1.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     1.2 +++ b/Hardware_Dependent/VMS__HW_measurement.h	Sat Mar 10 20:38:25 2012 -0800
     1.3 @@ -0,0 +1,49 @@
     1.4 +/*
     1.5 + *  Copyright 2009 OpenSourceStewardshipFoundation.org
     1.6 + *  Licensed under GNU General Public License version 2
     1.7 + *
     1.8 + * Author: seanhalle@yahoo.com
     1.9 + * 
    1.10 + */
    1.11 +
    1.12 +#ifndef _VMS__HW_MEASUREMENT_H
    1.13 +#define	_VMS__HW_MEASUREMENT_H
    1.14 +#define _GNU_SOURCE
    1.15 +
    1.16 +//===================  Macros to Capture Measurements  ======================
    1.17 +//
    1.18 +//===== RDTSC wrapper ===== 
    1.19 +//Also runs with x86_64 code
    1.20 +#define saveTSCLowHigh(lowHighIn) \
    1.21 +   asm volatile("RDTSC;                   \
    1.22 +                 movl %%eax, %0;          \
    1.23 +                 movl %%edx, %1;"         \
    1.24 +   /* outputs */ : "=m" (lowHighIn.lowHigh[0]), "=m" (lowHighIn.lowHigh[1])\
    1.25 +   /* inputs  */ :                        \
    1.26 +   /* clobber */ : "%eax", "%edx"         \
    1.27 +                );
    1.28 +
    1.29 +#define saveTimeStampCountInto(low, high) \
    1.30 +   asm volatile("RDTSC;                   \
    1.31 +                 movl %%eax, %0;          \
    1.32 +                 movl %%edx, %1;"         \
    1.33 +   /* outputs */ : "=m" (low), "=m" (high)\
    1.34 +   /* inputs  */ :                        \
    1.35 +   /* clobber */ : "%eax", "%edx"         \
    1.36 +                );
    1.37 +
    1.38 +#define saveLowTimeStampCountInto(low)    \
    1.39 +   asm volatile("RDTSC;                   \
    1.40 +                 movl %%eax, %0;"         \
    1.41 +   /* outputs */ : "=m" (low)             \
    1.42 +   /* inputs  */ :                        \
    1.43 +   /* clobber */ : "%eax", "%edx"         \
    1.44 +                );
    1.45 +
    1.46 +   //For code that calculates normalization-offset between TSC counts of
    1.47 +   // different cores.
    1.48 +//#define NUM_TSC_ROUND_TRIPS 10
    1.49 +
    1.50 +
    1.51 +#endif	/* */
    1.52 +

     2.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     2.2 +++ b/Hardware_Dependent/VMS__primitives.c	Sat Mar 10 20:38:25 2012 -0800
     2.3 @@ -0,0 +1,53 @@
     2.4 +/*
     2.5 + * This File contains all hardware dependent C code.
     2.6 + */
     2.7 +
     2.8 +
     2.9 +#include "../VMS.h"
    2.10 +
    2.11 +/*Set up the stack with __cdecl structure on it
    2.12 + * Except doing a trick for 64 bits, where put top-level fn pointer on
    2.13 + * stack, then call an assembly helper that copies it into a reg and
    2.14 + * jumps to it.  So, set the resumeInstrPtr to the helper-assembly.
    2.15 + *No need to save registers on old stack frame, because there's no old
    2.16 + * animator state to return to
    2.17 + * 
    2.18 + *This was factored into separate function because it's used stand-alone in
    2.19 + * some wrapper-libraries (but only "int" version, to warn users to check
    2.20 + * carefully that it's safe)
    2.21 + */
    2.22 +inline void
    2.23 +VMS_int__point_slaveVP_to_Fn( SlaveVP *slaveVP, TopLevelFnPtr fnPtr,
    2.24 +                            void    *dataParam)
    2.25 + { void  *stackPtr;
    2.26 +
    2.27 +// Start of Hardware dependent part           
    2.28 +   
    2.29 +    //Set slave's instr pointer to a helper Fn that copies params from stack
    2.30 +   slaveVP->resumeInstrPtr  = (TopLevelFnPtr)&startUpTopLevelFn;
    2.31 +   
    2.32 +    //fnPtr takes two params -- void *dataParam & void *animSlv
    2.33 +    // Stack grows *down*, so start it at highest stack addr, minus room
    2.34 +    // for 2 params + return addr. 
