File:  [gforth] / gforth / Attic / prims2y.fs
Revision 1.6: download - view: text, annotated - select for diffs
Sun Oct 5 20:14:09 2003 UTC (20 years, 6 months ago) by anton
Branches: MAIN
CVS tags: HEAD
more stack caching stuff

    1: \ converts primitives to, e.g., C code 
    2: 
    3: \ Copyright (C) 1995,1996,1997,1998,2000,2003 Free Software Foundation, Inc.
    4: 
    5: \ This file is part of Gforth.
    6: 
    7: \ Gforth is free software; you can redistribute it and/or
    8: \ modify it under the terms of the GNU General Public License
    9: \ as published by the Free Software Foundation; either version 2
   10: \ of the License, or (at your option) any later version.
   11: 
   12: \ This program is distributed in the hope that it will be useful,
   13: \ but WITHOUT ANY WARRANTY; without even the implied warranty of
   14: \ MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   15: \ GNU General Public License for more details.
   16: 
   17: \ You should have received a copy of the GNU General Public License
   18: \ along with this program; if not, write to the Free Software
   19: \ Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111, USA.
   20: 
   21: 
   22: \ This is not very nice (hard limits, no checking, assumes 1 chars = 1).
   23: \ And it grew even worse when it aged.
   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:
   40: \ add the store optimization for doubles
   41: \ regarding problem 1 above: It would be better (for over) to implement
   42: \ 	the alternative
   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.
   53: 
   54: \ for backwards compatibility, jaw
   55: require compat/strcomp.fs
   56: 
   57: warnings off
   58: 
   59: \ redefinitions of kernel words not present in gforth-0.6.1
   60: : latestxt lastcfa @ ;
   61: : latest last @ ;
   62: 
   63: [IFUNDEF] try
   64: include startup.fs
   65: [THEN]
   66: 
   67: : struct% struct ; \ struct is redefined in gray
   68: 
   69: warnings off
   70: \ warnings on
   71: 
   72: include ./gray.fs
   73: 128 constant max-effect \ number of things on one side of a stack effect
   74: 4 constant max-stacks  \ the max. number of stacks (including inst-stream).
   75: 255 constant maxchar
   76: maxchar 1+ constant eof-char
   77: #tab constant tab-char
   78: #lf constant nl-char
   79: 
   80: variable rawinput \ pointer to next character to be scanned
   81: variable endrawinput \ pointer to the end of the input (the char after the last)
   82: variable cookedinput \ pointer to the next char to be parsed
   83: variable line \ line number of char pointed to by input
   84: variable line-start \ pointer to start of current line (for error messages)
   85: 0 line !
   86: 2variable filename \ filename of original input file
   87: 0 0 filename 2!
   88: 2variable out-filename \ filename of the output file (for sync lines)
   89: 0 0 out-filename 2!
   90: 2variable f-comment
   91: 0 0 f-comment 2!
   92: variable skipsynclines \ are sync lines ("#line ...") invisible to the parser?
   93: skipsynclines on
   94: variable out-nls \ newlines in output (for output sync lines)
   95: 0 out-nls !
   96: variable store-optimization \ use store optimization?
   97: store-optimization off
   98: 
   99: variable include-skipped-insts
  100: \ does the threaded code for a combined instruction include the cells
  101: \ for the component instructions (true) or only the cells for the
  102: \ inline arguments (false)
  103: include-skipped-insts off
  104: 
  105: variable immarg \ values for immediate arguments (to be used in IMM_ARG macros)
  106: $12340000 immarg !
  107: 
  108: : th ( addr1 n -- addr2 )
  109:     cells + ;
  110: 
  111: : holds ( addr u -- )
  112:     \ like HOLD, but for a string
  113:     tuck + swap 0 +do
  114: 	1- dup c@ hold
  115:     loop
  116:     drop ;
  117: 
  118: : insert-wordlist { c-addr u wordlist xt -- }
  119:     \ adds name "addr u" to wordlist using defining word xt
  120:     \ xt may cause additional stack effects
  121:     get-current >r wordlist set-current
  122:     c-addr u nextname xt execute
  123:     r> set-current ;
  124: 
  125: : start ( -- addr )
  126:  cookedinput @ ;
  127: 
  128: : end ( addr -- addr u )
  129:  cookedinput @ over - ;
  130: 
  131: : print-error-line ( -- )
  132:     \ print the current line and position
  133:     line-start @ endrawinput @ over - 2dup nl-char scan drop nip ( start end )
  134:     over - type cr
  135:     line-start @ rawinput @ over - typewhite ." ^" cr ;
  136: 
  137: : ?print-error { f addr u -- }
  138:     f ?not? if
  139: 	outfile-id >r try
  140: 	    stderr to outfile-id
  141: 	    filename 2@ type ." :" line @ 0 .r ." : " addr u type cr
  142: 	    print-error-line
  143: 	    0
  144: 	recover endtry
  145: 	r> to outfile-id throw
  146: 	1 (bye) \ abort
  147:     endif ;
  148: 
  149: : quote ( -- )
  150:     [char] " emit ;
  151: 
  152: \ count output lines to generate sync lines for output
  153: 
  154: : count-nls ( addr u -- )
  155:     bounds u+do
  156: 	i c@ nl-char = negate out-nls +!
  157:     loop ;
  158: 
  159: :noname ( addr u -- )
  160:     2dup count-nls
  161:     defers type ;
  162: is type
  163: 
  164: variable output          \ xt ( -- ) of output word for simple primitives
  165: variable output-combined \ xt ( -- ) of output word for combined primitives
  166: 
  167: struct%
  168:     cell%    field stack-number \ the number of this stack
  169:     cell% 2* field stack-pointer \ stackpointer name
  170:     cell%    field stack-type \ name for default type of stack items
  171:     cell%    field stack-in-index-xt \ ( in-size item -- in-index )
  172:     cell%    field stack-access-transform \ ( nitem -- index )
  173: end-struct stack%
  174: 
  175: struct%
  176:  cell% 2* field item-name   \ name, excluding stack prefixes
  177:  cell%    field item-stack  \ descriptor for the stack used, 0 is default
  178:  cell%    field item-type   \ descriptor for the item type
  179:  cell%    field item-offset \ offset in stack items, 0 for the deepest element
  180:  cell%	  field item-first  \ true if this is the first occurence of the item
  181: end-struct item%
  182: 
  183: struct%
  184:     cell% 2* field type-c-name
  185:     cell%    field type-stack \ default stack
  186:     cell%    field type-size  \ size of type in stack items
  187:     cell%    field type-fetch \ xt of fetch code generator ( item -- )
  188:     cell%    field type-store \ xt of store code generator ( item -- )
  189: end-struct type%
  190: 
  191: struct%
  192:     cell%    field register-number
  193:     cell%    field register-type \ pointer to type
  194:     cell% 2* field register-name \ c name
  195: end-struct register%
  196: 
  197: struct%
  198:     cell% 2* field ss-registers  \ addr u; ss-registers[0] is TOS
  199:                                  \ 0 means: use memory
  200:     cell%    field ss-offset     \ stack pointer offset: sp[-offset] is TOS
  201: end-struct ss% \ stack-state
  202: 
  203: struct%
  204:     cell% max-stacks * field state-sss
  205: end-struct state%
  206: 
  207: variable next-stack-number 0 next-stack-number !
  208: create stacks max-stacks cells allot \ array of stacks
  209: 256 constant max-registers
  210: create registers max-registers cells allot \ array of registers
  211: variable nregisters 0 nregisters ! \ number of registers
  212: 
  213: : stack-in-index ( in-size item -- in-index )
  214:     item-offset @ - 1- ;
  215: 
  216: : inst-in-index ( in-size item -- in-index )
  217:     nip dup item-offset @ swap item-type @ type-size @ + 1- ;
  218: 
  219: : make-stack ( addr-ptr u1 type "stack-name" -- )
  220:     next-stack-number @ max-stacks < s" too many stacks" ?print-error
  221:     create stack% %allot >r
  222:     r@ stacks next-stack-number @ th !
  223:     next-stack-number @ r@ stack-number !
  224:     1 next-stack-number +!
  225:     r@ stack-type !
  226:     save-mem r@ stack-pointer 2! 
  227:     ['] stack-in-index r@ stack-in-index-xt !
  228:     ['] noop r@ stack-access-transform !
