Home My Page Projects Code Snippets Project Openings SML/NJ
Summary Activity Forums Tracker Lists Tasks Docs Surveys News SCM Files

SCM Repository

[smlnj] Diff of /sml/trunk/HISTORY
ViewVC logotype

Diff of /sml/trunk/HISTORY

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 916, Fri Aug 24 21:13:58 2001 UTC revision 1335, Tue May 27 21:54:23 2003 UTC
# Line 8  Line 8 
8  The form of an entry should be:  The form of an entry should be:
9    
10  Name:  Name:
11  Date:  Date: yyyy/mm/dd
12  Tag: <post-commit CVS tag>  Tag: <post-commit CVS tag>
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume (blume@tti-c.org)
17    Date: 2003/05/27 16:55:00 CDT
18    Tag: blume-20030527-polyeq
19    Description:
20    
21    Tried to eliminated most cases of polymorphic equality.
22    
23    ----------------------------------------------------------------------
24    Name: Matthias Blume (blume@tti-c.org)
25    Date: 2003/05/21 17:45:00 CDT
26    Tag: blume-20030517-complete
27    Description:
28    
29    Two changes:
30    
31      1. Added a flag for controlling whether non-exhaustive bindings will
32         be treated as errors (default is false).
33      2. Cleaned up the *entire* source tree so that CMB.make goes through
34         without a single non-exhaustive match- or bind warning.
35    
36    ----------------------------------------------------------------------
37    Name: Matthias Blume (blume@tti-c.org)
38    Date: 2003/05/17 10:20:00 CDT
39    Tag: blume-20030517-absyn
40    Description:
41    
42    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
43    
44       This mainly affects the quality of error messages.  However, some
45       of the code is now more straightforward than before. (Treatment of
46       the above four constructs in translate.sml is much simpler than
47       the "macro-expansion" that was going on before.  Plus, the mach-
48       compiler no longer gets invoked just to be able to compile an
49       if-expression.)
50    
51    2. The ErrorMsg.Error exception is now caught and absorbed by the
52       interactive loop.
53    
54    ----------------------------------------------------------------------
55    Name: Allen Leung
56    Date: 2003/05/16 13:05:00 CDT
57    Tag: leunga-20030516-cygwin-runtime
58    Description:
59    
60        Ported the runtime system to cygwin, which uses the unix
61    x86-unix bin files.  Missing/buggy features:
62    
63       o getnetbyname, getnetbyaddr: these functions seem to be missing in
64         the Cygwin library.
65       o Ctrl-C handling may be flaky.
66       o Windows system calls and Windows I/O are not supported.
67    
68       A new set of binfiles is located at:
69    
70         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
71    
72       This is only needed for bootstrapping the cygwin version of smlnj.
73       Other x86 versions can use the existing binfiles.
74    
75    ----------------------------------------------------------------------
76    Name: Matthias Blume
77    Date: 2003/04/08 15:42:00 CDT
78    Tag: blume-20030408-listpair
79    Description:
80    
81    1. Added a target 'mlrisc' to installer.
82    
83    2. Added missing elements to structure ListPair.
84    
85    ----------------------------------------------------------------------
86    Name: Allen Leung
87    Date: 2003/01/07 10:40:00 EST
88    Tag: leunga-20030107-int-rem
89    Description:
90    
91    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
92    The arguments to the SUBL instruction were swapped.
93    
94    ----------------------------------------------------------------------
95    Name: Matthias Blume
96    Date: 2002/12/12 16:25:00 EST
97    Tag: blume-20021212-risc-ra
98    Description:
99    
100    Fixed a serious bug in the rewrite code for FP spilling/reloading that
101    sent the RA into an infinite loop when floating point registers get
102    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
103    110.37 and 110.38.)
104    There was another set of potential problems related to the handling of
105    MLRISC annotations (but those did not yet cause real problems, apparently).
106    
107    ----------------------------------------------------------------------
108    Name: Matthias Blume
109    Date: 2002/12/06 22:40:00 EST
110    Tag: blume-20021206-cm-fileid
111    Description:
112    
113    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
114    This fixes the problem of CM getting confused by files that suddenly
115    change their identity (e.g., by getting unlinked and recreated by some
116    text editor such as vi).  There might be a better/cheaper/cleaner way
117    of doing this, but for now this will have to do.
118    
119    ----------------------------------------------------------------------
120    Name: Matthias Blume
121    Date: 2002/10/28 09:50:00 EST
122    Tag: blume-20021028-typecheck
123    Description:
124    
125    Exported structure Typecheck from $smlnj/viscomp/core.cm.
126    
127    ----------------------------------------------------------------------
128    Name: Matthias Blume
129    Date: 2002/10/17 09:10:00 EDT
130    Tag: Release_110_42
131    Description:
132    
133    In good old tradition, there has been a slight hiccup so that we have
134    to patch 110.42 after the fact.   The old release tag has been replaced
135    (see below).
136    
137    The change solves a problem with two competing approaches the
138    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
139    each other's way.
140    
141    This change only affects the runtime system code and the installer script.
142    (No new bootfiles.)
143    
144    ----------------------------------------------------------------------
145    Name: Matthias Blume
146    Date: 2002/10/16 12:00:00 EDT
147    Tag: Release_110_42_removed
148    Description:
149    
150    New working release.  New bootfiles.
151    
152    ----------------------------------------------------------------------
153    Name: Matthias Blume
154    Date: 2002/10/10 13:10:00 EDT
155    Tag: blume-20021010-ppc-divs
156    Description:
157    
158    The mltree operator DIVS must be implemented with an overflow check on
159    the PPC because the hardware indicates divide-by-zero using "overflow" as
160    well.
161    
162    ----------------------------------------------------------------------
163    Name: Matthias Blume
164    Date: 2002/07/23 11:45:00
165    Tag: blume-20020723-smlnj-home
166    Description:
167    
168    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
169    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
170    CM_PATHCONFIG is also set), the path configuration file is assumed
171    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
172    the entire tree to some other place and everything will "just work".
173    
174    (Companion commands such as ml-build and ml-makedepend also sense this
175    variable.)
176    
177    ----------------------------------------------------------------------
178    Name: Matthias Blume
179    Date: 2002/07/12 21:19:00 EDT
180    Tag: blume-20020712-liveness
181    Description:
182    
183    Exported two useful "step" functions from liveness module (MLRISC).
184    
185    ----------------------------------------------------------------------
186    Name: Matthias Blume
187    Date: 2002/07/05 16:00 EDT
188    Tag: Release_110_41
189    Description:
190    
191    New working release.  New bootfiles.
192    
193    ----------------------------------------------------------------------
194    Name: Matthias Blume
195    Date: 2002/07/05 10:25:00 EDT
196    Tag: blume-20020705-btimp
197    Description:
198    
199    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
200    other clients can set up backtracing support.
201    
202    ----------------------------------------------------------------------
203    Name: Matthias Blume
204    Date: 2002/06/25 17:23:00 EDT
205    Tag: blume-20020625-fpmax
206    Description:
207    
208    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
209    case.  (The sense of the isNaN test was reversed -- which made min and
210    max always return their first argument.)
211    
212    ----------------------------------------------------------------------
213    Name: Matthias Blume
214    Date: 2002/06/11
215    Tag: blume-20020611-unixpath
216    Description:
217    
218    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
219    
220    ----------------------------------------------------------------------
221    Name: Matthias Blume
222    Date: 2002/06/10 16:35:00 EDT
223    Tag: blume-20020610-ieeereal
224    Description:
225    
226    I back-ported my implementation of IEEEReal.fromString from the
227    idlbasis-devel branch so that we can test it.
228    
229    Another small change is that ppDec tries to give more information
230    than just "<sig>" in the case of functors.  However, this code is
231    broken in some mysterious way if the functor's body's signature
232    has not been declared by ascription but gets inferred from the
233    implementation.  This needs fixing...
234    
235    ----------------------------------------------------------------------
236    Name: Matthias Blume
237    Date: 2002/05/31
238    Tag: blume-20020531-btrace-mode
239    Description:
240    
241    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
242    the wayside when I switched over to using Controls everywhere.)
243    
244    ----------------------------------------------------------------------
245    Name: Lal George
246    Date: 2002/05/23 12:21:40 EDT
247    Tag: george-20020523-visual-labels
248    Description:
249    
250    Labels are now displayed in the graphical output to make
251    the fall-through and target blocks obvious.
252    
253    ----------------------------------------------------------------------
254    Name: Matthias Blume
255    Date: 2002/05/22 11:03:00 EDT
256    Tag: blume-20020522-shrink
257    Description:
258    
259    John tweaked yesterday's fix for 1131 to handle an out-of-memory
260    situation that comes up when allocating huge arrays.
261    
262    ----------------------------------------------------------------------
263    Name: Matthias Blume
264    Date: 2002/05/21 16:00:00 EDT
265    Tag: Release_110_40
266    Description:
267    
268    New working release (110.40).  New bootfiles.
269    
270    [Also: John Reppy fixed GC bug 1131.]
271    
272    ----------------------------------------------------------------------
273    Name: Matthias Blume
274    Date: 2002/05/21 12:35:00 EDT
275    Tag: blume-20020521-cmdoc
276    Description:
277    
278    CM documentation update.
279    
280    ----------------------------------------------------------------------
281    Name: Matthias Blume
282    Date: 2002/05/21 10:55:00 EDT
283    Tag: blume-20020521-misc
284    Description:
285    
286    - John tweaked runtime to be silent on heap export (except when
287      GC messages are on).
288    - I added a few more things (cross-compiling versions of CMB) to
289      config/preloads (as suggestions).
290    
291    ----------------------------------------------------------------------
292    Name: Matthias Blume
293    Date: 2002/05/20 22:25:00 EDT
294    Tag: blume-20020520-controls
295    Description:
296    
297    - Added ControlUtil structure to control-lib.cm.
298    - Use it throughout.
299    - Used Controls facility to define MLRISC controls (as opposed to
300      registering MLRISC control ref cells with Controls after the fact)
301    - Fixed messed-up controls priorities.
302    
303    * Removed again all the stuff from config/preloads that one wouldn't
304      be able to preload at the time the initial heap image is built.
305      (Many libraries, e.g., CML, do not exist yet at this time.  The
306       only libraries that can be preloaded via config/preloads are those
307       that come bundled with the bootfiles.)
308    
309    ----------------------------------------------------------------------
310    Name: Matthias Blume
311    Date: 2002/05/20 10:59:00 EDT
312    Tag: blume-20020520-preloads
313    Description:
314    
315    Added a lot of commented-out suggestions for things to be included
316    in config/preloads.