    2.35 +   stackPtr = 
    2.36 +     (void *)slaveVP->startOfStack + VIRT_PROCR_STACK_SIZE - 4*sizeof(void*);
    2.37 +  
    2.38 +    //setup __cdecl on stack
    2.39 +    //Normally, return Addr is in loc pointed to by stackPtr, but doing a
    2.40 +    // trick for 64 bit arch, where put ptr to top-level fn there instead,
    2.41 +    // and set resumeInstrPtr to a helper-fn that copies the top-level
    2.42 +    // fn ptr and params into registers.
    2.43 +    //Then, dataParam is at stackPtr + 8 bytes, & animating SlaveVP above
    2.44 +   *((SlaveVP**)stackPtr + 2 ) = slaveVP; //rightmost param
    2.45 +   *((void**)stackPtr + 1 ) = dataParam;  //next  param to left
    2.46 +   *((void**)stackPtr) = (void*)fnPtr;    //copied to reg by helper Fn
    2.47 +   
    2.48 +  
    2.49 +// end of Hardware dependent part           
    2.50 +   
    2.51 +      //core controller will switch to stack & frame pointers stored in slave,
    2.52 +      // suspend will save processor's stack and frame into slave
    2.53 +   slaveVP->stackPtr = slaveVP->startOfStack; 
    2.54 +   slaveVP->framePtr = slaveVP->startOfStack; 
    2.55 + }
    2.56 +

     3.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     3.2 +++ b/Hardware_Dependent/VMS__primitives.h	Sat Mar 10 20:38:25 2012 -0800
     3.3 @@ -0,0 +1,41 @@
     3.4 +/*
     3.5 + *  Copyright 2009 OpenSourceStewardshipFoundation.org
     3.6 + *  Licensed under GNU General Public License version 2
     3.7 + *
     3.8 + * Author: seanhalle@yahoo.com
     3.9 + * 
    3.10 + */
    3.11 +
    3.12 +#ifndef _VMS__PRIMITIVES_H
    3.13 +#define	_VMS__PRIMITIVES_H
    3.14 +#define _GNU_SOURCE
    3.15 +
    3.16 +void 
    3.17 +recordCoreCtlrReturnLabelAddr(void **returnAddress);
    3.18 +
    3.19 +void 
    3.20 +switchToSlv(SlaveVP *nextSlave);
    3.21 +
    3.22 +void 
    3.23 +switchToCoreCtlr(SlaveVP *nextSlave);
    3.24 +
    3.25 +void 
    3.26 +masterSwitchToCoreCtlr(SlaveVP *nextSlave);
    3.27 +
    3.28 +void 
    3.29 +startUpTopLevelFn();
    3.30 +
    3.31 +void *
    3.32 +asmTerminateCoreCtlr(SlaveVP *currSlv);
    3.33 +
    3.34 +#define flushRegisters() \
    3.35 +        asm volatile ("":::"%rbx", "%r12", "%r13","%r14","%r15")
    3.36 +
    3.37 +void
    3.38 +VMS_int__save_return_into_ptd_to_loc_then_do_ret(void *ptdToLoc);
    3.39 +
    3.40 +void
    3.41 +VMS_int__return_to_addr_in_ptd_to_loc(void *ptdToLoc);
    3.42 +
    3.43 +#endif	/* _VMS__HW_DEPENDENT_H */
    3.44 +

     4.1 --- /dev/null	Thu Jan 01 00:00:00 1970 +0000
     4.2 +++ b/Hardware_Dependent/VMS__primitives.s	Sat Mar 10 20:38:25 2012 -0800
     4.3 @@ -0,0 +1,167 @@
     4.4 +.data
     4.5 +
     4.6 +
     4.7 +.text
     4.8 +
     4.9 +//Save return label address for the coreCtlr to pointer
    4.10 +//Arguments: Pointer to variable holding address
    4.11 +.globl recordCoreCtlrReturnLabelAddr
    4.12 +recordCoreCtlrReturnLabelAddr:
    4.13 +    movq    $coreCtlrReturn, %rcx   #load label address
    4.14 +    movq    %rcx, (%rdi)           #save address to pointer
    4.15 +    ret
    4.16 +
    4.17 +
    4.18 +//Trick for 64 bit arch -- copies args from stack into regs, then does jmp to
    4.19 +// the top-level function, which was pointed to by the stack-ptr
    4.20 +.globl startUpTopLevelFn
    4.21 +startUpTopLevelFn:
    4.22 +    movq    %rdi      , %rsi #get second argument from first argument of switchSlv
    4.23 +    movq    0x08(%rsp), %rdi #get first argument from stack
    4.24 +    movq    (%rsp)    , %rax #get top-level function's addr from stack
    4.25 +    jmp     *%rax            #jump to the top-level function
    4.26 +
    4.27 +//Switches form CoreCtlr to either a normal Slv VP or the Master VP
    4.28 +//switch to VP's stack and frame ptr then jump to VP's next-instr-ptr
    4.29 +/* SlaveVP  offsets:
    4.30 + * 0x10  stackPtr
    4.31 + * 0x18 framePtr
    4.32 + * 0x20 resumeInstrPtr
    4.33 + * 0x30 coreCtlrFramePtr
    4.34 + * 0x38 coreCtlrStackPtr
    4.35 + *
    4.36 + * _VMSMasterEnv  offsets:
    4.37 + * 0x48 coreCtlrReturnPt
    4.38 + * 0x54 masterLock
    4.39 + */
    4.40 +.globl switchToSlv
    4.41 +switchToSlv:
    4.42 +    #SlaveVP in %rdi
    4.43 +    movq    %rsp      , 0x38(%rdi)   #save core ctlr stack pointer 
    4.44 +    movq    %rbp      , 0x30(%rdi)   #save core ctlr frame pointer
    4.45 +    movq    0x10(%rdi), %rsp         #restore stack pointer
    4.46 +    movq    0x18(%rdi), %rbp         #restore frame pointer
    4.47 +    movq    0x20(%rdi), %rax         #get jmp pointer
    4.48 +    jmp     *%rax                    #jmp to Slv
    4.49 +coreCtlrReturn:
    4.50 +    ret
    4.51 +
    4.52 +    
    4.53 +//switches to core controller. saves return address
    4.54 +/* SlaveVP  offsets:
    4.55 + * 0x10  stackPtr
    4.56 + * 0x18 framePtr
    4.57 + * 0x20 resumeInstrPtr
    4.58 + * 0x30 coreCtlrFramePtr
    4.59 + * 0x38 coreCtlrStackPtr
    4.60 + *
    4.61 + * _VMSMasterEnv  offsets:
    4.62 + * 0x48 coreCtlrReturnPt
    4.63 + * 0x54 masterLock
    4.64 + */
    4.65 +.globl switchToCoreCtlr
    4.66 +switchToCoreCtlr:
    4.67 +    #SlaveVP in %rdi
    4.68 +    movq    $SlvReturn, 0x20(%rdi)   #store return address
    4.69 +    movq    %rsp      , 0x10(%rdi)   #save stack pointer 
    4.70 +    movq    %rbp      , 0x18(%rdi)   #save frame pointer
    4.71 +    movq    0x38(%rdi), %rsp         #restore stack pointer
    4.72 +    movq    0x30(%rdi), %rbp         #restore frame pointer
    4.73 +    movq    $_VMSMasterEnv, %rcx
    4.74 +    movq    (%rcx)    , %rcx
    4.75 +    movq    0x48(%rcx), %rax         #get CoreCtlrStartPt
    4.76 +    jmp     *%rax                    #jmp to CoreCtlr
    4.77 +SlvReturn:
    4.78 +    ret
    4.79 +
    4.80 +
    4.81 +
    4.82 +//switches to core controller from master. saves return address
    4.83 +//Releases masterLock so the next MasterLoop can be executed
    4.84 +/* SlaveVP  offsets:
    4.85 + * 0x10  stackPtr
    4.86 + * 0x18 framePtr
    4.87 + * 0x20 resumeInstrPtr
    4.88 + * 0x30 coreCtlrFramePtr
    4.89 + * 0x38 coreCtlrStackPtr
    4.90 + *
    4.91 + * _VMSMasterEnv  offsets:
    4.92 + * 0x48 coreCtlrReturnPt
    4.93 + * 0x54 masterLock
    4.94 + */
    4.95 +.globl masterSwitchToCoreCtlr
    4.96 +masterSwitchToCoreCtlr:
    4.97 +    #SlaveVP in %rdi
    4.98 +    movq    $MasterReturn, 0x20(%rdi)   #store return address
    4.99 +    movq    %rsp      , 0x10(%rdi)   #save stack pointer 
   4.100 +    movq    %rbp      , 0x18(%rdi)   #save frame pointer
   4.101 +    movq    0x38(%rdi), %rsp         #restore stack pointer
   4.102 +    movq    0x30(%rdi), %rbp         #restore frame pointer