  229:     rdrop ;
  230: 
  231: : map-stacks { xt -- }
  232:     \ perform xt for all stacks
  233:     next-stack-number @ 0 +do
  234: 	stacks i th @ xt execute
  235:     loop ;
  236: 
  237: : map-stacks1 { xt -- }
  238:     \ perform xt for all stacks except inst-stream
  239:     next-stack-number @ 1 +do
  240: 	stacks i th @ xt execute
  241:     loop ;
  242: 
  243: \ stack items
  244: 
  245: : init-item ( addr u addr1 -- )
  246:     \ initialize item at addr1 with name addr u
  247:     \ !! remove stack prefix
  248:     dup item% %size erase
  249:     item-name 2! ;
  250: 
  251: : map-items { addr end xt -- }
  252:     \ perform xt for all items in array addr...end
  253:     end addr ?do
  254: 	i xt execute
  255:     item% %size +loop ;
  256: 
  257: \ types
  258: 
  259: : print-type-prefix ( type -- )
  260:     body> >head name>string type ;
  261: 
  262: \ various variables for storing stuff of one primitive
  263: 
  264: struct%
  265:     cell% 2* field prim-name
  266:     cell% 2* field prim-wordset
  267:     cell% 2* field prim-c-name
  268:     cell% 2* field prim-doc
  269:     cell% 2* field prim-c-code
  270:     cell% 2* field prim-forth-code
  271:     cell% 2* field prim-stack-string
  272:     cell%    field prim-num            \ ordinal number
  273:     cell%    field prim-items-wordlist \ unique items
  274:     item% max-effect * field prim-effect-in
  275:     item% max-effect * field prim-effect-out
  276:     cell%    field prim-effect-in-end
  277:     cell%    field prim-effect-out-end
  278:     cell% max-stacks * field prim-stacks-in  \ number of in items per stack
  279:     cell% max-stacks * field prim-stacks-out \ number of out items per stack
  280: end-struct prim%
  281: 
  282: : make-prim ( -- prim )
  283:     prim% %alloc { p }
  284:     s" " p prim-doc 2! s" " p prim-forth-code 2! s" " p prim-wordset 2!
  285:     p ;
  286: 
  287: 0 value prim     \ in combined prims either combined or a part
  288: 0 value combined \ in combined prims the combined prim
  289: variable in-part \ true if processing a part
  290:  in-part off
  291: 0 value state-in  \ state on entering prim
  292: 0 value state-out \ state on exiting prim
  293: 
  294: : prim-context ( ... p xt -- ... )
  295:     \ execute xt with prim set to p
  296:     prim >r
  297:     swap to prim
  298:     catch
  299:     r> to prim
  300:     throw ;
  301: 
  302: 1000 constant max-combined
  303: create combined-prims max-combined cells allot
  304: variable num-combined
  305: variable part-num \ current part number during process-combined
  306: 
  307: : map-combined { xt -- }
  308:     \ perform xt for all components of the current combined instruction
  309:     num-combined @ 0 +do
  310: 	combined-prims i th @ xt execute
  311:     loop ;
  312: 
  313: table constant combinations
  314:   \ the keys are the sequences of pointers to primitives
  315: 
  316: create current-depth max-stacks cells allot
  317: create max-depth     max-stacks cells allot
  318: create min-depth     max-stacks cells allot
  319: 
  320: create sp-update-in max-stacks cells allot
  321: \ where max-depth occured the first time
  322: create max-depths max-stacks max-combined 1+ * cells allot
  323: \ maximum depth at start of each part: array[parts] of array[stack]
  324: create max-back-depths max-stacks max-combined 1+ * cells allot
  325: \ maximun depth from end of the combination to the start of the each part
  326: 
  327: : s-c-max-depth ( nstack ncomponent -- addr )
  328:     max-stacks * + cells max-depths + ;
  329: 
  330: : s-c-max-back-depth ( nstack ncomponent -- addr )
  331:     max-stacks * + cells max-back-depths + ;
  332: 
  333: wordlist constant primitives
  334: 
  335: : create-prim ( prim -- )
  336:     dup prim-name 2@ primitives ['] constant insert-wordlist ;
  337: 
  338: : stack-in ( stack -- addr )
  339:     \ address of number of stack items in effect in
  340:     stack-number @ cells prim prim-stacks-in + ;
  341: 
  342: : stack-out ( stack -- addr )
  343:     \ address of number of stack items in effect out
  344:     stack-number @ cells prim prim-stacks-out + ;
  345: 
  346: \ global vars
  347: variable c-line
  348: 2variable c-filename
  349: variable name-line
  350: 2variable name-filename
  351: 2variable last-name-filename
  352: Variable function-number 0 function-number !
  353: Variable function-old 0 function-old !
  354: : function-diff ( n -- )
  355:     ." GROUPADD(" function-number @ function-old @ - 0 .r ." )" cr
  356:     function-number @ function-old ! ;
  357: : forth-fdiff ( -- )
  358:     function-number @ function-old @ - 0 .r ."  groupadd" cr
  359:     function-number @ function-old ! ;
  360: 
  361: \ a few more set ops
  362: 
  363: : bit-equivalent ( w1 w2 -- w3 )
  364:  xor invert ;
  365: 
  366: : complement ( set1 -- set2 )
  367:  empty ['] bit-equivalent binary-set-operation ;
  368: 
  369: \ forward declaration for inst-stream (breaks cycle in definitions)
  370: defer inst-stream-f ( -- stack )
  371: 
  372: \ stack access stuff
  373: 
  374: : normal-stack-access0 { n stack -- }
  375:     \ n has the ss-offset already applied (see ...-access1)
  376:     n stack stack-access-transform @ execute ." [" 0 .r ." ]" ;
  377: 
  378: : state-ss { stack state -- ss }
  379:     state state-sss stack stack-number @ th @ ;
  380: 
  381: : stack-reg { n stack state -- reg }
  382:     \ n is the index (TOS=0); reg is 0 if the access is to memory
  383:     stack state state-ss ss-registers 2@ n u> if ( addr ) \ in ss-registers?
  384: 	n th @
  385:     else
  386: 	drop 0
  387:     endif ;
  388: 
  389: : .reg ( reg -- )
  390:     register-name 2@ type ;
  391: 
  392: : stack-offset ( stack state -- n )
  393:     \ offset for stack in state
  394:     state-ss ss-offset @ ;
  395: 
  396: : normal-stack-access1 { n stack state -- }
  397:     n stack state stack-reg ?dup-if
  398: 	.reg exit
  399:     endif
  400:     stack stack-pointer 2@ type
  401:     n stack state stack-offset - stack normal-stack-access0 ;
  402: 
  403: : normal-stack-access ( n stack state -- )
  404:     over inst-stream-f = if
  405: 	." IMM_ARG(" normal-stack-access1 ." ," immarg ? ." )"
  406: 	1 immarg +!
  407:     else
  408: 	normal-stack-access1
  409:     endif ;
  410: 
  411: : stack-depth { stack -- n }
  412:     current-depth stack stack-number @ th @ ;
  413: 
  414: : part-stack-access { n stack -- }
  415:     \ print _<stack><x>, x=inst-stream? n : maxdepth-currentdepth-n-1
  416:     ." _" stack stack-pointer 2@ type
  417:     stack stack-number @ { stack# }
  418:     stack stack-depth n + { access-depth }
  419:     stack inst-stream-f = if
  420: 	access-depth
  421:     else
  422: 	combined prim-stacks-in stack# th @
  423: 	assert( dup max-depth stack# th @ = )
  424: 	access-depth - 1-
  425:     endif
  426:     0 .r ;
  427: 
  428: : part-stack-read { n stack -- }
  429:     stack stack-depth n + ( ndepth )
  430:     stack stack-number @ part-num @ s-c-max-depth @
  431: \    max-depth stack stack-number @ th @ ( ndepth nmaxdepth )
  432:     over <= if ( ndepth ) \ load from memory
  433: 	stack state-in normal-stack-access
  434:     else
  435: 	drop n stack part-stack-access
  436:     endif ;
  437: 
  438: : stack-diff ( stack -- n )
  439:     \ in-out
  440:     dup stack-in @ swap stack-out @ - ;
  441: 
  442: : part-stack-write { n stack -- }
  443:     stack stack-depth n +
  444:     stack stack-number @ part-num @ s-c-max-back-depth @
  445:     over <= if ( ndepth )
  446: 	stack combined ['] stack-diff prim-context -
  447: 	stack state-out normal-stack-access
  448:     else
  449: 	drop n stack part-stack-access
  450:     endif ;
  451: 
  452: : stack-read ( n stack -- )
  453:     \ print a stack access at index n of stack
  454:     in-part @ if
  455: 	part-stack-read
  456:     else
  457: 	state-in normal-stack-access
  458:     endif ;
  459: 
  460: : stack-write ( n stack -- )
  461:     \ print a stack access at index n of stack
  462:     in-part @ if
  463: 	part-stack-write
  464:     else
  465: 	state-out normal-stack-access
  466:     endif ;
  467: 
  468: : item-in-index { item -- n }
  469:     \ n is the index of item (in the in-effect)
  470:     item item-stack @ dup >r stack-in @ ( in-size r:stack )
  471:     item r> stack-in-index-xt @ execute ;
  472: 
  473: : item-stack-type-name ( item -- addr u )
  474:     item-stack @ stack-type @ type-c-name 2@ ;
  475: 
  476: : fetch-single ( item -- )
  477:     \ fetch a single stack item from its stack
  478:     >r
  479:     ." vm_" r@ item-stack-type-name type
  480:     ." 2" r@ item-type @ print-type-prefix ." ("
  481:     r@ item-in-index r@ item-stack @ stack-read ." ,"
  482:     r@ item-name 2@ type
  483:     ." );" cr
  484:     rdrop ; 
  485: 
  486: : fetch-double ( item -- )
  487:     \ fetch a double stack item from its stack
  488:     >r
  489:     ." vm_two"
  490:     r@ item-stack-type-name type ." 2"
  491:     r@ item-type @ print-type-prefix ." ("
  492:     r@ item-in-index r@ item-stack @ 2dup ." (Cell)" stack-read
  493:     ." , "                      -1 under+ ." (Cell)" stack-read
  494:     ." , " r@ item-name 2@ type
  495:     ." )" cr
  496:     rdrop ;
  497: 
  498: : same-as-in? ( item -- f )
  499:  \ f is true iff the offset and stack of item is the same as on input
  500:  >r
  501:  r@ item-first @ if
  502:      rdrop false exit
  503:  endif
  504:  r@ item-name 2@ prim prim-items-wordlist @ search-wordlist 0= abort" bug"
  505:  execute @
  506:  dup r@ =
  507:  if \ item first appeared in output
  508:    drop false
  509:  else
  510:    dup  item-stack  @ r@ item-stack  @ = 
  511:    swap item-offset @ r@ item-offset @ = and
  512:  endif
  513:  rdrop ;
  514: 
  515: : item-out-index ( item -- n )
  516:     \ n is the index of item (in the out-effect)
  517:     >r r@ item-stack @ stack-out @ r> item-offset @ - 1- ;
  518: 
  519: : really-store-single ( item -- )
  520:     >r
  521:     ." vm_"
  522:     r@ item-type @ print-type-prefix ." 2"
  523:     r@ item-stack-type-name type ." ("
  524:     r@ item-name 2@ type ." ,"
  525:     r@ item-out-index r@ item-stack @ stack-write ." );"
  526:     rdrop ;
  527: 
  528: : store-single { item -- }
  529:     item item-stack @ { stack }
  530:     store-optimization @ in-part @ 0= and item same-as-in? and
  531:     item item-in-index  stack state-in  stack-reg 0= and \  in in memory?
