Annotation of gforth/prims2x.fs, revision 1.165

1.16      anton       1: \ converts primitives to, e.g., C code 
                      2: 
1.163     anton       3: \ Copyright (C) 1995,1996,1997,1998,2000,2003,2004,2005,2006 Free Software Foundation, Inc.
1.16      anton       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
1.48      anton      19: \ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111, USA.
1.16      anton      20: 
                     21: 
1.71      anton      22: \ This is not very nice (hard limits, no checking, assumes 1 chars = 1).
                     23: \ And it grew even worse when it aged.
1.1       anton      24: 
                     25: \ Optimizations:
                     26: \ superfluous stores are removed. GCC removes the superfluous loads by itself
                     27: \ TOS and FTOS can be kept in register( variable)s.
                     28: \ 
                     29: \ Problems:
                     30: \ The TOS optimization is somewhat hairy. The problems by example:
                     31: \ 1) dup ( w -- w w ): w=TOS; sp-=1; sp[1]=w; TOS=w;
                     32: \    The store is not superfluous although the earlier opt. would think so
                     33: \    Alternatively:    sp[0]=TOS; w=TOS; sp-=1; TOS=w;
                     34: \ 2) ( -- .. ): sp[0] = TOS; ... /* This additional store is necessary */
                     35: \ 3) ( .. -- ): ... TOS = sp[0]; /* as well as this load */
                     36: \ 4) ( -- ): /* but here they are unnecessary */
                     37: \ 5) Words that call NEXT themselves have to be done very carefully.
                     38: \
                     39: \ To do:
1.8       pazsan     40: \ add the store optimization for doubles
1.1       anton      41: \ regarding problem 1 above: It would be better (for over) to implement
                     42: \      the alternative
1.80      anton      43: \ store optimization for combined instructions.
                     44: 
                     45: \ Design Uglyness:
                     46: 
                     47: \ - global state (values, variables) in connection with combined instructions.
                     48: 
                     49: \ - index computation is different for instruction-stream and the
                     50: \ stacks; there are two mechanisms for dealing with that
                     51: \ (stack-in-index-xt and a test for stack==instruction-stream); there
                     52: \ should be only one.
1.1       anton      53: 
1.159     anton      54: 
1.137     jwilke     55: \ for backwards compatibility, jaw
                     56: require compat/strcomp.fs
                     57: 
1.3       pazsan     58: warnings off
                     59: 
1.136     anton      60: \ redefinitions of kernel words not present in gforth-0.6.1
                     61: : latestxt lastcfa @ ;
                     62: : latest last @ ;
                     63: 
1.97      jwilke     64: [IFUNDEF] try
                     65: include startup.fs
                     66: [THEN]
                     67: 
1.49      anton      68: : struct% struct ; \ struct is redefined in gray
                     69: 
1.98      pazsan     70: warnings off
1.110     anton      71: \ warnings on
1.98      pazsan     72: 
1.39      jwilke     73: include ./gray.fs
1.133     anton      74: 128 constant max-effect \ number of things on one side of a stack effect
1.71      anton      75: 4 constant max-stacks  \ the max. number of stacks (including inst-stream).
1.1       anton      76: 255 constant maxchar
                     77: maxchar 1+ constant eof-char
1.17      anton      78: #tab constant tab-char
                     79: #lf constant nl-char
1.1       anton      80: 
1.18      anton      81: variable rawinput \ pointer to next character to be scanned
                     82: variable endrawinput \ pointer to the end of the input (the char after the last)
                     83: variable cookedinput \ pointer to the next char to be parsed
1.17      anton      84: variable line \ line number of char pointed to by input
1.65      anton      85: variable line-start \ pointer to start of current line (for error messages)
                     86: 0 line !
1.17      anton      87: 2variable filename \ filename of original input file
                     88: 0 0 filename 2!
1.111     anton      89: 2variable out-filename \ filename of the output file (for sync lines)
                     90: 0 0 out-filename 2!
1.25      pazsan     91: 2variable f-comment
                     92: 0 0 f-comment 2!
1.17      anton      93: variable skipsynclines \ are sync lines ("#line ...") invisible to the parser?
1.111     anton      94: skipsynclines on
                     95: variable out-nls \ newlines in output (for output sync lines)
                     96: 0 out-nls !
1.112     anton      97: variable store-optimization \ use store optimization?
                     98: store-optimization off
                     99: 
1.116     anton     100: variable include-skipped-insts
                    101: \ does the threaded code for a combined instruction include the cells
                    102: \ for the component instructions (true) or only the cells for the
                    103: \ inline arguments (false)
                    104: include-skipped-insts off
1.1       anton     105: 
1.159     anton     106: 2variable threaded-code-pointer-type \ type used for geninst etc.
                    107: s" Inst **" threaded-code-pointer-type 2!
                    108: 
1.121     anton     109: variable immarg \ values for immediate arguments (to be used in IMM_ARG macros)
                    110: $12340000 immarg !
                    111: 
1.72      anton     112: : th ( addr1 n -- addr2 )
                    113:     cells + ;
                    114: 
                    115: : holds ( addr u -- )
                    116:     \ like HOLD, but for a string
                    117:     tuck + swap 0 +do
                    118:        1- dup c@ hold
                    119:     loop
                    120:     drop ;
1.71      anton     121: 
1.82      anton     122: : insert-wordlist { c-addr u wordlist xt -- }
1.81      anton     123:     \ adds name "addr u" to wordlist using defining word xt
                    124:     \ xt may cause additional stack effects
                    125:     get-current >r wordlist set-current
                    126:     c-addr u nextname xt execute
                    127:     r> set-current ;
                    128: 
1.1       anton     129: : start ( -- addr )
1.18      anton     130:  cookedinput @ ;
1.1       anton     131: 
                    132: : end ( addr -- addr u )
1.18      anton     133:  cookedinput @ over - ;
1.1       anton     134: 
1.71      anton     135: : print-error-line ( -- )
                    136:     \ print the current line and position
                    137:     line-start @ endrawinput @ over - 2dup nl-char scan drop nip ( start end )
                    138:     over - type cr
                    139:     line-start @ rawinput @ over - typewhite ." ^" cr ;
                    140: 
1.165   ! anton     141: : print-error { addr u -- }
        !           142:     filename 2@ type ." :" line @ 0 .r ." : " addr u type cr
        !           143:     print-error-line ;
        !           144: 
1.71      anton     145: : ?print-error { f addr u -- }
                    146:     f ?not? if
1.165   ! anton     147:        addr u ['] print-error stderr outfile-execute
1.111     anton     148:        1 (bye) \ abort
1.71      anton     149:     endif ;
                    150: 
1.63      anton     151: : quote ( -- )
                    152:     [char] " emit ;
                    153: 
1.111     anton     154: \ count output lines to generate sync lines for output
                    155: 
                    156: : count-nls ( addr u -- )
                    157:     bounds u+do
                    158:        i c@ nl-char = negate out-nls +!
                    159:     loop ;
                    160: 
                    161: :noname ( addr u -- )
                    162:     2dup count-nls
                    163:     defers type ;
                    164: is type
                    165: 
1.72      anton     166: variable output          \ xt ( -- ) of output word for simple primitives
                    167: variable output-combined \ xt ( -- ) of output word for combined primitives
1.1       anton     168: 
1.49      anton     169: struct%
1.71      anton     170:     cell%    field stack-number \ the number of this stack
1.49      anton     171:     cell% 2* field stack-pointer \ stackpointer name
1.74      anton     172:     cell%    field stack-type \ name for default type of stack items
1.53      anton     173:     cell%    field stack-in-index-xt \ ( in-size item -- in-index )
1.126     anton     174:     cell%    field stack-access-transform \ ( nitem -- index )
1.49      anton     175: end-struct stack%
                    176: 
1.53      anton     177: struct%
                    178:  cell% 2* field item-name   \ name, excluding stack prefixes
                    179:  cell%    field item-stack  \ descriptor for the stack used, 0 is default
                    180:  cell%    field item-type   \ descriptor for the item type
                    181:  cell%    field item-offset \ offset in stack items, 0 for the deepest element
1.66      anton     182:  cell%   field item-first  \ true if this is the first occurence of the item
1.53      anton     183: end-struct item%
                    184: 
                    185: struct%
                    186:     cell% 2* field type-c-name
                    187:     cell%    field type-stack \ default stack
                    188:     cell%    field type-size  \ size of type in stack items
                    189:     cell%    field type-fetch \ xt of fetch code generator ( item -- )
                    190:     cell%    field type-store \ xt of store code generator ( item -- )
                    191: end-struct type%
                    192: 
1.144     anton     193: struct%
                    194:     cell%    field register-number
                    195:     cell%    field register-type \ pointer to type
                    196:     cell% 2* field register-name \ c name
                    197: end-struct register%
                    198: 
                    199: struct%
                    200:     cell% 2* field ss-registers  \ addr u; ss-registers[0] is TOS
                    201:                                  \ 0 means: use memory
                    202:     cell%    field ss-offset     \ stack pointer offset: sp[-offset] is TOS
                    203: end-struct ss% \ stack-state
                    204: 
                    205: struct%
1.161     anton     206:     cell%              field state-enabled
1.144     anton     207:     cell%              field state-number
                    208:     cell% max-stacks * field state-sss
                    209: end-struct state%
                    210: 
1.72      anton     211: variable next-stack-number 0 next-stack-number !
                    212: create stacks max-stacks cells allot \ array of stacks
1.144     anton     213: 256 constant max-registers
                    214: create registers max-registers cells allot \ array of registers
                    215: variable nregisters 0 nregisters ! \ number of registers
                    216: variable next-state-number 0 next-state-number ! \ next state number
1.72      anton     217: 
1.53      anton     218: : stack-in-index ( in-size item -- in-index )
                    219:     item-offset @ - 1- ;
                    220: 
                    221: : inst-in-index ( in-size item -- in-index )
                    222:     nip dup item-offset @ swap item-type @ type-size @ + 1- ;
                    223: 
1.92      anton     224: : make-stack ( addr-ptr u1 type "stack-name" -- )
                    225:     next-stack-number @ max-stacks < s" too many stacks" ?print-error
1.49      anton     226:     create stack% %allot >r
1.72      anton     227:     r@ stacks next-stack-number @ th !
1.92      anton     228:     next-stack-number @ r@ stack-number !
                    229:     1 next-stack-number +!
1.74      anton     230:     r@ stack-type !
1.53      anton     231:     save-mem r@ stack-pointer 2! 
1.126     anton     232:     ['] stack-in-index r@ stack-in-index-xt !
                    233:     ['] noop r@ stack-access-transform !