   4.103 +    movq    $_VMSMasterEnv, %rcx
   4.104 +    movq    (%rcx)    , %rcx
   4.105 +    movq    0x48(%rcx), %rax         #get CoreCtlr return pt
   4.106 +    movl    $0x0      , 0x54(%rcx)   #release lock
   4.107 +    jmp     *%rax                    #jmp to CoreCtlr
   4.108 +MasterReturn:
   4.109 +    ret
   4.110 +
   4.111 +
   4.112 +//Switch to terminateCoreCtlr
   4.113 +//therefor switch to coreCtlr context from master context
   4.114 +// no need to call because the stack is already set up for switchSlv
   4.115 +// and Slv is in %rdi
   4.116 +// and both functions have the same argument.
   4.117 +// do not save register of Slv because this function will never return
   4.118 +/* SlaveVP  offsets:
   4.119 + * 0x10  stackPtr
   4.120 + * 0x18 framePtr
   4.121 + * 0x20 resumeInstrPtr
   4.122 + * 0x30 coreCtlrFramePtr
   4.123 + * 0x38 coreCtlrStackPtr
   4.124 + *
   4.125 + * _VMSMasterEnv  offsets:
   4.126 + * 0x48 coreCtlrReturnPt
   4.127 + * 0x58 masterLock
   4.128 + */
   4.129 +.globl asmTerminateCoreCtlr
   4.130 +asmTerminateCoreCtlr:
   4.131 +    #SlaveVP in %rdi
   4.132 +    movq    0x38(%rdi), %rsp         #restore stack pointer
   4.133 +    movq    0x30(%rdi), %rbp         #restore frame pointer
   4.134 +    movq    $terminateCoreCtlr, %rax
   4.135 +    jmp     *%rax                    #jmp to CoreCtlr
   4.136 +
   4.137 +
   4.138 +/*
   4.139 + * This one for the sequential version is special. It discards the current stack
   4.140 + * and returns directly from the coreCtlr after VMS_WL__dissipate_slaveVP was called
   4.141 + */
   4.142 +.globl asmTerminateCoreCtlrSeq
   4.143 +asmTerminateCoreCtlrSeq:
   4.144 +    #SlaveVP in %rdi
   4.145 +    movq    0x38(%rdi), %rsp         #restore stack pointer
   4.146 +    movq    0x30(%rdi), %rbp         #restore frame pointer
   4.147 +    #argument is in %rdi
   4.148 +    call    VMS_int__dissipate_slaveVP
   4.149 +    movq    %rbp      , %rsp        #goto the coreCtlrs stack
   4.150 +    pop     %rbp        #restore the old framepointer
   4.151 +    ret                 #return from core controller
   4.152 +    
   4.153 +
   4.154 +//Takes the return addr off the stack and saves into the loc pointed to by
   4.155 +// by the parameter passed in via rdi.  Return addr is at 0x8(%rbp) for 64bit
   4.156 +.globl VMS_int__save_return_into_ptd_to_loc_then_do_ret
   4.157 +VMS_int__save_return_into_ptd_to_loc_then_do_ret:
   4.158 +    movq 0x8(%rbp),     %rax  #get ret address, rbp is the same as in the calling function
   4.159 +    movq     %rax,     (%rdi) #write ret addr into addr passed as param field
   4.160 +    ret
   4.161 +
   4.162 +
   4.163 +//Assembly code changes the return addr on the stack to the one
   4.164 +// pointed to by the parameter, then returns. Stack's return addr is at 0x8(%rbp)
   4.165 +.globl VMS_int__return_to_addr_in_ptd_to_loc
   4.166 +VMS_int__return_to_addr_in_ptd_to_loc:
   4.167 +    movq    (%rdi),    %rax  #get return addr from addr passed as param
   4.168 +    movq     %rax, 0x8(%rbp) #write return addr to the stack of the caller
   4.169 +    ret
   4.170 +

     5.1 --- a/VMS__HW_dependent.c	Sat Mar 10 20:35:38 2012 -0800
     5.2 +++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
     5.3 @@ -1,53 +0,0 @@
     5.4 -/*
     5.5 - * This File contains all hardware dependent C code.