317    
318    ----------------------------------------------------------------------
319    Name: Allen Leung
320    Date: 2002/05/18 14:20:00 EDT
321    Tag: leunga-20020518-mdl
322    Description:
323    
324    o Made the mdl tool stuff compile and run again.
325    o I've disabled all the stuff that depends on RTL specifications; they
326      are all badly broken anyway.
327    
328    ----------------------------------------------------------------------
329    Name: Matthias Blume
330    Date: 2002/05/17 16:49:00 EDT
331    Tag: blume-20020517-controls
332    Description:
333    
334    0. John Reppy made several modifications to the SML/NJ library.
335       In particular, there is a shiny new controls-lib.cm.
336    
337    1. Pushed new controls interface through compiler so that everything
338       compiles again.
339    
340    2. Added FormatComb and FORMAT_COMB to the CML version of the
341       SML/NJ library (so that CML compiles again).
342    
343    3. Modified init scripts because XXX_DEFAULT environment variables
344       are no longer with us.  (Boot-time initialization is now done
345       using the same environment variables that are also used for
346       startup-time initialization of controls.)
347    
348    ----------------------------------------------------------------------
349    Name: Lal George
350    Date: 2002/05/15 09:20:10 EDT
351    Tag: george-20020515-pseudo-op-decls
352    Description:
353    
354            All pseudo-ops emitted before the first segment declaration
355    such as TEXT, DATA, and BSS directives are assumed to be global
356    declarations and are emitted first in the assembly file. This is
357    useful in a number of situations where one has pseudo-ops that are not
358    specific to any segment, and also works around the constraint that one
359    cannot have client pseudo-ops in the TEXT segment.
360    
361            Because no segment is associated with these declarations it is
362    an error to allocate any space or objects before the first segment
363    directive and an exception will be raised. However, we cannot make
364    this check for client pseudo-ops.
365    
366            These top level declarations are a field in the CFG graph_info.
367    In theory you can continue to add to this field after the CFG has been
368    built -- provided you know what you are doing;-)
369    
370    ----------------------------------------------------------------------
371    Name: Matthias Blume
372    Date: 2002/05/13 16:40:00 EDT
373    Tag: blume-20020513-pp-etc
374    Description:
375    
376    A few minor bugfixes:
377    
378      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
379        (Bogus printouts for redefined bindings still occur.  Compiler
380        bug should no longer occur now.  We need to redo the prettyprinter
381        from scratch.)
382    
383      - CM pathname printer now also adds escape sequences for ( and )
384    
385      - commend and docu fixes for ml-nlffi
386    
387    ----------------------------------------------------------------------
388    Name: Matthias Blume
389    Date: 2002/05/10 16:40:00 EDT
390    Tag: blume-20020510-erg-textio
391    Description:
392    
393    Applied the following bugfix provided by Emden Gansner:
394    
395        Output is corrupted when outputSubstr is used rather than output.
396    
397        The problem occurs when a substring
398    
399            ss = (s, dataStart, dataLen)
400    
401        where dataStart > 0, fills a stream buffer with avail bytes left.
402        avail bytes of s, starting at index dataStart, are copied into the
403        buffer, the buffer is flushed, and then the remaining dataLen-avail
404        bytes of ss are copied into the beginning of the buffer. Instead of
405        starting this copy at index dataStart+avail in s, the current code
406        starts the copy at index avail.
407    
408        Fix:
409        In text-io-fn.sml, change line 695 from
410             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
411        to
412             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
413    
414    ----------------------------------------------------------------------
415    Name: Matthias Blume
416    Date: 2002/04/12 13:55:00 EDT
417    Tag: blume-20020412-assyntax
418    Description:
419    
420    1. Grabbed newer assyntax.h from the XFree86 project.
421    2. Fiddled with how to compile X86.prim.asm without warnings.
422    3. (Very) Minor cleanup in CM.
423    
424    ----------------------------------------------------------------------
425    Name: Matthias Blume
426    Date: 2002/04/01 (no joke!) 17:07:00 EST
427    Tag: blume-20020401-x86div
428    Description:
429    
430    Added full support for div/mod/rem/quot on the x86, using the machine
431    instruction's two results (without clumsily recomputing the remainder)
432    directly where appropriate.
433    
434    Some more extensive power-of-two support was added to the x86 instruction
435    selector (avoiding expensive divs, mods, and muls where they can be
436    replaced with cheaper shifts and masks).  However, this sort of thing
437    ought to be done earlier, e.g., within the CPS optimizer so that
438    all architectures benefit from it.
439    
440    The compiler compiles to a fixed point, but changes might be somewhat
441    fragile nevertheless.  Please, report any strange things that you might
442    see wrt. div/mod/quot/rem...
443    
444    ----------------------------------------------------------------------
445    Name: Matthias Blume
446    Date: 2002/03/29 17:22:00
447    Tag: blume-20020329-div
448    Description:
449    
450    Fixed my broken div/mod logic.  Unfortunately, this means that the
451    inline code for div/mod now has one more comparison than before.
452    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
453    The problem was with quotient = 0, because that alone does not tell
454    us which way the rounding went.  One then has to look at whether
455    remainder and divisor have the same sign...  :(
456    
457    Anyway, I replaced the bootfiles with fresh ones...
458    
459    ----------------------------------------------------------------------
460    Name: Matthias Blume
461    Date: 2002/03/29 14:10:00 EST
462    Tag: blume-20020329-inlprims
463    Description:
464    
465    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
466    
467    Primops have changed. This means that the bin/boot-file formats have
468    changed as well.
469    
470    To make sure that there is no confusion, I made a new version.
471    
472    
473    CHANGES:
474    
475    * removed REMT from mltree (remainder should never overflow).
476    
477    * added primops to deal with divisions of all flavors to the frontend
478    
479    * handled these primops all the way through so they map to their respective
480      MLRISC support
481    
482    * used these primops in the implementation of Int, Int32, Word, Word32
483    
484    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
485    
486    * parameterized INLMIN, INLMAX, and INLABS by a numkind
487    
488    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
489      floating point
490    
491    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
492      and Real (but Real.abs maps to a separate floating-point-only primop)
493    
494    
495    TODO items:
496    
497    * Hacked Alpha32 instruction selection, disabling the selection of REMx
498      instructions because the machine instruction encoder cannot handle
499      them.  (Hppa, PPC, and Sparc instruction selection did not handle
500      REM in the first place, and REM is supported by the x86 machine coder.)
501    
502    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
503      selection phase.  (The two can be streamlined because the hardware
504      delivers both quotient and remainder at the same time anyway.)
505    
506    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
507      (Currently the behavior is inconsistent both across architectures and
508      wrt. the draft Basis spec.)
509    
510    * Word8 should eventually be handled natively, too.
511    
512    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
513      as if there currently is no execution path that could trigger it in
514      SML/NJ.  (The assumptions underlying functions arith and promotable do not
515      hold for things like multiplication and division.)
516    
517    ----------------------------------------------------------------------
518    Name: Matthias Blume
519    Date: 2002/03/27 16:27:00 EST
520    Tag: blume-20020327-mlrisc-divisions
521    Description:
522    
523    Added support for all four division operations (ML's div, mod, quot,
524    and rem) to MLRISC.  In the course of doing so, I also rationalized
525    the naming (no more annoying switch-around of DIV and QUOT), by
526    parameterizing the operation by div_rounding_mode (which can be either
527    DIV_TO_ZERO or DIV_TO_NEGINF).
528    
529    The generic MLTreeGen functor takes care of compiling all four
530    operations down to only round-to-zero div.
531    
532    Missing pieces:
533    
534      * Doing something smarter than relying on MLTreeGen on architectures
535        like, e.g., the x86 where hardware division delivers both quotient and
536        remainder at the same time.  With this, the implementation of the
537        round-to-neginf operations could be further streamlined.
538    
539      * Remove inlining support for div/mod/rem from the frontend and replace it
540        with primops that get carried through to the backend.  Do this for all
541        int and word types.
542    
543    ----------------------------------------------------------------------
544    Name: Matthias Blume
545    Date: 2002/03/25 17:25:00 EST
546    Tag: blume-20020325-divmod
547    Description:
548    
549    I improved (hopefully without breaking them) the implementation of Int.div,
550    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
551    advantage of the following observations:
552    
553      Let  q = x quot y      r = x rem y
554           d = x div  y      m = x mod y
555    
556    where "quot" is the round-to-zero version of integer division that
557    hardware usually provides.  Then we have:
558    
559         r = x - q * y        where neither the * nor the - will overflow
560         d = if q >= 0 orelse x = q * y then q else q - 1
561                              where neither the * nor the - will overflow
562         m = if q >= 0 orelse r = 0 then r else r + y
563                              where the + will not overflow
564    
565    This results in substantial simplification of the generated code.
566    The following table shows the number of CFG nodes and edges generated
567    for
568            fun f (x, y) = x OPER y
569            (* with OPER \in div, mod, quot, rem *)
570    
571    
572        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
573        --------------------------------------------------------
574         div |         24 |         39 |         12 |         16
575         mod |         41 |         71 |         12 |         16
576        quot |          8 |         10 |          8 |         10
577         rem |         10 |         14 |          8 |         10
578    
579    
580    ----------------------------------------------------------------------
581    Name: Matthias Blume
582    Date: 2002/03/25 22:06:00 EST
583    Tag: blume-20020325-cprotobug
584    Description:
585    
586    Fixed a bug in cproto (c prototype decoder).
587    
588    ----------------------------------------------------------------------
589    Name: Matthias Blume
590    Date: 2002/03/25 16:00:00 EST
591    Tag: blume-20020325-raw-primops
592    Description:
593    
594    I did some cleanup to Allen's new primop code and
595    replaced yesterday's bootfiles with new ones.
596    (But they are stored in the same place.)
597    
598    ----------------------------------------------------------------------
599    Name: Matthias Blume
600    Date: 2002/03/24 22:40:00 EST
601    Tag: blume-20020324-bootfiles
602    Description:
603    
604    Made the bootfiles that Allen asked for.
605    
606    ----------------------------------------------------------------------
607    Name: Allen Leung
608    Date: 2002/03/23 15:50:00 EST
609    Tag: leunga-20020323-flint-cps-rcc-primops
610    Description:
611    
612      1. Changes to FLINT primops:
613    
614        (* make a call to a C-function;
615         * The primop carries C function prototype information and specifies
616         * which of its (ML-) arguments are floating point. C prototype
617         * information is for use by the backend, ML information is for
618         * use by the CPS converter. *)
619      | RAW_CCALL of { c_proto: CTypes.c_proto,
620                       ml_args: ccall_type list,
621                       ml_res_opt: ccall_type option,
622                       reentrant : bool
623                     } option
624       (* Allocate uninitialized storage on the heap.
