File:  [gforth] / gforth / prof-inline.fs
Revision 1.3: download - view: text, annotated - select for diffs
Mon Sep 6 13:44:56 2004 UTC (17 years, 1 month ago) by anton
Branches: MAIN
CVS tags: HEAD
continued prof-inline.fs

    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% field 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: variable last-colondef-profile \ pointer to the pp of last colon definition
   79: variable current-profile-point
   80: variable library-calls \ list of calls to library colon defs
   81: 
   82: \ list stuff
   83: 
   84: : map-list ( ... list xt -- ... )
   85:     { xt } begin { list }
   86: 	list while
   87: 	    list xt execute
   88: 	    list list-next @
   89:     repeat ;
   90: 
   91: : drop-1+ drop 1+ ;
   92: 
   93: : list-length ( list -- u )
   94:     0 swap ['] drop-1+ map-list ;
   95: 
   96: : insert-list ( listp listpp -- )
   97:     \ insert list node listp into list pointed to by listpp in front
   98:     tuck @ over list-next !
   99:     swap ! ;
  100: 
  101: : insert-list-end ( listp listppp -- )
  102:     \ insert list node listp into list, with listppp indicating the
  103:     \ position to insert at, and indicating the position behind the
  104:     \ new element afterwards.
  105:     2dup @ insert-list
  106:     swap list-next swap ! ;
  107: 
  108: \ calls
  109: 
  110: : new-call ( profile-point -- call )
  111:     calls% %alloc tuck calls-call ! ;
  112: 
  113: \ profile-point stuff   
  114: 
  115: : new-profile-point ( -- addr )
  116:     profile% %alloc >r
  117:     0. r@ profile-count 2!
  118:     current-sourcepos r@ profile-sourcepos 2!
  119:     >in @ r@ profile-char !
  120:     [ count-calls? ] [if]
  121: 	r@ profile-colondef? off
  122: 	0 r@ profile-calls !
  123: 	r@ profile-straight-line on
  124: 	0 r@ profile-calls-from !
  125:     [endif]
  126:     r@ next-profile-point-p insert-list-end
  127:     r@ current-profile-point !
  128:     r> ;
  129: 
  130: : print-profile ( -- )
  131:     profile-points @ begin
  132: 	dup while
  133: 	    dup >r
  134: 	    r@ profile-sourcepos 2@ .sourcepos ." :"
  135: 	    r@ profile-char @ 0 .r ." : "
  136: 	    r@ profile-count 2@ 0 d.r cr
  137: 	    r> list-next @
  138:     repeat
  139:     drop ;
  140: 
  141: : print-profile-coldef ( -- )
  142:     profile-points @ begin
  143: 	dup while
  144: 	    dup >r
  145: 	    r@ profile-colondef? @ if
  146: 		r@ profile-sourcepos 2@ .sourcepos ." :"
  147: 		r@ profile-char @ 3 .r ." : "
  148: 		r@ profile-count 2@ 10 d.r
  149: 		r@ profile-straight-line @ space 2 .r
  150: 		r@ profile-calls @ list-length 4 .r
  151: 		cr
  152: 	    endif
  153: 	    r> list-next @
  154:     repeat
  155:     drop ;
  156: 
  157: : 1= ( u -- f )
  158:     1 = ;
  159: 
  160: : 2= ( u -- f )
  161:     2 = ;
  162: 
  163: : 3= ( u -- f )
  164:     3 = ;
  165: 
  166: : 1u> ( u -- f )
  167:     1 u> ;
  168: 
  169: : call-count+ ( ud1 callp -- ud2 )
  170:     calls-call @ profile-count 2@ d+ ;
  171: 
  172: : add-calls ( ud-dyn-callee1 ud-dyn-caller1 u-stat1 xt-test profpp --
  173:               ud-dyn-callee2 ud-dyn-caller2 u-stat2 xt-test )
  174:     \ add the static and dynamic call counts to profpp up, if the
  175:     \ number of static calls to profpp satisfies xt-test ( u -- f )
  176:     { xt-test p }
  177:     p profile-colondef? @ if ( u-dyn1 u-stat1 )
  178: 	p profile-calls @ { calls }
  179: 	calls list-length { stat }
  180: 	stat xt-test execute if ( u-dyn u-stat )
  181: 	    stat + >r
  182: 	    0. calls ['] call-count+ map-list d+ 2>r
  183: 	    p profile-count 2@ d+
  184: 	    2r> r>
  185: 	endif
  186:     endif
  187:     xt-test ;
  188: 
  189: : print-stat-line ( xt -- )
  190:     >r 0. 0. 0 r> profile-points @ ['] add-calls map-list drop
  191:     ( ud-dyn-callee ud-dyn-caller u-stat )
  192:     7 u.r 12 ud.r 12 ud.r space ;
  193: 
  194: : print-statistics ( -- )
  195:     ."  static  dyn-caller  dyn-callee   condition" cr
  196:     ['] 0=  print-stat-line ." calls to coldefs with 0 callers" cr
  197:     ['] 1=  print-stat-line ." calls to coldefs with 1 callers" cr
  198:     ['] 2=  print-stat-line ." calls to coldefs with 2 callers" cr
  199:     ['] 3=  print-stat-line ." calls to coldefs with 3 callers" cr
  200:     ['] 1u> print-stat-line ." calls to coldefs with >1 callers" cr
  201:     ;
  202: 
  203: : dinc ( profilep -- )
  204:     \ increment double pointed to by d-addr
  205:     profile-count dup 2@ 1. d+ rot 2! ;
  206: 
  207: : profile-this ( -- )
  208:     new-profile-point POSTPONE literal POSTPONE dinc ;
  209: 
  210: \ Various words trigger PROFILE-THIS.  In order to avoid getting
  211: \ several calls to PROFILE-THIS from a compiling word (like ?EXIT), we
  212: \ just wait until the next word is parsed by the text interpreter (in
  213: \ compile state) and call PROFILE-THIS only once then.  The whole
  214: \ BEFORE-WORD hooking etc. is there for this.