     5.6 - */
     5.7 -
     5.8 -
     5.9 -#include "VMS.h"
    5.10 -
    5.11 -/*Set up the stack with __cdecl structure on it
    5.12 - * Except doing a trick for 64 bits, where put top-level fn pointer on
    5.13 - * stack, then call an assembly helper that copies it into a reg and
    5.14 - * jumps to it.  So, set the resumeInstrPtr to the helper-assembly.
    5.15 - *No need to save registers on old stack frame, because there's no old
    5.16 - * animator state to return to
    5.17 - * 
    5.18 - *This was factored into separate function because it's used stand-alone in
    5.19 - * some wrapper-libraries (but only "int" version, to warn users to check
    5.20 - * carefully that it's safe)
    5.21 - */
    5.22 -inline void
    5.23 -VMS_int__point_slaveVP_to_Fn( SlaveVP *slaveVP, TopLevelFnPtr fnPtr,
    5.24 -                            void    *dataParam)
    5.25 - { void  *stackPtr;
    5.26 -
    5.27 -// Start of Hardware dependent part           
    5.28 -   
    5.29 -    //Set slave's instr pointer to a helper Fn that copies params from stack
    5.30 -   slaveVP->resumeInstrPtr  = (TopLevelFnPtr)&startUpTopLevelFn;
    5.31 -   
    5.32 -    //fnPtr takes two params -- void *dataParam & void *animSlv
    5.33 -    // Stack grows *down*, so start it at highest stack addr, minus room
    5.34 -    // for 2 params + return addr. 
    5.35 -   stackPtr = 
    5.36 -     (void *)slaveVP->startOfStack + VIRT_PROCR_STACK_SIZE - 4*sizeof(void*);
    5.37 -  
    5.38 -    //setup __cdecl on stack
    5.39 -    //Normally, return Addr is in loc pointed to by stackPtr, but doing a
    5.40 -    // trick for 64 bit arch, where put ptr to top-level fn there instead,
    5.41 -    // and set resumeInstrPtr to a helper-fn that copies the top-level
    5.42 -    // fn ptr and params into registers.
    5.43 -    //Then, dataParam is at stackPtr + 8 bytes, & animating SlaveVP above
    5.44 -   *((SlaveVP**)stackPtr + 2 ) = slaveVP; //rightmost param
    5.45 -   *((void**)stackPtr + 1 ) = dataParam;  //next  param to left
    5.46 -   *((void**)stackPtr) = (void*)fnPtr;    //copied to reg by helper Fn
    5.47 -   
    5.48 -  
    5.49 -// end of Hardware dependent part           
    5.50 -   
    5.51 -      //core loop will switch to stack & frame pointers stored in slave,
    5.52 -      // suspend will save processor's stack and frame into slave
    5.53 -   slaveVP->stackPtr = slaveVP->startOfStack; 
    5.54 -   slaveVP->framePtr = slaveVP->startOfStack; 
    5.55 - }
    5.56 -

     6.1 --- a/VMS__HW_dependent.h	Sat Mar 10 20:35:38 2012 -0800
     6.2 +++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
     6.3 @@ -1,76 +0,0 @@
     6.4 -/*
     6.5 - *  Copyright 2009 OpenSourceStewardshipFoundation.org
     6.6 - *  Licensed under GNU General Public License version 2
     6.7 - *
     6.8 - * Author: seanhalle@yahoo.com
     6.9 - * 
    6.10 - */
    6.11 -
    6.12 -#ifndef _VMS__HW_DEPENDENT_H
    6.13 -#define	_VMS__HW_DEPENDENT_H
    6.14 -#define _GNU_SOURCE
    6.15 -
    6.16 -void 
    6.17 -recordCoreLoopReturnLabelAddr(void **returnAddress);
    6.18 -
    6.19 -void 
    6.20 -switchToSlv(SlaveVP *nextSlave);
    6.21 -
    6.22 -void 
    6.23 -switchToCoreLoop(SlaveVP *nextSlave);
    6.24 -
    6.25 -void 
    6.26 -masterSwitchToCoreLoop(SlaveVP *nextSlave);
    6.27 -
    6.28 -void 
    6.29 -startUpTopLevelFn();
    6.30 -
    6.31 -void *
    6.32 -asmTerminateCoreLoop(SlaveVP *currSlv);
    6.33 -
    6.34 -#define flushRegisters() \
    6.35 -        asm volatile ("":::"%rbx", "%r12", "%r13","%r14","%r15")
    6.36 -
    6.37 -void
    6.38 -VMS_int__save_return_into_ptd_to_loc_then_do_ret(void *ptdToLoc);
    6.39 -
    6.40 -void
    6.41 -VMS_int__return_to_addr_in_ptd_to_loc(void *ptdToLoc);
    6.42 -
    6.43 -//===================  Macros to Capture Measurements  ======================
    6.44 -//
    6.45 -//===== RDTSC wrapper ===== 
    6.46 -//Also runs with x86_64 code
    6.47 -#define saveTSCLowHigh(lowHighIn) \
    6.48 -   asm volatile("RDTSC;                   \
    6.49 -                 movl %%eax, %0;          \
    6.50 -                 movl %%edx, %1;"         \
    6.51 -   /* outputs */ : "=m" (lowHighIn.lowHigh[0]), "=m" (lowHighIn.lowHigh[1])\
    6.52 -   /* inputs  */ :                        \
    6.53 -   /* clobber */ : "%eax", "%edx"         \
    6.54 -                );
    6.55 -
    6.56 -#define saveTimeStampCountInto(low, high) \
    6.57 -   asm volatile("RDTSC;                   \
    6.58 -                 movl %%eax, %0;          \
    6.59 -                 movl %%edx, %1;"         \
    6.60 -   /* outputs */ : "=m" (low), "=m" (high)\
    6.61 -   /* inputs  */ :                        \
    6.62 -   /* clobber */ : "%eax", "%edx"         \
    6.63 -                );
    6.64 -
    6.65 -#define saveLowTimeStampCountInto(low)    \
    6.66 -   asm volatile("RDTSC;                   \
    6.67 -                 movl %%eax, %0;"         \
    6.68 -   /* outputs */ : "=m" (low)             \
    6.69 -   /* inputs  */ :                        \
    6.70 -   /* clobber */ : "%eax", "%edx"         \
    6.71 -                );
    6.72 -
    6.73 -   //For code that calculates normalization-offset between TSC counts of
    6.74 -   // different cores.
