Annotation of gforth/prof-inline.fs, revision 1.1

1.1     ! anton       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: struct
        !            49:     cell%    field profile-next
        !            50:     cell% 2* field profile-count
        !            51:     cell% 2* field profile-sourcepos
        !            52:     cell%    field profile-char \ character position in line
        !            53:     count-calls? [if]
        !            54:        cell% field profile-colondef? \ is this a colon definition start
        !            55:        cell% field profile-calls \ static calls to the colon def
        !            56:        cell% field profile-straight-line \ may contain calls, but no other CF
        !            57:        cell% field profile-calls-from \ static calls in the colon def
        !            58:     [endif]
        !            59: end-struct profile% \ profile point
        !            60: 
        !            61: variable profile-points \ linked list of profile%
        !            62: 0 profile-points !
        !            63: variable next-profile-point-p \ the address where the next pp will be stored
        !            64: profile-points next-profile-point-p !
        !            65: count-calls? [if]
        !            66:     variable last-colondef-profile \ pointer to the pp of last colon definition
        !            67: [endif]
        !            68:     
        !            69: : new-profile-point ( -- addr )
        !            70:     profile% %alloc >r
        !            71:     0. r@ profile-count 2!
        !            72:     current-sourcepos r@ profile-sourcepos 2!
        !            73:     >in @ r@ profile-char !
        !            74:     [ count-calls? ] [if]
        !            75:        r@ profile-colondef? off
        !            76:        0 r@ profile-calls !
        !            77:        r@ profile-straight-line on
        !            78:        0 r@ profile-calls-from !
        !            79:     [endif]
        !            80:     0 r@ profile-next !
        !            81:     r@ next-profile-point-p @ !
        !            82:     r@ profile-next next-profile-point-p !
        !            83:     r> ;
        !            84: 
        !            85: : print-profile ( -- )
        !            86:     profile-points @ begin
        !            87:        dup while
        !            88:            dup >r
        !            89:            r@ profile-sourcepos 2@ .sourcepos ." :"
        !            90:            r@ profile-char @ 0 .r ." : "
        !            91:            r@ profile-count 2@ 0 d.r cr
        !            92:            r> profile-next @
        !            93:     repeat
        !            94:     drop ;
        !            95: 
        !            96: : print-profile-coldef ( -- )
        !            97:     profile-points @ begin
        !            98:        dup while
        !            99:            dup >r
        !           100:            r@ profile-colondef? @ if
        !           101:                r@ profile-sourcepos 2@ .sourcepos ." :"
        !           102:                r@ profile-char @ 3 .r ." : "
        !           103:                r@ profile-count 2@ 10 d.r
        !           104:                r@ profile-straight-line @ space 2 .r
        !           105:                r@ profile-calls @ 4 .r
        !           106:                cr
        !           107:            endif
        !           108:            r> profile-next @
        !           109:     repeat
        !           110:     drop ;
        !           111: 
        !           112: : dinc ( profilep -- )
        !           113:     \ increment double pointed to by d-addr
        !           114:     profile-count dup 2@ 1. d+ rot 2! ;
        !           115: 
        !           116: : profile-this ( -- )
        !           117:     new-profile-point POSTPONE literal POSTPONE dinc ;
        !           118: 
        !           119: \ Various words trigger PROFILE-THIS.  In order to avoid getting
        !           120: \ several calls to PROFILE-THIS from a compiling word (like ?EXIT), we
        !           121: \ just wait until the next word is parsed by the text interpreter (in
        !           122: \ compile state) and call PROFILE-THIS only once then.  The whole
        !           123: \ BEFORE-WORD hooking etc. is there for this.
        !           124: 
        !           125: \ The reason that we do this is because we use the source position for
        !           126: \ the profiling information, and there's only one source position for
        !           127: \ ?EXIT.  If we used the threaded code position instead, we would see
        !           128: \ that ?EXIT compiles to several threaded-code words, and could use
        !           129: \ different profile points for them.  However, usually dealing with
        !           130: \ the source is more practical.