  532:     item item-out-index stack state-out stack-reg 0= and \ out in memory?
  533:     0= if
  534: 	item really-store-single cr
  535:     endif ;
  536: 
  537: : store-double ( item -- )
  538: \ !! store optimization is not performed, because it is not yet needed
  539:  >r
  540:  ." vm_"
  541:  r@ item-type @ print-type-prefix ." 2two"
  542:  r@ item-stack-type-name type ." ("
  543:  r@ item-name 2@ type ." , "
  544:  r@ item-out-index r@ item-stack @ 2dup stack-write
  545:  ." , "                       -1 under+ stack-write
  546:  ." )" cr
  547:  rdrop ;
  548: 
  549: : single ( -- xt1 xt2 n )
  550:     ['] fetch-single ['] store-single 1 ;
  551: 
  552: : double ( -- xt1 xt2 n )
  553:     ['] fetch-double ['] store-double 2 ;
  554: 
  555: : s, ( addr u -- )
  556: \ allocate a string
  557:  here swap dup allot move ;
  558: 
  559: wordlist constant prefixes
  560: 
  561: : declare ( addr "name" -- )
  562: \ remember that there is a stack item at addr called name
  563:  create , ;
  564: 
  565: : !default ( w addr -- )
  566:     dup @ if
  567: 	2drop \ leave nonzero alone
  568:     else
  569: 	!
  570:     endif ;
  571: 
  572: : create-type { addr u xt1 xt2 n stack -- } ( "prefix" -- )
  573:     \ describes a type
  574:     \ addr u specifies the C type name
  575:     \ stack effect entries of the type start with prefix
  576:     create type% %allot >r
  577:     addr u save-mem r@ type-c-name 2!
  578:     xt1   r@ type-fetch !
  579:     xt2   r@ type-store !
  580:     n     r@ type-size !
  581:     stack r@ type-stack !
  582:     rdrop ;
  583: 
  584: : type-prefix ( addr u xt1 xt2 n stack "prefix" -- )
  585:     get-current >r prefixes set-current
  586:     create-type r> set-current
  587: does> ( item -- )
  588:     \ initialize item
  589:     { item typ }
  590:     typ item item-type !
  591:     typ type-stack @ item item-stack !default
  592:     item item-name 2@ prim prim-items-wordlist @ search-wordlist 0= if
  593: 	item item-name 2@ nextname item declare
  594: 	item item-first on
  595: 	\ typ type-c-name 2@ type space type  ." ;" cr
  596:     else
  597: 	drop
  598: 	item item-first off
  599:     endif ;
  600: 
  601: : execute-prefix ( item addr1 u1 -- )
  602:     \ execute the word ( item -- ) associated with the longest prefix
  603:     \ of addr1 u1
  604:     0 swap ?do
  605: 	dup i prefixes search-wordlist
  606: 	if \ ok, we have the type ( item addr1 xt )
  607: 	    nip execute
  608: 	    UNLOOP EXIT
  609: 	endif
  610: 	-1 s+loop
  611:     \ we did not find a type, abort
  612:     false s" unknown prefix" ?print-error ;
  613: 
  614: : declaration ( item -- )
  615:     dup item-name 2@ execute-prefix ;
  616: 
  617: : declaration-list ( addr1 addr2 -- )
  618:     ['] declaration map-items ;
  619: 
  620: : declarations ( -- )
  621:  wordlist dup prim prim-items-wordlist ! set-current
  622:  prim prim-effect-in prim prim-effect-in-end @ declaration-list
  623:  prim prim-effect-out prim prim-effect-out-end @ declaration-list ;
  624: 
  625: : print-declaration { item -- }
  626:     item item-first @ if
  627: 	item item-type @ type-c-name 2@ type space
  628: 	item item-name 2@ type ." ;" cr
  629:     endif ;
  630: 
  631: : print-declarations ( -- )
  632:     prim prim-effect-in  prim prim-effect-in-end  @ ['] print-declaration map-items
  633:     prim prim-effect-out prim prim-effect-out-end @ ['] print-declaration map-items ;
  634:     
  635: : stack-prefix ( stack "prefix" -- )
  636:     get-current >r prefixes set-current
  637:     name tuck nextname create ( stack length ) 2,
  638:     r> set-current
  639: does> ( item -- )
  640:     2@ { item stack prefix-length }
  641:     item item-name 2@ prefix-length /string item item-name 2!
  642:     stack item item-stack !
  643:     item declaration ;
  644: 
  645: \ types pointed to by stacks for use in combined prims
  646: \ !! output-c-combined shouldn't use these names!
  647: : stack-type-name ( addr u "name" -- )
  648:     single 0 create-type ;
  649: 
  650: wordlist constant type-names \ this is here just to meet the requirement
  651:                     \ that a type be a word; it is never used for lookup
  652: 
  653: : define-type ( addr u -- xt )
  654:     \ define single type with name addr u, without stack
  655:     get-current type-names set-current >r
  656:     2dup nextname stack-type-name
  657:     r> set-current
  658:     latestxt ;
  659: 
  660: : stack ( "name" "stack-pointer" "type" -- )
  661:     \ define stack
  662:     name { d: stack-name }
  663:     name { d: stack-pointer }
  664:     name { d: stack-type }
  665:     stack-type define-type
  666:     stack-pointer rot >body stack-name nextname make-stack ;
  667: 
  668: stack inst-stream IP Cell
  669: ' inst-in-index inst-stream stack-in-index-xt !
  670: ' inst-stream <is> inst-stream-f
  671: \ !! initialize stack-in and stack-out
  672: 
  673: \ registers
  674: 
  675: : make-register ( type addr u -- )
  676:     \ define register with type TYPE and name ADDR U.
  677:     nregisters @ max-registers < s" too many registers" ?print-error
  678:     2dup nextname create register% %allot >r
  679:     r@ register-name 2!
  680:     r@ register-type !
  681:     nregisters @ r@ register-number !
  682:     1 nregisters +!
  683:     rdrop ;
  684: 
  685: : register ( "name" "type" -- )
  686:     \ define register
  687:     name { d: reg-name }
  688:     name { d: reg-type }
  689:     reg-type define-type >body
  690:     reg-name make-register ;
  691: 
  692: \ stack-states
  693: 
  694: : stack-state ( a-addr u uoffset "name" -- )
  695:     create ss% %allot >r
  696:     r@ ss-offset !
  697:     r@ ss-registers 2!
  698:     rdrop ;
  699: 
  700: 0 0 0 stack-state default-ss
  701: 
  702: \ state
  703: 
  704: : state ( "name" -- )
  705:     \ create a state initialized with default-sss
  706:     create state% %allot state-sss { sss }
  707:     max-stacks 0 ?do
  708: 	default-ss sss i th !
  709:     loop ;
  710: 
  711: : set-ss ( ss stack state -- )
  712:     state-sss swap stack-number @ th ! ;
  713: 
  714: \ offset computation
  715: \ the leftmost (i.e. deepest) item has offset 0
  716: \ the rightmost item has the highest offset
  717: 
  718: : compute-offset { item xt -- }
  719:     \ xt specifies in/out; update stack-in/out and set item-offset
  720:     item item-type @ type-size @
  721:     item item-stack @ xt execute dup @ >r +!
