13 18:30:05 2012 -0700
     6.3 @@ -7,7 +7,7 @@
     6.4  #include <stdio.h>
     6.5  #include <stdlib.h>
     6.6  
     6.7 -#include "../../C_Libraries/Queue_impl/PrivateQueue.h"
     6.8 +#include "Queue_impl/PrivateQueue.h"
     6.9  #include "SSR.h"
    6.10  #include "SSR_Request_Handlers.h"
    6.11  
    6.12 @@ -25,21 +25,22 @@
    6.13  handleCreate( VMSReqst *req, SlaveVP *requestingPr, SSRSemEnv *semEnv  );
    6.14  
    6.15  
    6.16 -//============================== Scheduler ==================================
    6.17 +//============================== Assigner ==================================
    6.18  //
    6.19 -/*For SSR, scheduling a slave simply takes the next work-unit off the
    6.20 +/*For SSR, assigning a slave simply takes the next work-unit off the
    6.21   * ready-to-go work-unit queue and assigns it to the slaveToSched.
    6.22 - *If the ready-to-go work-unit queue is empty, then nothing to schedule
    6.23 + *If the ready-to-go work-unit queue is empty, then nothing to assign
    6.24   * to the slave -- return FALSE to let Master loop know scheduling that
    6.25   * slave failed.
    6.26   */
    6.27 -char __Scheduler[] = "FIFO Scheduler"; //Gobal variable for name in saved histogram
    6.28 -
    6.29  SlaveVP *
    6.30 -SSR__schedule_slaveVP( void *_semEnv, int coreNum, int slotNum )
    6.31 +SSR__assign_slaveVP( void *_semEnv, int coreNum, SchedSlot *slot )
    6.32   { SlaveVP   *schedPr;
    6.33     SSRSemEnv *semEnv;
    6.34 -
    6.35 +   int32      slotNum;
    6.36 +   
    6.37 +   slotNum = slot->slotIdx;
    6.38 +   
    6.39     semEnv  = (SSRSemEnv *)_semEnv;
    6.40  
    6.41     schedPr = readPrivQ( semEnv->readyVPQs[coreNum] );
    6.42 @@ -48,36 +49,36 @@
    6.43         schedPr = semEnv->idlePr[coreNum][slotNum];
    6.44       //things that would normally happen in resume(), but these VPs never go there
    6.45       #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
    6.46 -        schedPr->numTimesScheduled++; //Somewhere here!
    6.47 +        schedPr->numTimesAssigned++; //Somewhere here!
    6.48          Unit newu;
    6.49          newu.vp = schedPr->slaveID;
    6.50 -        newu.task = schedPr->numTimesScheduled;
    6.51 +        newu.task = schedPr->numTimesAssigned;
    6.52          addToListOfArrays(Unit,newu,semEnv->unitList);
    6.53     
    6.54 -        if (schedPr->numTimesScheduled > 1){
    6.55 +        if (schedPr->numTimesAssigned > 1){
    6.56                  Dependency newd;
    6.57                  newd.from_vp = schedPr->slaveID;
    6.58 -                newd.from_task = schedPr->numTimesScheduled - 1;
    6.59 +                newd.from_task = schedPr->numTimesAssigned - 1;
    6.60                  newd.to_vp = schedPr->slaveID;
    6.61 -                newd.to_task = schedPr->numTimesScheduled;
    6.62 +                newd.to_task = schedPr->numTimesAssigned;
    6.63                  addToListOfArrays(Dependency, newd ,semEnv->ctlDependenciesList);  
    6.64          }
    6.65        #endif
    6.66     }
    6.67     #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
    6.68     if (schedPr) {
    6.69 -        //schedPr->numTimesScheduled++;
    6.70 +        //schedPr->numTimesAssigned++;
    6.71          Unit prev_in_slot = semEnv->last_in_slot[coreNum * NUM_SCHED_SLOTS + slotNum];
    6.72          if(prev_in_slot.vp != 0){
    6.73                  Dependency newd;
    6.74                  newd.from_vp = prev_in_slot.vp;
    6.75                  newd.from_task = prev_in_slot.task;
    6.76                  newd.to_vp = schedPr->slaveID;
    6.77 -                newd.to_task = schedPr->numTimesScheduled;
    6.78 +                newd.to_task = schedPr->numTimesAssigned;
    6.79                  addToListOfArrays(Dependency,newd,semEnv->hwArcs);   
    6.80          }
    6.81          prev_in_slot.vp = schedPr->slaveID;
    6.82 -        prev_in_slot.task = schedPr->numTimesScheduled;
    6.83 +        prev_in_slot.task = schedPr->numTimesAssigned;
    6.84          semEnv->last_in_slot[coreNum * NUM_SCHED_SLOTS + slotNum] = prev_in_slot;        
    6.85     }
    6.86     #endif
    6.87 @@ -91,7 +92,7 @@
    6.88   * Upon send, check the hash to see if a receive is waiting.
    6.89   * Upon receive, check hash to see if a send has already happened.
    6.90   * When other is not there, put in.  When other is there, the comm.
    6.91 - *  completes, which means the receiver P gets scheduled and
    6.92 + *  completes, which means the receiver P gets assigned and
    6.93   *  picks up right after the receive request.  So make the work-unit
    6.94   *  and put it into the queue of work-units ready to go.
    6.95   * Other request is create a new Processor, with the function to run in the
    6.96 @@ -194,7 +195,7 @@
    6.97   */
    6.98    SlaveVP *
    6.99  SSR__create_procr_helper( TopLevelFnPtr fnPtr, void *initData,
   6.100 -                          SSRSemEnv *semEnv,    int32 coreToScheduleOnto )
   6.101 +                          SSRSemEnv *semEnv,    int32 coreToAssignOnto )
   6.102   { SlaveVP    *newPr;
   6.103     SSRSemData   *semData;
   6.104  
   6.105 @@ -215,7 +216,7 @@
   6.106  
   6.107     #else
   6.108  
   6.109 -   if(coreToScheduleOnto < 0 || coreToScheduleOnto >= NUM_CORES )
   6.110 +   if(coreToAssignOnto < 0 || coreToAssignOnto >= NUM_CORES )
   6.111      {    //out-of-range, so round-robin assignment
   6.112        newPr->coreAnimatedBy = semEnv->nextCoreToGetNewPr;
   6.113  
   6.114 @@ -225,7 +226,7 @@
   6.115            semEnv->nextCoreToGetNewPr += 1;
   6.116      }
   6.117     else //core num in-range, so use it
   6.118 -    { newPr->coreAnimatedBy = coreToScheduleOnto;
   6.119 +    { newPr->coreAnimatedBy = coreToAssignOnto;
   6.120      }
   6.121     #endif
   6.122     //========================================================================
   6.123 @@ -243,14 +244,14 @@
   6.124     semReq = VMS_PI__take_sem_reqst_from( req );
   6.125   
   6.126     newPr = SSR__create_procr_helper( semReq->fnPtr, semReq->initData, semEnv,
   6.127 -                                     semReq->coreToScheduleOnto );
   6.128 +                                     semReq->coreToAssignOnto );
   6.129     
   6.130     DEBUG_Print1(dbgRqstHdlr,"(new VP: %d)\n",newPr->slaveID)
   6.131  
   6.132     #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
   6.133     Dependency newd;
   6.134     newd.from_vp = requestingPr->slaveID;
   6.135 -   newd.from_task = requestingPr->numTimesScheduled;
   6.136 +   newd.from_task = requestingPr->numTimesAssigned;
   6.137     newd.to_vp = newPr->slaveID;
   6.138     newd.to_task = 1;
   6.139     //addToListOfArraysDependency(newd,semEnv->commDependenciesList);  
   6.140 @@ -277,18 +278,18 @@
   6.141  */
   6.142     #endif
   6.143     #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
   6.144 -   procr->numTimesScheduled++; //Somewhere here!