625        * The record is meant to hold short-lived C objects, i.e., they
626        * are not ML pointers.  With the tag, the representation is
627        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
628        *)
629      | RAW_RECORD of {tag:bool,sz:int}
630      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
631    
632      2.  These CPS primops are now overloaded:
633    
634           rawload of {kind:numkind}
635           rawstore of {kind:numkind}
636    
637          The one argument form is:
638    
639             rawload {kind} address
640    
641          The two argument form is:
642    
643             rawload {kind} [ml object, byte-offset]
644    
645      3. RAW_CCALL/RCC now takes two extra arguments:
646    
647         a. The first is whether the C call is reentrant, i.e., whether
648            ML state should be saved and restored.
649         b. The second argument is a string argument specifying the name of
650            library and the C function.
651    
652         These things are currently not handled in the code generator, yet.
653    
654      4. In CProto,
655    
656         An encoding type of "bool" means "ml object" and is mapped into
657         C prototype of PTR.  Note that "bool" is different than "string",
658         even though "string" is also mapped into PTR, because "bool"
659         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
660    
661      5. Pickler/unpicker
662    
663         Changed to handle RAW_RECORD and newest RAW_CCALL
664    
665      6. MLRiscGen,
666    
667         1. Changed to handle the new rawload/rawstore/rawrecord operators.
668         2. Code for handling C Calls has been moved to a new module CPSCCalls,
669            in the file CodeGen/cpscompile/cps-c-calls.sml
670    
671      7. Added the conditional move operator
672    
673             condmove of branch
674    
675         to cps.  Generation of this is still buggy so it is currently
676         disabled.
677    
678    ----------------------------------------------------------------------
679    Name: Lal George
680    Date: 2002/03/22 14:18:25 EST
681    Tag: george-20020322-cps-branch-prob
682    Description:
683    
684    Implemented the Ball-Larus branch prediction-heuristics, and
685    incorporated graphical viewers for control flow graphs.
686    
687    Ball-Larus Heuristics:
688    ---------------------
689    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
690    
691    By design it uses the Dempster-Shafer theory for combining
692    probabilities.  For example, in the function:
693    
694        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
695    
696    the ball-larus heuristics predicts that the n=0 is unlikely
697    (OH-heuristic), and the 'then' branch is unlikely because of the
698    RH-heuristic -- giving the 'then' branch an even lower combined
699    probability using the Dempster-Shafer theory.
700    
701    Finally, John Reppy's loop analysis in MLRISC, further lowers the
702    probability of the 'then' branch because of the loop in the else
703    branch.
704    
705    
706    Graphical Viewing:
707    ------------------
708    I merely plugged in Allen's graphical viewers into the compiler. The
709    additional code is not much. At the top level, saying:
710    
711            Control.MLRISC.getFlag "cfg-graphical-view" := true;
712    
713    will display the graphical view of the control flow graph just before
714    back-patching.  daVinci must be in your path for this to work. If
715    daVinci is not available, then the default viewer can be changed
716    using:
717    
718            Control.MLRISC.getString "viewer"
719    
720    which can be set to "dot" or "vcg" for the corresponding viewers. Of
721    course, these viewers must be in your path.
722    
723    The above will display the compilation unit at the level of clusters,
724    many of which are small, boring, and un-interesting. Also setting:
725    
726            Control.MLRISC.getInt "cfg-graphical-view_size"
727    
728    will display clusters that are larger than the value set by the above.
729    
730    
731    ----------------------------------------------------------------------
732    Name: Matthias Blume
733    Date: 2002/03/21 22:20:00 EST
734    Tag: blume-20020321-kmp-bugfix
735    Description:
736    
737    Changed the interface to the KMP routine in PreString and fixed
738    a minor bug in one place where it was used.
739    
740    ----------------------------------------------------------------------
741    Name: Allen Leung
742    Date: 2002/03/21 20:30:00 EST
743    Tag: leunga-20020321-cfg
744    Description:
745    
746      Fixed a potential problem in cfg edge splitting.
747    
748    ----------------------------------------------------------------------
749    Name: Allen Leung
750    Date: 2002/03/21 17:15:00 EST
751    Tag: leunga-20020321-x86-fp-cfg
752    Description:
753    
754      1. Recoded the buggy parts of x86-fp.
755    
756         a. All the block reordering code has been removed.
757            We now depend on the block placement phases to do this work.
758    
759         b. Critical edge splitting code has been simplified and moved into the
760            CFG modules, as where they belong.
761    
762         Both of these were quite buggy and complex.  The code is now much, much
763         simpler.
764    
765      2. X86 backend.
766    
767         a. Added instructions for 64-bit support.  Instruction selection for
768            64-bit has not been committed, however, since that
769            requires changes to MLTREE which haven't been approved by
770            Lal and John.
771    
772         b. Added support for FUCOMI and FUCOMIP when generating code for
773            PentiumPro and above.  We only generate these instructions in
774            the fast-fp mode.
775    
776         c. Added cases for JP and JNP in X86FreqProps.
777    
778      3. CFG
779    
780         CFG now has a bunch of methods for edge splitting and merging.
781    
782      4. Machine description.
783    
784         John's simplification of MLTREE_BASIS.fcond broke a few machine
785         description things:
786    
787         rtl-build.{sig,sml} and hppa.mdl fixed.
788    
789         NOTE: the machine description stuff in the repository is still broken.
790               Again, I can't put my fixes in because that involves
791               changes to MLTREE.
792    
793    ----------------------------------------------------------------------
794    Name: Matthias Blume
795    Date: 2002/03/20 15:55:00 EST
796    Tag: blume-20020320-kmp
797    Description:
798    
799    Implemented Knuth-Morris-Pratt string matching in PreString and used
800    it for String.isSubstring, Substring.isSubstring, and
801    Substring.position.
802    
803    (Might need some stress-testing.  Simple examples worked fine.)
804    
805    ----------------------------------------------------------------------
806    Name: Matthias Blume
807    Date: 2002/03/19 16:37:00 EST
808    Tag: blume-20020319-witnesses
809    Description:
810    
811    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
812    
813    This implements a generic mechanism for changing constness qualifiers
814    anywhere within big C types without resorting to outright "casts".
815    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
816    modify the constness at the outermost level.)
817    The implementation of "convert" is based on the idea of "witness"
818    values -- values that are not used by the operation but whose types
819    "testify" to their applicability.  On the implementation side, "convert"
820    is simply a projection (returning its second curried argument).  With
821    cross-module inlining, it should not result in any machine code being
822    generated.
823    
824    ----------------------------------------------------------------------
825    Name: Matthias Blume
826    Date: 2002/03/15 16:40:00 EST
827    Tag: blume-20020315-basis
828    Description:
829    
830    Provided (preliminary?) implementations for
831    
832      {String,Substring}.{concatWith,isSuffix,isSubstring}
833    
834    and
835    
836      Substring.full
837    
838    Those are in the Basis spec but they were missing in SML/NJ.
839    
840    ----------------------------------------------------------------------
841    Name: Matthias Blume
842    Date: 2002/03/14 21:30:00 EST
843    Tag: blume-20020314-controls
844    Description:
845    
846    Controls:
847    ---------
848    
849    1. Factored out the recently-added Controls : CONTROLS stuff and put
850       it into its own library $/controls-lib.cm.  The source tree for
851       this is under src/smlnj-lib/Controls.
852    
853    2. Changed the names of types and functions in this interface, so they
854       make a bit more "sense":
855    
856          module -> registry
857          'a registry -> 'a group
858    
859    3. The interface now deals in ref cells only.  The getter/setter interface
860       is (mostly) gone.
861    
862    4. Added a function that lets one register an already-existing ref cell.
863    
864    5. Made the corresponding modifications to the rest of the code so that
865       everything compiles again.
866    
867    6. Changed the implementation of Controls.MLRISC back to something closer
868       to the original.  In particular, this module (and therefore MLRISC)
869       does not depend on Controls.  There now is some link-time code in
870       int-sys.sml that registers the MLRISC controls with the Controls
871       module.
872    
873    CM:
874    ---
875    
876      * One can now specify the lambda-split aggressiveness in init.cmi.
877    
878    ----------------------------------------------------------------------
879    Name: Allen Leung
880    Date: 2002/03/13 17:30:00 EST
881    Tag: leunga-20020313-x86-fp-unary
882    Description:
883    
884    Bug fix for:
885    
886    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
887    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
888    > - fun f(x,(y,z)) = Real.~ y;
889    > [autoloading]
890    > [autoloading done]
891    >       fchsl   (%eax), 184(%esp)
892    > Error: MLRisc bug: X86MCEmitter.emitInstr
893    >
894    > uncaught exception Error
895    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
896    
897    The problem was that the code generator did not generate any fp registers
898    in this case, and the ra didn't know that it needed to run the X86FP phase to
899    translate the pseudo fp instruction.   This only happened with unary fp
900    operators in certain situations.
901    
902    ----------------------------------------------------------------------
903    Name: Matthias Blume
904    Date: 2002/03/13 14:00:00 EST
905    Tag: blume-20020313-overload-etc
906    Description:
907    
908    1. Added _overload as a synonym for overload for backward compatibility.
909       (Control.overloadKW must be true for either version to be accepted.)
910    
911    2. Fixed bug in install script that caused more things to be installed
912       than what was requested in config/targets.
913    
914    3. Made CM aware of the (_)overload construct so that autoloading
915       works.
916    
917    ----------------------------------------------------------------------
918    Name: Matthias Blume
919    Date: 2002/03/12 22:03:00 EST
920    Tag: blume-20020312-url
921    Description:
922    
923    Forgot to update BOOT and srcarchiveurl.
924    
925    ----------------------------------------------------------------------
926    Name: Matthias Blume
927    Date: 2002/03/12 17:30:00 EST
928    Tag: blume-20020312-version110392
929    Description:
930    
931    Yet another version number bump (because of small changes to the
932    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
933    
934    Changes:
935    
936      The new pid generation scheme described a few weeks ago was overly
937      complicated.  I implemented a new mechanism that is simpler and
938      provides a bit more "stability":  Once CM has seen a compilation
939      unit, it keeps its identity constant (as long as you do not delete
940      those crucial CM/GUID/* files).  This means that when you change
941      an interface, compile, then go back to the old interface, and
942      compile again, you arrive at the original pid.
943    
944      There now also is a mechanism that instructs CM to use the plain
945      environment hash as a module's pid (effectively making its GUID
946      the empty string).  For this, "noguid" must be specified as an
947      option to the .sml file in question within its .cm file.
