Return to prof-inline.fs CVS log

Up to [gforth] / gforth

changed default for GFORTHSYSTEMPREFIX to use $COMSPEC
prof-inline.fs changes

    1: \ get some data on potential (partial) inlining
    2: 
    3: \ Copyright (C) 2004 Free Software Foundation, Inc.
    4: 
    5: \ This file is part of Gforth.
    6: 
    7: \ Gforth is free software; you can redistribute it and/or
    8: \ modify it under the terms of the GNU General Public License
    9: \ as published by the Free Software Foundation; either version 2
   10: \ of the License, or (at your option) any later version.
   11: 
   12: \ This program is distributed in the hope that it will be useful,
   13: \ but WITHOUT ANY WARRANTY; without even the implied warranty of
   14: \ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   15: \ GNU General Public License for more details.
   16: 
   17: \ You should have received a copy of the GNU General Public License
   18: \ along with this program; if not, write to the Free Software
   19: \ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111, USA.
   20: 
   21: 
   22: \ relies on some Gforth internals
   23: 
   24: \ !! assumption: each file is included only once; otherwise you get
   25: \ the counts for just one of the instances of the file.  This can be
   26: \ fixed by making sure that every source position occurs only once as
   27: \ a profile point.
   28: 
   29: true constant count-calls? \ do some profiling of colon definitions etc.
   30: 
   31: \ for true COUNT-CALLS?:
   32: 
   33: \ What data do I need for evaluating the effectiveness of (partial) inlining?
   34: 
   35: \ static and dynamic counts of everything:
   36: 
   37: \ original BB length (histogram and average)
   38: \ BB length with partial inlining (histogram and average)
   39: \   since we cannot partially inline library calls, we use a parameter
   40: \   that represents the amount of partial inlining we can expect there.
   41: \ number of tail calls (original and after partial inlining)
   42: \ number of calls (original and after partial inlining)
   43: \ reason for BB end: call, return, execute, branch
   44: 
   45: \ how many static calls are there to a word?  How many of the dynamic
   46: \ calls call just a single word?
   47: 
   48: \ how much does inlining called-once words help?
   49: \ how much does inlining words without control flow help?
   50: \ how much does partial inlining help?
   51: \ what's the overlap?
   52: \ optimizing return-to-returns (tail calls), return-to-calls, call-to-calls
   53: 
   54: struct
   55:     cell% list-next
   56: end-struct list%
   57: 
   58: list%
   59:     cell% 2* field profile-count
   60:     cell% 2* field profile-sourcepos
   61:     cell%    field profile-char \ character position in line
   62:     count-calls? [if]
   63: 	cell% field profile-colondef? \ is this a colon definition start
   64: 	cell% field profile-calls \ static calls to the colon def (calls%)
   65: 	cell% field profile-straight-line \ may contain calls, but no other CF
   66: 	cell% field profile-calls-from \ static calls in the colon def
   67:     [endif]
   68: end-struct profile% \ profile point
   69: 
   70: list%
   71:     cell% field calls%-call \ ptr to profile point of bb containing the call
   72: end-struct calls%
   73: 
   74: variable profile-points \ linked list of profile%
   75: 0 profile-points !
   76: variable next-profile-point-p \ the address where the next pp will be stored
   77: profile-points next-profile-point-p !
   78: count-calls? [if]
   79:     variable last-colondef-profile \ pointer to the pp of last colon definition
   80: [endif]
   81: 
   82: \ list stuff
   83: 
   84: 
   85: 
   86: \ profile-point stuff   
   87: 
   88: : new-profile-point ( -- addr )
   89:     profile% %alloc >r
   90:     0. r@ profile-count 2!
   91:     current-sourcepos r@ profile-sourcepos 2!
   92:     >in @ r@ profile-char !
   93:     [ count-calls? ] [if]
   94: 	r@ profile-colondef? off
   95: 	0 r@ profile-calls !
   96: 	r@ profile-straight-line on
   97: 	0 r@ profile-calls-from !
   98:     [endif]
   99:     0 r@ list-next !
  100:     r@ next-profile-point-p @ !
  101:     r@ list-next next-profile-point-p !
  102:     r> ;
  103: 
  104: : print-profile ( -- )
  105:     profile-points @ begin
  106: 	dup while
  107: 	    dup >r
  108: 	    r@ profile-sourcepos 2@ .sourcepos ." :"
  109: 	    r@ profile-char @ 0 .r ." : "
  110: 	    r@ profile-count 2@ 0 d.r cr
  111: 	    r> list-next @
  112:     repeat
  113:     drop ;
  114: 
  115: : print-profile-coldef ( -- )
  116:     profile-points @ begin
  117: 	dup while