   6.145 +   procr->numTimesAssigned++; //Somewhere here!
   6.146     Unit newu;
   6.147     newu.vp = procr->slaveID;
   6.148 -   newu.task = procr->numTimesScheduled;
   6.149 +   newu.task = procr->numTimesAssigned;
   6.150     addToListOfArrays(Unit,newu,semEnv->unitList);
   6.151     
   6.152 -   if (procr->numTimesScheduled > 1){
   6.153 +   if (procr->numTimesAssigned > 1){
   6.154          Dependency newd;
   6.155          newd.from_vp = procr->slaveID;
   6.156 -        newd.from_task = procr->numTimesScheduled - 1;
   6.157 +        newd.from_task = procr->numTimesAssigned - 1;
   6.158          newd.to_vp = procr->slaveID;
   6.159 -        newd.to_task = procr->numTimesScheduled;
   6.160 +        newd.to_task = procr->numTimesAssigned;
   6.161          addToListOfArrays(Dependency, newd ,semEnv->ctlDependenciesList);  
   6.162     }
   6.163     #endif

     7.1 --- a/SSR_Request_Handlers.c	Tue Mar 13 10:04:14 2012 -0700
     7.2 +++ b/SSR_Request_Handlers.c	Tue Mar 13 18:30:05 2012 -0700
     7.3 @@ -129,9 +129,9 @@
     7.4         #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
     7.5          Dependency newd;
     7.6          newd.from_vp = sendPr->slaveID;
     7.7 -        newd.from_task = sendPr->numTimesScheduled;
     7.8 +        newd.from_task = sendPr->numTimesAssigned;
     7.9          newd.to_vp = receivePr->slaveID;
    7.10 -        newd.to_task = receivePr->numTimesScheduled +1;
    7.11 +        newd.to_task = receivePr->numTimesAssigned +1;
    7.12          //(newd,semEnv->commDependenciesList);  
    7.13          addToListOfArrays(Dependency,newd,semEnv->dynDependenciesList);  
    7.14                  int32 groupId = semReq->msgType;
    7.15 @@ -143,10 +143,10 @@
    7.16          }
    7.17          Unit u;
    7.18          u.vp = sendPr->slaveID;
    7.19 -        u.task = sendPr->numTimesScheduled;
    7.20 +        u.task = sendPr->numTimesAssigned;
    7.21          addToListOfArrays(Unit,u,semEnv->ntonGroups[groupId]->senders);
    7.22          u.vp = receivePr->slaveID;
    7.23 -        u.task = receivePr->numTimesScheduled +1;
    7.24 +        u.task = receivePr->numTimesAssigned +1;
    7.25          addToListOfArrays(Unit,u,semEnv->ntonGroups[groupId]->receivers);
    7.26         #endif
    7.27  
    7.28 @@ -209,9 +209,9 @@
    7.29        #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
    7.30          Dependency newd;
    7.31          newd.from_vp = sendPr->slaveID;
    7.32 -        newd.from_task = sendPr->numTimesScheduled;
    7.33 +        newd.from_task = sendPr->numTimesAssigned;
    7.34          newd.to_vp = receivePr->slaveID;
    7.35 -        newd.to_task = receivePr->numTimesScheduled +1;
    7.36 +        newd.to_task = receivePr->numTimesAssigned +1;
    7.37          //addToListOfArraysDependency(newd,semEnv->commDependenciesList);  
    7.38          addToListOfArrays(Dependency,newd,semEnv->commDependenciesList);   
    7.39        #endif 
    7.40 @@ -318,9 +318,9 @@
    7.41         #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
    7.42          Dependency newd;
    7.43          newd.from_vp = sendPr->slaveID;
    7.44 -        newd.from_task = sendPr->numTimesScheduled;
    7.45 +        newd.from_task = sendPr->numTimesAssigned;
    7.46          newd.to_vp = receivePr->slaveID;
    7.47 -        newd.to_task = receivePr->numTimesScheduled +1;
    7.48 +        newd.to_task = receivePr->numTimesAssigned +1;
    7.49          //addToListOfArraysDependency(newd,semEnv->commDependenciesList);  
    7.50          addToListOfArrays(Dependency,newd,semEnv->dynDependenciesList); 
    7.51          int32 groupId = semReq->msgType;
    7.52 @@ -332,10 +332,10 @@
    7.53          }
    7.54          Unit u;
    7.55          u.vp = sendPr->slaveID;
    7.56 -        u.task = sendPr->numTimesScheduled;
    7.57 +        u.task = sendPr->numTimesAssigned;
    7.58          addToListOfArrays(Unit,u,semEnv->ntonGroups[groupId]->senders);
    7.59          u.vp = receivePr->slaveID;
    7.60 -        u.task = receivePr->numTimesScheduled +1;
    7.61 +        u.task = receivePr->numTimesAssigned +1;
    7.62          addToListOfArrays(Unit,u,semEnv->ntonGroups[groupId]->receivers);
    7.63         #endif
    7.64        
    7.65 @@ -378,9 +378,9 @@
    7.66        #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
    7.67          Dependency newd;
    7.68          newd.from_vp = sendPr->slaveID;
    7.69 -        newd.from_task = sendPr->numTimesScheduled;
    7.70 +        newd.from_task = sendPr->numTimesAssigned;
    7.71          newd.to_vp = receivePr->slaveID;
    7.72 -        newd.to_task = receivePr->numTimesScheduled +1;
    7.73 +        newd.to_task = receivePr->numTimesAssigned +1;
    7.74          //addToListOfArraysDependency(newd,semEnv->commDependenciesList);  
    7.75          addToListOfArrays(Dependency,newd,semEnv->commDependenciesList);    
    7.76        #endif  

     8.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     8.2 +++ b/SSR_asm.s	Tue Mar 13 18:30:05 2012 -0700
     8.3 @@ -0,0 +1,21 @@
     8.4 +
     8.5 +//Assembly code takes the return addr off the stack and saves
     8.6 +// into the singleton.  The first field in the singleton is the