948      This is most useful for code that is being generated by tools such
949      as ml-nlffigen (because during development programmers tend to
950      erase the tool's entire output directory tree including CM's cached
951      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
952      revert to the old, broken behavior of SML/NJ, but in specific cases
953      where there is no danger of interface confusion, its use is ok
954      (I think).
955    
956      ml-nlffigen by default generates "noguid" annotations.  They can be
957      turned off by specifying -guid in its command line.
958    
959    ----------------------------------------------------------------------
960    Name: Lal George
961    Date: 2002/03/12 12 14:42:36 EST
962    Tag: george-20020312-frequency-computation
963    Description:
964    
965    Integrated jump chaining and static block frequency into the
966    compiler. More details and numbers later.
967    
968    ----------------------------------------------------------------------
969    Name: Lal George
970    Date: 2002/03/11 11 22:38:53 EST
971    Tag: george-20020311-jump-chain-elim
972    Description:
973    
974    Tested the jump chain elimination on all architectures (except the
975    hppa).  This is on by default right now and is profitable for the
976    alpha and x86, however, it may not be profitable for the sparc and ppc
977    when compiling the compiler.
978    
979    The gc test will typically jump to a label at the end of the cluster,
980    where there is another jump to an external cluster containing the actual
981    code to invoke gc. This is to allow factoring of common gc invocation
982    sequences. That is to say, we generate:
983    
984            f:
985               testgc
986               ja   L1      % jump if above to L1
987    
988            L1:
989               jmp L2
990    
991    
992    After jump chain elimination the 'ja L1' instructions is converted to
993    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
994    up being implemented in their long form (if L2 is far away) using:
995    
996            jbe     L3      % jump if below or equal to L3
997            jmp     L2
998         L3:
999            ...
1000    
1001    
1002    For large compilation units L2  may be far away.
1003    
1004    
1005    ----------------------------------------------------------------------
1006    Name: Matthias Blume
1007    Date: 2002/03/11 13:30:00 EST
1008    Tag: blume-20020311-mltreeeval
1009    Description:
1010    
1011    A functor parameter was missing.
1012    
1013    ----------------------------------------------------------------------
1014    Name: Allen Leung
1015    Date: 2002/03/11 10:30:00 EST
1016    Tag: leunga-20020311-runtime-string0
1017    Description:
1018    
1019       The representation of the empty string now points to a
1020    legal null terminated C string instead of unit.  It is now possible
1021    to convert an ML string into C string with InlineT.CharVector.getData.
1022    This compiles into one single machine instruction.
1023    
1024    ----------------------------------------------------------------------
1025    Name: Allen Leung
1026    Date: 2002/03/10 23:55:00 EST
1027    Tag: leunga-20020310-x86-call
1028    Description:
1029    
1030       Added machine generation for CALL instruction (relative displacement mode)
1031    
1032    ----------------------------------------------------------------------
1033    Name: Matthias Blume
1034    Date: 2002/03/08 16:05:00
1035    Tag: blume-20020308-entrypoints
1036    Description:
1037    
1038    Version number bumped to 110.39.1.  NEW BOOTFILES!
1039    
1040    Entrypoints: non-zero offset into a code object where execution should begin.
1041    
1042    - Added the notion of an entrypoint to CodeObj.
1043    - Added reading/writing of entrypoint info to Binfile.
1044    - Made runtime system bootloader aware of entrypoints.
1045    - Use the address of the label of the first function given to mlriscGen
1046      as the entrypoint.  This address is currently always 0, but it will
1047      not be 0 once we turn on block placement.
1048    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1049      with entry points) from mlriscGen.
1050    
1051    ----------------------------------------------------------------------
1052    Name: Allen Leung
1053    Date: 2002/03/07 20:45:00 EST
1054    Tag: leunga-20020307-x86-cmov
1055    Description:
1056    
1057       Bug fixes for CMOVcc on x86.
1058    
1059       1. Added machine code generation for CMOVcc
1060       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1061       3. CMOVcc cannot have an immediate operand as argument.
1062    
1063    ----------------------------------------------------------------------
1064    Name: Matthias Blume
1065    Date: 2002/03/07 16:15:00 EST
1066    Tag: blume-20020307-controls
1067    Description:
1068    
1069    This is a very large but mostly boring patch which makes (almost)
1070    every tuneable compiler knob (i.e., pretty much everything under
1071    Control.* plus a few other things) configurable via both the command
1072    line and environment variables in the style CM did its configuration
1073    until now.
1074    
1075    Try starting sml with '-h' (or, if you are brave, '-H')
1076    
1077    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1078    implements the underlying generic mechanism.
1079    
1080    The interface to some of the existing such facilities has changed somewhat.
1081    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1082    (The getFoo interface is still there for backward-compatibility, but its
1083    use is deprecated.)
1084    
1085    The ml-build script passes -Cxxx=yyy command-line arguments through so
1086    that one can now twiddle the compiler settings when using this "batch"
1087    compiler.
1088    
1089    TODO items:
1090    
1091    We should go through and throw out all controls that are no longer
1092    connected to anything.  Moreover, we should go through and provide
1093    meaningful (and correct!) documentation strings for those controls
1094    that still are connected.
1095    
1096    Currently, multiple calls to Controls.new are accepted (only the first
1097    has any effect).  Eventually we should make sure that every control
1098    is being made (via Controls.new) exactly once.  Future access can then
1099    be done using Controls.acc.
1100    
1101    Finally, it would probably be a good idea to use the getter-setter
1102    interface to controls rather than ref cells.  For the time being, both
1103    styles are provided by the Controls module, but getter-setter pairs are
1104    better if thread-safety is of any concern because they can be wrapped.
1105    
1106    *****************************************
1107    
1108    One bug fix: The function blockPlacement in three of the MLRISC
1109    backpatch files used to be hard-wired to one of two possibilities at
1110    link time (according to the value of the placementFlag).  But (I
1111    think) it should rather sense the flag every time.
1112    
1113    *****************************************
1114    
1115    Other assorted changes (by other people who did not supply a HISTORY entry):
1116    
1117    1. the cross-module inliner now works much better (Monnier)
1118    2. representation of weights, frequencies, and probabilities in MLRISC
1119       changed in preparation of using those for weighted block placement
1120       (Reppy, George)
1121    
1122    ----------------------------------------------------------------------
1123    Name: Lal George
1124    Date: 2002/03/07 14:44:24 EST 2002
1125    Tag: george-20020307-weighted-block-placement
1126    
1127    Tested the weighted block placement optimization on all architectures
1128    (except the hppa) using AMPL to generate the block and edge frequencies.
1129    Changes were required in the machine properties to correctly
1130    categorize trap instructions. There is an MLRISC flag
1131    "weighted-block-placement" that can be used to enable weighted block
1132    placement, but this will be ineffective without block/edge
1133    frequencies (coming soon).
1134    
1135    
1136    ----------------------------------------------------------------------
1137    Name: Lal George
1138    Date: 2002/03/05 17:24:48 EST
1139    Tag: george-20020305-linkage-cluster
1140    
1141    In order to support the block placement optimization, a new cluster
1142    is generated as the very first cluster (called the linkage cluster).
1143    It contains a single jump to the 'real' entry point for the compilation
1144    unit. Block placement has no effect on the linkage cluster itself, but
1145    all the other clusters  have full freedom in the manner in which they
1146    reorder blocks or functions.
1147    
1148    On the x86 the typical linkage code that is generated is:
1149       ----------------------
1150            .align 2
1151       L0:
1152            addl    $L1-L0, 72(%esp)
1153            jmp     L1
1154    
1155    
1156            .align  2
1157       L1:
1158       ----------------------
1159    
1160    72(%esp) is the memory location for the stdlink register. This
1161    must contain the address of the CPS function being called. In the
1162    above example, it contains the address of  L0; before
1163    calling L1 (the real entry point for the compilation unit), it
1164    must contain the address for L1, and hence
1165    
1166            addl $L1-L0, 72(%esp)
1167    
1168    I have tested this on all architectures except the hppa.The increase
1169    in code size is of course negligible
1170    
1171    ----------------------------------------------------------------------
1172    Name: Allen Leung
1173    Date: 2002/03/03 13:20:00 EST
1174    Tag: leunga-20020303-mlrisc-tools
1175    
1176      Added #[ ... ] expressions to mlrisc tools
1177    
1178    ----------------------------------------------------------------------
1179    Name: Matthias Blume
1180    Date: 2002/02/27 12:29:00 EST
1181    Tag: blume-20020227-cdebug
1182    Description:
1183    
1184    - made types in structure C and C_Debug to be equal
1185    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1186    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1187    
1188    ----------------------------------------------------------------------
1189    Name: Matthias Blume
1190    Date: 2002/02/26 12:00:00 EST
1191    Tag: blume-20020226-ffi
1192    Description:
1193    
1194    1. Fixed a minor bug in CM's "noweb" tool:
1195       If numbering is turned off, then truly don't number (i.e., do not
1196       supply the -L option to noweb).  The previous behavior was to supply
1197       -L'' -- which caused noweb to use the "default" line numbering scheme.
1198       Thanks to Chris Richards for pointing this out (and supplying the fix).
1199    
1200    2. Once again, I reworked some aspects of the FFI:
1201    
1202       A. The incomplete/complete type business:
1203    
1204       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1205         gone!
1206       - ML types representing an incomplete type are now *equal* to
1207         ML types representing their corresponding complete types (just like
1208         in C).  This is still safe because ml-nlffigen will not generate
1209         RTTI for incomplete types, nor will it generate functions that
1210         require access to such RTTI.   But when ML code generated from both
1211         incomplete and complete versions of the C type meet, the ML types
1212         are trivially interoperable.
1213    
1214         NOTE:  These changes restore the full generality of the translation
1215         (which was previously lost when I eliminated functorization)!
1216    
1217       B. Enum types:
1218    
1219       - Structure C now has a type constructor "enum" that is similar to
1220         how the "su" constructor works.  However, "enum" is not a phantom
1221         type because each "T enum" has values (and is isomorphic to
1222         MLRep.Signed.int).
1223       - There are generic access operations for enum objects (using
1224         MLRep.Signed.int).
1225       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1226         * The structure contains the definition of type "mlrep" (the ML-side
1227         representation type of the enum).  Normally, mlrep is the same
1228         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1229         then mlrep will be defined as a datatype -- thus facilitating
1230         pattern matching on mlrep values.
1231         ("-ec" will be suppressed if there are duplicate values in an
1232          enumeration.)
1233         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1234         will be generated for each C enum constant xxx.
1235         * Conversion functions m2i and i2m convert between mlrep and
1236         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1237         * Coversion functions c and ml convert between mlrep and "tag enum".