                    234:     rdrop ;
1.49      anton     235: 
1.92      anton     236: : map-stacks { xt -- }
1.157     anton     237:     \ perform xt ( stack -- ) for all stacks
1.118     anton     238:     next-stack-number @ 0 +do
                    239:        stacks i th @ xt execute
                    240:     loop ;
                    241: 
                    242: : map-stacks1 { xt -- }
1.157     anton     243:     \ perform xt ( stack -- ) for all stacks except inst-stream
1.92      anton     244:     next-stack-number @ 1 +do
                    245:        stacks i th @ xt execute
                    246:     loop ;
                    247: 
1.49      anton     248: \ stack items
                    249: 
                    250: : init-item ( addr u addr1 -- )
                    251:     \ initialize item at addr1 with name addr u
1.156     anton     252:     \ the stack prefix is removed by the stack-prefix
1.49      anton     253:     dup item% %size erase
                    254:     item-name 2! ;
                    255: 
1.64      anton     256: : map-items { addr end xt -- }
                    257:     \ perform xt for all items in array addr...end
                    258:     end addr ?do
                    259:        i xt execute
                    260:     item% %size +loop ;
                    261: 
1.77      anton     262: \ types
                    263: 
                    264: : print-type-prefix ( type -- )
                    265:     body> >head name>string type ;
                    266: 
1.49      anton     267: \ various variables for storing stuff of one primitive
1.1       anton     268: 
1.69      anton     269: struct%
                    270:     cell% 2* field prim-name
                    271:     cell% 2* field prim-wordset
                    272:     cell% 2* field prim-c-name
1.146     anton     273:     cell% 2* field prim-c-name-orig \ for reprocessed prims, the original name
1.69      anton     274:     cell% 2* field prim-doc
                    275:     cell% 2* field prim-c-code
                    276:     cell% 2* field prim-forth-code
                    277:     cell% 2* field prim-stack-string
1.82      anton     278:     cell%    field prim-num            \ ordinal number
1.75      anton     279:     cell%    field prim-items-wordlist \ unique items
1.69      anton     280:     item% max-effect * field prim-effect-in
                    281:     item% max-effect * field prim-effect-out
                    282:     cell%    field prim-effect-in-end
                    283:     cell%    field prim-effect-out-end
1.71      anton     284:     cell% max-stacks * field prim-stacks-in  \ number of in items per stack
                    285:     cell% max-stacks * field prim-stacks-out \ number of out items per stack
1.156     anton     286:     cell% max-stacks * field prim-stacks-sync \ sync flag per stack
1.69      anton     287: end-struct prim%
                    288: 
1.70      anton     289: : make-prim ( -- prim )
                    290:     prim% %alloc { p }
                    291:     s" " p prim-doc 2! s" " p prim-forth-code 2! s" " p prim-wordset 2!
                    292:     p ;
                    293: 
1.79      anton     294: 0 value prim     \ in combined prims either combined or a part
                    295: 0 value combined \ in combined prims the combined prim
                    296: variable in-part \ true if processing a part
                    297:  in-part off
1.144     anton     298: 0 value state-in  \ state on entering prim
                    299: 0 value state-out \ state on exiting prim
                    300: 0 value state-default  \ canonical state at bb boundaries
1.79      anton     301: 
1.118     anton     302: : prim-context ( ... p xt -- ... )
                    303:     \ execute xt with prim set to p
                    304:     prim >r
                    305:     swap to prim
                    306:     catch
                    307:     r> to prim
                    308:     throw ;
                    309: 
1.146     anton     310: : prim-c-name-2! ( c-addr u -- )
                    311:     2dup prim prim-c-name 2! prim prim-c-name-orig 2! ;
                    312: 
1.79      anton     313: 1000 constant max-combined
                    314: create combined-prims max-combined cells allot
                    315: variable num-combined
1.118     anton     316: variable part-num \ current part number during process-combined
1.79      anton     317: 
1.114     anton     318: : map-combined { xt -- }
                    319:     \ perform xt for all components of the current combined instruction
                    320:     num-combined @ 0 +do
                    321:        combined-prims i th @ xt execute
                    322:     loop ;
                    323: 
1.81      anton     324: table constant combinations
                    325:   \ the keys are the sequences of pointers to primitives
                    326: 
1.79      anton     327: create current-depth max-stacks cells allot
                    328: create max-depth     max-stacks cells allot
                    329: create min-depth     max-stacks cells allot
1.69      anton     330: 
1.118     anton     331: create sp-update-in max-stacks cells allot
                    332: \ where max-depth occured the first time
                    333: create max-depths max-stacks max-combined 1+ * cells allot
1.119     anton     334: \ maximum depth at start of each part: array[parts] of array[stack]
                    335: create max-back-depths max-stacks max-combined 1+ * cells allot
                    336: \ maximun depth from end of the combination to the start of the each part
1.118     anton     337: 
                    338: : s-c-max-depth ( nstack ncomponent -- addr )
                    339:     max-stacks * + cells max-depths + ;
                    340: 
1.119     anton     341: : s-c-max-back-depth ( nstack ncomponent -- addr )
                    342:     max-stacks * + cells max-back-depths + ;
                    343: 
1.71      anton     344: wordlist constant primitives
                    345: 
                    346: : create-prim ( prim -- )
1.82      anton     347:     dup prim-name 2@ primitives ['] constant insert-wordlist ;
1.71      anton     348: 
                    349: : stack-in ( stack -- addr )
                    350:     \ address of number of stack items in effect in
                    351:     stack-number @ cells prim prim-stacks-in + ;
                    352: 
                    353: : stack-out ( stack -- addr )
                    354:     \ address of number of stack items in effect out
                    355:     stack-number @ cells prim prim-stacks-out + ;
                    356: 
1.158     anton     357: : stack-prim-stacks-sync ( stack -- addr )
                    358:     prim prim-stacks-sync swap stack-number @ th ;
                    359: 
1.69      anton     360: \ global vars
1.17      anton     361: variable c-line
                    362: 2variable c-filename
                    363: variable name-line
                    364: 2variable name-filename
                    365: 2variable last-name-filename
1.30      pazsan    366: Variable function-number 0 function-number !
1.140     pazsan    367: Variable function-old 0 function-old !
                    368: : function-diff ( n -- )
                    369:     ." GROUPADD(" function-number @ function-old @ - 0 .r ." )" cr
                    370:     function-number @ function-old ! ;
                    371: : forth-fdiff ( -- )
                    372:     function-number @ function-old @ - 0 .r ."  groupadd" cr
                    373:     function-number @ function-old ! ;
1.1       anton     374: 
                    375: \ a few more set ops
                    376: 
                    377: : bit-equivalent ( w1 w2 -- w3 )
                    378:  xor invert ;
                    379: 
                    380: : complement ( set1 -- set2 )
                    381:  empty ['] bit-equivalent binary-set-operation ;
                    382: 
1.121     anton     383: \ forward declaration for inst-stream (breaks cycle in definitions)
                    384: defer inst-stream-f ( -- stack )
                    385: 
1.80      anton     386: \ stack access stuff
1.79      anton     387: 
1.126     anton     388: : normal-stack-access0 { n stack -- }
1.144     anton     389:     \ n has the ss-offset already applied (see ...-access1)
1.126     anton     390:     n stack stack-access-transform @ execute ." [" 0 .r ." ]" ;
1.144     anton     391: 
                    392: : state-ss { stack state -- ss }
                    393:     state state-sss stack stack-number @ th @ ;
                    394: 
                    395: : stack-reg { n stack state -- reg }
                    396:     \ n is the index (TOS=0); reg is 0 if the access is to memory
                    397:     stack state state-ss ss-registers 2@ n u> if ( addr ) \ in ss-registers?
                    398:        n th @
1.49      anton     399:     else
1.144     anton     400:        drop 0
1.49      anton     401:     endif ;
1.1       anton     402: 
1.144     anton     403: : .reg ( reg -- )
                    404:     register-name 2@ type ;
                    405: 
                    406: : stack-offset ( stack state -- n )
                    407:     \ offset for stack in state
                    408:     state-ss ss-offset @ ;
                    409: 
                    410: : normal-stack-access1 { n stack state -- }
                    411:     n stack state stack-reg ?dup-if
                    412:        .reg exit
                    413:     endif
                    414:     stack stack-pointer 2@ type
                    415:     n stack state stack-offset - stack normal-stack-access0 ;
                    416: 
                    417: : normal-stack-access ( n stack state -- )
                    418:     over inst-stream-f = if
1.121     anton     419:        ." IMM_ARG(" normal-stack-access1 ." ," immarg ? ." )"
                    420:        1 immarg +!
                    421:     else
                    422:        normal-stack-access1
                    423:     endif ;
1.80      anton     424: 
1.118     anton     425: : stack-depth { stack -- n }
                    426:     current-depth stack stack-number @ th @ ;
                    427: 
1.79      anton     428: : part-stack-access { n stack -- }
1.80      anton     429:     \ print _<stack><x>, x=inst-stream? n : maxdepth-currentdepth-n-1
1.79      anton     430:     ." _" stack stack-pointer 2@ type
                    431:     stack stack-number @ { stack# }
1.118     anton     432:     stack stack-depth n + { access-depth }
1.80      anton     433:     stack inst-stream-f = if
                    434:        access-depth
                    435:     else
                    436:        combined prim-stacks-in stack# th @
                    437:        assert( dup max-depth stack# th @ = )
                    438:        access-depth - 1-
                    439:     endif
1.79      anton     440:     0 .r ;
                    441: 
1.118     anton     442: : part-stack-read { n stack -- }
                    443:     stack stack-depth n + ( ndepth )
                    444:     stack stack-number @ part-num @ s-c-max-depth @
                    445: \    max-depth stack stack-number @ th @ ( ndepth nmaxdepth )
                    446:     over <= if ( ndepth ) \ load from memory
1.144     anton     447:        stack state-in normal-stack-access
1.118     anton     448:     else
                    449:        drop n stack part-stack-access
                    450:     endif ;
                    451: 
1.119     anton     452: : stack-diff ( stack -- n )
                    453:     \ in-out
                    454:     dup stack-in @ swap stack-out @ - ;
                    455: 
                    456: : part-stack-write { n stack -- }
                    457:     stack stack-depth n +
                    458:     stack stack-number @ part-num @ s-c-max-back-depth @
                    459:     over <= if ( ndepth )
                    460:        stack combined ['] stack-diff prim-context -
1.144     anton     461:        stack state-out normal-stack-access
1.119     anton     462:     else
                    463:        drop n stack part-stack-access
                    464:     endif ;
1.118     anton     465: 
                    466: : stack-read ( n stack -- )
                    467:     \ print a stack access at index n of stack
                    468:     in-part @ if
                    469:        part-stack-read
                    470:     else
1.144     anton     471:        state-in normal-stack-access
1.118     anton     472:     endif ;
                    473: 
                    474: : stack-write ( n stack -- )
1.79      anton     475:     \ print a stack access at index n of stack
                    476:     in-part @ if
1.118     anton     477:        part-stack-write
1.79      anton     478:     else
1.144     anton     479:        state-out normal-stack-access
1.79      anton     480:     endif ;
                    481: 
1.53      anton     482: : item-in-index { item -- n }
1.49      anton     483:     \ n is the index of item (in the in-effect)
1.53      anton     484:     item item-stack @ dup >r stack-in @ ( in-size r:stack )
                    485:     item r> stack-in-index-xt @ execute ;
1.1       anton     486: 
1.78      anton     487: : item-stack-type-name ( item -- addr u )
                    488:     item-stack @ stack-type @ type-c-name 2@ ;
                    489: 
1.1       anton     490: : fetch-single ( item -- )
1.106     anton     491:     \ fetch a single stack item from its stack
                    492:     >r
                    493:     ." vm_" r@ item-stack-type-name type
                    494:     ." 2" r@ item-type @ print-type-prefix ." ("
1.118     anton     495:     r@ item-in-index r@ item-stack @ stack-read ." ,"
1.106     anton     496:     r@ item-name 2@ type
                    497:     ." );" cr
                    498:     rdrop ; 
1.1       anton     499: 
                    500: : fetch-double ( item -- )
1.106     anton     501:     \ fetch a double stack item from its stack
                    502:     >r
                    503:     ." vm_two"
                    504:     r@ item-stack-type-name type ." 2"
                    505:     r@ item-type @ print-type-prefix ." ("
1.162     anton     506:     r@ item-in-index r@ item-stack @ 2dup stack-read
                    507:     ." , "                      -1 under+ stack-read
1.106     anton     508:     ." , " r@ item-name 2@ type
                    509:     ." )" cr
                    510:     rdrop ;
1.1       anton     511: 
1.49      anton     512: : same-as-in? ( item -- f )
1.158     anton     513:     \ f is true iff the offset and stack of item is the same as on input
                    514:     >r
                    515:     r@ item-stack @ stack-prim-stacks-sync @ if
                    516:        rdrop false exit
                    517:     endif
                    518:     r@ item-first @ if
                    519:        rdrop false exit
                    520:     endif
                    521:     r@ item-name 2@ prim prim-items-wordlist @ search-wordlist 0= abort" bug"
                    522:     execute @
                    523:     dup r@ =
                    524:     if \ item first appeared in output
                    525:        drop false
                    526:     else
                    527:        dup  item-stack  @ r@ item-stack  @ = 
                    528:        swap item-offset @ r@ item-offset @ = and
                    529:     endif
                    530:     rdrop ;
1.1       anton     531: 
1.49      anton     532: : item-out-index ( item -- n )
1.144     anton     533:     \ n is the index of item (in the out-effect)
1.49      anton     534:     >r r@ item-stack @ stack-out @ r> item-offset @ - 1- ;
1.31      pazsan    535: 
1.1       anton     536: : really-store-single ( item -- )
1.106     anton     537:     >r
                    538:     ." vm_"
                    539:     r@ item-type @ print-type-prefix ." 2"
                    540:     r@ item-stack-type-name type ." ("
                    541:     r@ item-name 2@ type ." ,"
1.118     anton     542:     r@ item-out-index r@ item-stack @ stack-write ." );"
1.106     anton     543:     rdrop ;
1.1       anton     544: 
1.144     anton     545: : store-single { item -- }
                    546:     item item-stack @ { stack }
                    547:     store-optimization @ in-part @ 0= and item same-as-in? and
1.147     anton     548:     item item-in-index  stack state-in  stack-reg       \  in reg/mem
                    549:     item item-out-index stack state-out stack-reg = and \ out reg/mem
1.144     anton     550:     0= if
                    551:        item really-store-single cr
                    552:     endif ;
1.1       anton     553: 
                    554: : store-double ( item -- )
                    555: \ !! store optimization is not performed, because it is not yet needed
                    556:  >r
1.78      anton     557:  ." vm_"
                    558:  r@ item-type @ print-type-prefix ." 2two"
                    559:  r@ item-stack-type-name type ." ("
                    560:  r@ item-name 2@ type ." , "
1.118     anton     561:  r@ item-out-index r@ item-stack @ 2dup stack-write
                    562:  ." , "                       -1 under+ stack-write
1.106     anton     563:  ." )" cr
1.1       anton     564:  rdrop ;
                    565: 
1.54      anton     566: : single ( -- xt1 xt2 n )
                    567:     ['] fetch-single ['] store-single 1 ;
1.1       anton     568: 
1.54      anton     569: : double ( -- xt1 xt2 n )
                    570:     ['] fetch-double ['] store-double 2 ;
1.1       anton     571: 
                    572: : s, ( addr u -- )
                    573: \ allocate a string
                    574:  here swap dup allot move ;
                    575: 
1.50      anton     576: wordlist constant prefixes
                    577: 
                    578: : declare ( addr "name" -- )
                    579: \ remember that there is a stack item at addr called name
                    580:  create , ;
                    581: 
                    582: : !default ( w addr -- )
                    583:     dup @ if
                    584:        2drop \ leave nonzero alone
                    585:     else
                    586:        !