    6.75 -//#define NUM_TSC_ROUND_TRIPS 10
    6.76 -
    6.77 -
    6.78 -#endif	/* _VMS__HW_DEPENDENT_H */
    6.79 -

     7.1 --- a/VMS__HW_dependent.s	Sat Mar 10 20:35:38 2012 -0800
     7.2 +++ /dev/null	Thu Jan 01 00:00:00 1970 +0000
     7.3 @@ -1,167 +0,0 @@
     7.4 -.data
     7.5 -
     7.6 -
     7.7 -.text
     7.8 -
     7.9 -//Save return label address for the coreLoop to pointer
    7.10 -//Arguments: Pointer to variable holding address
    7.11 -.globl saveCoreLoopReturnAddr
    7.12 -recordCoreLoopReturnLabelAddr:
    7.13 -    movq    $coreLoopReturn, %rcx   #load label address
    7.14 -    movq    %rcx, (%rdi)           #save address to pointer
    7.15 -    ret
    7.16 -
    7.17 -
    7.18 -//Trick for 64 bit arch -- copies args from stack into regs, then does jmp to
    7.19 -// the top-level function, which was pointed to by the stack-ptr
    7.20 -.globl startUpTopLevelFn
    7.21 -startUpTopLevelFn:
    7.22 -    movq    %rdi      , %rsi #get second argument from first argument of switchSlv
    7.23 -    movq    0x08(%rsp), %rdi #get first argument from stack
    7.24 -    movq    (%rsp)    , %rax #get top-level function's addr from stack
    7.25 -    jmp     *%rax            #jump to the top-level function
    7.26 -
    7.27 -//Switches form CoreLoop to either a normal Slv VP or the Master VP
    7.28 -//switch to VP's stack and frame ptr then jump to VP's next-instr-ptr
    7.29 -/* SlaveVP  offsets:
    7.30 - * 0x10  stackPtr
    7.31 - * 0x18 framePtr
    7.32 - * 0x20 resumeInstrPtr
    7.33 - * 0x30 coreLoopFramePtr
    7.34 - * 0x38 coreLoopStackPtr
    7.35 - *
    7.36 - * _VMSMasterEnv  offsets:
    7.37 - * 0x48 coreLoopReturnPt
    7.38 - * 0x54 masterLock
    7.39 - */
    7.40 -.globl switchToSlv
    7.41 -switchToSlv:
    7.42 -    #SlaveVP in %rdi
    7.43 -    movq    %rsp      , 0x38(%rdi)   #save core loop stack pointer 
    7.44 -    movq    %rbp      , 0x30(%rdi)   #save core loop frame pointer
    7.45 -    movq    0x10(%rdi), %rsp         #restore stack pointer
    7.46 -    movq    0x18(%rdi), %rbp         #restore frame pointer
    7.47 -    movq    0x20(%rdi), %rax         #get jmp pointer
    7.48 -    jmp     *%rax                    #jmp to Slv
    7.49 -coreLoopReturn:
    7.50 -    ret
    7.51 -
    7.52 -    
    7.53 -//switches to core loop. saves return address
    7.54 -/* SlaveVP  offsets:
    7.55 - * 0x10  stackPtr
    7.56 - * 0x18 framePtr