  722:     r> item item-offset ! ;
  723: 
  724: : compute-offset-in ( addr1 addr2 -- )
  725:     ['] stack-in compute-offset ;
  726: 
  727: : compute-offset-out ( addr1 addr2 -- )
  728:     ['] stack-out compute-offset ;
  729: 
  730: : compute-offsets ( -- )
  731:     prim prim-stacks-in  max-stacks cells erase
  732:     prim prim-stacks-out max-stacks cells erase
  733:     prim prim-effect-in  prim prim-effect-in-end  @ ['] compute-offset-in  map-items
  734:     prim prim-effect-out prim prim-effect-out-end @ ['] compute-offset-out map-items
  735:     inst-stream stack-out @ 0= s" # can only be on the input side" ?print-error ;
  736: 
  737: : process-simple ( -- )
  738:     prim prim { W^ key } key cell
  739:     combinations ['] constant insert-wordlist
  740:     declarations compute-offsets
  741:     output @ execute ;
  742: 
  743: : stack-state-items ( stack state -- n )
  744:     state-ss ss-registers 2@ nip ;
  745: 
  746: : unused-stack-items { stack -- n-in n-out }
  747:     \ n-in  are the stack items in state-in  not used    by prim
  748:     \ n-out are the stack items in state-out not written by prim
  749:     stack state-in  stack-state-items stack stack-in  @ - 0 max
  750:     stack state-out stack-state-items stack stack-out @ - 0 max ;
  751: 
  752: : spill-stack { stack -- }
  753:     \ spill regs of state-in that are not used by prim and are not in state-out
  754:     stack state-in stack-offset { offset }
  755:     stack state-in stack-state-items ( items )
  756:     dup stack unused-stack-items - - +do
  757: 	\ loop through the bottom items
  758: 	stack stack-pointer 2@ type
  759: 	i offset - stack normal-stack-access0 ."  = "
  760: 	i stack state-in normal-stack-access1 ." ;" cr
  761:     loop ;
  762: 
  763: : spill-state ( -- )
  764:     ['] spill-stack map-stacks1 ;
  765: 
  766: : fill-stack { stack -- }
  767:     stack state-out stack-offset { offset }
  768:     stack state-out stack-state-items ( items )
  769:     dup stack unused-stack-items - + +do
  770: 	\ loop through the bottom items
  771: 	i stack state-out normal-stack-access1 ."  = "
  772: 	stack stack-pointer 2@ type
  773: 	i offset - stack normal-stack-access0 ." ;" cr
  774:     loop ;
  775: 
  776: : fill-state ( -- )
  777:     \ !! inst-stream for prefetching?
  778:     ['] fill-stack map-stacks1 ;
  779: 
  780: : fetch ( addr -- )
  781:     dup item-type @ type-fetch @ execute ;
  782: 
  783: : fetches ( -- )
  784:     prim prim-effect-in prim prim-effect-in-end @ ['] fetch map-items ;
  785: 
  786: : reg-reg-move ( reg-from reg-to -- )
  787:     2dup = if
  788: 	2drop
  789:     else
  790: 	.reg ."  = " .reg ." ;" cr
  791:     endif ;
  792: 
  793: : stack-bottom-reg { n stack state -- reg }
  794:     stack state stack-state-items n - 1- stack state stack-reg ;
  795: 
  796: : stack-moves { stack -- }
  797:     \ generate moves between registers in state-in/state-out that are
  798:     \ not spilled or consumed/produced by prim.
  799:     \ !! this works only for a simple stack cache, not e.g., for
  800:     \ rotating stack caches, or registers shared between stacks (the
  801:     \ latter would also require a change in interface)
  802:     \ !! maybe place this after NEXT_P1?
  803:     stack unused-stack-items 2dup < if ( n-in n-out )
  804: 	\ move registers from 0..n_in-1 to n_out-n_in..n_out-1
  805: 	over - { diff } ( n-in )
  806: 	-1 swap 1- -do
  807: 	    i stack state-in stack-bottom-reg ( reg-from )
  808: 	    i diff + stack state-out stack-bottom-reg reg-reg-move
  809: 	1 -loop
  810:     else
  811: 	\ move registers from n_in-n_out..n_in-1 to 0..n_out-1
  812: 	swap over - { diff } ( n-out )
  813: 	0 +do
  814: 	    i diff + stack state-in stack-bottom-reg ( reg-from )
  815: 	    i stack state-out stack-bottom-reg reg-reg-move
  816: 	loop
  817:     endif ;
  818: 
  819: : stack-update-transform ( n1 stack -- n2 )
  820:     \ n2 is the number by which the stack pointer should be
  821:     \ incremented to pop n1 items
  822:     stack-access-transform @ dup >r execute
  823:     0 r> execute - ;
  824: 
  825: : stack-pointer-update { stack -- }
  826:     \ and moves
  827:     \ stacks grow downwards
  828:     stack stack-diff ( in-out )
  829:     stack state-in  stack-offset -
  830:     stack state-out stack-offset + ( [in-in_offset]-[out-out_offset] )
  831:     ?dup-if \ this check is not necessary, gcc would do this for us
  832: 	stack inst-stream = if
  833: 	    ." INC_IP(" 0 .r ." );" cr
  834: 	else
  835: 	    stack stack-pointer 2@ type ."  += "
  836: 	    stack stack-update-transform 0 .r ." ;" cr
  837: 	endif
  838:     endif
  839:     stack stack-moves ;
  840: 
  841: : stack-pointer-updates ( -- )
  842:     ['] stack-pointer-update map-stacks ;
  843: 
  844: : store ( item -- )
  845: \ f is true if the item should be stored
  846: \ f is false if the store is probably not necessary
  847:  dup item-type @ type-store @ execute ;
  848: 
  849: : stores ( -- )
  850:     prim prim-effect-out prim prim-effect-out-end @ ['] store map-items ;
  851: 
  852: : print-debug-arg { item -- }
  853:     ." fputs(" quote space item item-name 2@ type ." =" quote ." , vm_out); "
  854:     ." printarg_" item item-type @ print-type-prefix
  855:     ." (" item item-name 2@ type ." );" cr ;
  856:     
  857: : print-debug-args ( -- )
  858:     ." #ifdef VM_DEBUG" cr
  859:     ." if (vm_debug) {" cr
  860:     prim prim-effect-in prim prim-effect-in-end @ ['] print-debug-arg map-items
  861: \    ." fputc('\n', vm_out);" cr
  862:     ." }" cr
  863:     ." #endif" cr ;
  864: 
  865: : print-debug-result { item -- }
  866:     item item-first @ if
  867: 	item print-debug-arg
  868:     endif ;
  869: 
  870: : print-debug-results ( -- )
  871:     cr
  872:     ." #ifdef VM_DEBUG" cr
  873:     ." if (vm_debug) {" cr
  874:     ." fputs(" quote ."  -- " quote ." , vm_out); "
  875:     prim prim-effect-out prim prim-effect-out-end @ ['] print-debug-result map-items
  876:     ." fputc('\n', vm_out);" cr
  877:     ." }" cr
  878:     ." #endif" cr ;
  879: 
  880: : output-super-end ( -- )
  881:     prim prim-c-code 2@ s" SET_IP" search if
  882: 	." SUPER_END;" cr
  883:     endif
  884:     2drop ;
  885: 
  886: : output-nextp2 ( -- )
  887:     ." NEXT_P2;" cr ;
  888: 
  889: variable tail-nextp2 \ xt to execute for printing NEXT_P2 in INST_TAIL
  890: ' output-nextp2 tail-nextp2 !
  891: 
  892: : output-label2 ( -- )
  893:     ." LABEL2(" prim prim-c-name 2@ type ." )" cr
  894:     ." NEXT_P2;" cr ;
  895: 
  896: : output-c-tail1 { xt -- }
  897:     \ the final part of the generated C code, with xt printing LABEL2 or not.