                    587:     endif ;
                    588: 
                    589: : create-type { addr u xt1 xt2 n stack -- } ( "prefix" -- )
1.49      anton     590:     \ describes a type
                    591:     \ addr u specifies the C type name
                    592:     \ stack effect entries of the type start with prefix
                    593:     create type% %allot >r
                    594:     addr u save-mem r@ type-c-name 2!
                    595:     xt1   r@ type-fetch !
                    596:     xt2   r@ type-store !
                    597:     n     r@ type-size !
                    598:     stack r@ type-stack !
                    599:     rdrop ;
1.1       anton     600: 
1.105     anton     601: : type-prefix ( addr u xt1 xt2 n stack "prefix" -- )
1.94      anton     602:     get-current >r prefixes set-current
                    603:     create-type r> set-current
1.50      anton     604: does> ( item -- )
                    605:     \ initialize item
                    606:     { item typ }
                    607:     typ item item-type !
                    608:     typ type-stack @ item item-stack !default
1.75      anton     609:     item item-name 2@ prim prim-items-wordlist @ search-wordlist 0= if
1.66      anton     610:        item item-name 2@ nextname item declare
                    611:        item item-first on
                    612:        \ typ type-c-name 2@ type space type  ." ;" cr
1.50      anton     613:     else
                    614:        drop
1.66      anton     615:        item item-first off
1.50      anton     616:     endif ;
                    617: 
                    618: : execute-prefix ( item addr1 u1 -- )
                    619:     \ execute the word ( item -- ) associated with the longest prefix
                    620:     \ of addr1 u1
                    621:     0 swap ?do
                    622:        dup i prefixes search-wordlist
                    623:        if \ ok, we have the type ( item addr1 xt )
                    624:            nip execute
                    625:            UNLOOP EXIT
                    626:        endif
                    627:        -1 s+loop
1.158     anton     628:        \ we did not find a type, abort
                    629:        abort
1.81      anton     630:     false s" unknown prefix" ?print-error ;
1.1       anton     631: 
                    632: : declaration ( item -- )
1.50      anton     633:     dup item-name 2@ execute-prefix ;
1.1       anton     634: 
1.64      anton     635: : declaration-list ( addr1 addr2 -- )
                    636:     ['] declaration map-items ;
                    637: 
                    638: : declarations ( -- )
1.75      anton     639:  wordlist dup prim prim-items-wordlist ! set-current
1.69      anton     640:  prim prim-effect-in prim prim-effect-in-end @ declaration-list
                    641:  prim prim-effect-out prim prim-effect-out-end @ declaration-list ;
1.64      anton     642: 
1.66      anton     643: : print-declaration { item -- }
                    644:     item item-first @ if
                    645:        item item-type @ type-c-name 2@ type space
                    646:        item item-name 2@ type ." ;" cr
                    647:     endif ;
                    648: 
                    649: : print-declarations ( -- )
1.69      anton     650:     prim prim-effect-in  prim prim-effect-in-end  @ ['] print-declaration map-items
                    651:     prim prim-effect-out prim prim-effect-out-end @ ['] print-declaration map-items ;
1.66      anton     652:     
1.51      anton     653: : stack-prefix ( stack "prefix" -- )
1.94      anton     654:     get-current >r prefixes set-current
1.51      anton     655:     name tuck nextname create ( stack length ) 2,
1.94      anton     656:     r> set-current
1.51      anton     657: does> ( item -- )
                    658:     2@ { item stack prefix-length }
                    659:     item item-name 2@ prefix-length /string item item-name 2!
                    660:     stack item item-stack !
                    661:     item declaration ;
1.73      anton     662: 
1.157     anton     663: : set-prim-stacks-sync ( stack -- )
                    664:     stack-prim-stacks-sync on ;
                    665: 
                    666: : clear-prim-stacks-sync ( stack -- )
                    667:     stack-prim-stacks-sync off ;
                    668: 
                    669: 
1.156     anton     670: get-current prefixes set-current
                    671: : ... ( item -- )
                    672:     \ this "prefix" ensures that the appropriate stack is synced with memory
                    673:     dup item-name 2@ s" ..." str= 0= abort" '...' must end the item name"
                    674:     item-stack @ dup if
1.157     anton     675:        set-prim-stacks-sync
1.156     anton     676:     else \ prefixless "..." syncs all stacks
1.158     anton     677:        drop ['] set-prim-stacks-sync map-stacks1
1.156     anton     678:     endif ;
                    679: set-current
                    680: 
                    681: create ...-item ( -- addr ) \ just used for letting stack-prefixes work on it
                    682: item% %allot                \ stores the stack temporarily until used by ...
                    683: 
                    684: : init-item1 ( addr1 addr u -- addr2 )
                    685:     \ initialize item at addr1 with name addr u, next item is at addr2
                    686:     \ !! make sure that any mention of "..." is only stack-prefixed
                    687:     2dup s" ..." search nip nip if ( addr1 addr u )
                    688:        0 ...-item item-stack ! \ initialize to prefixless
                    689:        2dup ...-item item-name 2!
                    690:        ...-item rot rot execute-prefix ( addr1 )
                    691:     else
                    692:        2 pick init-item item% %size +
                    693:     endif ;
                    694: 
1.74      anton     695: \ types pointed to by stacks for use in combined prims
1.83      anton     696: \ !! output-c-combined shouldn't use these names!
1.92      anton     697: : stack-type-name ( addr u "name" -- )
                    698:     single 0 create-type ;
                    699: 
1.93      anton     700: wordlist constant type-names \ this is here just to meet the requirement
                    701:                     \ that a type be a word; it is never used for lookup
1.83      anton     702: 
1.144     anton     703: : define-type ( addr u -- xt )
                    704:     \ define single type with name addr u, without stack
                    705:     get-current type-names set-current >r
                    706:     2dup nextname stack-type-name
                    707:     r> set-current
                    708:     latestxt ;
                    709: 
1.93      anton     710: : stack ( "name" "stack-pointer" "type" -- )
                    711:     \ define stack
                    712:     name { d: stack-name }
                    713:     name { d: stack-pointer }
                    714:     name { d: stack-type }
1.144     anton     715:     stack-type define-type
                    716:     stack-pointer rot >body stack-name nextname make-stack ;
1.93      anton     717: 
                    718: stack inst-stream IP Cell
1.73      anton     719: ' inst-in-index inst-stream stack-in-index-xt !
1.80      anton     720: ' inst-stream <is> inst-stream-f
1.73      anton     721: \ !! initialize stack-in and stack-out
1.1       anton     722: 
1.144     anton     723: \ registers
                    724: 
                    725: : make-register ( type addr u -- )
                    726:     \ define register with type TYPE and name ADDR U.
                    727:     nregisters @ max-registers < s" too many registers" ?print-error
                    728:     2dup nextname create register% %allot >r
                    729:     r@ register-name 2!
                    730:     r@ register-type !
                    731:     nregisters @ r@ register-number !
                    732:     1 nregisters +!
                    733:     rdrop ;
                    734: 
                    735: : register ( "name" "type" -- )
                    736:     \ define register
                    737:     name { d: reg-name }
                    738:     name { d: reg-type }
                    739:     reg-type define-type >body
                    740:     reg-name make-register ;
                    741: 
                    742: \ stack-states
                    743: 
                    744: : stack-state ( a-addr u uoffset "name" -- )
                    745:     create ss% %allot >r
                    746:     r@ ss-offset !
                    747:     r@ ss-registers 2!
                    748:     rdrop ;
                    749: 
                    750: 0 0 0 stack-state default-ss
                    751: 
                    752: \ state
                    753: 
                    754: : state ( "name" -- )
                    755:     \ create a state initialized with default-sss
                    756:     create state% %allot { s }
1.161     anton     757:     s state-enabled on
1.144     anton     758:     next-state-number @ s state-number ! 1 next-state-number +!
                    759:     max-stacks 0 ?do
                    760:        default-ss s state-sss i th !
                    761:     loop ;
                    762: 
1.161     anton     763: : state-disable ( state -- )
                    764:     state-enabled off ;
                    765: 
                    766: : state-enabled? ( state -- f )
                    767:     state-enabled @ ;
                    768: 
1.149     anton     769: : .state ( state -- )
                    770:     0 >body - >name .name ;
                    771: 
1.144     anton     772: : set-ss ( ss stack state -- )
                    773:     state-sss swap stack-number @ th ! ;
                    774: 
1.1       anton     775: \ offset computation
                    776: \ the leftmost (i.e. deepest) item has offset 0
                    777: \ the rightmost item has the highest offset
                    778: 
1.49      anton     779: : compute-offset { item xt -- }
                    780:     \ xt specifies in/out; update stack-in/out and set item-offset
                    781:     item item-type @ type-size @
                    782:     item item-stack @ xt execute dup @ >r +!