     8.7 +// "endInstrAddr" field, and the return addr is at 0x4(%ebp)
     8.8 +.globl asm_save_ret_to_singleton
     8.9 +asm_save_ret_to_singleton:
    8.10 +    movq 0x8(%rbp),     %rax   #get ret address, ebp is the same as in the calling function
    8.11 +    movq     %rax,     (%rdi) #write ret addr to endInstrAddr field
    8.12 +    ret
    8.13 +
    8.14 +
    8.15 +//Assembly code changes the return addr on the stack to the one
    8.16 +// saved into the singleton by the end-singleton-fn
    8.17 +//The stack's return addr is at 0x4(%%ebp)
    8.18 +.globl asm_write_ret_from_singleton
    8.19 +asm_write_ret_from_singleton:
    8.20 +    movq    (%rdi),    %rax  #get endInstrAddr field
    8.21 +    movq      %rax,    0x8(%rbp) #write return addr to the stack of the caller
    8.22 +    ret
    8.23 +
    8.24 +

     9.1 --- a/SSR_lib.c	Tue Mar 13 10:04:14 2012 -0700
     9.2 +++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
     9.3 @@ -1,807 +0,0 @@
     9.4 -/*
     9.5 - * Copyright 2010  OpenSourceCodeStewardshipFoundation
     9.6 - *
     9.7 - * Licensed under BSD
     9.8 - */
     9.9 -
    9.10 -#include <stdio.h>
    9.11 -#include <stdlib.h>
    9.12 -#include <malloc.h>
    9.13 -
    9.14 -#include "../VMS_impl/VMS.h"
    9.15 -#include "SSR.h"
    9.16 -#include "../../C_Libraries/Queue_impl/PrivateQueue.h"
    9.17 -#include "../../C_Libraries/Hash_impl/PrivateHash.h"
    9.18 -#include "SSR.h"
    9.19 -#include "SSR_Counter_Recording.h"
    9.20 -
    9.21 -//==========================================================================
    9.22 -
    9.23 -void
    9.24 -SSR__init();
    9.25 -
    9.26 -void
    9.27 -SSR__init_Seq();
    9.28 -
    9.29 -void
    9.30 -SSR__init_Helper();
    9.31 -//==========================================================================
    9.32 -
    9.33 -
    9.34 -/*TODO: Q: dealing with library f()s and DKU vs WT vs FoR
    9.35 - * (still want to do FoR, with time-lines as syntax, could be super cool)
    9.36 - * A: thinking pin the coreCtlrs for all of BLIS -- let Master arbitrate
    9.37 - * among library, DKU, WT, FoR -- all the patterns in terms of virtual
    9.38 - * processors (or equivalently work-units), so Master picks which virt procr
    9.39 - * from which portions of app (DKU, WT, FoR) onto which sched slots
    9.40 - *Might even do hierarchy of masters -- group of sched slots for each core
    9.41 - * has its own master, that keeps generated work local
    9.42 - * single-reader-single-writer sync everywhere -- no atomic primitives
    9.43 - * Might have the different schedulers talk to each other, to negotiate
    9.44 - * larger-grain sharing of resources, according to predicted critical
    9.45 - * path, and expansion of work
    9.46 - */
    9.47 -
    9.48 -
    9.49 -
    9.50 -//===========================================================================
    9.51 -
    9.52 -
    9.53 -/*These are the library functions *called in the application*
    9.54 - * 
    9.55 - *There's a pattern for the outside sequential code to interact with the
    9.56 - * VMS_HW code.
    9.57 - *The VMS_HW system is inside a boundary..  every SSR system is in its
    9.58 - * own directory that contains the functions for each of the processor types.
    9.59 - * One of the processor types is the "seed" processor that starts the
    9.60 - * cascade of creating all the processors that do the work.
    9.61 - *So, in the directory is a file called "EntryPoint.c" that contains the
    9.62 - * function, named appropriately to the work performed, that the outside
    9.63 - * sequential code calls.  This function follows a pattern:
    9.64 - *1) it calls SSR__init()
    9.65 - *2) it creates the initial data for the seed processor, which is passed
    9.66 - *    in to the function
    9.67 - *3) it creates the seed SSR processor, with the data to start it with.
    9.68 - *4) it calls startSSRThenWaitUntilWorkDone
    9.69 - *5) it gets the returnValue from the transfer struc and returns that
    9.70 - *    from the function
    9.71 - *
    9.72 - *For now, a new SSR system has to be created via SSR__init every
    9.73 - * time an entry point function is called -- later, might add letting the
    9.74 - * SSR system be created once, and let all the entry points just reuse
    9.75 - * it -- want to be as simple as possible now, and see by using what makes
    9.76 - * sense for later..
    9.77 - */
    9.78 -
    9.79 -
    9.80 -
    9.81 -//===========================================================================
    9.82 -
    9.83 -/*This is the "border crossing" function -- the thing that crosses from the
    9.84 - * outside world, into the VMS_HW world.  It initializes and starts up the
    9.85 - * VMS system, then creates one processor from the specified function and
    9.86 - * puts it into the readyQ.  From that point, that one function is resp.
    9.87 - * for creating all the other processors, that then create others, and so
    9.88 - * forth.
    9.89 - *When all the processors, including the seed, have dissipated, then this
    9.90 - * function returns.  The results will have been written by side-effect via
    9.91 - * pointers read from, or written into initData.