1238         * Access functions (get/set) fetch and store mlrep values.
1239       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1240         enumerations are merged into one single enumeration represented by
1241         structure E_'.
1242    
1243    ----------------------------------------------------------------------
1244    Name: Allen Leung
1245    Date: 2002/02/25 04:45:00 EST
1246    Tag: leunga-20020225-cps-spill
1247    
1248    This is a new implementation of the CPS spill phase.
1249    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1250    In case of problems, replace it with the old file spill.sml
1251    
1252    The current compiler runs into some serious performance problems when
1253    constructing a large record.  This can happen when we try to compile a
1254    structure with many items.  Even a very simple structure like the following
1255    makes the compiler slow down.
1256    
1257        structure Foo = struct
1258           val x_1 = 0w1 : Word32.int
1259           val x_2 = 0w2 : Word32.int
1260           val x_3 = 0w3 : Word32.int
1261           ...
1262           val x_N = 0wN : Word32.int
1263        end
1264    
1265    The following table shows the compile time, from N=1000 to N=4000,
1266    with the old compiler:
1267    
1268    N
1269    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1270           MLRISC ra                               0.06u  0.00s  0.05g
1271              (spills = 0 reloads = 0)
1272           TOTAL                                   0.63u  0.07s  0.21g
1273    
1274    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1275           MLRISC ra                               5.68u  0.59s  3.93g
1276              (spills = 0 reloads = 0)
1277           TOTAL                                   14.71u  0.99s  4.81g
1278    
1279    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1280           MLRISC ra                               5.54u  0.65s  3.91g
1281              (spills = 543 reloads = 1082)
1282           TOTAL                                   65.40u  3.13s  6.00g
1283    
1284    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1285           MLRISC ra                               0.80u  0.10s  0.55g
1286              (spills = 42 reloads = 84)
1287           TOTAL                                   129.42u  5.10s  4.13g
1288    
1289    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1290           MLRISC ra                               2.69u  0.27s  1.38g
1291              (spills = 62 reloads = 124)
1292           TOTAL                                   682.48u  19.61s  13.99g
1293    
1294    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1295           MLRISC ra                               4.96u  0.27s  2.72g
1296              (spills = 85 reloads = 170)
1297           TOTAL                                   2375.26u  57.21s  48.00g
1298    
1299    As you can see the old cps spill module suffers from some serious
1300    performance problem.  But since I cannot decipher the old code fully,
1301    instead of patching the problems up, I'm reimplementing it
1302    with a different algorithm.  The new code is more modular,
1303    smaller when compiled, and substantially faster
1304    (O(n log n) time and O(n) space).  Timing of the new spill module:
1305    
1306    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1307          MLRISC ra                               0.25u  0.02s  0.15g
1308             (spills=1 reloads=3)
1309          TOTAL                                   7.74u  0.34s  1.62g
1310    
1311    Implementation details:
1312    
1313    As far as I can tell, the purpose of the CPS spill module is to make sure the
1314    number of live variables at any program point (the bandwidth)
1315    does not exceed a certain limit, which is determined by the
1316    size of the spill area.
1317    
1318    When the bandwidth is too large, we decrease the register pressure by
1319    packing live variables into spill records.  How we achieve this is
1320    completely different than what we did in the old code.
1321    
1322    First, there is something about the MLRiscGen code generator
1323    that we should be aware of:
1324    
1325    o MLRiscGen performs code motion!
1326    
1327       In particular, it will move floating point computations and
1328       address computations involving only the heap pointer to
1329       their use sites (if there is only a single use).
1330       What this means is that if we have a CPS record construction
1331       statement
1332    
1333           RECORD(k,vl,w,e)
1334    
1335       we should never count the new record address w as live if w
1336       has only one use (which is often the case).
1337    
1338       We should do something similar to floating point, but the transformation
1339       there is much more complex, so I won't deal with that.
1340    
1341    Secondly, there are now two new cps primops at our disposal:
1342    
1343     1. rawrecord of record_kind option
1344        This pure operator allocates some uninitialized storage from the heap.
1345        There are two forms:
1346    
1347         rawrecord NONE [INT n]  allocates a tagless record of length n
1348         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1349                                     and initializes the tag.
1350    
1351     2. rawupdate of cty
1352          rawupdate cty (v,i,x)
1353          Assigns to x to the ith component of record v.
1354          The storelist is not updated.
1355    
1356    We use these new primops for both spilling and increment record construction.
1357    
1358     1. Spilling.
1359    
1360        This is implemented with a linear scan algorithm (but generalized
1361        to trees).  The algorithm will create a single spill record at the
1362        beginning of the cps function and use rawupdate to spill to it,
1363        and SELECT or SELp to reload from it.  So both spills and reloads
1364        are fine-grain operations.  In contrast, in the old algorithm
1365        "spills" have to be bundled together in records.
1366    
1367        Ideally, we should sink the spill record construction to where
1368        it is needed.  We can even split the spill record into multiple ones
1369        at the places where they are needed.  But CPS is not a good
1370        representation for global code motion, so I'll keep it simple and
1371        am not attempting this.
1372    
1373     2. Incremental record construction (aka record splitting).
1374    
1375        Long records with many component values which are simulatenously live
1376        (recall that single use record addresses are not considered to
1377         be live) are constructed with rawrecord and rawupdate.
1378        We allocate space on the heap with rawrecord first, then gradually
1379        fill it in with rawupdate.  This is the technique suggested to me
1380        by Matthias.
1381    
1382        Some restrictions on when this is applicable:
1383        1. It is not a VECTOR record.  The code generator currently does not handle
1384           this case. VECTOR record uses double indirection like arrays.
1385        2. All the record component values are defined in the same "basic block"
1386           as the record constructor.  This is to prevent speculative
1387           record construction.
1388    
1389    ----------------------------------------------------------------------
1390    Name: Allen Leung
1391    Date: 2002/02/22 01:02:00 EST
1392    Tag: leunga-20020222-mlrisc-tools
1393    
1394    Minor bug fixes in the parser and rewriter
1395    
1396    ----------------------------------------------------------------------
1397    Name: Allen Leung
1398    Date: 2002/02/21 20:20:00 EST
1399    Tag: leunga-20020221-peephole
1400    
1401    Regenerated the peephole files.  Some contained typos in the specification
1402    and some didn't compile because of pretty printing bugs in the old version
1403    of 'nowhere'.
1404    
1405    ----------------------------------------------------------------------
1406    Name: Allen Leung
1407    Date: 2002/02/19 20:20:00 EST
1408    Tag: leunga-20020219-mlrisc-tools
1409    Description:
1410    
1411       Minor bug fixes to the mlrisc-tools library:
1412    
1413       1.  Fixed up parsing colon suffixed keywords
1414       2.  Added the ability to shut the error messages up
1415       3.  Reimplemented the pretty printer and fixed up/improved
1416           the pretty printing of handle and -> types.
1417       4.  Fixed up generation of literal symbols in the nowhere tool.
1418       5.  Added some SML keywords to to sml.sty
1419    
1420    ----------------------------------------------------------------------
1421    Name: Matthias Blume
1422    Date: 2002/02/19 16:20:00 EST
1423    Tag: blume-20020219-cmffi
1424    Description:
1425    
1426    A wild mix of changes, some minor, some major:
1427    
1428    * All C FFI-related libraries are now anchored under $c:
1429        $/c.cm      --> $c/c.cm
1430        $/c-int.cm  --> $c/internals/c-int.cm
1431        $/memory.cm --> $c/memory/memory.cm
1432    
1433    * "make" tool (in CM) now treats its argument pathname slightly
1434      differently:
1435        1. If the native expansion is an absolute name, then before invoking
1436           the "make" command on it, CM will apply OS.Path.mkRelative
1437           (with relativeTo = OS.FileSys.getDir()) to it.
1438        2. The argument will be passed through to subsequent phases of CM
1439           processing without "going native".  In particular, if the argument
1440           was an anchored path, then "make" will not lose track of that anchor.
1441    
1442    * Compiler backends now "know" their respective C calling conventions
1443      instead of having to be told about it by ml-nlffigen.  This relieves
1444      ml-nlffigen from one of its burdens.
1445    
1446    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1447    
1448    * Export C_DEBUG and C_Debug from $c/c.cm.
1449    
1450    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1451      subtyping relationship between incomplete pointers and their complete
1452      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1453      with the convention of instantiating 'o with ('t, 'c) obj whenever
1454      the pointer target type is complete.  In the incomplete case, 'o
1455      will be instantiated with some "'c iobj" -- a type obtained by
1456      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1457    
1458      Operations that work on both incomplete and complete pointer types are
1459      typed as taking an 'o ptr while operations that require the target to
1460      be known are typed as taking some ('t, 'c) obj ptr.
1461    
1462      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1463      where void is an eqtype without any values.  This makes it possible
1464      to work on voidptr values using functions meant to operate on light
1465      incomplete pointers.
1466    
1467    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1468      been vastly simplified.
1469    
1470    ----------------------------------------------------------------------
1471    Name: Matthias Blume
1472    Date: 2002/02/19 10:48:00 EST
1473    Tag: blume-20020219-pqfix
1474    Description:
1475    
1476    Applied Chris Okasaki's bug fix for priority queues.
1477    
1478    ----------------------------------------------------------------------
1479    Name: Matthias Blume
1480    Date: 2002/02/15 17:05:00
1481    Tag: Release_110_39
1482    Description:
1483    
1484    Last-minute retagging is becoming a tradition... :-(
1485    
1486    This is the working release 110.39.
1487    
1488    ----------------------------------------------------------------------
1489    Name: Matthias Blume
1490    Date: 2002/02/15 16:00:00 EST
1491    Tag: Release_110_39-orig
1492    Description:
1493    
1494    Working release 110.39.  New bootfiles.
1495    
1496    (Update: There was a small bug in the installer so it wouldn't work
1497    with all shells.  So I retagged. -Matthias)
1498    
1499    ----------------------------------------------------------------------
1500    Name: Matthias Blume
1501    Date: 2002/02/15 14:17:00 EST
1502    Tag: blume-20020215-showbindings
1503    Description:
1504    
1505    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1506    the latter can be useful for exploring what bindings are available at
1507    the interactive prompt.  (The first function returns only the list
1508    of symbols that are really bound, the second prints those but also the
1509    ones that CM's autoloading mechanism knows about.)
1510    
1511    ----------------------------------------------------------------------
1512    Name: Matthias Blume
1513    Date: 2002/02/15 12:08:00 EST
1514    Tag: blume-20020215-iptrs
1515    Description:
1516    
1517    Two improvements to ml-nlffigen:
1518    
1519      1. Write files only if they do not exist or if their current contents
1520         do not coincide with what's being written.  (That is, avoid messing
1521         with the time stamps unless absolutely necessary.)