                    783:     r> item item-offset ! ;
                    784: 
1.64      anton     785: : compute-offset-in ( addr1 addr2 -- )
                    786:     ['] stack-in compute-offset ;
                    787: 
                    788: : compute-offset-out ( addr1 addr2 -- )
                    789:     ['] stack-out compute-offset ;
1.49      anton     790: 
1.1       anton     791: : compute-offsets ( -- )
1.132     anton     792:     prim prim-stacks-in  max-stacks cells erase
                    793:     prim prim-stacks-out max-stacks cells erase
1.69      anton     794:     prim prim-effect-in  prim prim-effect-in-end  @ ['] compute-offset-in  map-items
                    795:     prim prim-effect-out prim prim-effect-out-end @ ['] compute-offset-out map-items
1.81      anton     796:     inst-stream stack-out @ 0= s" # can only be on the input side" ?print-error ;
                    797: 
1.156     anton     798: : init-simple { prim -- }
                    799:     \ much of the initialization is elsewhere
1.157     anton     800:     ['] clear-prim-stacks-sync map-stacks ;
1.156     anton     801: 
1.81      anton     802: : process-simple ( -- )
                    803:     prim prim { W^ key } key cell
1.82      anton     804:     combinations ['] constant insert-wordlist
1.81      anton     805:     declarations compute-offsets
1.82      anton     806:     output @ execute ;
1.49      anton     807: 
1.144     anton     808: : stack-state-items ( stack state -- n )
                    809:     state-ss ss-registers 2@ nip ;
                    810: 
                    811: : unused-stack-items { stack -- n-in n-out }
                    812:     \ n-in  are the stack items in state-in  not used    by prim
                    813:     \ n-out are the stack items in state-out not written by prim
                    814:     stack state-in  stack-state-items stack stack-in  @ - 0 max
                    815:     stack state-out stack-state-items stack stack-out @ - 0 max ;
                    816: 
1.157     anton     817: : spill-stack-items { stack -- u }
                    818:     \ there are u items to spill in stack
                    819:     stack unused-stack-items
                    820:     stack stack-prim-stacks-sync @ if
                    821:        drop 0
                    822:     endif
                    823:     swap - ;
                    824: 
1.144     anton     825: : spill-stack { stack -- }
                    826:     \ spill regs of state-in that are not used by prim and are not in state-out
                    827:     stack state-in stack-offset { offset }
                    828:     stack state-in stack-state-items ( items )
1.157     anton     829:     dup stack spill-stack-items + +do
1.144     anton     830:        \ loop through the bottom items
                    831:        stack stack-pointer 2@ type
                    832:        i offset - stack normal-stack-access0 ."  = "
                    833:        i stack state-in normal-stack-access1 ." ;" cr
                    834:     loop ;
1.1       anton     835: 
1.144     anton     836: : spill-state ( -- )
                    837:     ['] spill-stack map-stacks1 ;
1.49      anton     838: 
1.157     anton     839: : fill-stack-items { stack -- u }
                    840:     \ there are u items to fill in stack
                    841:     stack unused-stack-items
                    842:     stack stack-prim-stacks-sync @ if
                    843:        swap drop 0 swap
                    844:     endif
                    845:     - ;
                    846: 
1.144     anton     847: : fill-stack { stack -- }
                    848:     stack state-out stack-offset { offset }
                    849:     stack state-out stack-state-items ( items )
1.157     anton     850:     dup stack fill-stack-items + +do
1.144     anton     851:        \ loop through the bottom items
                    852:        i stack state-out normal-stack-access1 ."  = "
                    853:        stack stack-pointer 2@ type
                    854:        i offset - stack normal-stack-access0 ." ;" cr
                    855:     loop ;
1.1       anton     856: 
1.144     anton     857: : fill-state ( -- )
1.53      anton     858:     \ !! inst-stream for prefetching?
1.144     anton     859:     ['] fill-stack map-stacks1 ;
1.49      anton     860: 
                    861: : fetch ( addr -- )
1.72      anton     862:     dup item-type @ type-fetch @ execute ;
1.1       anton     863: 
                    864: : fetches ( -- )
1.69      anton     865:     prim prim-effect-in prim prim-effect-in-end @ ['] fetch map-items ;
1.49      anton     866: 
1.144     anton     867: : reg-reg-move ( reg-from reg-to -- )
                    868:     2dup = if
                    869:        2drop
                    870:     else
                    871:        .reg ."  = " .reg ." ;" cr
                    872:     endif ;
                    873: 
                    874: : stack-bottom-reg { n stack state -- reg }
                    875:     stack state stack-state-items n - 1- stack state stack-reg ;
                    876: 
                    877: : stack-moves { stack -- }
                    878:     \ generate moves between registers in state-in/state-out that are
                    879:     \ not spilled or consumed/produced by prim.
                    880:     \ !! this works only for a simple stack cache, not e.g., for
                    881:     \ rotating stack caches, or registers shared between stacks (the
                    882:     \ latter would also require a change in interface)
                    883:     \ !! maybe place this after NEXT_P1?
                    884:     stack unused-stack-items 2dup < if ( n-in n-out )
                    885:        \ move registers from 0..n_in-1 to n_out-n_in..n_out-1
                    886:        over - { diff } ( n-in )
                    887:        -1 swap 1- -do
                    888:            i stack state-in stack-bottom-reg ( reg-from )
                    889:            i diff + stack state-out stack-bottom-reg reg-reg-move
                    890:        1 -loop
                    891:     else
                    892:        \ move registers from n_in-n_out..n_in-1 to 0..n_out-1
                    893:        swap over - { diff } ( n-out )
                    894:        0 +do
                    895:            i diff + stack state-in stack-bottom-reg ( reg-from )
                    896:            i stack state-out stack-bottom-reg reg-reg-move
                    897:        loop
                    898:     endif ;
                    899: 
1.126     anton     900: : stack-update-transform ( n1 stack -- n2 )
                    901:     \ n2 is the number by which the stack pointer should be
                    902:     \ incremented to pop n1 items
                    903:     stack-access-transform @ dup >r execute
                    904:     0 r> execute - ;
                    905: 
1.157     anton     906: : update-stack-pointer { stack n -- }
                    907:     n if \ this check is not necessary, gcc would do this for us
1.118     anton     908:        stack inst-stream = if
1.157     anton     909:            ." INC_IP(" n 0 .r ." );" cr
1.118     anton     910:        else
1.126     anton     911:            stack stack-pointer 2@ type ."  += "
1.157     anton     912:            n stack stack-update-transform 0 .r ." ;" cr
1.118     anton     913:        endif
1.157     anton     914:     endif ;
                    915: 
                    916: : stack-pointer-update { stack -- }
                    917:     \ and moves
                    918:     \ stacks grow downwards
                    919:     stack stack-prim-stacks-sync @ if
                    920:        stack stack-in @
                    921:        stack state-in  stack-offset -
                    922:        stack swap update-stack-pointer
                    923:     else
                    924:        stack stack-diff ( in-out )
                    925:        stack state-in  stack-offset -
                    926:        stack state-out stack-offset + ( [in-in_offset]-[out-out_offset] )
                    927:        stack swap update-stack-pointer
                    928:        stack stack-moves
                    929:     endif ;
1.55      anton     930: 
1.1       anton     931: : stack-pointer-updates ( -- )
1.92      anton     932:     ['] stack-pointer-update map-stacks ;
1.1       anton     933: 
1.157     anton     934: : stack-pointer-update2 { stack -- }
                    935:     stack stack-prim-stacks-sync @ if
                    936:        stack state-out stack-offset
                    937:        stack stack-out @ -
                    938:        stack swap update-stack-pointer
                    939:     endif ;
                    940: 
                    941: : stack-pointer-updates2 ( -- )
                    942:     \ update stack pointers after C code, where necessary
                    943:     ['] stack-pointer-update2 map-stacks ;
                    944: 
1.1       anton     945: : store ( item -- )
                    946: \ f is true if the item should be stored
                    947: \ f is false if the store is probably not necessary
1.49      anton     948:  dup item-type @ type-store @ execute ;
1.1       anton     949: 
                    950: : stores ( -- )
1.69      anton     951:     prim prim-effect-out prim prim-effect-out-end @ ['] store map-items ;
1.8       pazsan    952: 
1.91      anton     953: : print-debug-arg { item -- }
                    954:     ." fputs(" quote space item item-name 2@ type ." =" quote ." , vm_out); "
                    955:     ." printarg_" item item-type @ print-type-prefix
                    956:     ." (" item item-name 2@ type ." );" cr ;
                    957:     
                    958: : print-debug-args ( -- )
                    959:     ." #ifdef VM_DEBUG" cr
                    960:     ." if (vm_debug) {" cr
                    961:     prim prim-effect-in prim prim-effect-in-end @ ['] print-debug-arg map-items
                    962: \    ." fputc('\n', vm_out);" cr
                    963:     ." }" cr
                    964:     ." #endif" cr ;
                    965: 
                    966: : print-debug-result { item -- }
                    967:     item item-first @ if
                    968:        item print-debug-arg
                    969:     endif ;
                    970: 
                    971: : print-debug-results ( -- )
                    972:     cr
                    973:     ." #ifdef VM_DEBUG" cr
                    974:     ." if (vm_debug) {" cr
                    975:     ." fputs(" quote ."  -- " quote ." , vm_out); "
                    976:     prim prim-effect-out prim prim-effect-out-end @ ['] print-debug-result map-items
                    977:     ." fputc('\n', vm_out);" cr
                    978:     ." }" cr
                    979:     ." #endif" cr ;
                    980: 
1.86      anton     981: : output-super-end ( -- )
                    982:     prim prim-c-code 2@ s" SET_IP" search if
                    983:        ." SUPER_END;" cr
                    984:     endif
                    985:     2drop ;
                    986: 
1.145     anton     987: 
                    988: defer output-nextp0
                    989: :noname ( -- )
                    990:     ." NEXT_P0;" cr ;
                    991: is output-nextp0
                    992: 
                    993: defer output-nextp1
                    994: :noname ( -- )
                    995:     ." NEXT_P1;" cr ;
                    996: is output-nextp1
                    997: 
1.124     anton     998: : output-nextp2 ( -- )
                    999:     ." NEXT_P2;" cr ;
                   1000: 
                   1001: variable tail-nextp2 \ xt to execute for printing NEXT_P2 in INST_TAIL
                   1002: ' output-nextp2 tail-nextp2 !
                   1003: 
1.120     anton    1004: : output-label2 ( -- )
1.121     anton    1005:     ." LABEL2(" prim prim-c-name 2@ type ." )" cr
1.153     anton    1006:     ." NEXT_P1_5;" cr
                   1007:     ." LABEL3(" prim prim-c-name 2@ type ." )" cr
                   1008:     ." DO_GOTO;" cr ;
1.120     anton    1009: 
                   1010: : output-c-tail1 { xt -- }
                   1011:     \ the final part of the generated C code, with xt printing LABEL2 or not.