    9.92 - *
    9.93 - *NOTE: no Threads should exist in the outside program that might touch
    9.94 - * any of the data reachable from initData passed in to here
    9.95 - */
    9.96 -void
    9.97 -SSR__create_seed_procr_and_do_work( TopLevelFnPtr fnPtr, void *initData )
    9.98 - { SSRSemEnv *semEnv;
    9.99 -   SlaveVP *seedPr;
   9.100 -
   9.101 -   #ifdef DEBUG__TURN_ON_SEQUENTIAL_MODE
   9.102 -   SSR__init_Seq();  //debug sequential exe
   9.103 -   #else
   9.104 -   SSR__init();      //normal multi-thd
   9.105 -   #endif
   9.106 -   semEnv = _VMSMasterEnv->semanticEnv;
   9.107 -
   9.108 -      //SSR starts with one processor, which is put into initial environ,
   9.109 -      // and which then calls create() to create more, thereby expanding work
   9.110 -   seedPr = SSR__create_procr_helper( fnPtr, initData,
   9.111 -                                      semEnv, semEnv->nextCoreToGetNewPr++ );
   9.112 -
   9.113 -   resume_slaveVP( seedPr, semEnv );
   9.114 -   
   9.115 -   #ifdef DEBUG__TURN_ON_SEQUENTIAL_MODE
   9.116 -   VMS_SS__start_the_work_then_wait_until_done_Seq();  //debug sequential exe
   9.117 -   #else
   9.118 -   VMS_SS__start_the_work_then_wait_until_done();      //normal multi-thd
   9.119 -   #endif
   9.120 -
   9.121 -   SSR__cleanup_after_shutdown();
   9.122 - }
   9.123 -
   9.124 -
   9.125 -int32
   9.126 -SSR__giveMinWorkUnitCycles( float32 percentOverhead )
   9.127 - {
   9.128 -   return MIN_WORK_UNIT_CYCLES;
   9.129 - }
   9.130 -
   9.131 -int32
   9.132 -SSR__giveIdealNumWorkUnits()
   9.133 - {
   9.134 -   return NUM_SCHED_SLOTS * NUM_CORES;
   9.135 - }
   9.136 -
   9.137 -int32
   9.138 -SSR__give_number_of_cores_to_schedule_onto()
   9.139 - {
   9.140 -   return NUM_CORES;
   9.141 - }
   9.142 -
   9.143 -/*For now, use TSC -- later, make these two macros with assembly that first
   9.144 - * saves jump point, and second jumps back several times to get reliable time
   9.145 - */
   9.146 -void
   9.147 -SSR__start_primitive()
   9.148 - { saveLowTimeStampCountInto( ((SSRSemEnv *)(_VMSMasterEnv->semanticEnv))->
   9.149 -                              primitiveStartTime );
   9.150 - }
   9.151 -
   9.152 -/*Just quick and dirty for now -- make reliable later
   9.153 - * will want this to jump back several times -- to be sure cache is warm
   9.154 - * because don't want comm time included in calc-time measurement -- and
   9.155 - * also to throw out any "weird" values due to OS interrupt or TSC rollover
   9.156 - */
   9.157 -int32
   9.158 -SSR__end_primitive_and_give_cycles()
   9.159 - { int32 endTime, startTime;
   9.160 -   //TODO: fix by repeating time-measurement
   9.161 -   saveLowTimeStampCountInto( endTime );
   9.162 -   startTime =((SSRSemEnv*)(_VMSMasterEnv->semanticEnv))->primitiveStartTime;
   9.163 -   return (endTime - startTime);
   9.164 - }
   9.165 -
   9.166 -//===========================================================================
   9.167 -
   9.168 -/*Initializes all the data-structures for a SSR system -- but doesn't
   9.169 - * start it running yet!
   9.170 - *
   9.171 - *This runs in the main thread -- before VMS starts up
   9.172 - * 
   9.173 - *This sets up the semantic layer over the VMS system
   9.174 - *
   9.175 - *First, calls VMS_Setup, then creates own environment, making it ready
   9.176 - * for creating the seed processor and then starting the work.
   9.177 - */
   9.178 -void
   9.179 -SSR__init()
   9.180 - {
   9.181 -   VMS_SS__init();
   9.182 -      //masterEnv, a global var, now is partially set up by init_VMS
   9.183 -      // after this, have VMS_int__malloc and VMS_int__free available
   9.184 -
   9.185 -   SSR__init_Helper();
   9.186 - }
   9.187 -
   9.188 -#ifdef DEBUG__TURN_ON_SEQUENTIAL_MODE
   9.189 -void
   9.190 -SSR__init_Seq()
   9.191 - {
   9.192 -   VMS_SS__init_Seq();
   9.193 -   flushRegisters();
   9.194 -      //masterEnv, a global var, now is partially set up by init_VMS
   9.195 -
   9.196 -   SSR__init_Helper();
   9.197 - }
   9.198 -#endif
   9.199 -
   9.200 -void idle_fn(void* data, SlaveVP *animatingSlv){
   9.201 -    while(1){
   9.202 -        VMS_int__suspend_slaveVP_and_send_req(animatingSlv);
   9.203 -    }
   9.204 -}
   9.205 -
   9.206 -void
   9.207 -SSR__init_Helper()
   9.208 - { SSRSemEnv       *semanticEnv;
   9.209 -   PrivQueueStruc **readyVPQs;
   9.210 -   int              coreIdx, i, j;
   9.211 - 
   9.212 -      //Hook up the semantic layer's plug-ins to the Master virt procr
   9.213 -   _VMSMasterEnv->requestHandler = &SSR__Request_Handler;
   9.214 -   _VMSMasterEnv->slaveAssigner  = &SSR__schedule_slaveVP;
   9.215 -   #ifdef HOLISTIC__TURN_ON_PERF_COUNTERS
   9.216 -   _VMSMasterEnv->counterHandler = &SSR__counter_handler;
   9.217 -   #endif
   9.218 -
   9.219 -      //create the semantic layer's environment (all its data) and add to
   9.220 -      // the master environment
   9.221 -   semanticEnv = VMS_int__malloc( sizeof( SSRSemEnv ) );
   9.222 -   _VMSMasterEnv->semanticEnv = semanticEnv;
   9.223 -   
   9.224 -   #ifdef HOLISTIC__TURN_ON_PERF_COUNTERS
   9.225 -   SSR__init_counter_data_structs();
   9.226 -   #endif
   9.227 -   for(i=0;i<NUM_CORES;++i){
   9.228 -       for(j=0;j<NUM_SCHED_SLOTS;++j){
   9.229 -           semanticEnv->idlePr[i][j] = VMS_int__create_slaveVP(&idle_fn,NULL);
   9.230 -           semanticEnv->idlePr[i][j]->coreAnimatedBy = i;
   9.231 -       }
   9.232 -   }
   9.233 -
   9.234 -   #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
   9.235 -   semanticEnv->unitList = makeListOfArrays(sizeof(Unit),128);
   9.236 -   semanticEnv->ctlDependenciesList = makeListOfArrays(sizeof(Dependency),128);
   9.237 -   semanticEnv->commDependenciesList = makeListOfArrays(sizeof(Dependency),128);
   9.238 -   semanticEnv->dynDependenciesList = makeListOfArrays(sizeof(Dependency),128);
   9.239 -   semanticEnv->ntonGroupsInfo = makePrivDynArrayOfSize((void***)&(semanticEnv->ntonGroups),8);
   9.240 -   
   9.241 -   semanticEnv->hwArcs = makeListOfArrays(sizeof(Dependency),128);
   9.242 -   memset(semanticEnv->last_in_slot,0,sizeof(NUM_CORES * NUM_SCHED_SLOTS * sizeof(Unit)));
   9.243 -   #endif
   9.244 -
   9.245 -      //create the ready queue, hash tables used for pairing send to receive
   9.246 -      // and so forth
   9.247 -      //TODO: add hash tables for pairing sends with receives, and
   9.248 -      // initialize the data ownership system
   9.249 -   readyVPQs = VMS_int__malloc( NUM_CORES * sizeof(PrivQueueStruc *) );
   9.250 -
   9.251 -   for( coreIdx = 0; coreIdx < NUM_CORES; coreIdx++ )
   9.252 -    {
   9.253 -      readyVPQs[ coreIdx ] = makeVMSQ();
   9.254 -    }
   9.255 -   
   9.256 -   semanticEnv->readyVPQs = readyVPQs;
   9.257 -   
   9.258 -   semanticEnv->nextCoreToGetNewPr = 0;
   9.259 -   semanticEnv->numSlaveVP = 0;
   9.260 -   
   9.261 -   semanticEnv->commHashTbl  = makeHashTable( 1<<16, &VMS_int__free );//start big
   9.262 -
   9.263 -   //TODO: bug -- turn these arrays into dyn arrays to eliminate limit
   9.264 -   //semanticEnv->singletonHasBeenExecutedFlags = makeDynArrayInfo( );