  898:     output-super-end
  899:     print-debug-results
  900:     ." NEXT_P1;" cr
  901:     stores
  902:     fill-state 
  903:     xt execute ;
  904: 
  905: : output-c-tail1-no-stores { xt -- }
  906:     \ the final part of the generated C code for combinations
  907:     output-super-end
  908:     ." NEXT_P1;" cr
  909:     fill-state 
  910:     xt execute ;
  911: 
  912: : output-c-tail ( -- )
  913:     tail-nextp2 @ output-c-tail1 ;
  914: 
  915: : output-c-tail2 ( -- )
  916:     ['] output-label2 output-c-tail1 ;
  917: 
  918: : output-c-tail-no-stores ( -- )
  919:     tail-nextp2 @ output-c-tail1-no-stores ;
  920: 
  921: : output-c-tail2-no-stores ( -- )
  922:     ['] output-label2 output-c-tail1-no-stores ;
  923: 
  924: : type-c-code ( c-addr u xt -- )
  925:     \ like TYPE, but replaces "INST_TAIL;" with tail code produced by xt
  926:     { xt }
  927:     ." {" cr
  928:     ." #line " c-line @ . quote c-filename 2@ type quote cr
  929:     begin ( c-addr1 u1 )
  930: 	2dup s" INST_TAIL;" search
  931:     while ( c-addr1 u1 c-addr3 u3 )
  932: 	2dup 2>r drop nip over - type
  933: 	xt execute
  934: 	2r> 10 /string
  935: 	\ !! resync #line missing
  936:     repeat
  937:     2drop type
  938:     ." #line " out-nls @ 2 + . quote out-filename 2@ type quote cr
  939:     ." }" cr ;
  940: 
  941: : print-entry ( -- )
  942:     ." LABEL(" prim prim-c-name 2@ type ." )" ;
  943:     
  944: : output-c ( -- ) 
  945:     print-entry ."  /* " prim prim-name 2@ type ."  ( " prim prim-stack-string 2@ type ." ) */" cr
  946:     ." /* " prim prim-doc 2@ type ."  */" cr
  947:     ." NAME(" quote prim prim-name 2@ type quote ." )" cr \ debugging
  948:     ." {" cr
  949:     ." DEF_CA" cr
  950:     print-declarations
  951:     ." NEXT_P0;" cr
  952:     spill-state
  953:     fetches
  954:     print-debug-args
  955:     stack-pointer-updates
  956:     prim prim-c-code 2@ ['] output-c-tail type-c-code
  957:     output-c-tail2
  958:     ." }" cr
  959:     cr
  960: ;
  961: 
  962: : disasm-arg { item -- }
  963:     item item-stack @ inst-stream = if
  964: 	." {" cr
  965: 	item print-declaration
  966: 	item fetch
  967: 	item print-debug-arg
  968: 	." }" cr
  969:     endif ;
  970: 
  971: : disasm-args ( -- )
  972:     prim prim-effect-in prim prim-effect-in-end @ ['] disasm-arg map-items ;
  973: 
  974: : output-disasm ( -- )
  975:     \ generate code for disassembling VM instructions
  976:     ." if (VM_IS_INST(*ip, " function-number @ 0 .r ." )) {" cr
  977:     ."   fputs(" quote prim prim-name 2@ type quote ." , vm_out);" cr
  978:     disasm-args
  979:     ."   ip += " inst-stream stack-in @ 1+ 0 .r ." ;" cr
  980:     ."   goto _endif_;" cr
  981:     ." }" cr ;
  982: 
  983: : output-profile ( -- )
  984:     \ generate code for postprocessing the VM block profile stuff
  985:     ." if (VM_IS_INST(*ip, " function-number @ 0 .r ." )) {" cr
  986:     ."   add_inst(b, " quote prim prim-name 2@ type quote ." );" cr
  987:     ."   ip += " inst-stream stack-in @ 1+ 0 .r ." ;" cr
  988:     prim prim-c-code 2@  s" SET_IP"    search nip nip
  989:     prim prim-c-code 2@  s" SUPER_END" search nip nip or if
  990: 	."   return;" cr
  991:     else
  992: 	."   goto _endif_;" cr
  993:     endif
  994:     ." }" cr ;
  995: 
  996: : output-profile-part ( p )
  997:     ."   add_inst(b, " quote
  998:     prim-name 2@ type
  999:     quote ." );" cr ;
 1000:     
 1001: : output-profile-combined ( -- )
 1002:     \ generate code for postprocessing the VM block profile stuff
 1003:     ." if (VM_IS_INST(*ip, " function-number @ 0 .r ." )) {" cr
 1004:     ['] output-profile-part map-combined
 1005:     ."   ip += " inst-stream stack-in @ 1+ 0 .r ." ;" cr
 1006:     combined-prims num-combined @ 1- th @ prim-c-code 2@  s" SET_IP"    search nip nip
 1007:     combined-prims num-combined @ 1- th @ prim-c-code 2@  s" SUPER_END" search nip nip or if
 1008: 	."   return;" cr
 1009:     else
 1010: 	."   goto _endif_;" cr
 1011:     endif
 1012:     ." }" cr ;
 1013: 
 1014: : prim-branch? { prim -- f }
 1015:     \ true if prim is a branch or super-end
 1016:     prim prim-c-code 2@  s" SET_IP" search nip nip 0<> ;
 1017: 
 1018: : output-superend ( -- )
 1019:     \ output flag specifying whether the current word ends a dynamic superinst
 1020:     prim prim-branch?
 1021:     prim prim-c-code 2@  s" SUPER_END" search nip nip 0<> or
 1022:     prim prim-c-code 2@  s" SUPER_CONTINUE" search nip nip 0= and
 1023:     negate 0 .r ." , /* " prim prim-name 2@ type ."  */" cr ;
 1024: 
 1025: : gen-arg-parm { item -- }
 1026:     item item-stack @ inst-stream = if
 1027: 	." , " item item-type @ type-c-name 2@ type space
 1028: 	item item-name 2@ type
 1029:     endif ;
 1030: 
 1031: : gen-args-parm ( -- )
 1032:     prim prim-effect-in prim prim-effect-in-end @ ['] gen-arg-parm map-items ;
 1033: 
 1034: : gen-arg-gen { item -- }
 1035:     item item-stack @ inst-stream = if
 1036: 	."   genarg_" item item-type @ print-type-prefix
 1037:         ." (ctp, " item item-name 2@ type ." );" cr
 1038:     endif ;
 1039: 
 1040: : gen-args-gen ( -- )
 1041:     prim prim-effect-in prim prim-effect-in-end @ ['] gen-arg-gen map-items ;
 1042: 
 1043: : output-gen ( -- )
 1044:     \ generate C code for generating VM instructions
 1045:     ." void gen_" prim prim-c-name 2@ type ." (Inst **ctp" gen-args-parm ." )" cr
 1046:     ." {" cr
 1047:     ."   gen_inst(ctp, vm_prim[" function-number @ 0 .r ." ]);" cr
 1048:     gen-args-gen
 1049:     ." }" cr ;
 1050: 
 1051: : stack-used? { stack -- f }
 1052:     stack stack-in @ stack stack-out @ or 0<> ;
 1053: 
 1054: : output-funclabel ( -- )
 1055:   ." &I_" prim prim-c-name 2@ type ." ," cr ;
 1056: 
 1057: : output-forthname ( -- )
 1058:   '" emit prim prim-name 2@ type '" emit ." ," cr ;
 1059: 
 1060: \  : output-c-func ( -- )
 1061: \  \ used for word libraries
 1062: \      ." Cell * I_" prim prim-c-name 2@ type ." (Cell *SP, Cell **FP)      /* " prim prim-name 2@ type
 1063: \      ."  ( " prim prim-stack-string 2@ type ."  ) */" cr
 1064: \      ." /* " prim prim-doc 2@ type ."  */" cr
 1065: \      ." NAME(" quote prim prim-name 2@ type quote ." )" cr
 1066: \      \ debugging
 1067: \      ." {" cr
 1068: \      print-declarations
 1069: \      \ !! don't know what to do about that
 1070: \      inst-stream  stack-used? IF ." Cell *ip=IP;" cr THEN
 1071: \      data-stack   stack-used? IF ." Cell *sp=SP;" cr THEN
 1072: \      fp-stack     stack-used? IF ." Cell *fp=*FP;" cr THEN
 1073: \      return-stack stack-used? IF ." Cell *rp=*RP;" cr THEN
 1074: \      spill-state
 1075: \      fetches
 1076: \      stack-pointer-updates
 1077: \      fp-stack   stack-used? IF ." *FP=fp;" cr THEN
 1078: \      ." {" cr
 1079: \      ." #line " c-line @ . quote c-filename 2@ type quote cr
 1080: \      prim prim-c-code 2@ type
 1081: \      ." }" cr
 1082: \      stores
 1083: \      fill-state
 1084: \      ." return (sp);" cr
 1085: \      ." }" cr
 1086: \      cr ;
 1087: 
 1088: : output-label ( -- )  
 1089:     ." INST_ADDR(" prim prim-c-name 2@ type ." )," cr ;
 1090: 
 1091: : output-alias ( -- ) 
 1092:     ( primitive-number @ . ." alias " ) ." Primitive " prim prim-name 2@ type cr ;
 1093: 
 1094: : output-c-prim-num ( -- )
 1095:     ." N_" prim prim-c-name 2@ type ." ," cr ;
 1096: 
 1097: : output-forth ( -- )  
 1098:     prim prim-forth-code @ 0=
 1099:     IF    	\ output-alias
 1100: 	\ this is bad for ec: an alias is compiled if tho word does not exist!
 1101: 	\ JAW
 1102:     ELSE  ." : " prim prim-name 2@ type ."   ( "
 1103: 	prim prim-stack-string 2@ type ." )" cr
 1104: 	prim prim-forth-code 2@ type cr
 1105:     THEN ;
 1106: 
 1107: : output-tag-file ( -- )
 1108:     name-filename 2@ last-name-filename 2@ compare if
 1109: 	name-filename 2@ last-name-filename 2!