1.86      anton    1012:     output-super-end
1.91      anton    1013:     print-debug-results
1.145     anton    1014:     output-nextp1
1.157     anton    1015:     stack-pointer-updates2
1.52      anton    1016:     stores
1.144     anton    1017:     fill-state 
1.121     anton    1018:     xt execute ;
1.108     anton    1019: 
1.155     anton    1020: : output-c-vm-jump-tail ( -- )
                   1021:     \ !! this functionality not yet implemented for superinstructions
                   1022:     output-super-end
                   1023:     print-debug-results
                   1024:     stores
                   1025:     fill-state 
                   1026:     ." LABEL2(" prim prim-c-name 2@ type ." )" cr
                   1027:     ." LABEL3(" prim prim-c-name 2@ type ." )" cr
                   1028:     ." DO_GOTO;" cr ;
                   1029: 
1.120     anton    1030: : output-c-tail1-no-stores { xt -- }
                   1031:     \ the final part of the generated C code for combinations
                   1032:     output-super-end
1.145     anton    1033:     output-nextp1
1.144     anton    1034:     fill-state 
1.121     anton    1035:     xt execute ;
1.120     anton    1036: 
                   1037: : output-c-tail ( -- )
1.124     anton    1038:     tail-nextp2 @ output-c-tail1 ;
1.52      anton    1039: 
1.108     anton    1040: : output-c-tail2 ( -- )
1.155     anton    1041:     prim prim-c-code 2@ s" VM_JUMP(" search nip nip if
                   1042:        output-c-vm-jump-tail
                   1043:     else
                   1044:        ['] output-label2 output-c-tail1
                   1045:     endif ;
1.120     anton    1046: 
                   1047: : output-c-tail-no-stores ( -- )
1.124     anton    1048:     tail-nextp2 @ output-c-tail1-no-stores ;
1.119     anton    1049: 
                   1050: : output-c-tail2-no-stores ( -- )
1.120     anton    1051:     ['] output-label2 output-c-tail1-no-stores ;
1.108     anton    1052: 
1.85      anton    1053: : type-c-code ( c-addr u xt -- )
1.109     anton    1054:     \ like TYPE, but replaces "INST_TAIL;" with tail code produced by xt
1.85      anton    1055:     { xt }
1.111     anton    1056:     ." {" cr
                   1057:     ." #line " c-line @ . quote c-filename 2@ type quote cr
1.52      anton    1058:     begin ( c-addr1 u1 )
1.109     anton    1059:        2dup s" INST_TAIL;" search
1.52      anton    1060:     while ( c-addr1 u1 c-addr3 u3 )
                   1061:        2dup 2>r drop nip over - type
1.85      anton    1062:        xt execute
1.109     anton    1063:        2r> 10 /string
1.52      anton    1064:        \ !! resync #line missing
                   1065:     repeat
1.111     anton    1066:     2drop type
                   1067:     ." #line " out-nls @ 2 + . quote out-filename 2@ type quote cr
                   1068:     ." }" cr ;
1.63      anton    1069: 
1.72      anton    1070: : print-entry ( -- )
1.109     anton    1071:     ." LABEL(" prim prim-c-name 2@ type ." )" ;
1.154     pazsan   1072: 
                   1073: : prim-type ( addr u -- )
                   1074:     \ print out a primitive, but avoid "*/"
                   1075:     2dup s" */" search  nip nip  IF
                   1076:        bounds ?DO  I c@ dup '* = IF  drop 'x  THEN  emit  LOOP
                   1077:     ELSE  type  THEN ;
                   1078: 
1.157     anton    1079: : output-c ( -- )
1.154     pazsan   1080:     print-entry ."  /* " prim prim-name 2@ prim-type
1.149     anton    1081:     ."  ( " prim prim-stack-string 2@ type ." ) "
                   1082:     state-in .state ." -- " state-out .state ."  */" cr
1.111     anton    1083:     ." /* " prim prim-doc 2@ type ."  */" cr
                   1084:     ." NAME(" quote prim prim-name 2@ type quote ." )" cr \ debugging
                   1085:     ." {" cr
                   1086:     ." DEF_CA" cr
                   1087:     print-declarations
1.145     anton    1088:     output-nextp0
1.144     anton    1089:     spill-state
1.111     anton    1090:     fetches
                   1091:     print-debug-args
                   1092:     stack-pointer-updates
                   1093:     prim prim-c-code 2@ ['] output-c-tail type-c-code
                   1094:     output-c-tail2
                   1095:     ." }" cr
                   1096:     cr
1.1       anton    1097: ;
                   1098: 
1.56      anton    1099: : disasm-arg { item -- }
                   1100:     item item-stack @ inst-stream = if
1.107     anton    1101:        ." {" cr
                   1102:        item print-declaration
                   1103:        item fetch
                   1104:        item print-debug-arg
                   1105:        ." }" cr
1.56      anton    1106:     endif ;
                   1107: 
                   1108: : disasm-args ( -- )
1.69      anton    1109:     prim prim-effect-in prim prim-effect-in-end @ ['] disasm-arg map-items ;
1.56      anton    1110: 
                   1111: : output-disasm ( -- )
                   1112:     \ generate code for disassembling VM instructions
1.106     anton    1113:     ." if (VM_IS_INST(*ip, " function-number @ 0 .r ." )) {" cr
1.69      anton    1114:     ."   fputs(" quote prim prim-name 2@ type quote ." , vm_out);" cr
1.56      anton    1115:     disasm-args
                   1116:     ."   ip += " inst-stream stack-in @ 1+ 0 .r ." ;" cr
1.91      anton    1117:     ."   goto _endif_;" cr
                   1118:     ." }" cr ;
1.56      anton    1119: 
1.86      anton    1120: : output-profile ( -- )
                   1121:     \ generate code for postprocessing the VM block profile stuff
1.87      anton    1122:     ." if (VM_IS_INST(*ip, " function-number @ 0 .r ." )) {" cr
1.104     anton    1123:     ."   add_inst(b, " quote prim prim-name 2@ type quote ." );" cr
1.86      anton    1124:     ."   ip += " inst-stream stack-in @ 1+ 0 .r ." ;" cr
                   1125:     prim prim-c-code 2@  s" SET_IP"    search nip nip
                   1126:     prim prim-c-code 2@  s" SUPER_END" search nip nip or if
                   1127:        ."   return;" cr
1.91      anton    1128:     else
                   1129:        ."   goto _endif_;" cr
1.86      anton    1130:     endif
1.91      anton    1131:     ." }" cr ;
1.86      anton    1132: 
1.114     anton    1133: : output-profile-part ( p )
                   1134:     ."   add_inst(b, " quote
                   1135:     prim-name 2@ type
                   1136:     quote ." );" cr ;
                   1137:     
1.104     anton    1138: : output-profile-combined ( -- )
                   1139:     \ generate code for postprocessing the VM block profile stuff
                   1140:     ." if (VM_IS_INST(*ip, " function-number @ 0 .r ." )) {" cr
1.114     anton    1141:     ['] output-profile-part map-combined
1.104     anton    1142:     ."   ip += " inst-stream stack-in @ 1+ 0 .r ." ;" cr
                   1143:     combined-prims num-combined @ 1- th @ prim-c-code 2@  s" SET_IP"    search nip nip
                   1144:     combined-prims num-combined @ 1- th @ prim-c-code 2@  s" SUPER_END" search nip nip or if
                   1145:        ."   return;" cr
                   1146:     else
                   1147:        ."   goto _endif_;" cr
                   1148:     endif
                   1149:     ." }" cr ;
                   1150: 
1.143     anton    1151: : prim-branch? { prim -- f }
                   1152:     \ true if prim is a branch or super-end
                   1153:     prim prim-c-code 2@  s" SET_IP" search nip nip 0<> ;
                   1154: 
1.103     anton    1155: : output-superend ( -- )
                   1156:     \ output flag specifying whether the current word ends a dynamic superinst
1.143     anton    1157:     prim prim-branch?
                   1158:     prim prim-c-code 2@  s" SUPER_END" search nip nip 0<> or
1.103     anton    1159:     prim prim-c-code 2@  s" SUPER_CONTINUE" search nip nip 0= and
1.154     pazsan   1160:     negate 0 .r ." , /* " prim prim-name 2@ prim-type ."  */" cr ;
1.103     anton    1161: 
1.60      anton    1162: : gen-arg-parm { item -- }
                   1163:     item item-stack @ inst-stream = if
                   1164:        ." , " item item-type @ type-c-name 2@ type space
                   1165:        item item-name 2@ type
                   1166:     endif ;
                   1167: 
                   1168: : gen-args-parm ( -- )
1.69      anton    1169:     prim prim-effect-in prim prim-effect-in-end @ ['] gen-arg-parm map-items ;
1.60      anton    1170: 
                   1171: : gen-arg-gen { item -- }
                   1172:     item item-stack @ inst-stream = if
                   1173:        ."   genarg_" item item-type @ print-type-prefix
                   1174:         ." (ctp, " item item-name 2@ type ." );" cr
                   1175:     endif ;
                   1176: 
                   1177: : gen-args-gen ( -- )
1.69      anton    1178:     prim prim-effect-in prim prim-effect-in-end @ ['] gen-arg-gen map-items ;
1.60      anton    1179: 
                   1180: : output-gen ( -- )
                   1181:     \ generate C code for generating VM instructions
1.159     anton    1182:     ." void gen_" prim prim-c-name 2@ type ." ("
                   1183:     threaded-code-pointer-type 2@ type ." ctp" gen-args-parm ." )" cr
1.60      anton    1184:     ." {" cr
1.159     anton    1185:     ."   gen_inst(ctp, " function-number @ 0 .r ." );" cr
1.60      anton    1186:     gen-args-gen
1.68      anton    1187:     ." }" cr ;
1.60      anton    1188: 
1.49      anton    1189: : stack-used? { stack -- f }
                   1190:     stack stack-in @ stack stack-out @ or 0<> ;
1.44      jwilke   1191: 
1.30      pazsan   1192: : output-funclabel ( -- )
1.69      anton    1193:   ." &I_" prim prim-c-name 2@ type ." ," cr ;
1.30      pazsan   1194: 
                   1195: : output-forthname ( -- )
1.69      anton    1196:   '" emit prim prim-name 2@ type '" emit ." ," cr ;
1.30      pazsan   1197: 
1.92      anton    1198: \  : output-c-func ( -- )
                   1199: \  \ used for word libraries
                   1200: \      ." Cell * I_" prim prim-c-name 2@ type ." (Cell *SP, Cell **FP)      /* " prim prim-name 2@ type
                   1201: \      ."  ( " prim prim-stack-string 2@ type ."  ) */" cr
                   1202: \      ." /* " prim prim-doc 2@ type ."  */" cr
                   1203: \      ." NAME(" quote prim prim-name 2@ type quote ." )" cr
                   1204: \      \ debugging
                   1205: \      ." {" cr
                   1206: \      print-declarations
                   1207: \      \ !! don't know what to do about that
                   1208: \      inst-stream  stack-used? IF ." Cell *ip=IP;" cr THEN
                   1209: \      data-stack   stack-used? IF ." Cell *sp=SP;" cr THEN
                   1210: \      fp-stack     stack-used? IF ." Cell *fp=*FP;" cr THEN
                   1211: \      return-stack stack-used? IF ." Cell *rp=*RP;" cr THEN
1.144     anton    1212: \      spill-state
1.92      anton    1213: \      fetches
                   1214: \      stack-pointer-updates
                   1215: \      fp-stack   stack-used? IF ." *FP=fp;" cr THEN
                   1216: \      ." {" cr
                   1217: \      ." #line " c-line @ . quote c-filename 2@ type quote cr
                   1218: \      prim prim-c-code 2@ type
                   1219: \      ." }" cr
                   1220: \      stores
1.144     anton    1221: \      fill-state
1.92      anton    1222: \      ." return (sp);" cr
                   1223: \      ." }" cr
                   1224: \      cr ;
1.30      pazsan   1225: 
1.43      jwilke   1226: : output-label ( -- )  
1.127     anton    1227:     ." INST_ADDR(" prim prim-c-name 2@ type ." )," cr ;
1.1       anton    1228: 
1.43      jwilke   1229: : output-alias ( -- ) 
1.69      anton    1230:     ( primitive-number @ . ." alias " ) ." Primitive " prim prim-name 2@ type cr ;
1.1       anton    1231: 
1.148     anton    1232: defer output-c-prim-num ( -- )
                   1233: 
                   1234: :noname ( -- )
1.141     anton    1235:     ." N_" prim prim-c-name 2@ type ." ," cr ;
1.148     anton    1236: is output-c-prim-num
1.114     anton    1237: 
1.43      jwilke   1238: : output-forth ( -- )  
1.69      anton    1239:     prim prim-forth-code @ 0=
1.30      pazsan   1240:     IF         \ output-alias
1.28      jwilke   1241:        \ this is bad for ec: an alias is compiled if tho word does not exist!