   9.265 -   //semanticEnv->transactionStrucs = makeDynArrayInfo( );
   9.266 -   for( i = 0; i < NUM_STRUCS_IN_SEM_ENV; i++ )
   9.267 -    {
   9.268 -      semanticEnv->fnSingletons[i].endInstrAddr      = NULL;
   9.269 -      semanticEnv->fnSingletons[i].hasBeenStarted    = FALSE;
   9.270 -      semanticEnv->fnSingletons[i].hasFinished       = FALSE;
   9.271 -      semanticEnv->fnSingletons[i].waitQ             = makeVMSQ();
   9.272 -      semanticEnv->transactionStrucs[i].waitingVPQ   = makeVMSQ();
   9.273 -    }
   9.274 - }
   9.275 -
   9.276 -
   9.277 -/*Frees any memory allocated by SSR__init() then calls VMS_int__shutdown
   9.278 - */
   9.279 -void
   9.280 -SSR__cleanup_after_shutdown()
   9.281 - { SSRSemEnv *semanticEnv;
   9.282 -   
   9.283 -   semanticEnv = _VMSMasterEnv->semanticEnv;
   9.284 -
   9.285 -   #ifdef HOLISTIC__TURN_ON_OBSERVE_UCC
   9.286 -   //UCC
   9.287 -   FILE* output;
   9.288 -   int n;
   9.289 -   char filename[255];    
   9.290 -    for(n=0;n<255;n++)
   9.291 -    {
   9.292 -        sprintf(filename, "./counters/UCC.%d",n);
   9.293 -        output = fopen(filename,"r");
   9.294 -        if(output)
   9.295 -        {
   9.296 -            fclose(output);
   9.297 -        }else{
   9.298 -            break;
   9.299 -        }
   9.300 -    }
   9.301 -   if(n<255){
   9.302 -    printf("Saving UCC to File: %s ...\n", filename);
   9.303 -    output = fopen(filename,"w+");
   9.304 -    if(output!=NULL){
   9.305 -        set_dependency_file(output);
   9.306 -        //fprintf(output,"digraph Dependencies {\n");
   9.307 -        //set_dot_file(output);
   9.308 -        //FIXME:  first line still depends on counters being enabled, replace w/ unit struct!
   9.309 -        //forAllInDynArrayDo(_VMSMasterEnv->counter_history_array_info, &print_dot_node_info );
   9.310 -        forAllInListOfArraysDo(semanticEnv->unitList, &print_unit_to_file);
   9.311 -        forAllInListOfArraysDo( semanticEnv->commDependenciesList, &print_comm_dependency_to_file );
   9.312 -        forAllInListOfArraysDo( semanticEnv->ctlDependenciesList, &print_ctl_dependency_to_file );
   9.313 -        forAllInDynArrayDo(semanticEnv->ntonGroupsInfo,&print_nton_to_file);
   9.314 -        //fprintf(output,"}\n");
   9.315 -        fflush(output);
   9.316 -
   9.317 -    } else
   9.318 -        printf("Opening UCC file failed. Please check that folder \"counters\" exists in run directory and has write permission.\n");
   9.319 -   } else {
   9.320 -       printf("Could not open UCC file, please clean \"counters\" folder. (Must contain less than 255 files.)\n");
   9.321 -   }
   9.322 -   //Loop Graph
   9.323 -   for(n=0;n<255;n++)
   9.324 -    {
   9.325 -        sprintf(filename, "./counters/LoopGraph.%d",n);
   9.326 -        output = fopen(filename,"r");
   9.327 -        if(output)
   9.328 -        {
   9.329 -            fclose(output);
   9.330 -        }else{
   9.331 -            break;
   9.332 -        }
   9.333 -    }
   9.334 -   if(n<255){
   9.335 -    printf("Saving LoopGraph to File: %s ...\n", filename);
   9.336 -    output = fopen(filename,"w+");
   9.337 -    if(output!=NULL){
   9.338 -        set_dependency_file(output);
   9.339 -        //fprintf(output,"digraph Dependencies {\n");
   9.340 -        //set_dot_file(output);
   9.341 -        //FIXME:  first line still depends on counters being enabled, replace w/ unit struct!
   9.342 -        //forAllInDynArrayDo(_VMSMasterEnv->counter_history_array_info, &print_dot_node_info );
   9.343 -        forAllInListOfArraysDo( semanticEnv->unitList, &print_unit_to_file );
   9.344 -        forAllInListOfArraysDo( semanticEnv->commDependenciesList, &print_comm_dependency_to_file );
   9.345 -        forAllInListOfArraysDo( semanticEnv->ctlDependenciesList, &print_ctl_dependency_to_file );
   9.346 -        forAllInListOfArraysDo( semanticEnv->dynDependenciesList, &print_dyn_dependency_to_file );
   9.347 -        forAllInListOfArraysDo( semanticEnv->hwArcs, &print_hw_dependency_to_file );
   9.348 -        //fprintf(output,"}\n");
   9.349 -        fflush(output);
   9.350 -
   9.351 -    } else
   9.352 -        printf("Opening LoopGraph file failed. Please check that folder \"counters\" exists in run directory and has write permission.\n");
   9.353 -   } else {
   9.354 -       printf("Could not open LoopGraph file, please clean \"counters\" folder. (Must contain less than 255 files.)\n");
   9.355 -   }
   9.356 -   
   9.357 -   
   9.358 -   freeListOfArrays(semanticEnv->unitList);
   9.359 -   freeListOfArrays(semanticEnv->commDependenciesList);
   9.360 -   freeListOfArrays(semanticEnv->ctlDependenciesList);
   9.361 -   freeListOfArrays(semanticEnv->dynDependenciesList);
   9.362 -   
   9.363 -   #endif
   9.364 -#ifdef HOLISTIC__TURN_ON_PERF_COUNTERS    
   9.365 -    for(n=0;n<255;n++)
   9.366 -    {
   9.367 -        sprintf(filename, "./counters/Counters.%d.csv",n);
   9.368 -        output = fopen(filename,"r");
   9.369 -        if(output)
   9.370 -        {
   9.371 -            fclose(output);
   9.372 -        }else{
   9.373 -            break;
   9.374 -        }
   9.375 -    }
   9.376 -    if(n<255){
   9.377 -    printf("Saving Counter measurements to File: %s ...\n", filename);
   9.378 -    output = fopen(filename,"w+");
   9.379 -    if(output!=NULL){
   9.380 -        set_counter_file(output);
   9.381 -        int i;
   9.382 -        for(i=0;i<NUM_CORES;i++){
   9.383 -            forAllInListOfArraysDo( semanticEnv->counterList[i], &print_counter_events_to_file );
   9.384 -            fflush(output);
   9.385 -        }
   9.386 -
   9.387 -    } else
   9.388 -        printf("Opening UCC file failed. Please check that folder \"counters\" exists in run directory and has write permission.\n");
   9.389 -   } else {
   9.390 -       printf("Could not open UCC file, please clean \"counters\" folder. (Must contain less than 255 files.)\n");
   9.391 -   }
   9.392 -    
   9.393 -#endif
   9.394 -/* It's all allocated inside VMS's big chunk -- that's about to be freed, so
   9.395 - *  nothing to do here
   9.396 -   
   9.397 -
   9.398 -   for( coreIdx = 0; coreIdx < NUM_CORES; coreIdx++ )
   9.399 -    {
   9.400 -      VMS_int__free( semanticEnv->readyVPQs[coreIdx]->startOfData );
   9.401 -      VMS_int__free( semanticEnv->readyVPQs[coreIdx] );
   9.402 -    }
   9.403 -   VMS_int__free( semanticEnv->readyVPQs );
   9.404 -   
   9.405 -   freeHashTable( semanticEnv->commHashTbl );
   9.406 -   VMS_int__free( _VMSMasterEnv->semanticEnv );
   9.407 - */
   9.408 -   VMS_SS__cleanup_at_end_of_shutdown();
   9.409 - }
   9.410 -
   9.411 -
   9.412 -//===========================================================================
   9.413 -
   9.414 -/*
   9.415 - */
   9.416 -  SlaveVP *
   9.417 -SSR__create_procr_with( TopLevelFnPtr fnPtr,   void *initData,
   9.418 -                        SlaveVP *creatingPr )
   9.419 - { SSRSemReq reqData;
   9.420 -
   9.421 -      //the semantic request data is on the stack and disappears when this
   9.422 -      // call returns -- it's guaranteed to remain in the VP's stack for as
   9.423 -      // long as the VP is suspended.