 1110: 	#ff emit cr
 1111: 	name-filename 2@ type
 1112: 	." ,0" cr
 1113:     endif ;
 1114: 
 1115: : output-tag ( -- )
 1116:     output-tag-file
 1117:     prim prim-name 2@ 1+ type
 1118:     127 emit
 1119:     space prim prim-name 2@ type space
 1120:     1 emit
 1121:     name-line @ 0 .r
 1122:     ." ,0" cr ;
 1123: 
 1124: : output-vi-tag ( -- )
 1125:     name-filename 2@ type #tab emit
 1126:     prim prim-name 2@ type #tab emit
 1127:     ." /^" prim prim-name 2@ type ."  *(/" cr ;
 1128: 
 1129: [IFDEF] documentation
 1130: : register-doc ( -- )
 1131:     prim prim-name 2@ documentation ['] create insert-wordlist
 1132:     prim prim-name 2@ 2,
 1133:     prim prim-stack-string 2@ condition-stack-effect 2,
 1134:     prim prim-wordset 2@ 2,
 1135:     prim prim-c-name 2@ condition-pronounciation 2,
 1136:     prim prim-doc 2@ 2, ;
 1137: [THEN]
 1138: 
 1139: 
 1140: \ combining instructions
 1141: 
 1142: \ The input should look like this:
 1143: 
 1144: \ lit_+ = lit +
 1145: 
 1146: \ The output should look like this:
 1147: 
 1148: \  I_lit_+:
 1149: \  {
 1150: \  DEF_CA
 1151: \  Cell _x_ip0;
 1152: \  Cell _x_sp0;
 1153: \  Cell _x_sp1;
 1154: \  NEXT_P0;
 1155: \  _x_ip0 = (Cell) IPTOS;
 1156: \  _x_sp0 = (Cell) spTOS;
 1157: \  INC_IP(1);
 1158: \  /* sp += 0; */
 1159: \  /* lit ( #w -- w ) */
 1160: \  /*  */
 1161: \  NAME("lit")
 1162: \  {
 1163: \  Cell w;
 1164: \  w = (Cell) _x_ip0;
 1165: \  #ifdef VM_DEBUG
 1166: \  if (vm_debug) {
 1167: \  fputs(" w=", vm_out); printarg_w (w);
 1168: \  fputc('\n', vm_out);
 1169: \  }
 1170: \  #endif
 1171: \  {
 1172: \  #line 136 "./prim"
 1173: \  }
 1174: \  _x_sp1 = (Cell)w;
 1175: \  }
 1176: \  I_plus:	/* + ( n1 n2 -- n ) */
 1177: \  /*  */
 1178: \  NAME("+")
 1179: \  {
 1180: \  DEF_CA
 1181: \  Cell n1;
 1182: \  Cell n2;
 1183: \  Cell n;
 1184: \  NEXT_P0;
 1185: \  n1 = (Cell) _x_sp0;
 1186: \  n2 = (Cell) _x_sp1;
 1187: \  #ifdef VM_DEBUG
 1188: \  if (vm_debug) {
 1189: \  fputs(" n1=", vm_out); printarg_n (n1);
 1190: \  fputs(" n2=", vm_out); printarg_n (n2);
 1191: \  fputc('\n', vm_out);
 1192: \  }
 1193: \  #endif
 1194: \  {
 1195: \  #line 516 "./prim"
 1196: \  n = n1+n2;
 1197: \  }
 1198: \  _x_sp0 = (Cell)n;
 1199: \  }
 1200: \  NEXT_P1;
 1201: \  spTOS = (Cell)_x_sp0;
 1202: \  NEXT_P2;
 1203: 
 1204: : init-combined ( -- )
 1205:     prim to combined
 1206:     0 num-combined !
 1207:     current-depth max-stacks cells erase
 1208:     include-skipped-insts @ current-depth 0 th !
 1209:     max-depth     max-stacks cells erase
 1210:     min-depth     max-stacks cells erase
 1211:     prim prim-effect-in  prim prim-effect-in-end  !
 1212:     prim prim-effect-out prim prim-effect-out-end ! ;
 1213: 
 1214: : max! ( n addr -- )
 1215:     tuck @ max swap ! ;
 1216: 
 1217: : min! ( n addr -- )
 1218:     tuck @ min swap ! ;
 1219: 
 1220: : inst-stream-adjustment ( nstack -- n )
 1221:     \ number of stack items to add for each part
 1222:     0= include-skipped-insts @ and negate ;
 1223: 
 1224: : add-depths { p -- }
 1225:     \ combine stack effect of p with *-depths
 1226:     max-stacks 0 ?do
 1227: 	current-depth i th @
 1228: 	p prim-stacks-in  i th @ + i inst-stream-adjustment +
 1229: 	dup max-depth i th max!
 1230: 	p prim-stacks-out i th @ -
 1231: 	dup min-depth i th min!
 1232: 	current-depth i th !
 1233:     loop ;
 1234: 
 1235: : copy-maxdepths ( n -- )
 1236:     max-depth max-depths rot max-stacks * th max-stacks cells move ;
 1237: 
 1238: : add-prim ( addr u -- )
 1239:     \ add primitive given by "addr u" to combined-prims
 1240:     primitives search-wordlist s" unknown primitive" ?print-error
 1241:     execute { p }
 1242:     p combined-prims num-combined @ th !
 1243:     num-combined @ copy-maxdepths
 1244:     1 num-combined +!
 1245:     p add-depths
 1246:     num-combined @ copy-maxdepths ;
 1247: 
 1248: : compute-effects { q -- }
 1249:     \ compute the stack effects of q from the depths
 1250:     max-stacks 0 ?do
 1251: 	max-depth i th @ dup
 1252: 	q prim-stacks-in i th !
 1253: 	current-depth i th @ -
 1254: 	q prim-stacks-out i th !
 1255:     loop ;
 1256: 
 1257: : make-effect-items { stack# items effect-endp -- }
 1258:     \ effect-endp points to a pointer to the end of the current item-array
 1259:     \ and has to be updated
 1260:     stacks stack# th @ { stack }
 1261:     items 0 +do
 1262: 	effect-endp @ { item }
 1263: 	i 0 <# #s stack stack-pointer 2@ holds [char] _ hold #> save-mem
 1264: 	item item-name 2!
 1265: 	stack item item-stack !
 1266: 	stack stack-type @ item item-type !
 1267: 	i item item-offset !
 1268: 	item item-first on
 1269: 	item% %size effect-endp +!
 1270:     loop ;
 1271: 
 1272: : init-effects { q -- }
 1273:     \ initialize effects field for FETCHES and STORES
 1274:     max-stacks 0 ?do
 1275: 	i q prim-stacks-in  i th @ q prim-effect-in-end  make-effect-items
 1276: 	i q prim-stacks-out i th @ q prim-effect-out-end make-effect-items
 1277:     loop ;
 1278: 
 1279: : compute-stack-max-back-depths ( stack -- )
 1280:     stack-number @ { stack# }
 1281:     current-depth stack# th @ dup
 1282:     dup stack# num-combined @ s-c-max-back-depth !
 1283:     -1 num-combined @ 1- -do ( max-depth current-depth )
 1284: 	combined-prims i th @ { p }
 1285: 	p prim-stacks-out stack# th @ +
 1286: 	dup >r max r>
 1287: 	over stack# i s-c-max-back-depth !
 1288: 	p prim-stacks-in stack# th @ -
 1289: 	stack# inst-stream-adjustment -
 1290:     1 -loop
 1291:     assert( dup stack# inst-stream-adjustment negate = )
 1292:     assert( over max-depth stack# th @ = )
 1293:     2drop ;
 1294: 
 1295: : compute-max-back-depths ( -- )
 1296:     \ compute max-back-depths.
 1297:     \ assumes that current-depths is correct for the end of the combination
 1298:     ['] compute-stack-max-back-depths map-stacks ;
 1299: 
 1300: : process-combined ( -- )
 1301:     combined combined-prims num-combined @ cells
 1302:     combinations ['] constant insert-wordlist
 1303:     combined-prims num-combined @ 1- th ( last-part )
 1304:     @ prim-c-code 2@ prim prim-c-code 2! \ used by output-super-end
 1305:     prim compute-effects
 1306:     prim init-effects
 1307:     compute-max-back-depths
 1308:     output-combined perform ;
 1309: 
 1310: \ reprocessing (typically to generate versions for another cache states)
 1311: 
 1312: variable reprocessed-num 0 reprocessed-num !
 1313: 
 1314: : new-name ( -- c-addr u )
 1315:     reprocessed-num @ 0
 1316:     1 reprocessed-num +!
 1317:     <# #s 'p hold '_ hold #> save-mem ;
 1318: 
 1319: : reprocess-simple ( prim -- )
 1320:     to prim
 1321:     new-name prim prim-c-name 2!