                   1242:        \ JAW
1.69      anton    1243:     ELSE  ." : " prim prim-name 2@ type ."   ( "
                   1244:        prim prim-stack-string 2@ type ." )" cr
                   1245:        prim prim-forth-code 2@ type cr
1.30      pazsan   1246:     THEN ;
1.10      anton    1247: 
1.17      anton    1248: : output-tag-file ( -- )
1.117     pazsan   1249:     name-filename 2@ last-name-filename 2@ compare if
1.17      anton    1250:        name-filename 2@ last-name-filename 2!
                   1251:        #ff emit cr
                   1252:        name-filename 2@ type
                   1253:        ." ,0" cr
                   1254:     endif ;
                   1255: 
                   1256: : output-tag ( -- )
                   1257:     output-tag-file
1.69      anton    1258:     prim prim-name 2@ 1+ type
1.17      anton    1259:     127 emit
1.69      anton    1260:     space prim prim-name 2@ type space
1.17      anton    1261:     1 emit
                   1262:     name-line @ 0 .r
                   1263:     ." ,0" cr ;
                   1264: 
1.100     pazsan   1265: : output-vi-tag ( -- )
                   1266:     name-filename 2@ type #tab emit
                   1267:     prim prim-name 2@ type #tab emit
                   1268:     ." /^" prim prim-name 2@ type ."  *(/" cr ;
                   1269: 
1.10      anton    1270: [IFDEF] documentation
                   1271: : register-doc ( -- )
1.82      anton    1272:     prim prim-name 2@ documentation ['] create insert-wordlist
1.69      anton    1273:     prim prim-name 2@ 2,
                   1274:     prim prim-stack-string 2@ condition-stack-effect 2,
                   1275:     prim prim-wordset 2@ 2,
                   1276:     prim prim-c-name 2@ condition-pronounciation 2,
1.82      anton    1277:     prim prim-doc 2@ 2, ;
1.10      anton    1278: [THEN]
1.67      anton    1279: 
                   1280: 
1.69      anton    1281: \ combining instructions
                   1282: 
                   1283: \ The input should look like this:
                   1284: 
                   1285: \ lit_+ = lit +
                   1286: 
                   1287: \ The output should look like this:
                   1288: 
                   1289: \  I_lit_+:
                   1290: \  {
                   1291: \  DEF_CA
                   1292: \  Cell _x_ip0;
                   1293: \  Cell _x_sp0;
                   1294: \  Cell _x_sp1;
                   1295: \  NEXT_P0;
                   1296: \  _x_ip0 = (Cell) IPTOS;
                   1297: \  _x_sp0 = (Cell) spTOS;
                   1298: \  INC_IP(1);
                   1299: \  /* sp += 0; */
                   1300: \  /* lit ( #w -- w ) */
                   1301: \  /*  */
                   1302: \  NAME("lit")
                   1303: \  {
                   1304: \  Cell w;
                   1305: \  w = (Cell) _x_ip0;
                   1306: \  #ifdef VM_DEBUG
                   1307: \  if (vm_debug) {
                   1308: \  fputs(" w=", vm_out); printarg_w (w);
                   1309: \  fputc('\n', vm_out);
                   1310: \  }
                   1311: \  #endif
                   1312: \  {
                   1313: \  #line 136 "./prim"
                   1314: \  }
                   1315: \  _x_sp1 = (Cell)w;
                   1316: \  }
                   1317: \  I_plus:     /* + ( n1 n2 -- n ) */
                   1318: \  /*  */
                   1319: \  NAME("+")
                   1320: \  {
                   1321: \  DEF_CA
                   1322: \  Cell n1;
                   1323: \  Cell n2;
                   1324: \  Cell n;
                   1325: \  NEXT_P0;
                   1326: \  n1 = (Cell) _x_sp0;
                   1327: \  n2 = (Cell) _x_sp1;
                   1328: \  #ifdef VM_DEBUG
                   1329: \  if (vm_debug) {
                   1330: \  fputs(" n1=", vm_out); printarg_n (n1);
                   1331: \  fputs(" n2=", vm_out); printarg_n (n2);
                   1332: \  fputc('\n', vm_out);
                   1333: \  }
                   1334: \  #endif
                   1335: \  {
                   1336: \  #line 516 "./prim"
                   1337: \  n = n1+n2;
                   1338: \  }
                   1339: \  _x_sp0 = (Cell)n;
                   1340: \  }
                   1341: \  NEXT_P1;
                   1342: \  spTOS = (Cell)_x_sp0;
                   1343: \  NEXT_P2;
                   1344: 
1.71      anton    1345: : init-combined ( -- )
1.79      anton    1346:     prim to combined
1.71      anton    1347:     0 num-combined !
                   1348:     current-depth max-stacks cells erase
1.116     anton    1349:     include-skipped-insts @ current-depth 0 th !
1.72      anton    1350:     max-depth     max-stacks cells erase
                   1351:     min-depth     max-stacks cells erase
                   1352:     prim prim-effect-in  prim prim-effect-in-end  !
                   1353:     prim prim-effect-out prim prim-effect-out-end ! ;
1.71      anton    1354: 
                   1355: : max! ( n addr -- )
                   1356:     tuck @ max swap ! ;
                   1357: 
1.72      anton    1358: : min! ( n addr -- )
                   1359:     tuck @ min swap ! ;
                   1360: 
1.119     anton    1361: : inst-stream-adjustment ( nstack -- n )
                   1362:     \ number of stack items to add for each part
                   1363:     0= include-skipped-insts @ and negate ;
1.116     anton    1364: 
1.71      anton    1365: : add-depths { p -- }
                   1366:     \ combine stack effect of p with *-depths
                   1367:     max-stacks 0 ?do
1.72      anton    1368:        current-depth i th @
1.119     anton    1369:        p prim-stacks-in  i th @ + i inst-stream-adjustment +
1.72      anton    1370:        dup max-depth i th max!
                   1371:        p prim-stacks-out i th @ -
                   1372:        dup min-depth i th min!
                   1373:        current-depth i th !
1.71      anton    1374:     loop ;
                   1375: 
1.118     anton    1376: : copy-maxdepths ( n -- )
                   1377:     max-depth max-depths rot max-stacks * th max-stacks cells move ;
                   1378: 
1.71      anton    1379: : add-prim ( addr u -- )
                   1380:     \ add primitive given by "addr u" to combined-prims
                   1381:     primitives search-wordlist s" unknown primitive" ?print-error
                   1382:     execute { p }
1.72      anton    1383:     p combined-prims num-combined @ th !
1.118     anton    1384:     num-combined @ copy-maxdepths
1.71      anton    1385:     1 num-combined +!
1.118     anton    1386:     p add-depths
                   1387:     num-combined @ copy-maxdepths ;
1.71      anton    1388: 
                   1389: : compute-effects { q -- }
                   1390:     \ compute the stack effects of q from the depths
                   1391:     max-stacks 0 ?do
1.72      anton    1392:        max-depth i th @ dup
                   1393:        q prim-stacks-in i th !
                   1394:        current-depth i th @ -
                   1395:        q prim-stacks-out i th !
                   1396:     loop ;
                   1397: 
                   1398: : make-effect-items { stack# items effect-endp -- }
                   1399:     \ effect-endp points to a pointer to the end of the current item-array
                   1400:     \ and has to be updated
                   1401:     stacks stack# th @ { stack }
                   1402:     items 0 +do
                   1403:        effect-endp @ { item }
                   1404:        i 0 <# #s stack stack-pointer 2@ holds [char] _ hold #> save-mem
                   1405:        item item-name 2!
                   1406:        stack item item-stack !
1.74      anton    1407:        stack stack-type @ item item-type !
1.72      anton    1408:        i item item-offset !
                   1409:        item item-first on
                   1410:        item% %size effect-endp +!
                   1411:     loop ;
                   1412: 
                   1413: : init-effects { q -- }
                   1414:     \ initialize effects field for FETCHES and STORES
                   1415:     max-stacks 0 ?do
                   1416:        i q prim-stacks-in  i th @ q prim-effect-in-end  make-effect-items
                   1417:        i q prim-stacks-out i th @ q prim-effect-out-end make-effect-items
1.71      anton    1418:     loop ;
                   1419: 
1.119     anton    1420: : compute-stack-max-back-depths ( stack -- )
                   1421:     stack-number @ { stack# }
                   1422:     current-depth stack# th @ dup
                   1423:     dup stack# num-combined @ s-c-max-back-depth !
                   1424:     -1 num-combined @ 1- -do ( max-depth current-depth )
                   1425:        combined-prims i th @ { p }
                   1426:        p prim-stacks-out stack# th @ +
                   1427:        dup >r max r>
                   1428:        over stack# i s-c-max-back-depth !
                   1429:        p prim-stacks-in stack# th @ -
                   1430:        stack# inst-stream-adjustment -
                   1431:     1 -loop
                   1432:     assert( dup stack# inst-stream-adjustment negate = )
                   1433:     assert( over max-depth stack# th @ = )
                   1434:     2drop ;
                   1435: 
                   1436: : compute-max-back-depths ( -- )
                   1437:     \ compute max-back-depths.
                   1438:     \ assumes that current-depths is correct for the end of the combination
                   1439:     ['] compute-stack-max-back-depths map-stacks ;
                   1440: 
1.71      anton    1441: : process-combined ( -- )
1.81      anton    1442:     combined combined-prims num-combined @ cells
1.82      anton    1443:     combinations ['] constant insert-wordlist
1.86      anton    1444:     combined-prims num-combined @ 1- th ( last-part )
                   1445:     @ prim-c-code 2@ prim prim-c-code 2! \ used by output-super-end
1.72      anton    1446:     prim compute-effects
                   1447:     prim init-effects
1.119     anton    1448:     compute-max-back-depths
1.72      anton    1449:     output-combined perform ;
                   1450: 
1.144     anton    1451: \ reprocessing (typically to generate versions for another cache states)
                   1452: \ !! use prim-context
                   1453: 
                   1454: variable reprocessed-num 0 reprocessed-num !
                   1455: 
                   1456: : new-name ( -- c-addr u )
                   1457:     reprocessed-num @ 0
                   1458:     1 reprocessed-num +!
                   1459:     <# #s 'p hold '_ hold #> save-mem ;
                   1460: 
                   1461: : reprocess-simple ( prim -- )
                   1462:     to prim
                   1463:     new-name prim prim-c-name 2!