    7.57 - * 0x20 resumeInstrPtr
    7.58 - * 0x30 coreLoopFramePtr
    7.59 - * 0x38 coreLoopStackPtr
    7.60 - *
    7.61 - * _VMSMasterEnv  offsets:
    7.62 - * 0x48 coreLoopReturnPt
    7.63 - * 0x54 masterLock
    7.64 - */
    7.65 -.globl switchToCoreLoop
    7.66 -switchToCoreLoop:
    7.67 -    #SlaveVP in %rdi
    7.68 -    movq    $SlvReturn, 0x20(%rdi)   #store return address
    7.69 -    movq    %rsp      , 0x10(%rdi)   #save stack pointer 
    7.70 -    movq    %rbp      , 0x18(%rdi)   #save frame pointer
    7.71 -    movq    0x38(%rdi), %rsp         #restore stack pointer
    7.72 -    movq    0x30(%rdi), %rbp         #restore frame pointer
    7.73 -    movq    $_VMSMasterEnv, %rcx
    7.74 -    movq    (%rcx)    , %rcx
    7.75 -    movq    0x48(%rcx), %rax         #get CoreLoopStartPt
    7.76 -    jmp     *%rax                    #jmp to CoreLoop
    7.77 -SlvReturn:
    7.78 -    ret
    7.79 -
    7.80 -
    7.81 -
    7.82 -//switches to core loop from master. saves return address
    7.83 -//Releases masterLock so the next MasterLoop can be executed
    7.84 -/* SlaveVP  offsets:
    7.85 - * 0x10  stackPtr
    7.86 - * 0x18 framePtr
    7.87 - * 0x20 resumeInstrPtr
    7.88 - * 0x30 coreLoopFramePtr
    7.89 - * 0x38 coreLoopStackPtr
    7.90 - *
    7.91 - * _VMSMasterEnv  offsets:
    7.92 - * 0x48 coreLoopReturnPt
    7.93 - * 0x54 masterLock
    7.94 - */
    7.95 -.globl masterSwitchToCoreLoop
    7.96 -masterSwitchToCoreLoop:
    7.97 -    #SlaveVP in %rdi
    7.98 -    movq    $MasterReturn, 0x20(%rdi)   #store return address
    7.99 -    movq    %rsp      , 0x10(%rdi)   #save stack pointer 
   7.100 -    movq    %rbp      , 0x18(%rdi)   #save frame pointer
   7.101 -    movq    0x38(%rdi), %rsp         #restore stack pointer
   7.102 -    movq    0x30(%rdi), %rbp         #restore frame pointer
   7.103 -    movq    $_VMSMasterEnv, %rcx
   7.104 -    movq    (%rcx)    , %rcx
   7.105 -    movq    0x48(%rcx), %rax         #get CoreLoopStartPt
   7.106 -    movl    $0x0      , 0x54(%rcx)   #release lock
   7.107 -    jmp     *%rax                    #jmp to CoreLoop
   7.108 -MasterReturn:
   7.109 -    ret
   7.110 -
   7.111 -
   7.112 -//Switch to terminateCoreLoop
   7.113 -//therefor switch to coreLoop context from master context
   7.114 -// no need to call because the stack is already set up for switchSlv
   7.115 -// and Slv is in %rdi
   7.116 -// and both functions have the same argument.