 1322:     output @ execute ;
 1323: 
 1324: : lookup-prim ( c-addr u -- prim )
 1325:     primitives search-wordlist 0= -13 and throw execute ;
 1326: 
 1327: : state-prim1 { in-state out-state prim -- }
 1328:     in-state  to state-in
 1329:     out-state to state-out
 1330:     prim reprocess-simple ;
 1331: 
 1332: : state-prim ( in-state out-state "name" -- )
 1333:     parse-word lookup-prim state-prim1 ;
 1334: 
 1335: \ reprocessing with default states
 1336: 
 1337: \ This is a simple scheme and should be generalized
 1338: \ assumes we only cache one stack and use simple states for that
 1339: 
 1340: 0 value cache-stack  \ stack that we cache
 1341: 2variable cache-states \ states of the cache, starting with the empty state
 1342: 
 1343: : compute-default-state-out ( n-in -- n-out )
 1344:     \ for the current prim
 1345:     cache-stack stack-in @ - 0 max
 1346:     cache-stack stack-out @ + cache-states 2@ nip 1- min ;
 1347: 
 1348: : gen-prim-states ( prim -- )
 1349:     to prim
 1350:     cache-states 2@ swap { states } ( nstates )
 1351:     cache-stack stack-in @ +do
 1352: 	states i th @
 1353: 	states i compute-default-state-out th @
 1354: 	prim state-prim1
 1355:     loop ;
 1356: 
 1357: : prim-states ( "name" -- )
 1358:     parse-word lookup-prim gen-prim-states ;
 1359: 
 1360: \ C output
 1361: 
 1362: : print-item { n stack -- }
 1363:     \ print nth stack item name
 1364:     stack stack-type @ type-c-name 2@ type space
 1365:     ." MAYBE_UNUSED _" stack stack-pointer 2@ type n 0 .r ;
 1366: 
 1367: : print-declarations-combined ( -- )
 1368:     max-stacks 0 ?do
 1369: 	max-depth i th @ min-depth i th @ - 0 +do
 1370: 	    i stacks j th @ print-item ." ;" cr
 1371: 	loop
 1372:     loop ;
 1373: 
 1374: : part-fetches ( -- )
 1375:     fetches ;
 1376: 
 1377: : part-output-c-tail ( -- )
 1378:     print-debug-results
 1379:     stores ;
 1380: 
 1381: : output-combined-tail ( -- )
 1382:     part-output-c-tail
 1383:     in-part @ >r in-part off
 1384:     combined ['] output-c-tail-no-stores prim-context
 1385:     r> in-part ! ;
 1386: 
 1387: : part-stack-pointer-updates ( -- )
 1388:     next-stack-number @ 0 +do
 1389: 	i part-num @ 1+ s-c-max-depth @ dup
 1390: 	i num-combined @ s-c-max-depth @ =    \ final depth
 1391: 	swap i part-num @ s-c-max-depth @ <> \ just reached now
 1392: 	part-num @ 0= \ first part
 1393: 	or and if
 1394: 	    stacks i th @ stack-pointer-update
 1395: 	endif
 1396:     loop ;
 1397: 
 1398: : output-part ( p -- )
 1399:     to prim
 1400:     ." /* " prim prim-name 2@ type ."  ( " prim prim-stack-string 2@ type ." ) */" cr
 1401:     ." NAME(" quote prim prim-name 2@ type quote ." )" cr \ debugging
 1402:     ." {" cr
 1403:     print-declarations
 1404:     part-fetches
 1405:     print-debug-args
 1406:     combined ['] part-stack-pointer-updates prim-context
 1407:     1 part-num +!
 1408:     prim add-depths \ !! right place?
 1409:     prim prim-c-code 2@ ['] output-combined-tail type-c-code
 1410:     part-output-c-tail
 1411:     ." }" cr ;
 1412: 
 1413: : output-parts ( -- )
 1414:     prim >r in-part on
 1415:     current-depth max-stacks cells erase
 1416:     0 part-num !
 1417:     ['] output-part map-combined
 1418:     in-part off
 1419:     r> to prim ;
 1420: 
 1421: : output-c-combined ( -- )
 1422:     print-entry cr
 1423:     \ debugging messages just in parts
 1424:     ." {" cr
 1425:     ." DEF_CA" cr
 1426:     print-declarations-combined
 1427:     ." NEXT_P0;" cr
 1428:     spill-state
 1429:     \ fetches \ now in parts
 1430:     \ print-debug-args
 1431:     \ stack-pointer-updates now in parts
 1432:     output-parts
 1433:     output-c-tail2-no-stores
 1434:     ." }" cr
 1435:     cr ;
 1436: 
 1437: : output-forth-combined ( -- )
 1438: ;
 1439: 
 1440: 
 1441: \ peephole optimization rules
 1442: 
 1443: \ data for a simple peephole optimizer that always tries to combine
 1444: \ the currently compiled instruction with the last one.
 1445: 
 1446: \ in order for this to work as intended, shorter combinations for each
 1447: \ length must be present, and the longer combinations must follow
 1448: \ shorter ones (this restriction may go away in the future).
 1449:   
 1450: : output-peephole ( -- )
 1451:     combined-prims num-combined @ 1- cells combinations search-wordlist
 1452:     s" the prefix for this superinstruction must be defined earlier" ?print-error
 1453:     ." {"
 1454:     execute prim-num @ 5 .r ." ,"
 1455:     combined-prims num-combined @ 1- th @ prim-num @ 5 .r ." ,"
 1456:     combined prim-num @ 5 .r ." }, /* "
 1457:     combined prim-c-name 2@ type ."  */"
 1458:     cr ;
 1459: 
 1460: 
 1461: \ cost and superinstruction data for a sophisticated combiner (e.g.,
 1462: \ shortest path)
 1463: 
 1464: \ This is intended as initializer for a structure like this
 1465: 
 1466: \  struct cost {
 1467: \    int loads;       /* number of stack loads */
 1468: \    int stores;      /* number of stack stores */
 1469: \    int updates;     /* number of stack pointer updates */
 1470: \    int offset;      /* offset into super2 table */
 1471: \    int length;      /* number of components */
 1472: \  };
 1473: 
 1474: \ How do you know which primitive or combined instruction this
 1475: \ structure refers to?  By the order of cost structures, as in most
 1476: \ other cases.
 1477: 
 1478: : super2-length ( -- n )
 1479:     combined if
 1480: 	num-combined @
 1481:     else
 1482: 	1
 1483:     endif ;
 1484: 
 1485: : compute-costs { p -- nloads nstores nupdates }
 1486:     \ compute the number of loads, stores, and stack pointer updates
 1487:     \ of a primitive or combined instruction; does not take TOS
 1488:     \ caching into account
 1489:     0 max-stacks 0 +do
 1490: 	p prim-stacks-in i th @ +
 1491:     loop
 1492:     super2-length 1- - \ don't count instruction fetches of subsumed insts
 1493:     0 max-stacks 0 +do
 1494: 	p prim-stacks-out i th @ +
 1495:     loop
 1496:     0 max-stacks 1 +do \ don't count ip updates, therefore "1 +do"
 1497: 	p prim-stacks-in i th @ p prim-stacks-out i th @ <> -
 1498:     loop ;
 1499: 
 1500: : output-num-part ( p -- )
 1501:     ." N_" prim-c-name 2@ type ." ," ;
 1502:     \ prim-num @ 4 .r ." ," ;
 1503: 
 1504: : output-name-comment ( -- )
 1505:     ."  /* " prim prim-name 2@ type ."  */" ;
 1506: 
 1507: variable offset-super2  0 offset-super2 ! \ offset into the super2 table
 1508: 
 1509: : output-costs-prefix ( -- )
 1510:     ." {" prim compute-costs
 1511:     rot 2 .r ." ," swap 2 .r ." ," 2 .r ." , "
 1512:     prim prim-branch? negate . ." ," ;
 1513: 
 1514: : output-costs-gforth-simple ( -- )
 1515:     output-costs-prefix
 1516:     prim output-num-part
 1517:     1 2 .r ." },"
 1518:     output-name-comment
 1519:     cr ;
 1520: 
 1521: : output-costs-gforth-combined ( -- )
 1522:     output-costs-prefix
 1523:     ." N_START_SUPER+" offset-super2 @ 5 .r ." ,"
 1524:     super2-length dup 2 .r ." }," offset-super2 +!
 1525:     output-name-comment
 1526:     cr ;
 1527: 
 1528: : output-costs ( -- )
 1529:     \ description of superinstructions and simple instructions
 1530:     ." {" prim compute-costs
 1531:     rot 2 .r ." ," swap 2 .r ." ," 2 .r ." ,"
 1532:     offset-super2 @ 5 .r ." ,"
 1533:     super2-length dup 2 .r ." }," offset-super2 +!
 1534:     output-name-comment
 1535:     cr ;
 1536: 
 1537: : output-super2 ( -- )
 1538:     \ table of superinstructions without requirement for existing prefixes
 1539:     combined if
 1540: 	['] output-num-part map-combined 
 1541:     else
 1542: 	prim output-num-part
 1543:     endif
 1544:     output-name-comment
 1545:     cr ;   
 1546: 
 1547: \ the parser
 1548: 
 1549: eof-char max-member \ the whole character set + EOF
 1550: 
 1551: : getinput ( -- n )
 1552:  rawinput @ endrawinput @ =
 1553:  if
 1554:    eof-char
 1555:  else
 1556:    cookedinput @ c@
 1557:  endif ;
 1558: 
 1559: :noname ( n -- )
 1560:  dup bl > if
 1561:   emit space
 1562:  else
 1563:   .
 1564:  endif ;
 1565: print-token !
 1566: 
 1567: : testchar? ( set -- f )
 1568:  getinput member? ;
 1569: ' testchar? test-vector !
 1570: 
 1571: : checksynclines ( -- )
 1572:     \ when input points to a newline, check if the next line is a
 1573:     \ sync line.  If it is, perform the appropriate actions.
 1574:     rawinput @ begin >r
 1575: 	s" #line " r@ over compare if
 1576: 	    rdrop 1 line +! EXIT
 1577: 	endif
 1578: 	0. r> 6 chars + 20 >number drop >r drop line ! r> ( c-addr )
 1579: 	dup c@ bl = if
 1580: 	    char+ dup c@ [char] " <> 0= s" sync line syntax" ?print-error
 1581: 	    char+ dup 100 [char] " scan drop swap 2dup - save-mem filename 2!
 1582: 	    char+
 1583: 	endif
 1584: 	dup c@ nl-char <> 0= s" sync line syntax" ?print-error
 1585: 	skipsynclines @ if
 1586: 	    char+ dup rawinput !
 1587: 	    rawinput @ c@ cookedinput @ c!