1522    
1523      2. Implement a "repository" mechanism for generated files related
1524         to "incomplete pointer types".   See the README file for details.
1525    
1526    ----------------------------------------------------------------------
1527    Name: Matthias Blume
1528    Date: 2002/02/14 11:50:00 EST
1529    Tag: blume-20020214-quote
1530    Description:
1531    
1532    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1533    because of the new and improved tag generation scheme.  (Thanks to Allen
1534    Leung for pointing it out.)
1535    
1536    ----------------------------------------------------------------------
1537    Name: Lal George
1538    Date: 2002/02/14 09:55:27 EST 2002
1539    Tag: george-20020214-isabelle-bug
1540    Description:
1541    
1542    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1543    of Isabelle on the x86.
1544    
1545    From Allen:
1546    -----------
1547     I've found the problem:
1548    
1549         in ra-core.sml, I use the counter "blocked" to keep track of the
1550         true number of elements in the freeze queue.  When the counter goes
1551         to zero, I skip examining the queue.  But I've messed up the
1552         bookkeeping in combine():
1553    
1554             else ();
1555             case !ucol of
1556               PSEUDO => (if !cntv > 0 then
1557                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1558                                        ^^^^^^^^^^^^^^^^^^^^^^^
1559                      moveu := mergeMoveList(!movev, !moveu)
1560                     )
1561                  else ();
1562    
1563         combine() is called to coalesce two nodes u and v.
1564         I think I was thinking that if the move counts of u and v are both
1565         greater than zero then after they are coalesced then one node is
1566         removed from the freeze queue.  Apparently I was thinking that
1567         both u and v are of low degree, but that's clearly not necessarily true.
1568    
1569    
1570    02/12/2002:
1571        Here's the patch.  HOL now compiles.
1572    
1573        I don't know how this impact on performance (compile
1574        time or runtime).  This bug caused the RA (especially on the x86)
1575        to go thru the potential spill phase when there are still nodes on the
1576        freeze queue.
1577    
1578    
1579    
1580    
1581    ----------------------------------------------------------------------
1582    Name: Matthias Blume
1583    Date: 2002/02/13 22:40:00 EST
1584    Tag: blume-20020213-fptr-rtti
1585    Description:
1586    
1587    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1588    updates.
1589    
1590    ----------------------------------------------------------------------
1591    Name: Matthias Blume
1592    Date: 2002/02/13 16:41:00 EST
1593    Tag: blume-20020213-cmlpq
1594    Description:
1595    
1596    Added new priority queue export symbols (which have just been added to
1597    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1598    not compile and the installer would choke.)
1599    
1600    ----------------------------------------------------------------------
1601    Name: Matthias Blume
1602    Date: 2002/02/13 16:15:00 EST
1603    Tag: blume-20020213-various
1604    Description:
1605    
1606    1. More tweaks to ml-nlffigen:
1607    
1608       - better internal datastructures (resulting in slight speedup)
1609       - "-match" option requires exact match
1610       - "localized" gensym counters (untagged structs/unions nested within
1611         other structs/unions or within typedefs get a fresh counter; their
1612         tag will be prefixed by a concatenation of their parents' tags)
1613       - bug fixes (related to calculation of transitive closure of types
1614         to be included in the output)
1615    
1616    2. Minor Basis updates:
1617    
1618       - added implementations for List.collate and Option.app
1619    
1620    ----------------------------------------------------------------------
1621    Name: Matthias Blume
1622    Date: 2002/02/11 15:55:00 EST
1623    Tag: blume-20020211-gensym
1624    Description:
1625    
1626    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1627    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1628    names (ML structure names that correspond to unnamed C structs, unions,
1629    and enums), so that separate runs of ml-nlffigen do not clash.
1630    
1631    ----------------------------------------------------------------------
1632    Name: Matthias Blume
1633    Date: 2002/02/11 12:05:00 EST
1634    Tag: blume-20020211-gensml
1635    Description:
1636    
1637    A quick fix for a problem with GenSML (in the pgraph-util library):
1638    Make generation of toplevel "local" optional.  (Strictly speaking,
1639    signature definitions within "local" are not legal SML.)
1640    
1641    Other than that: updates to INSTALL and cm/TODO.
1642    
1643    ----------------------------------------------------------------------
1644    Name: Matthias Blume
1645    Date: 2002/02/08 15:00:00 EST
1646    Tag: blume-20020208-uniquepid
1647    Description:
1648    
1649    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1650    
1651    1. The installer (config/install.sh) has gotten smarter:
1652    
1653         - Configuration options are a bit easier to specify now
1654           (in config/targets).
1655         - Bug in recognizing .tar.bz2 files fixed.
1656         - Installer automatically resolves dependencies between
1657           configuration options (e.g., if you ask for eXene, you will
1658           also get cml -- regardless whether you asked for it or not).
1659         - Installer can run in "quieter mode" by setting the environment
1660           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1661           "completely silent", though.
1662         - Build HashCons library as part of smlnj-lib.
1663    
1664    2. A new scheme for assigning persistent identifiers to compilation
1665       units (and, by extension, to types etc.) has been put into place.
1666       This fixes a long-standing bug where types and even dynamic values
1667       can get internally confused, thereby compromising type safety
1668       (abstraction) and dynamic correctness.  See
1669    
1670         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1671    
1672       for an example of how things could go wrong until now.
1673    
1674       The downside of the new scheme is that pids are not quite as
1675       persistent as they used to be: CM will generate a fresh pid
1676       for every compilation unit that it thinks it sees for the first
1677       time.  That means that if you compile starting from a clean, fresh
1678       source tree at two different times, you end up with different
1679       binaries.
1680    
1681       Cutoff recompilation, however, has not been compromised because
1682       CM keeps pid information in special caches between runs.
1683    
1684    ----------------------------------------------------------------------
1685    Name: Lal George
1686    Date: 2002/02/07 15:34:13 EST 2002
1687    Tag: <none>
1688    Description:
1689    
1690    Compilers that generate assembly code may produce  global labels
1691    whose value is resolved at link time. The various peephole optimization
1692    modules did not take this in account.
1693    
1694    TODO. The Labels.addrOf function should really return an option
1695    type so that clients are forced to deal with this issue, rather
1696    than an exception being raised.
1697    
1698    ----------------------------------------------------------------------
1699    Name: Lal George
1700    Date: 2002/02/06 13:55:02 EST
1701    Tag: george-20020206-ra-breakup
1702    Description:
1703    
1704    1. A bug fix from Allen.
1705    
1706        A typo causes extra fstp %st(0)'s to be generated at compensation
1707        edges, which might cause stack underflow traps at runtime.  This
1708        occurs in fft where there are extraneous fstps right before the 'into'
1709        trap instruction (in this case they are harmless since none of the
1710        integers overflow.)
1711    
1712    2. Pulled out various utility modules that were embedded in the modules
1713       of the register allocator. I need these modules for other purposes, but
1714       they are not complete enough to put into a library (just yet).
1715    ----------------------------------------------------------------------
1716    Name: Matthias Blume
1717    Date: 2002/01/31 16:05:00 EST
1718    Tag: blume-20020131-sparc-ccalls
1719    Description:
1720    
1721    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1722       of extra stack space by mistake.  Fixed.
1723    
1724    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1725    
1726    ----------------------------------------------------------------------
1727    Name: Allen Leung
1728    Date: 2002/01/30
1729    Tag: leunga-20020130-nowhere-bug-fix
1730    Description:
1731    
1732       MLRISC bug fixes:
1733       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1734       2. MachineInt.fromString was negating its value.
1735    
1736    ----------------------------------------------------------------------
1737    Name: Matthias Blume
1738    Date: 2002/01/29
1739    Tag: blume-20020129-INSTALL
1740    Description:
1741    
1742    - Added somewhat detailed installation instructions (file INSTALL).
1743    - Fixed curl-detection bug in config/install.sh.
1744    - It is now possible to select the URL getter using the URLGETTER
1745      environment variable:
1746    
1747          not set / "unknown"      --> automatic detection (script tries wget,
1748                                       curl, and lynx)
1749          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1750                                       how to properly invoke them)
1751          other                    --> use $URLGETTER directly, it must take
1752                                       precisely two command-line arguments
1753                                       (source URL and destination file name)
1754    
1755    ----------------------------------------------------------------------
1756    Name: Matthias Blume
1757    Date: 2002/01/28
1758    Tag: blume-20020128-sparc-ccalls
1759    Description:
1760    
1761    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1762    - Make use of the allocParam argument in sparc-c-calls.
1763    
1764    ----------------------------------------------------------------------
1765    Name: Matthias Blume
1766    Date: 2002/01/28
1767    Tag: blume-20020128-allocParam
1768    Description:
1769    
1770    John Reppy:  Changes c-calls API to accept client-callback for
1771    allocating extra stack space.
1772    me: Corresponding changes to mlriscGen (using a dummy argument that
1773        does not change the current behavior).
1774    
1775    ----------------------------------------------------------------------
1776    Name: Matthias Blume
1777    Date: 2002/01/28 12:00:00
1778    Tag: Release_110_38
1779    Description:
1780    
1781    This time for real!!!
1782    
1783    ----------------------------------------------------------------------
1784    Name: Matthias Blume
1785    Date: 2002/01/28 10:56:00 EST
1786    Tag: blume-20020128-retraction
1787    Description:
1788    
1789    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1790       with blume-Release_110_38-retracted.)
1791    
1792    1. Fixed a problem with incorrect rounding modes in real64.sml.
1793       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1794    
1795    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1796       and enums fixed.  The naming of corresponding ML identifiers should
1797       now be consistent again.
1798    
1799    ----------------------------------------------------------------------
1800    Name: Allen Leung
1801    Date: 2002/01/27
1802    Tag: leunga-20020127-nowhere
1803    Description:
1804    
1805       Added a target called nowhere in the configuration scripts.
1806       Enabling this will build the MLRISC 'nowhere' tool (for translating
1807       programs with where-clauses into legal SML code) during installation.
1808    
1809    ----------------------------------------------------------------------
1810    Name: Matthias Blume
1811    Date: 2002/01/25 21:27:00 EST
1812    Tag: blume-Release_110_38-retracted
1813    Description:
1814    
1815    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1816    
1817    README will be added later.
1818    
1819    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1820    (Release_110_38) removed.  Reason: Last-minute bug fixes.