   7.117 -// do not save register of Slv because this function will never return
   7.118 -/* SlaveVP  offsets:
   7.119 - * 0x10  stackPtr
   7.120 - * 0x18 framePtr
   7.121 - * 0x20 resumeInstrPtr
   7.122 - * 0x30 coreLoopFramePtr
   7.123 - * 0x38 coreLoopStackPtr
   7.124 - *
   7.125 - * _VMSMasterEnv  offsets:
   7.126 - * 0x48 coreLoopReturnPt
   7.127 - * 0x58 masterLock
   7.128 - */
   7.129 -.globl asmTerminateCoreLoop
   7.130 -asmTerminateCoreLoop:
   7.131 -    #SlaveVP in %rdi
   7.132 -    movq    0x38(%rdi), %rsp         #restore stack pointer
   7.133 -    movq    0x30(%rdi), %rbp         #restore frame pointer
   7.134 -    movq    $terminateCoreLoop, %rax
   7.135 -    jmp     *%rax                    #jmp to CoreLoop
   7.136 -
   7.137 -
   7.138 -/*
   7.139 - * This one for the sequential version is special. It discards the current stack
   7.140 - * and returns directly from the coreLoop after VMS_WL__dissipate_slaveVP was called
   7.141 - */
   7.142 -.globl asmTerminateCoreLoopSeq
   7.143 -asmTerminateCoreLoopSeq:
   7.144 -    #SlaveVP in %rdi
   7.145 -    movq    0x38(%rdi), %rsp         #restore stack pointer
   7.146 -    movq    0x30(%rdi), %rbp         #restore frame pointer
   7.147 -    #argument is in %rdi
   7.148 -    call    VMS_int__dissipate_slaveVP
   7.149 -    movq    %rbp      , %rsp        #goto the coreLoops stack
   7.150 -    pop     %rbp        #restore the old framepointer
   7.151 -    ret                 #return from core loop
   7.152 -    
   7.153 -
   7.154 -//Takes the return addr off the stack and saves into the loc pointed to by
   7.155 -// by the parameter passed in via rdi.  Return addr is at 0x8(%rbp) for 64bit
   7.156 -.globl VMS_int__save_return_into_ptd_to_loc_then_do_ret
   7.157 -VMS_int__save_return_into_ptd_to_loc_then_do_ret:
   7.158 -    movq 0x8(%rbp),     %rax  #get ret address, rbp is the same as in the calling function
   7.159 -    movq     %rax,     (%rdi) #write ret addr into addr passed as param field
   7.160 -    ret
   7.161 -
   7.162 -
   7.163 -//Assembly code changes the return addr on the stack to the one
   7.164 -// pointed to by the parameter, then returns. Stack's return addr is at 0x8(%rbp)
   7.165 -.globl VMS_int__return_to_addr_in_ptd_to_loc
   7.166 -VMS_int__return_to_addr_in_ptd_to_loc:
   7.167 -    movq    (%rdi),    %rax  #get return addr from addr passed as param
   7.168 -    movq     %rax, 0x8(%rbp) #write return addr to the stack of the caller
   7.169 -    ret
   7.170 -