   9.424 -   reqData.reqType            = 0; //know type because in a VMS create req
   9.425 -   reqData.coreToScheduleOnto = -1; //means round-robin schedule
   9.426 -   reqData.fnPtr              = fnPtr;
   9.427 -   reqData.initData           = initData;
   9.428 -   reqData.sendPr             = creatingPr;
   9.429 -
   9.430 -   VMS_WL__send_create_slaveVP_req( &reqData, creatingPr );
   9.431 -
   9.432 -   return creatingPr->dataRetFromReq;
   9.433 - }
   9.434 -
   9.435 -  SlaveVP *
   9.436 -SSR__create_procr_with_affinity( TopLevelFnPtr fnPtr, void *initData,
   9.437 -                        SlaveVP *creatingPr,  int32  coreToScheduleOnto )
   9.438 - { SSRSemReq  reqData;
   9.439 -
   9.440 -      //the semantic request data is on the stack and disappears when this
   9.441 -      // call returns -- it's guaranteed to remain in the VP's stack for as
   9.442 -      // long as the VP is suspended.
   9.443 -   reqData.reqType            = 0; //know type because in a VMS create req
   9.444 -   reqData.coreToScheduleOnto = coreToScheduleOnto;
   9.445 -   reqData.fnPtr              = fnPtr;
   9.446 -   reqData.initData           = initData;
   9.447 -   reqData.sendPr             = creatingPr;
   9.448 -
   9.449 -   VMS_WL__send_create_slaveVP_req( &reqData, creatingPr );
   9.450 -
   9.451 -   return creatingPr->dataRetFromReq;
   9.452 - }
   9.453 -
   9.454 -
   9.455 -  void
   9.456 -SSR__dissipate_procr( SlaveVP *procrToDissipate )
   9.457 - {
   9.458 -   VMS_WL__send_dissipate_req( procrToDissipate );
   9.459 - }
   9.460 -
   9.461 -
   9.462 -//===========================================================================
   9.463 -
   9.464 -void *
   9.465 -SSR__malloc_to( int32 sizeToMalloc, SlaveVP *owningPr )
   9.466 - { SSRSemReq reqData;
   9.467 -
   9.468 -   reqData.reqType      = malloc_req;
   9.469 -   reqData.sendPr       = owningPr;
   9.470 -   reqData.sizeToMalloc = sizeToMalloc;
   9.471 -
   9.472 -   VMS_WL__send_sem_request( &reqData, owningPr );
   9.473 -
   9.474 -   return owningPr->dataRetFromReq;
   9.475 - }
   9.476 -
   9.477 -
   9.478 -/*Sends request to Master, which does the work of freeing
   9.479 - */
   9.480 -void
   9.481 -SSR__free( void *ptrToFree, SlaveVP *owningPr )
   9.482 - { SSRSemReq reqData;
   9.483 -
   9.484 -   reqData.reqType      = free_req;
   9.485 -   reqData.sendPr       = owningPr;
   9.486 -   reqData.ptrToFree    = ptrToFree;
   9.487 -
   9.488 -   VMS_WL__send_sem_request( &reqData, owningPr );
   9.489 - }
   9.490 -
   9.491 -
   9.492 -void
   9.493 -SSR__transfer_ownership_of_from_to( void *data, SlaveVP *oldOwnerSlv,
   9.494 -                                                  SlaveVP *newOwnerPr )
   9.495 - {
   9.496 -   //TODO: put in the ownership system that automatically frees when no
   9.497 -   // owners of data left -- will need keeper for keeping data around when
   9.498 -   // future created processors might need it but don't exist yet
   9.499 - }
   9.500 -
   9.501 -
   9.502 -void
   9.503 -SSR__add_ownership_by_to( SlaveVP *newOwnerSlv, void *data )
   9.504 - {
   9.505 -
   9.506 - }
   9.507 -
   9.508 -
   9.509 -void
   9.510 -SSR__remove_ownership_by_from( SlaveVP *loserSlv, void *dataLosing )
   9.511 - {
   9.512 -
   9.513 - }
   9.514 -
   9.515 -
   9.516 -/*Causes the SSR system to remove internal ownership, so data won't be
   9.517 - * freed when SSR shuts down, and will persist in the external program.
   9.518 - *
   9.519 - *Must be called from the processor that currently owns the data.
   9.520 - *
   9.521 - *IMPL: Transferring ownership touches two different virtual processor's
   9.522 - * state -- which means it has to be done carefully -- the VMS rules for
   9.523 - * semantic layers say that a work-unit is only allowed to touch the
   9.524 - * virtual processor it is part of, and that only a single work-unit per
   9.525 - * virtual processor be scheduled to a slave at a time.  So, this has to
   9.526 - * modify the virtual processor that owns the work-unit that called this
   9.527 - * function, then create a request to have the other processor modified.
   9.528 - *However, in this case, the TO processor is the outside, and transfers
   9.529 - * are only allowed to be called by the giver-upper, so can mark caller of
   9.530 - * this function as no longer owner, and return -- done.