1821    
1822    ----------------------------------------------------------------------
1823    Name: Matthias Blume
1824    Date: 2002/01/25
1825    Tag: blume-20020125-ffi
1826    Description:
1827    
1828    A large number of tweaks and improvements to ml-nlffi-lib and
1829    ml-nlffigen:
1830    
1831       - ML represenation types have been streamlined
1832       - getter and setter functions work with concrete values, not abstract
1833         ones where possible
1834       - ml-nlffigen command line more flexible (see README file there)
1835       - some bugs have been fixed (hopefully)
1836    
1837    ----------------------------------------------------------------------
1838    Name: Lal George
1839    Date: 2002/01/24
1840    Tag: george-20020124-risc-ra-interface
1841    Description:
1842    
1843       There is a dramatic simplification in the interface to the
1844       register allocator for RISC architectures as a result of making
1845       parallel copy instructions explicit.
1846    
1847    ----------------------------------------------------------------------
1848    Name: Matthias Blume
1849    Date: 2002/01/22
1850    Tag: blume-20020122-x86-ccalls
1851    Description:
1852    
1853    Bug fix for c-calls on x86 (having to do with how char- and
1854    short-arguments are being handled).
1855    
1856    ----------------------------------------------------------------------
1857    Name: Matthias Blume
1858    Date: 2002/01/21
1859    Tag: blume-20020121-ff
1860    Description:
1861    
1862    Another day of fiddling with the FFI...
1863    
1864    1. Bug fix/workaround:  CKIT does not complain about negative array
1865       dimensions, so ml-nlffigen has to guard itself against this possibility.
1866       (Otherwise a negative dimension would send it into an infinite loop.)
1867    
1868    2. Some of the abstract types (light objects, light pointers, most "base"
1869       types) in structure C are now eqtypes.
1870    
1871    3. Added constructors and test functions for NULL function pointers.
1872    
1873    ----------------------------------------------------------------------
1874    Name: Matthias Blume
1875    Date: 2002/01/18
1876    Tag: blume-20020118-ready-for-new-release
1877    Description:
1878    
1879    Made config/srcarchiveurl point to a new place.  (Will provide boot
1880    files shortly.)
1881    
1882    Maybe we christen this to be 110.38?
1883    
1884    ----------------------------------------------------------------------
1885    Name: Matthias Blume
1886    Date: 2002/01/18
1887    Tag: blume-20020118-more-ffifiddle
1888    Description:
1889    
1890    Today's FFI fiddling:
1891    
1892      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
1893        C.Set.* that use concrete (MLRep.*) arguments and results instead
1894        of abstract ones.
1895    
1896      - Provided word-style bit operations etc. for "int" representation
1897        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1898        and Long.
1899    
1900    ----------------------------------------------------------------------
1901    Name: Matthias Blume
1902    Date: 2002/01/18
1903    Tag: blume-20020118-use-x86-fp
1904    Description:
1905    
1906    Now that x86-fast-fp seems to be working, I turned it back on again
1907    by default.  (Seems to work fine now, even with the FFI.)
1908    
1909    Other than that, I added some documentation about the FFI to
1910    src/ml-nlffigen/README and updated the FFI test examples in
1911    src/ml-nlffi-lib/Tests/*.
1912    
1913    ----------------------------------------------------------------------
1914    Name: Allen Leung
1915    Date: 2002/01/17
1916    Tag: leunga-20020117-x86-fast-fp-call
1917    Description:
1918    
1919       1. Fixed a problem with handling return fp values when x86's fast fp
1920          mode is turned on.
1921    
1922       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
1923          of %f32.
1924    
1925       3. Added a constructor INT32lit to the ast of MLRISC tools.
1926    
1927    ----------------------------------------------------------------------
1928    Name: Matthias Blume
1929    Date: 2002/01/16
1930    Tag: blume-20020116-ffifiddle
1931    Description:
1932    
1933    More fiddling with the FFI interface:
1934    
1935     - Make constness 'c instead of rw wherever possible.  This eliminates
1936       the need for certain explicit coercions.  (However, due to ML's
1937       value polymorphism, there will still be many cases where explicit
1938       coercions are necessary.  Phantom types are not the whole answer
1939       to modeling a subtyping relationship in ML.)
1940    
1941     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
1942    
1943     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
1944       clumsy workaround for varargs problem.)
1945    
1946    ----------------------------------------------------------------------
1947    Name: Lal George
1948    Date: 2002/01/15
1949    Tag: <none>
1950    Description:
1951    
1952    1. Since COPY instructions are no longer native to the architecture,
1953       a generic functor can be used to implement the expandCopies function.
1954    
1955    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1956       TEXT segment.
1957    
1958    ----------------------------------------------------------------------
1959    Name: Matthias Blume
1960    Date: 2002/01/15
1961    Tag: blume-20020115-ffiupdates
1962    Description:
1963    
1964    1. Fix for bug resulting in single-precision float values being returned
1965       incorrectly from FFI calls.
1966    
1967    2. Small modifications to C FFI API:
1968    
1969        - memory-allocation routines return straight objects (no options)
1970          and raise an exception in out-of-memory situations
1971        - unsafe extensions to cast between function pointers and pointers
1972          from/to ints
1973        - added structure C_Debug as an alternative to structure C where
1974          pointer-dereferencing (|*| and |*!) always check for null-pointers
1975        - added open_lib' to DynLinkage;  open_lib' works like open_lib
1976          but also takes a (possibly empty) list of existing library handles
1977          that the current library depends on
1978    
1979    ----------------------------------------------------------------------
1980    Name: Matthias Blume
1981    Date: 2002/01/10
1982    Tag: blume-20020110-newffigen
1983    Description:
1984    
1985    1. Updates to portable graph code.
1986    
1987    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
1988       more scalable now so that even huge interfaces such as the one
1989       for GTK compile in finite time and space. :-)
1990       See src/ml-nlffigen/README for details on what's new.
1991    
1992    ----------------------------------------------------------------------
1993    Name: Lal George
1994    Date: 2001/01/09 14:31:35 EST 2002
1995    Tag: george-20011206-rm-native-copy
1996    Description:
1997    
1998            Removed the native COPY and FCOPY instructions
1999            from all the architectures and replaced it with the
2000            explicit COPY instruction from the previous commit.
2001    
2002            It is now possible to simplify many of the optimizations
2003            modules that manipulate copies. This has not been
2004            done in this change.
2005    
2006    ----------------------------------------------------------------------
2007    Name: Lal George
2008    Date: 2001/12/06 16:50:13 EST 2001
2009    Tag: george-20011206-mlrisc-instruction
2010    Description:
2011    
2012    Changed the representation of instructions from being fully abstract
2013    to being partially concrete. That is to say:
2014    
2015      from
2016            type instruction
2017    
2018      to
2019            type instr                              (* machine instruction *)
2020    
2021            datatype instruction =
2022                LIVE of {regs: C.cellset, spilled: C.cellset}
2023              | KILL of {regs: C.cellset, spilled: C.cellset}
2024              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
2025              | ANNOTATION of {i: instruction, a: Annotations.annotation}
2026              | INSTR of instr
2027    
2028    This makes the handling of certain special instructions that appear on
2029    all architectures easier and uniform.
2030    
2031    LIVE and KILL say that a list of registers are live or killed at the
2032    program point where they appear. No spill code is generated when an
2033    element of the 'regs' field is spilled, but the register is moved to
2034    the 'spilled' (which is present, more for debugging than anything else).
2035    
2036    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
2037    We used to generate:
2038    
2039            DEFFREG f1
2040            f1 := f2 + f3
2041            trapb
2042    
2043    but now generate:
2044    
2045            f1 := f2 + f3
2046            trapb
2047            LIVE {regs=[f1,f2,f3], spilled=[]}
2048    
2049    Furthermore, the DEFFREG (hack) required that all floating point instruction
2050    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
2051    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
2052    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
2053    this architecture).
2054    
2055    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
2056    all the architectures. This will result in further simplification of the
2057    register allocator that must be aware of them for coalescing purposes, and
2058    will also simplify certain aspects of the machine description that provides
2059    callbacks related to parallel copies.
2060    
2061    ANNOTATION should be obvious, and now INSTR represents the honest to God
2062    machine instruction set!
2063    
2064    The <arch>/instructions/<arch>Instr.sml files define certain utility
2065    functions for making porting easier -- essentially converting upper case
2066    to lower case. All machine instructions (of type instr) are in upper case,
2067    and the lower case form generates an MLRISC instruction. For example on
2068    the alpha we have:
2069    
2070      datatype instr =
2071         LDA of {r:cell, b:cell, d:operand}
2072       | ...
2073    
2074      val lda : {r:cell, b:cell, d:operand} -> instruction
2075        ...
2076    
2077    where lda is just (INSTR o LDA), etc.
2078    
2079    ----------------------------------------------------------------------
2080    Name: Matthias Blume
2081    Date: 2001/11/22 21:40:00 EST
2082    Tag: Release_110_37
2083    Description:
2084    
2085    Release 110.37.  This time for real.
2086    
2087    ----------------------------------------------------------------------
2088    Name: Matthias Blume
2089    Date: 2001/11/21 16:35:00 EST
2090    Tag: blume-20011121-foot-in-mouth
2091    Description:
2092    
2093    Removed the "Release_110_37" tag because of a serious bug.
2094    This will be re-tagged once the bug is fixed.
2095    
2096    ----------------------------------------------------------------------
2097    Name: Matthias Blume
2098    Date: 2001/11/21 16:14:00 EST
2099    Tag: blume-20011121-forgottenfile
2100    Description:
2101    
2102    Forgot to add a file.  (Just a .tex-file -- part of
2103    the CM manual source.)
2104    
2105    ----------------------------------------------------------------------
2106    Name: Matthias Blume
2107    Date: 2001/11/21 16:10:00 EST
2108    Tag: blume-20011121-invalid_110_37
2109    Description:
2110    
2111    Note: I removed the original tag "Release_110_37" from this commit
2112          because we found a serious bug in all non-x86 backends.
2113          - Matthias
2114    
2115    1. Modifications to the SML/NJ code generator and to the runtime system
2116       so that code object name strings are directly inserted into code
2117       objects at code generation time.  The only business the runtime system
2118       has with this is now to read the name strings on occasions.
2119       (The encoding of the name string has also changed somewhat.)
2120    
2121    2. CM now implements a simple "set calculus" for specifying export lists.
2122       In particular, it is now possible to refer to the export lists of
2123       other libraries/groups/sources and form unions as well as differences.
2124       See the latest CM manual for details.
2125    
2126    3. An separate notion of "proxy" libraries has again be eliminated from
2127       CM's model.  (Proxy libraries are now simply a special case of using
2128       the export list calculus.)
2129    
2130    4. Some of the existing libraries now take advantage of the new set
2131       calculus.