  118: 	    dup >r
  119: 	    r@ profile-colondef? @ if
  120: 		r@ profile-sourcepos 2@ .sourcepos ." :"
  121: 		r@ profile-char @ 3 .r ." : "
  122: 		r@ profile-count 2@ 10 d.r
  123: 		r@ profile-straight-line @ space 2 .r
  124: 		r@ profile-calls @ 4 .r
  125: 		cr
  126: 	    endif
  127: 	    r> list-next @
  128:     repeat
  129:     drop ;
  130: 
  131: : dinc ( profilep -- )
  132:     \ increment double pointed to by d-addr
  133:     profile-count dup 2@ 1. d+ rot 2! ;
  134: 
  135: : profile-this ( -- )
  136:     new-profile-point POSTPONE literal POSTPONE dinc ;
  137: 
  138: \ Various words trigger PROFILE-THIS.  In order to avoid getting
  139: \ several calls to PROFILE-THIS from a compiling word (like ?EXIT), we
  140: \ just wait until the next word is parsed by the text interpreter (in
  141: \ compile state) and call PROFILE-THIS only once then.  The whole
  142: \ BEFORE-WORD hooking etc. is there for this.
  143: 
  144: \ The reason that we do this is because we use the source position for
  145: \ the profiling information, and there's only one source position for
  146: \ ?EXIT.  If we used the threaded code position instead, we would see
  147: \ that ?EXIT compiles to several threaded-code words, and could use
  148: \ different profile points for them.  However, usually dealing with
  149: \ the source is more practical.
  150: 
  151: \ Another benefit is that we can ask for profiling anywhere in a
  152: \ control-flow word (even before it compiles its own stuff).
  153: 
  154: \ Potential problem: Consider "COMPILING ] [" where COMPILING compiles
  155: \ a whole colon definition (and triggers our profiler), but during the
  156: \ compilation of the colon definition there is no parsing.  Afterwards
  157: \ you get interpret state at first (no profiling, either), but after
  158: \ the "]" you get parsing in compile state, and PROFILE-THIS gets
  159: \ called (and compiles code that is never executed).  It would be
  160: \ better if we had a way of knowing whether we are in a colon def or
  161: \ not (and used that knowledge instead of STATE).
  162: 
  163: Defer before-word-profile ( -- )
  164: ' noop IS before-word-profile
  165: 
  166: : before-word1 ( -- )
  167:     before-word-profile defers before-word ;
  168: 
  169: ' before-word1 IS before-word
  170: 
  171: : profile-this-compiling ( -- )
  172:     state @ if
  173: 	profile-this
  174: 	['] noop IS before-word-profile
  175:     endif ;
  176: 
  177: : cock-profiler ( -- )
  178:     \ as in cock the gun - pull the trigger
  179:     ['] profile-this-compiling IS before-word-profile
  180:     [ count-calls? ] [if] \ we are at a non-colondef profile point
  181: 	last-colondef-profile @ profile-straight-line off
  182:     [endif]
  183: ;
  184: 
  185: : hook-profiling-into ( "name" -- )
  186:     \ make (deferred word) "name" call cock-profiler, too
  187:     ' >body >r :noname
  188:     POSTPONE cock-profiler
  189:     r@ @ compile, \ old hook behaviour
  190:     POSTPONE ;
  191:     r> ! ; \ change hook behaviour
  192: 
  193: : note-execute ( -- )
  194:     \ end of BB due to execute
  195: ;
  196: 
  197: : note-call ( addr -- )
  198:     \ addr is the body address of a called colon def or does handler
  199:     dup 3 cells + @ ['] dinc >body = if
  200: 	1 over  cell+ @ profile-calls +!
  201:     endif
  202:     drop ;
  203:     
  204: : prof-compile, ( xt -- )
  205:     dup >does-code if
  206: 	dup >does-code note-call
  207:     then
  208:     dup >code-address CASE
  209: 	docol:   OF dup >body note-call ENDOF
  210: 	dodefer: OF note-execute ENDOF
  211: 	dofield: OF >body @ ['] lit+ peephole-compile, , EXIT ENDOF
  212: 	\ dofield: OF >body @ POSTPONE literal ['] + peephole-compile, EXIT ENDOF
  213: 	\ code words and ;code-defined words (code words could be optimized):
  214: 	dup in-dictionary? IF drop POSTPONE literal ['] execute peephole-compile, EXIT THEN
  215:     ENDCASE
  216:     DEFERS compile, ;
  217: 
  218: \ hook-profiling-into then-like
  219: \ \ hook-profiling-into if-like    \ subsumed by other-control-flow
  220: \ \ hook-profiling-into ahead-like \ subsumed by other-control-flow
  221: \ hook-profiling-into other-control-flow
  222: \ hook-profiling-into begin-like
  223: \ hook-profiling-into again-like
  224: \ hook-profiling-into until-like
  225: 
  226: : :-hook-profile ( -- )
  227:     defers :-hook
  228:     next-profile-point-p @
  229:     profile-this
  230:     @ dup last-colondef-profile !
  231:     profile-colondef? on ;
  232: 
  233: ' :-hook-profile IS :-hook
  234: ' prof-compile, IS compile,