  215: 
  216: \ The reason that we do this is because we use the source position for
  217: \ the profiling information, and there's only one source position for
  218: \ ?EXIT.  If we used the threaded code position instead, we would see
  219: \ that ?EXIT compiles to several threaded-code words, and could use
  220: \ different profile points for them.  However, usually dealing with
  221: \ the source is more practical.
  222: 
  223: \ Another benefit is that we can ask for profiling anywhere in a
  224: \ control-flow word (even before it compiles its own stuff).
  225: 
  226: \ Potential problem: Consider "COMPILING ] [" where COMPILING compiles
  227: \ a whole colon definition (and triggers our profiler), but during the
  228: \ compilation of the colon definition there is no parsing.  Afterwards
  229: \ you get interpret state at first (no profiling, either), but after
  230: \ the "]" you get parsing in compile state, and PROFILE-THIS gets
  231: \ called (and compiles code that is never executed).  It would be
  232: \ better if we had a way of knowing whether we are in a colon def or
  233: \ not (and used that knowledge instead of STATE).
  234: 
  235: Defer before-word-profile ( -- )
  236: ' noop IS before-word-profile
  237: 
  238: : before-word1 ( -- )
  239:     before-word-profile defers before-word ;
  240: 
  241: ' before-word1 IS before-word
  242: 
  243: : profile-this-compiling ( -- )
  244:     state @ if
  245: 	profile-this
  246: 	['] noop IS before-word-profile
  247:     endif ;
  248: 
  249: : cock-profiler ( -- )
  250:     \ as in cock the gun - pull the trigger
  251:     ['] profile-this-compiling IS before-word-profile
  252:     [ count-calls? ] [if] \ we are at a non-colondef profile point
  253: 	last-colondef-profile @ profile-straight-line off
  254:     [endif]
  255: ;
  256: 
  257: : hook-profiling-into ( "name" -- )
  258:     \ make (deferred word) "name" call cock-profiler, too
  259:     ' >body >r :noname
  260:     POSTPONE cock-profiler
  261:     r@ @ compile, \ old hook behaviour
  262:     POSTPONE ;
  263:     r> ! ; \ change hook behaviour
  264: 
  265: : note-execute ( -- )
  266:     \ end of BB due to execute
  267: ;
  268: 
  269: : note-call ( addr -- )
  270:     \ addr is the body address of a called colon def or does handler
  271:     dup 3 cells + @ ['] dinc >body = if ( addr )
  272: 	current-profile-point @ new-call over cell+ @ profile-calls insert-list
  273:     endif
  274:     drop ;
  275:     
  276: : prof-compile, ( xt -- )
  277:     dup >does-code if
  278: 	dup >does-code note-call
  279:     then
  280:     dup >code-address CASE
  281: 	docol:   OF dup >body note-call ENDOF
  282: 	dodefer: OF note-execute ENDOF
  283: 	dofield: OF >body @ ['] lit+ peephole-compile, , EXIT ENDOF
  284: 	\ dofield: OF >body @ POSTPONE literal ['] + peephole-compile, EXIT ENDOF
  285: 	\ code words and ;code-defined words (code words could be optimized):
  286: 	dup in-dictionary? IF drop POSTPONE literal ['] execute peephole-compile, EXIT THEN
  287:     ENDCASE
  288:     DEFERS compile, ;
  289: 
  290: \ hook-profiling-into then-like
  291: \ \ hook-profiling-into if-like    \ subsumed by other-control-flow
  292: \ \ hook-profiling-into ahead-like \ subsumed by other-control-flow
  293: \ hook-profiling-into other-control-flow
  294: \ hook-profiling-into begin-like
  295: \ hook-profiling-into again-like
  296: \ hook-profiling-into until-like
  297: 
  298: : :-hook-profile ( -- )
  299:     defers :-hook
  300:     next-profile-point-p @
  301:     profile-this
  302:     @ dup last-colondef-profile !
  303:     profile-colondef? on ;
  304: 
  305: ' :-hook-profile IS :-hook
  306: ' prof-compile, IS compile,

FreeBSD-CVSweb <freebsd-cvsweb@FreeBSD.org>