 1588: 	endif
 1589:     again ;
 1590: 
 1591: : ?nextchar ( f -- )
 1592:     s" syntax error, wrong char" ?print-error
 1593:     rawinput @ endrawinput @ <> if
 1594: 	rawinput @ c@
 1595: 	1 chars rawinput +!
 1596: 	1 chars cookedinput +!
 1597: 	nl-char = if
 1598: 	    checksynclines
 1599: 	    rawinput @ line-start !
 1600: 	endif
 1601: 	rawinput @ c@
 1602: 	cookedinput @ c!
 1603:     endif ;
 1604: 
 1605: : charclass ( set "name" -- )
 1606:  ['] ?nextchar terminal ;
 1607: 
 1608: : .. ( c1 c2 -- set )
 1609:  ( creates a set that includes the characters c, c1<=c<=c2 )
 1610:  empty copy-set
 1611:  swap 1+ rot do
 1612:   i over add-member
 1613:  loop ;
 1614: 
 1615: : ` ( -- terminal ) ( use: ` c )
 1616:  ( creates anonymous terminal for the character c )
 1617:  char singleton ['] ?nextchar make-terminal ;
 1618: 
 1619: char a char z ..  char A char Z ..  union char _ singleton union  charclass letter
 1620: char 0 char 9 ..					charclass digit
 1621: bl singleton tab-char over add-member			charclass white
 1622: nl-char singleton eof-char over add-member complement	charclass nonl
 1623: nl-char singleton eof-char over add-member
 1624:     char : over add-member complement                   charclass nocolonnl
 1625: nl-char singleton eof-char over add-member
 1626:     char } over add-member complement                   charclass nobracenl
 1627: bl 1+ maxchar .. char \ singleton complement intersection
 1628:                                                         charclass nowhitebq
 1629: bl 1+ maxchar ..                                        charclass nowhite
 1630: char " singleton eof-char over add-member complement	charclass noquote
 1631: nl-char singleton					charclass nl
 1632: eof-char singleton					charclass eof
 1633: nl-char singleton eof-char over add-member		charclass nleof
 1634: 
 1635: (( letter (( letter || digit )) **
 1636: )) <- c-ident ( -- )
 1637: 
 1638: (( ` # ?? (( letter || digit || ` : )) ++
 1639: )) <- stack-ident ( -- )
 1640: 
 1641: (( nowhitebq nowhite ** ))
 1642: <- forth-ident ( -- )
 1643: 
 1644: Variable forth-flag
 1645: Variable c-flag
 1646: 
 1647: (( (( ` e || ` E )) {{ start }} nonl ** 
 1648:    {{ end evaluate }}
 1649: )) <- eval-comment ( ... -- ... )
 1650: 
 1651: (( (( ` f || ` F )) {{ start }} nonl ** 
 1652:    {{ end forth-flag @ IF type cr ELSE 2drop THEN }}
 1653: )) <- forth-comment ( -- )
 1654: 
 1655: (( (( ` c || ` C )) {{ start }} nonl ** 
 1656:    {{ end c-flag @ IF type cr ELSE 2drop THEN }}
 1657: )) <- c-comment ( -- )
 1658: 
 1659: (( ` - nonl ** {{ 
 1660: 	forth-flag @ IF forth-fdiff ." [ELSE]" cr THEN
 1661: 	c-flag @ IF
 1662: 	    function-diff
 1663: 	    ." #else /* " function-number @ 0 .r ."  */" cr THEN }}
 1664: )) <- else-comment
 1665: 
 1666: (( ` + {{ start }} nonl ** {{ end
 1667: 	dup
 1668: 	IF	c-flag @
 1669: 	    IF
 1670: 		function-diff
 1671: 		." #ifdef HAS_" bounds ?DO  I c@ toupper emit  LOOP cr
 1672: 		THEN
 1673: 		forth-flag @
 1674: 		IF  forth-fdiff  ." has? " type ."  [IF]"  cr THEN
 1675: 	ELSE	2drop
 1676: 	    c-flag @      IF
 1677: 		function-diff  ." #endif" cr THEN
 1678: 	    forth-flag @  IF  forth-fdiff  ." [THEN]"  cr THEN
 1679: 	THEN }}
 1680: )) <- if-comment
 1681: 
 1682: (( (( ` g || ` G )) {{ start }} nonl **
 1683:    {{ end
 1684:       forth-flag @ IF  forth-fdiff  ." group " type cr  THEN
 1685:       c-flag @     IF  function-diff
 1686: 	  ." GROUP(" type ." , " function-number @ 0 .r ." )" cr  THEN }}
 1687: )) <- group-comment
 1688: 
 1689: (( (( eval-comment || forth-comment || c-comment || else-comment || if-comment || group-comment )) ?? nonl ** )) <- comment-body
 1690: 
 1691: (( ` \ comment-body nleof )) <- comment ( -- )
 1692: 
 1693: (( {{ start }} stack-ident {{ end 2 pick init-item item% %size + }} white ** )) **
 1694: <- stack-items
 1695: 
 1696: (( {{ prim prim-effect-in }}  stack-items {{ prim prim-effect-in-end ! }}
 1697:    ` - ` - white **
 1698:    {{ prim prim-effect-out }} stack-items {{ prim prim-effect-out-end ! }}
 1699: )) <- stack-effect ( -- )
 1700: 
 1701: (( {{ prim create-prim }}
 1702:    ` ( white ** {{ start }} stack-effect {{ end prim prim-stack-string 2! }} ` ) white **
 1703:    (( {{ start }} forth-ident {{ end prim prim-wordset 2! }} white **
 1704:       (( {{ start }}  c-ident {{ end prim prim-c-name 2! }} )) ??
 1705:    )) ??  nleof
 1706:    (( ` " ` "  {{ start }} (( noquote ++ ` " )) ++ {{ end 1- prim prim-doc 2! }} ` " white ** nleof )) ??
 1707:    {{ skipsynclines off line @ c-line ! filename 2@ c-filename 2! start }}
 1708:    (( (( ` { nonl ** nleof (( (( nobracenl {{ line @ drop }} nonl ** )) ?? nleof )) ** ` } white ** nleof white ** ))
 1709:    || (( nocolonnl nonl **  nleof white ** )) ** ))
 1710:    {{ end prim prim-c-code 2! skipsynclines on }}
 1711:    (( ` :  white ** nleof
 1712:       {{ start }} (( nonl ++  nleof white ** )) ++ {{ end prim prim-forth-code 2! }}
 1713:    )) ?? {{ process-simple }}
 1714:    nleof
 1715: )) <- simple-primitive ( -- )
 1716: 
 1717: (( {{ init-combined }}
 1718:    ` = white ** (( {{ start }} forth-ident {{ end add-prim }} white ** )) ++
 1719:    nleof {{ process-combined }}
 1720: )) <- combined-primitive
 1721: 
 1722: (( {{ make-prim to prim 0 to combined
 1723:       line @ name-line ! filename 2@ name-filename 2!
 1724:       function-number @ prim prim-num !
 1725:       start }} [ifdef] vmgen c-ident [else] forth-ident [then] {{ end
 1726:       2dup prim prim-name 2! prim prim-c-name 2! }}  white **
 1727:    (( ` / white ** {{ start }} c-ident {{ end prim prim-c-name 2! }} white ** )) ??
 1728:    (( simple-primitive || combined-primitive ))
 1729:    {{ 1 function-number +! }}
 1730: )) <- primitive ( -- )
 1731: 
 1732: (( (( comment || primitive || nl white ** )) ** eof ))
 1733: parser primitives2something
 1734: warnings @ [IF]
 1735: .( parser generated ok ) cr
 1736: [THEN]
 1737: 
 1738: 
 1739: \ run with gforth-0.5.0 (slurp-file is missing)
 1740: [IFUNDEF] slurp-file
 1741: : slurp-file ( c-addr1 u1 -- c-addr2 u2 )
 1742:     \ c-addr1 u1 is the filename, c-addr2 u2 is the file's contents
 1743:     r/o bin open-file throw >r
 1744:     r@ file-size throw abort" file too large"
 1745:     dup allocate throw swap
 1746:     2dup r@ read-file throw over <> abort" could not read whole file"
 1747:     r> close-file throw ;
 1748: [THEN]
 1749: 
 1750: : primfilter ( addr u -- )
 1751:     \ process the string at addr u
 1752:     over dup rawinput ! dup line-start ! cookedinput !
 1753:     + endrawinput !
 1754:     checksynclines
 1755:     primitives2something ;    
 1756: 
 1757: : unixify ( c-addr u1 -- c-addr u2 )
 1758:     \ delete crs from the string
 1759:     bounds tuck tuck ?do ( c-addr1 )
 1760: 	i c@ dup #cr <> if
 1761: 	    over c! char+
 1762: 	else
 1763: 	    drop
 1764: 	endif
 1765:     loop
 1766:     over - ;
 1767: 
 1768: : process-file ( addr u xt-simple x-combined -- )
 1769:     output-combined ! output !
 1770:     save-mem 2dup filename 2!
 1771:     slurp-file unixify
 1772:     warnings @ if
 1773: 	." ------------ CUT HERE -------------" cr  endif
 1774:     primfilter ;
 1775: 
 1776: \  : process      ( xt -- )
 1777: \      bl word count rot
 1778: \      process-file ;

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