   9.531 - */
   9.532 -void
   9.533 -SSR__transfer_ownership_to_outside( void *data )
   9.534 - {
   9.535 -   //TODO: removeAllOwnersFrom( data );
   9.536 - }
   9.537 -
   9.538 -
   9.539 -//===========================================================================
   9.540 -
   9.541 -void
   9.542 -SSR__send_of_type_to( SlaveVP *sendPr, void *msg, const int type,
   9.543 -                        SlaveVP *receivePr)
   9.544 - { SSRSemReq  reqData;
   9.545 -
   9.546 -   reqData.receivePr = receivePr;
   9.547 -   reqData.sendPr    = sendPr;
   9.548 -   reqData.reqType   = send_type;
   9.549 -   reqData.msgType   = type;
   9.550 -   reqData.msg       = msg;
   9.551 -   reqData.nextReqInHashEntry = NULL;
   9.552 -
   9.553 -      //On ownership -- remove inside the send and let ownership sit in limbo
   9.554 -      // as a potential in an entry in the hash table, when this receive msg
   9.555 -      // gets paired to a send, the ownership gets added to the receivePr --
   9.556 -      // the next work-unit in the receivePr's trace will have ownership.
   9.557 -   VMS_WL__send_sem_request( &reqData, sendPr );
   9.558 -
   9.559 -      //When come back from suspend, no longer own data reachable from msg
   9.560 -      //TODO: release ownership here
   9.561 - }
   9.562 -
   9.563 -void
   9.564 -SSR__send_from_to( void *msg, SlaveVP *sendPr, SlaveVP *receivePr )
   9.565 - { SSRSemReq  reqData;
   9.566 -
   9.567 -      //hash on the receiver, 'cause always know it, but sometimes want to
   9.568 -      // receive from anonymous sender
   9.569 -
   9.570 -   reqData.receivePr = receivePr;
   9.571 -   reqData.sendPr    = sendPr;
   9.572 -   reqData.reqType   = send_from_to;
   9.573 -   reqData.msg       = msg;
   9.574 -   reqData.nextReqInHashEntry = NULL;
   9.575 -
   9.576 -   VMS_WL__send_sem_request( &reqData, sendPr );
   9.577 - }
   9.578 -
   9.579 -
   9.580 -//===========================================================================
   9.581 -
   9.582 -void *
   9.583 -SSR__receive_any_to( SlaveVP *receivePr )
   9.584 - {
   9.585 -
   9.586 - }
   9.587 -
   9.588 -void *
   9.589 -SSR__receive_type_to( const int type, SlaveVP *receivePr )
   9.590 - { 
   9.591 -   SSRSemReq  reqData;
   9.592 -
   9.593 -   reqData.receivePr = receivePr;
   9.594 -   reqData.reqType   = receive_type;
   9.595 -   reqData.msgType   = type;
   9.596 -   reqData.nextReqInHashEntry = NULL;
   9.597 -
   9.598 -   VMS_WL__send_sem_request( &reqData, receivePr );
   9.599 -   
   9.600 -   return receivePr->dataRetFromReq;
   9.601 - }
   9.602 -
   9.603 -
   9.604 -
   9.605 -/*Call this at point receiving virt pr wants in-coming data.
   9.606 - * 
   9.607 - *The reason receivePr must call this is that it modifies the receivPr
   9.608 - * loc structure directly -- and the VMS rules state a virtual processor
   9.609 - * loc structure can only be modified by itself.
   9.610 - */
   9.611 -void *
   9.612 -SSR__receive_from_to( SlaveVP *sendPr, SlaveVP *receivePr )
   9.613 - { SSRSemReq  reqData;
   9.614 -
   9.615 -      //hash on the receiver, 'cause always know it, but sometimes want to
   9.616 -      // receive from anonymous sender
   9.617 -
   9.618 -   reqData.receivePr = receivePr;
   9.619 -   reqData.sendPr    = sendPr;
   9.620 -   reqData.reqType   = receive_from_to;
   9.621 -   reqData.nextReqInHashEntry = NULL;
   9.622 -
   9.623 -   VMS_WL__send_sem_request( &reqData, receivePr );
   9.624 -
   9.625 -   return receivePr->dataRetFromReq;
   9.626 - }
   9.627 -
   9.628 -
   9.629 -//===========================================================================
   9.630 -//
   9.631 -/*A function singleton is a function whose body executes exactly once, on a
   9.632 - * single core, no matter how many times the fuction is called and no
   9.633 - * matter how many cores or the timing of cores calling it.
   9.634 - *
   9.635 - *A data singleton is a ticket attached to data.  That ticket can be used
   9.636 - * to get the data through the function exactly once, no matter how many
   9.637 - * times the data is given to the function, and no matter the timing of
   9.638 - * trying to get the data through from different cores.
   9.639 - */
   9.640 -
   9.641 -/*asm function declarations*/
   9.642 -void asm_save_ret_to_singleton(SSRSingleton *singletonPtrAddr);
   9.643 -void asm_write_ret_from_singleton(SSRSingleton *singletonPtrAddr);
   9.644 -
   9.645 -/*Fn singleton uses ID as index into array of singleton structs held in the
   9.646 - * semantic environment.
   9.647 - */
   9.648 -void
   9.649 -SSR__start_fn_singleton( int32 singletonID,   SlaveVP *animPr )
   9.650 - {
   9.651 -   SSRSemReq  reqData;
   9.652 -
   9.653 -      //
   9.654 -   reqData.reqType     = singleton_fn_start;
   9.655 -   reqData.singletonID = singletonID;
   9.656 -
   9.657 -   VMS_WL__send_sem_request( &reqData, animPr );
   9.658 -   if( animPr->dataRetFromReq ) //will be 0 or addr of label in end singleton
   9.659 -    {
   9.660 -       SSRSemEnv *semEnv = VMS_int__give_sem_env_for( animPr );
   9.661 -       asm_write_ret_from_singleton(&(semEnv->fnSingletons[ singletonID]));
   9.662 -    }
   9.663 - }
   9.664 -
   9.665 -/*Data singleton hands addr of loc holding a pointer to a singleton struct.
   9.666 - * The start_data_singleton makes the structure and puts its addr into the
   9.667 - * location.
   9.668 - */
   9.669 -void
   9.670 -SSR__start_data_singleton( SSRSingleton **singletonAddr,  SlaveVP *animPr )
   9.671 - {
   9.672 -   SSRSemReq  reqData;
   9.673 -
   9.674 -   if( *singletonAddr && (*singletonAddr)->hasFinished )
   9.675 -       goto JmpToEndSingleton;
   9.676 -   
   9.677 -   reqData.reqType          = singleton_data_start;
   9.678 -   reqData.singletonPtrAddr = singletonAddr;
   9.679 -
   9.680 -   VMS_WL__send_sem_request( &reqData, animPr );
   9.681 -   if( animPr->dataRetFromReq ) //either 0 or end singleton's return addr
   9.682 -    {    //Assembly code changes the return addr on the stack to the one
   9.683 -         // saved into the singleton by the end-singleton-fn
   9.684 -         //The return addr is at 0x4(%%ebp)
   9.685 -        JmpToEndSingleton:
   9.686 -          asm_write_ret_from_singleton(*singletonAddr);
   9.687 -    }
   9.688 -   //now, simply return
   9.689 -   //will exit either from the start singleton call or the end-singleton call
   9.690 - }
   9.691 -
   9.692 -/*Uses ID as index into array of flags.  If flag already set, resumes from
   9.693 - * end-label.  Else, sets flag and resumes normally.