                   1464:     output @ execute ;
                   1465: 
                   1466: : lookup-prim ( c-addr u -- prim )
                   1467:     primitives search-wordlist 0= -13 and throw execute ;
                   1468: 
                   1469: : state-prim1 { in-state out-state prim -- }
                   1470:     in-state out-state state-default dup d= ?EXIT
1.161     anton    1471:     in-state state-enabled? out-state state-enabled? and 0= ?EXIT
1.144     anton    1472:     in-state  to state-in
                   1473:     out-state to state-out
                   1474:     prim reprocess-simple ;
                   1475: 
                   1476: : state-prim ( in-state out-state "name" -- )
                   1477:     parse-word lookup-prim state-prim1 ;
                   1478: 
                   1479: \ reprocessing with default states
                   1480: 
                   1481: \ This is a simple scheme and should be generalized
                   1482: \ assumes we only cache one stack and use simple states for that
                   1483: 
                   1484: 0 value cache-stack  \ stack that we cache
                   1485: 2variable cache-states \ states of the cache, starting with the empty state
                   1486: 
                   1487: : compute-default-state-out ( n-in -- n-out )
                   1488:     \ for the current prim
                   1489:     cache-stack stack-in @ - 0 max
1.157     anton    1490:     cache-stack stack-prim-stacks-sync @ if
                   1491:        drop 0
                   1492:     endif
1.144     anton    1493:     cache-stack stack-out @ + cache-states 2@ nip 1- min ;
                   1494: 
                   1495: : gen-prim-states ( prim -- )
                   1496:     to prim
                   1497:     cache-states 2@ swap { states } ( nstates )
                   1498:     cache-stack stack-in @ +do
                   1499:        states i th @
                   1500:        states i compute-default-state-out th @
                   1501:        prim state-prim1
                   1502:     loop ;
                   1503: 
                   1504: : prim-states ( "name" -- )
                   1505:     parse-word lookup-prim gen-prim-states ;
                   1506: 
                   1507: : gen-branch-states ( prim -- )
                   1508:     \ generate versions that produce state-default; useful for branches
                   1509:     to prim
                   1510:     cache-states 2@ swap { states } ( nstates )
                   1511:     cache-stack stack-in @ +do
                   1512:        states i th @ state-default prim state-prim1
                   1513:     loop ;
                   1514: 
                   1515: : branch-states ( out-state "name" -- )
                   1516:     parse-word lookup-prim gen-branch-states ;
                   1517: 
                   1518: \ producing state transitions
                   1519: 
                   1520: : gen-transitions ( "name" -- )
                   1521:     parse-word lookup-prim { prim }
                   1522:     cache-states 2@ { states nstates }
                   1523:     nstates 0 +do
                   1524:        nstates 0 +do
                   1525:            i j <> if
                   1526:                states i th @ states j th @ prim state-prim1
                   1527:            endif
                   1528:        loop
                   1529:     loop ;
                   1530: 
1.72      anton    1531: \ C output
                   1532: 
                   1533: : print-item { n stack -- }
                   1534:     \ print nth stack item name
1.79      anton    1535:     stack stack-type @ type-c-name 2@ type space
1.142     anton    1536:     ." MAYBE_UNUSED _" stack stack-pointer 2@ type n 0 .r ;
1.72      anton    1537: 
                   1538: : print-declarations-combined ( -- )
                   1539:     max-stacks 0 ?do
                   1540:        max-depth i th @ min-depth i th @ - 0 +do
                   1541:            i stacks j th @ print-item ." ;" cr
                   1542:        loop
                   1543:     loop ;
1.79      anton    1544: 
                   1545: : part-fetches ( -- )
                   1546:     fetches ;
                   1547: 
                   1548: : part-output-c-tail ( -- )
1.91      anton    1549:     print-debug-results
1.85      anton    1550:     stores ;
                   1551: 
                   1552: : output-combined-tail ( -- )
                   1553:     part-output-c-tail
                   1554:     in-part @ >r in-part off
1.119     anton    1555:     combined ['] output-c-tail-no-stores prim-context
1.118     anton    1556:     r> in-part ! ;
                   1557: 
                   1558: : part-stack-pointer-updates ( -- )
1.123     anton    1559:     next-stack-number @ 0 +do
1.118     anton    1560:        i part-num @ 1+ s-c-max-depth @ dup
                   1561:        i num-combined @ s-c-max-depth @ =    \ final depth
                   1562:        swap i part-num @ s-c-max-depth @ <> \ just reached now
                   1563:        part-num @ 0= \ first part
                   1564:        or and if
                   1565:            stacks i th @ stack-pointer-update
                   1566:        endif
                   1567:     loop ;
1.79      anton    1568: 
                   1569: : output-part ( p -- )
                   1570:     to prim
1.154     pazsan   1571:     ." /* " prim prim-name 2@ prim-type ."  ( " prim prim-stack-string 2@ type ." ) */" cr
1.79      anton    1572:     ." NAME(" quote prim prim-name 2@ type quote ." )" cr \ debugging
                   1573:     ." {" cr
                   1574:     print-declarations
                   1575:     part-fetches
                   1576:     print-debug-args
1.118     anton    1577:     combined ['] part-stack-pointer-updates prim-context
                   1578:     1 part-num +!
1.79      anton    1579:     prim add-depths \ !! right place?
1.85      anton    1580:     prim prim-c-code 2@ ['] output-combined-tail type-c-code
1.79      anton    1581:     part-output-c-tail
                   1582:     ." }" cr ;
                   1583: 
1.74      anton    1584: : output-parts ( -- )
1.79      anton    1585:     prim >r in-part on
                   1586:     current-depth max-stacks cells erase
1.118     anton    1587:     0 part-num !
1.114     anton    1588:     ['] output-part map-combined
1.79      anton    1589:     in-part off
1.74      anton    1590:     r> to prim ;
                   1591: 
1.72      anton    1592: : output-c-combined ( -- )
                   1593:     print-entry cr
1.74      anton    1594:     \ debugging messages just in parts
1.72      anton    1595:     ." {" cr
                   1596:     ." DEF_CA" cr
                   1597:     print-declarations-combined
1.145     anton    1598:     output-nextp0
1.144     anton    1599:     spill-state
1.118     anton    1600:     \ fetches \ now in parts
1.74      anton    1601:     \ print-debug-args
1.118     anton    1602:     \ stack-pointer-updates now in parts
1.74      anton    1603:     output-parts
1.119     anton    1604:     output-c-tail2-no-stores
1.74      anton    1605:     ." }" cr
                   1606:     cr ;
1.72      anton    1607: 
                   1608: : output-forth-combined ( -- )
1.81      anton    1609: ;
                   1610: 
                   1611: 
1.83      anton    1612: \ peephole optimization rules
1.81      anton    1613: 
1.114     anton    1614: \ data for a simple peephole optimizer that always tries to combine
                   1615: \ the currently compiled instruction with the last one.
                   1616: 
1.81      anton    1617: \ in order for this to work as intended, shorter combinations for each
                   1618: \ length must be present, and the longer combinations must follow
                   1619: \ shorter ones (this restriction may go away in the future).
                   1620:   
1.83      anton    1621: : output-peephole ( -- )
1.81      anton    1622:     combined-prims num-combined @ 1- cells combinations search-wordlist
1.114     anton    1623:     s" the prefix for this superinstruction must be defined earlier" ?print-error
1.82      anton    1624:     ." {"
                   1625:     execute prim-num @ 5 .r ." ,"
                   1626:     combined-prims num-combined @ 1- th @ prim-num @ 5 .r ." ,"
                   1627:     combined prim-num @ 5 .r ." }, /* "
                   1628:     combined prim-c-name 2@ type ."  */"
                   1629:     cr ;
                   1630: 
1.114     anton    1631: 
1.115     anton    1632: \ cost and superinstruction data for a sophisticated combiner (e.g.,
                   1633: \ shortest path)
1.114     anton    1634: 
                   1635: \ This is intended as initializer for a structure like this
                   1636: 
1.116     anton    1637: \  struct cost {
1.146     anton    1638: \    char loads;       /* number of stack loads */
                   1639: \    char stores;      /* number of stack stores */
                   1640: \    char updates;     /* number of stack pointer updates */
                   1641: \    char branch;      /* is it a branch (SET_IP) */
                   1642: \    char state_in;    /* state on entry */
                   1643: \    char state_out;   /* state on exit */
                   1644: \    short offset;     /* offset into super2 table */
                   1645: \    char length;      /* number of components */
1.114     anton    1646: \  };
                   1647: 
1.115     anton    1648: \ How do you know which primitive or combined instruction this
                   1649: \ structure refers to?  By the order of cost structures, as in most
                   1650: \ other cases.
                   1651: 
1.139     anton    1652: : super2-length ( -- n )
                   1653:     combined if
                   1654:        num-combined @
                   1655:     else
                   1656:        1
                   1657:     endif ;
                   1658: 
1.115     anton    1659: : compute-costs { p -- nloads nstores nupdates }
                   1660:     \ compute the number of loads, stores, and stack pointer updates
                   1661:     \ of a primitive or combined instruction; does not take TOS
1.139     anton    1662:     \ caching into account
1.115     anton    1663:     0 max-stacks 0 +do
                   1664:        p prim-stacks-in i th @ +
                   1665:     loop
1.139     anton    1666:     super2-length 1- - \ don't count instruction fetches of subsumed insts
1.115     anton    1667:     0 max-stacks 0 +do
                   1668:        p prim-stacks-out i th @ +
                   1669:     loop
1.139     anton    1670:     0 max-stacks 1 +do \ don't count ip updates, therefore "1 +do"
1.115     anton    1671:        p prim-stacks-in i th @ p prim-stacks-out i th @ <> -
                   1672:     loop ;
1.114     anton    1673: 
                   1674: : output-num-part ( p -- )
1.146     anton    1675:     ." N_" prim-c-name-orig 2@ type ." ," ;
1.141     anton    1676:     \ prim-num @ 4 .r ." ," ;
1.138     anton    1677: 
                   1678: : output-name-comment ( -- )
1.154     pazsan   1679:     ."  /* " prim prim-name 2@ prim-type ."  */" ;
1.138     anton    1680: 
                   1681: variable offset-super2  0 offset-super2 ! \ offset into the super2 table
1.141     anton    1682: 
1.143     anton    1683: : output-costs-prefix ( -- )
                   1684:     ." {" prim compute-costs
                   1685:     rot 2 .r ." ," swap 2 .r ." ," 2 .r ." , "
1.144     anton    1686:     prim prim-branch? negate . ." ,"
                   1687:     state-in  state-number @ 2 .r ." ,"
1.151     anton    1688:     state-out state-number @ 2 .r ." ,"
                   1689:     inst-stream stack-in @ 1 .r ." ,"
                   1690: ;
1.143     anton    1691: 
1.141     anton    1692: : output-costs-gforth-simple ( -- )
1.143     anton    1693:     output-costs-prefix
1.141     anton    1694:     prim output-num-part
1.151     anton    1695:     1 2 .r ." },"
1.141     anton    1696:     output-name-comment
                   1697:     cr ;
                   1698: 
                   1699: : output-costs-gforth-combined ( -- )
1.143     anton    1700:     output-costs-prefix
1.141     anton    1701:     ." N_START_SUPER+" offset-super2 @ 5 .r ." ,"
1.151     anton    1702:     super2-length dup 2 .r ." }," offset-super2 +!
1.141     anton    1703:     output-name-comment
                   1704:     cr ;
1.138     anton    1705: 
1.150     anton    1706: \  : output-costs ( -- )
                   1707: \      \ description of superinstructions and simple instructions
                   1708: \      ." {" prim compute-costs
                   1709: \      rot 2 .r ." ," swap 2 .r ." ," 2 .r ." ,"
                   1710: \      offset-super2 @ 5 .r ." ,"
                   1711: \      super2-length dup 2 .r ." ," offset-super2 +!
                   1712: \      inst-stream stack-in @ 1 .r ." },"
                   1713: \      output-name-comment
                   1714: \      cr ;
1.138     anton    1715: 
1.146     anton    1716: : output-super2-simple ( -- )
                   1717:     prim prim-c-name 2@ prim prim-c-name-orig 2@ d= if
1.138     anton    1718:        prim output-num-part
1.146     anton    1719:        output-name-comment
                   1720:        cr
                   1721:     endif ;   
                   1722:   
                   1723: : output-super2-combined ( -- )
                   1724:     ['] output-num-part map-combined 
1.138     anton    1725:     output-name-comment
                   1726:     cr ;   
1.69      anton    1727: 
1.67      anton    1728: \ the parser
                   1729: 
                   1730: eof-char max-member \ the whole character set + EOF
                   1731: 
                   1732: : getinput ( -- n )
                   1733:  rawinput @ endrawinput @ =
                   1734:  if
                   1735:    eof-char
                   1736:  else
                   1737:    cookedinput @ c@
                   1738:  endif ;
                   1739: 
                   1740: :noname ( n -- )
                   1741:  dup bl > if
                   1742:   emit space
                   1743:  else
                   1744:   .
                   1745:  endif ;
                   1746: print-token !