        !           131: 
        !           132: \ Another benefit is that we can ask for profiling anywhere in a
        !           133: \ control-flow word (even before it compiles its own stuff).
        !           134: 
        !           135: \ Potential problem: Consider "COMPILING ] [" where COMPILING compiles
        !           136: \ a whole colon definition (and triggers our profiler), but during the
        !           137: \ compilation of the colon definition there is no parsing.  Afterwards
        !           138: \ you get interpret state at first (no profiling, either), but after
        !           139: \ the "]" you get parsing in compile state, and PROFILE-THIS gets
        !           140: \ called (and compiles code that is never executed).  It would be
        !           141: \ better if we had a way of knowing whether we are in a colon def or
        !           142: \ not (and used that knowledge instead of STATE).
        !           143: 
        !           144: Defer before-word-profile ( -- )
        !           145: ' noop IS before-word-profile
        !           146: 
        !           147: : before-word1 ( -- )
        !           148:     before-word-profile defers before-word ;
        !           149: 
        !           150: ' before-word1 IS before-word
        !           151: 
        !           152: : profile-this-compiling ( -- )
        !           153:     state @ if
        !           154:        profile-this
        !           155:        ['] noop IS before-word-profile
        !           156:     endif ;
        !           157: 
        !           158: : cock-profiler ( -- )
        !           159:     \ as in cock the gun - pull the trigger
        !           160:     ['] profile-this-compiling IS before-word-profile
        !           161:     [ count-calls? ] [if] \ we are at a non-colondef profile point
        !           162:        last-colondef-profile @ profile-straight-line off
        !           163:     [endif]
        !           164: ;
        !           165: 
        !           166: : hook-profiling-into ( "name" -- )
        !           167:     \ make (deferred word) "name" call cock-profiler, too
        !           168:     ' >body >r :noname
        !           169:     POSTPONE cock-profiler
        !           170:     r@ @ compile, \ old hook behaviour
        !           171:     POSTPONE ;
        !           172:     r> ! ; \ change hook behaviour
        !           173: 
        !           174: : note-execute ( -- )
        !           175:     \ end of BB due to execute
        !           176: ;
        !           177: 
        !           178: : note-call ( addr -- )
        !           179:     \ addr is the body address of a called colon def or does handler
        !           180:     dup 3 cells + @ ['] dinc >body = if
        !           181:        1 over  cell+ @ profile-calls +!
        !           182:     endif
        !           183:     drop ;
        !           184:     
        !           185: : prof-compile, ( xt -- )
        !           186:     dup >does-code if
        !           187:        dup >does-code note-call
        !           188:     then
        !           189:     dup >code-address CASE
        !           190:        docol:   OF dup >body note-call ENDOF
        !           191:        dodefer: OF note-execute ENDOF
        !           192:        dofield: OF >body @ ['] lit+ peephole-compile, , EXIT ENDOF
        !           193:        \ dofield: OF >body @ POSTPONE literal ['] + peephole-compile, EXIT ENDOF
        !           194:        \ code words and ;code-defined words (code words could be optimized):
        !           195:        dup in-dictionary? IF drop POSTPONE literal ['] execute peephole-compile, EXIT THEN
        !           196:     ENDCASE
        !           197:     DEFERS compile, ;
        !           198: 
        !           199: \ hook-profiling-into then-like
        !           200: \ \ hook-profiling-into if-like    \ subsumed by other-control-flow
        !           201: \ \ hook-profiling-into ahead-like \ subsumed by other-control-flow
        !           202: \ hook-profiling-into other-control-flow
        !           203: \ hook-profiling-into begin-like
        !           204: \ hook-profiling-into again-like
        !           205: \ hook-profiling-into until-like
        !           206: 
        !           207: : :-hook-profile ( -- )
        !           208:     defers :-hook
        !           209:     next-profile-point-p @
        !           210:     profile-this
        !           211:     @ dup last-colondef-profile !
        !           212:     profile-colondef? on ;
        !           213: 
        !           214: ' :-hook-profile IS :-hook
        !           215: ' prof-compile, IS compile,

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