   9.694 - *
   9.695 - *Note, this call cannot be inlined because the instr addr at the label
   9.696 - * inside is shared by all invocations of a given singleton ID.
   9.697 - */
   9.698 -void
   9.699 -SSR__end_fn_singleton( int32 singletonID, SlaveVP *animPr )
   9.700 - {
   9.701 -   SSRSemReq  reqData;
   9.702 -
   9.703 -      //don't need this addr until after at least one singleton has reached
   9.704 -      // this function
   9.705 -   SSRSemEnv *semEnv = VMS_int__give_sem_env_for( animPr );
   9.706 -   asm_write_ret_from_singleton(&(semEnv->fnSingletons[ singletonID]));
   9.707 -
   9.708 -   reqData.reqType     = singleton_fn_end;
   9.709 -   reqData.singletonID = singletonID;
   9.710 -
   9.711 -   VMS_WL__send_sem_request( &reqData, animPr );
   9.712 -
   9.713 -EndSingletonInstrAddr:
   9.714 -   return;
   9.715 - }
   9.716 -
   9.717 -void
   9.718 -SSR__end_data_singleton(  SSRSingleton **singletonPtrAddr, SlaveVP *animPr )
   9.719 - {
   9.720 -   SSRSemReq  reqData;
   9.721 -
   9.722 -      //don't need this addr until after singleton struct has reached
   9.723 -      // this function for first time
   9.724 -      //do assembly that saves the return addr of this fn call into the
   9.725 -      // data singleton -- that data-singleton can only be given to exactly
   9.726 -      // one instance in the code of this function.  However, can use this
   9.727 -      // function in different places for different data-singletons.
   9.728 -//   (*(singletonAddr))->endInstrAddr =  &&EndDataSingletonInstrAddr;
   9.729 -
   9.730 -
   9.731 -   asm_save_ret_to_singleton(*singletonPtrAddr);
   9.732 -
   9.733 -   reqData.reqType          = singleton_data_end;
   9.734 -   reqData.singletonPtrAddr = singletonPtrAddr;
   9.735 -
   9.736 -   VMS_WL__send_sem_request( &reqData, animPr );
   9.737 - }
   9.738 -
   9.739 -/*This executes the function in the masterVP, so it executes in isolation
   9.740 - * from any other copies -- only one copy of the function can ever execute
   9.741 - * at a time.
   9.742 - *
   9.743 - *It suspends to the master, and the request handler takes the function
   9.744 - * pointer out of the request and calls it, then resumes the VP.
   9.745 - *Only very short functions should be called this way -- for longer-running
   9.746 - * isolation, use transaction-start and transaction-end, which run the code
   9.747 - * between as work-code.
   9.748 - */
   9.749 -void
   9.750 -SSR__animate_short_fn_in_isolation( PtrToAtomicFn ptrToFnToExecInMaster,
   9.751 -                                    void *data, SlaveVP *animPr )
   9.752 - {
   9.753 -   SSRSemReq  reqData;
   9.754 -
   9.755 -      //
   9.756 -   reqData.reqType          = atomic;
   9.757 -   reqData.fnToExecInMaster = ptrToFnToExecInMaster;
   9.758 -   reqData.dataForFn        = data;
   9.759 -
   9.760 -   VMS_WL__send_sem_request( &reqData, animPr );
   9.761 - }
   9.762 -
   9.763 -
   9.764 -/*This suspends to the master.
   9.765 - *First, it looks at the VP's data, to see the highest transactionID that VP
   9.766 - * already has entered.  If the current ID is not larger, it throws an
   9.767 - * exception stating a bug in the code.  Otherwise it puts the current ID
   9.768 - * there, and adds the ID to a linked list of IDs entered -- the list is
   9.769 - * used to check that exits are properly ordered.
   9.770 - *Next it is uses transactionID as index into an array of transaction
   9.771 - * structures.
   9.772 - *If the "VP_currently_executing" field is non-null, then put requesting VP
   9.773 - * into queue in the struct.  (At some point a holder will request
   9.774 - * end-transaction, which will take this VP from the queue and resume it.)
   9.775 - *If NULL, then write requesting into the field and resume.
   9.776 - */
   9.777 -void
   9.778 -SSR__start_transaction( int32 transactionID, SlaveVP *animPr )
   9.779 - {
   9.780 -   SSRSemReq  reqData;
   9.781 -
   9.782 -      //
   9.783 -   reqData.sendPr      = animPr;
   9.784 -   reqData.reqType     = trans_start;
   9.785 -   reqData.transID     = transactionID;
   9.786 -
   9.787 -   VMS_WL__send_sem_request( &reqData, animPr );
   9.788 - }
   9.789 -
   9.790 -/*This suspends to the master, then uses transactionID as index into an
   9.791 - * array of transaction structures.
   9.792 - *It looks at VP_currently_executing to be sure it's same as requesting VP.
   9.793 - * If different, throws an exception, stating there's a bug in the code.
   9.794 - *Next it looks at the queue in the structure.
   9.795 - *If it's empty, it sets VP_currently_executing field to NULL and resumes.
   9.796 - *If something in, gets it, sets VP_currently_executing to that VP, then
   9.797 - * resumes both.
   9.798 - */
   9.799 -void
   9.800 -SSR__end_transaction( int32 transactionID, SlaveVP *animPr )
   9.801 - {
   9.802 -   SSRSemReq  reqData;
   9.803 -
   9.804 -      //
   9.805 -   reqData.sendPr      = animPr;
   9.806 -   reqData.reqType     = trans_end;
   9.807 -   reqData.transID     = transactionID;
   9.808 -
   9.809 -   VMS_WL__send_sem_request( &reqData, animPr );
   9.810 - }

    10.1 --- a/dependency.c	Tue Mar 13 10:04:14 2012 -0700
    10.2 +++ b/dependency.c	Tue Mar 13 18:30:05 2012 -0700
    10.3 @@ -1,5 +1,5 @@
    10.4  #include "dependency.h"
    10.5 -#include "../VMS_impl/VMS.h"
    10.6 +#include "VMS_impl/VMS.h"
    10.7  
    10.8  Dependency* new_dependency(int from_vp, int from_task, int to_vp, int to_task){
    10.9      Dependency* newDep = (Dependency*) VMS_int__malloc(sizeof(Dependency));