                   1747: 
                   1748: : testchar? ( set -- f )
                   1749:  getinput member? ;
                   1750: ' testchar? test-vector !
                   1751: 
1.130     anton    1752: : checksynclines ( -- )
1.67      anton    1753:     \ when input points to a newline, check if the next line is a
                   1754:     \ sync line.  If it is, perform the appropriate actions.
1.131     anton    1755:     rawinput @ begin >r
1.130     anton    1756:        s" #line " r@ over compare if
                   1757:            rdrop 1 line +! EXIT
                   1758:        endif
                   1759:        0. r> 6 chars + 20 >number drop >r drop line ! r> ( c-addr )
                   1760:        dup c@ bl = if
                   1761:            char+ dup c@ [char] " <> 0= s" sync line syntax" ?print-error
                   1762:            char+ dup 100 [char] " scan drop swap 2dup - save-mem filename 2!
                   1763:            char+
                   1764:        endif
                   1765:        dup c@ nl-char <> 0= s" sync line syntax" ?print-error
                   1766:        skipsynclines @ if
1.131     anton    1767:            char+ dup rawinput !
1.130     anton    1768:            rawinput @ c@ cookedinput @ c!
                   1769:        endif
                   1770:     again ;
1.67      anton    1771: 
                   1772: : ?nextchar ( f -- )
1.71      anton    1773:     s" syntax error, wrong char" ?print-error
1.67      anton    1774:     rawinput @ endrawinput @ <> if
                   1775:        rawinput @ c@
                   1776:        1 chars rawinput +!
                   1777:        1 chars cookedinput +!
                   1778:        nl-char = if
1.130     anton    1779:            checksynclines
1.67      anton    1780:            rawinput @ line-start !
                   1781:        endif
1.130     anton    1782:        rawinput @ c@
                   1783:        cookedinput @ c!
1.67      anton    1784:     endif ;
                   1785: 
                   1786: : charclass ( set "name" -- )
                   1787:  ['] ?nextchar terminal ;
                   1788: 
                   1789: : .. ( c1 c2 -- set )
                   1790:  ( creates a set that includes the characters c, c1<=c<=c2 )
                   1791:  empty copy-set
                   1792:  swap 1+ rot do
                   1793:   i over add-member
                   1794:  loop ;
                   1795: 
                   1796: : ` ( -- terminal ) ( use: ` c )
                   1797:  ( creates anonymous terminal for the character c )
                   1798:  char singleton ['] ?nextchar make-terminal ;
                   1799: 
                   1800: char a char z ..  char A char Z ..  union char _ singleton union  charclass letter
                   1801: char 0 char 9 ..                                       charclass digit
                   1802: bl singleton tab-char over add-member                  charclass white
                   1803: nl-char singleton eof-char over add-member complement  charclass nonl
                   1804: nl-char singleton eof-char over add-member
                   1805:     char : over add-member complement                   charclass nocolonnl
1.110     anton    1806: nl-char singleton eof-char over add-member
                   1807:     char } over add-member complement                   charclass nobracenl
1.67      anton    1808: bl 1+ maxchar .. char \ singleton complement intersection
                   1809:                                                         charclass nowhitebq
                   1810: bl 1+ maxchar ..                                        charclass nowhite
                   1811: char " singleton eof-char over add-member complement   charclass noquote
                   1812: nl-char singleton                                      charclass nl
                   1813: eof-char singleton                                     charclass eof
1.79      anton    1814: nl-char singleton eof-char over add-member             charclass nleof
1.67      anton    1815: 
                   1816: (( letter (( letter || digit )) **
                   1817: )) <- c-ident ( -- )
                   1818: 
1.158     anton    1819: (( ` . ` . ` .
                   1820: )) <- sync-stack ( -- )
                   1821: 
                   1822: (( ` # ?? (( letter || digit || ` : )) ++ sync-stack ??
                   1823: || sync-stack
1.67      anton    1824: )) <- stack-ident ( -- )
                   1825: 
                   1826: (( nowhitebq nowhite ** ))
                   1827: <- forth-ident ( -- )
                   1828: 
                   1829: Variable forth-flag
                   1830: Variable c-flag
                   1831: 
                   1832: (( (( ` e || ` E )) {{ start }} nonl ** 
                   1833:    {{ end evaluate }}
                   1834: )) <- eval-comment ( ... -- ... )
                   1835: 
                   1836: (( (( ` f || ` F )) {{ start }} nonl ** 
                   1837:    {{ end forth-flag @ IF type cr ELSE 2drop THEN }}
                   1838: )) <- forth-comment ( -- )
                   1839: 
                   1840: (( (( ` c || ` C )) {{ start }} nonl ** 
                   1841:    {{ end c-flag @ IF type cr ELSE 2drop THEN }}
                   1842: )) <- c-comment ( -- )
                   1843: 
                   1844: (( ` - nonl ** {{ 
1.140     pazsan   1845:        forth-flag @ IF forth-fdiff ." [ELSE]" cr THEN
                   1846:        c-flag @ IF
                   1847:            function-diff
                   1848:            ." #else /* " function-number @ 0 .r ."  */" cr THEN }}
1.67      anton    1849: )) <- else-comment
                   1850: 
                   1851: (( ` + {{ start }} nonl ** {{ end
                   1852:        dup
                   1853:        IF      c-flag @
1.140     pazsan   1854:            IF
                   1855:                function-diff
                   1856:                ." #ifdef HAS_" bounds ?DO  I c@ toupper emit  LOOP cr
1.67      anton    1857:                THEN
                   1858:                forth-flag @
1.140     pazsan   1859:                IF  forth-fdiff  ." has? " type ."  [IF]"  cr THEN
1.67      anton    1860:        ELSE    2drop
1.140     pazsan   1861:            c-flag @      IF
                   1862:                function-diff  ." #endif" cr THEN
                   1863:            forth-flag @  IF  forth-fdiff  ." [THEN]"  cr THEN
1.67      anton    1864:        THEN }}
                   1865: )) <- if-comment
                   1866: 
1.98      pazsan   1867: (( (( ` g || ` G )) {{ start }} nonl **
                   1868:    {{ end
1.140     pazsan   1869:       forth-flag @ IF  forth-fdiff  ." group " type cr  THEN
                   1870:       c-flag @     IF  function-diff
                   1871:          ." GROUP(" type ." , " function-number @ 0 .r ." )" cr  THEN }}
1.98      pazsan   1872: )) <- group-comment
                   1873: 
                   1874: (( (( eval-comment || forth-comment || c-comment || else-comment || if-comment || group-comment )) ?? nonl ** )) <- comment-body
1.67      anton    1875: 
1.79      anton    1876: (( ` \ comment-body nleof )) <- comment ( -- )
1.67      anton    1877: 
1.156     anton    1878: (( {{ start }} stack-ident {{ end init-item1 }} white ** )) **
                   1879: <- stack-items ( addr1 -- addr2 )
1.67      anton    1880: 
1.69      anton    1881: (( {{ prim prim-effect-in }}  stack-items {{ prim prim-effect-in-end ! }}
1.67      anton    1882:    ` - ` - white **
1.69      anton    1883:    {{ prim prim-effect-out }} stack-items {{ prim prim-effect-out-end ! }}
1.67      anton    1884: )) <- stack-effect ( -- )
                   1885: 
1.157     anton    1886: (( {{ prim create-prim prim init-simple }}
1.69      anton    1887:    ` ( white ** {{ start }} stack-effect {{ end prim prim-stack-string 2! }} ` ) white **
                   1888:    (( {{ start }} forth-ident {{ end prim prim-wordset 2! }} white **
1.146     anton    1889:       (( {{ start }}  c-ident {{ end 2dup prim-c-name-2! }} )) ??
1.79      anton    1890:    )) ??  nleof
                   1891:    (( ` " ` "  {{ start }} (( noquote ++ ` " )) ++ {{ end 1- prim prim-doc 2! }} ` " white ** nleof )) ??
1.110     anton    1892:    {{ skipsynclines off line @ c-line ! filename 2@ c-filename 2! start }}
                   1893:    (( (( ` { nonl ** nleof (( (( nobracenl {{ line @ drop }} nonl ** )) ?? nleof )) ** ` } white ** nleof white ** ))
                   1894:    || (( nocolonnl nonl **  nleof white ** )) ** ))
                   1895:    {{ end prim prim-c-code 2! skipsynclines on }}
1.79      anton    1896:    (( ` :  white ** nleof
                   1897:       {{ start }} (( nonl ++  nleof white ** )) ++ {{ end prim prim-forth-code 2! }}
1.81      anton    1898:    )) ?? {{ process-simple }}
1.79      anton    1899:    nleof
1.69      anton    1900: )) <- simple-primitive ( -- )
                   1901: 
1.71      anton    1902: (( {{ init-combined }}
1.89      anton    1903:    ` = white ** (( {{ start }} forth-ident {{ end add-prim }} white ** )) ++
1.79      anton    1904:    nleof {{ process-combined }}
1.69      anton    1905: )) <- combined-primitive
                   1906: 
1.79      anton    1907: (( {{ make-prim to prim 0 to combined
1.69      anton    1908:       line @ name-line ! filename 2@ name-filename 2!
1.82      anton    1909:       function-number @ prim prim-num !
1.110     anton    1910:       start }} [ifdef] vmgen c-ident [else] forth-ident [then] {{ end
1.146     anton    1911:       2dup prim prim-name 2! prim-c-name-2! }}  white **
                   1912:    (( ` / white ** {{ start }} c-ident {{ end prim-c-name-2! }} white ** )) ??
1.138     anton    1913:    (( simple-primitive || combined-primitive ))
                   1914:    {{ 1 function-number +! }}
1.67      anton    1915: )) <- primitive ( -- )
                   1916: 
                   1917: (( (( comment || primitive || nl white ** )) ** eof ))
                   1918: parser primitives2something
                   1919: warnings @ [IF]
                   1920: .( parser generated ok ) cr
                   1921: [THEN]
                   1922: 
1.95      jwilke   1923: 
1.97      jwilke   1924: \ run with gforth-0.5.0 (slurp-file is missing)
1.95      jwilke   1925: [IFUNDEF] slurp-file
                   1926: : slurp-file ( c-addr1 u1 -- c-addr2 u2 )
                   1927:     \ c-addr1 u1 is the filename, c-addr2 u2 is the file's contents
                   1928:     r/o bin open-file throw >r
                   1929:     r@ file-size throw abort" file too large"
                   1930:     dup allocate throw swap
                   1931:     2dup r@ read-file throw over <> abort" could not read whole file"
                   1932:     r> close-file throw ;
                   1933: [THEN]
                   1934: 
1.69      anton    1935: : primfilter ( addr u -- )
                   1936:     \ process the string at addr u
                   1937:     over dup rawinput ! dup line-start ! cookedinput !
                   1938:     + endrawinput !
1.130     anton    1939:     checksynclines
1.69      anton    1940:     primitives2something ;    
1.8       pazsan   1941: 
1.130     anton    1942: : unixify ( c-addr u1 -- c-addr u2 )
                   1943:     \ delete crs from the string
                   1944:     bounds tuck tuck ?do ( c-addr1 )
                   1945:        i c@ dup #cr <> if
                   1946:            over c! char+
                   1947:        else
                   1948:            drop
                   1949:        endif
                   1950:     loop
                   1951:     over - ;
                   1952: 
1.72      anton    1953: : process-file ( addr u xt-simple x-combined -- )
                   1954:     output-combined ! output !
1.61      anton    1955:     save-mem 2dup filename 2!
1.130     anton    1956:     slurp-file unixify
1.17      anton    1957:     warnings @ if
                   1958:        ." ------------ CUT HERE -------------" cr  endif
1.69      anton    1959:     primfilter ;
1.30      pazsan   1960: 
1.72      anton    1961: \  : process      ( xt -- )
                   1962: \      bl word count rot
                   1963: \      process-file ;

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