2132       (Notice that not all libraries have been converted because some
2133       of the existing .cm-files are supposed to be backward compatible
2134       with 110.0.x.)
2135    
2136    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2137       "General.exnMessage"!  The former relies on a certain hook to be
2138       initialized, and that often does not happen in the stand-alone case.)
2139    
2140    ----------------------------------------------------------------------
2141    Name: Lal George
2142    Date: 2001/11/21  13:56:18 EST
2143    Tag: george-2001121-pseudo-ops
2144    Description:
2145    
2146      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2147      ought to never be any question of incompatabilities with
2148      pseudo-op syntax expected by host assemblers.
2149    
2150      For now, only modules supporting GAS syntax are implemented
2151      but more should follow, such as MASM, and vendor assembler
2152      syntax, e.g. IBM as, Sun as, etc.
2153    
2154    ----------------------------------------------------------------------
2155    Name: Matthias Blume
2156    Date: 2001/11/14 11:52:00 EST
2157    Tag: blume-20011114-srcname
2158    Description:
2159    
2160    1. Routed the name of the current source file to mlriscgen where it
2161       should be directly emitted into the code object.  (This last part
2162       is yet to be done.)
2163    
2164    2. Some cleanup of the pgraph code to make it match the proposal that
2165       I put out the other day.  (The proposal notwithstanding, things are
2166       still in flux here.)
2167    
2168    ----------------------------------------------------------------------
2169    Name: Lal George
2170    Date: 2001/11/14 09:44:04 EST
2171    Tag:
2172    Description:
2173    
2174      Fix for a backpatching bug reported by Allen.
2175    
2176      Because the boundary between short and long span-dependent
2177      instructions is +/- 128, there are an astounding number of
2178      span-dependent instructions whose size is over estimated.
2179    
2180      Allen came up with the idea of letting the size of span
2181      dependent instructions be non-monotonic, for a maxIter
2182      number of times, after which the size must be monotonically
2183      increasing.
2184    
2185      This table shows the number of span-dependent instructions
2186      whose size was over-estimated as a function of maxIter, for the
2187      file Parse/parse/ml.grm.sml:
2188    
2189         maxIter            # of instructions:
2190            10                      687
2191            20                      438
2192            30                      198
2193            40                        0
2194    
2195      In compiling the compiler, there is no significant difference in
2196      compilation speed between maxIter=10 and maxIter=40. Actually,
2197      my measurements showed that maxIter=40 was a tad faster than
2198      maxIter=10! Also 96% of the  files in the compiler reach a fix
2199      point within 13 iterations, so fixing maxIter at 40, while high,
2200      is okay.
2201    
2202    ----------------------------------------------------------------------
2203    Name: Matthias Blume
2204    Date: 2001/10/31 15:25:00 EST
2205    Tag: blume-20011031-pgraph
2206    Description:
2207    
2208    CKIT:
2209    * Changed the "Function" constructor of type Ast.ctype to carry optional
2210      argument identifiers.
2211    * Changed the return type of TypeUtil.getFunction accordingly.
2212    * Type equality ignores the argument names.
2213    * TypeUtil.composite tries to preserve argument names but gives up quickly
2214      if there is a mismatch.
2215    
2216    installation script:
2217    * attempts to use "curl" if available (unless "wget" is available as well)
2218    
2219    CM:
2220    * has an experimental implementation of "portable graphs" which I will
2221      soon propose as an implementation-independent library format
2222    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2223    
2224    NLFFI-LIB:
2225    * some cleanup (all cosmetic)
2226    
2227    NLFFIGEN:
2228    * temporarily disabled the mechanism that suppresses ML output for
2229      C definitions whose identifiers start with an underscore character
2230    * generate val bindings for enum constants
2231    * user can request that only one style (light or heavy) is being used;
2232      default is to use both (command-line arguments: -heavy and -light)
2233    * fixed bug in handling of function types involving incomplete pointers
2234    * generate ML entry points that take record arguments (i.e., using
2235      named arguments) for C functions that have a prototype with named
2236      arguments
2237      (see changes to CKIT)
2238    
2239    ----------------------------------------------------------------------
2240    Name: Allen Leung
2241    Date: 2001/10/27 20:34:00 EDT
2242    Tag: leunga-20011027-x86-fast-fp-call
2243    Description:
2244    
2245       Fixed the bug described in blume-20010920-slowfp.
2246    
2247       The fix involves
2248          1. generating FCOPYs in FSTP in ia32-svid
2249          2. marking a CALL with the appropriate annotation
2250    
2251    ----------------------------------------------------------------------
2252    Name: Matthias Blume
2253    Date: 2001/10/16 11:32:00 EDT
2254    Tag: blume-20011016-netbsd
2255    Description:
2256    
2257    Underscore patch from Chris Richards (fixing problem with compiling
2258    runtime system under recent NetBSD).
2259    
2260    ----------------------------------------------------------------------
2261    Name: Allen Leung
2262    Date: 2001/10/12 17:18:32 EDT 2001
2263    Tag: leung-20011012-x86-printflowgraph
2264    Description:
2265    
2266    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2267    
2268    ----------------------------------------------------------------------
2269    Name: Lal George
2270    Date: 2001/10/11 23:51:34 EDT
2271    Tag: george-20011011-too-many-instrs
2272    Description:
2273    
2274    The representation of a program point never expected to see more
2275    than 65536 instructions in a basic block!
2276    
2277    ----------------------------------------------------------------------
2278    Name: Lal George
2279    Date: 2001/10/09 09:41:37 EDT
2280    Tag: george-20011008-mlrisc-labels
2281    Description:
2282    
2283    Changed the machine description files to support printing of
2284    local and global labels in assembly code, based on host assembler
2285    conventions.
2286    
2287    ----------------------------------------------------------------------
2288    Name: Matthias Blume
2289    Date: 2001/09/25 15:25:00 EDT
2290    Tag: blume-20010925-exninfo
2291    Description:
2292    
2293    I provided a non-hook implementation of exnName (at the toplevel) and
2294    made the "dummy" implementation of exnMessage (at the toplevel) more
2295    useful: if nothing gets "hooked in", then at least you are going to
2296    see the exception name and a message indicating why you don't see more.
2297    
2298    [For the time being, programs that need exnMessage and want to use
2299    ml-build should either use General.exnMessage (strongly recommended) or
2300    refer to structure General at some other point so that CM sees a
2301    static dependency.]
2302    
2303    [Similar remarks go for "print" and "use":  If you want to use their
2304    functionality in stand-alone programs generated by ml-build, then use
2305    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2306    
2307    ----------------------------------------------------------------------
2308    Name: Matthias Blume
2309    Date: 2001/09/20 17:28:00 EDT
2310    Tag: blume-20010920-slowfp
2311    Description:
2312    
2313    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2314    
2315    ----------------------------------------------------------------------
2316    Name: Matthias Blume
2317    Date: 2001/09/20 17:20:00 EDT
2318    Tag: blume-20010920-canonicalpaths
2319    Description:
2320    
2321    0. Updated the BOOT file (something that I forgot to do earlier).
2322    
2323    1. Small internal change to CM so that it avoids "/../" in filenames
2324       as much as possible (but only where it is safe).
2325    
2326    2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2327       that arguments that contain delimiters are passed through correctly.
2328       This change also means that all "special" arguments of the form
2329       @SMLxxx... must come first.
2330    
2331    3. Changed install script to put relative anchor names for tool commands
2332       into pathconfig.
2333    
2334    ----------------------------------------------------------------------
2335    Name: Matthias Blume
2336    Date: 2001/09/18 15:35:00 EDT
2337    Tag: blume-20010918-readme11036
2338    Description:
2339    
2340    Added README files.
2341    
2342    ----------------------------------------------------------------------
2343    Name: Matthias Blume
2344    Date: 2001/09/18 11:45:00 EDT
2345    Tag: Release_110_36 (retag)
2346    Description:
2347    
2348    Fixed mistake in config/preloads. Retagged as 110.36.
2349    
2350    ----------------------------------------------------------------------
2351    Name: Matthias Blume
2352    Date: 2001/09/18 09:40:00 EDT
2353    Tag: Release_110_36_orig (tag changed)
2354    Description:
2355    
2356    New version (110.36).  New bootfiles.
2357    
2358    ----------------------------------------------------------------------
2359    Name: Matthias Blume
2360    Date: 2001/09/14 16:15:00 EDT
2361    Tag: blume-20010914-x86fastfp
2362    Description:
2363    
2364    John committed some changes that Allen made, in particular a (hopefully)
2365    correctly working version of the x86-fp module.
2366    
2367    I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
2368    flag to "true".  Everything seems to compile to a fixpoint ok, and
2369    "mandelbrot" speeds up by about 15%.
2370    
2371    ----------------------------------------------------------------------
2372    Name: Matthias Blume
2373    Date: 2001/09/13 11:20:00 EDT
2374    Tag: blume-20010913-minimal
2375    Description:
2376    
2377    1. Stefan Monnier's patch to fix a miscompilation problem that
2378       was brought to light by John Reppy's work on Moby.
2379    
2380    2. Implemented a minimal "structure Compiler" that contains just
2381       "version" and "architecture".  The minimal version will be
2382       available when the full version is not.  This is for backward-
2383       compatibility with code that wants to test Compiler.version.
2384    
2385    ----------------------------------------------------------------------
2386    Name: Matthias Blume
2387    Date: 2001/08/28 14:03:00 EDT
2388    Tag: blume-20010828-ml-lex
2389    Description:
2390    
2391    Fix for bug 1581, received from Neophytos Michael.
2392    
2393    ----------------------------------------------------------------------
2394    Name: Matthias Blume
2395    Date: 2001/08/27 11:20:00 EDT
2396    Tag: blume-20010827-readme11035
2397    Description:
2398    
2399    Fleshed out the README file for 110.35.
2400    
2401    ----------------------------------------------------------------------
2402  Name: Matthias Blume  Name: Matthias Blume
2403  Date: 2001/08/24 17:10:00 EDT  Date: 2001/08/24 17:10:00 EDT
2404  Tag: Release_110_35  Tag: Release_110_35
# Line 2923  Line 5309 
5309       elaborator). There were a lot of changes during my "linkpath" trials       elaborator). There were a lot of changes during my "linkpath" trials
5310       that could have been reverted to their original state but weren't.       that could have been reverted to their original state but weren't.
5311       Please, don't be too harsh on me for messing with this code a bit more       Please, don't be too harsh on me for messing with this code a bit more
5312       than what was strictly necessary...  (I _did_ resist the tempation       than what was strictly necessary...  (I _did_ resist the temptation
5313       of doing any "global reformatting" to avoid an untimely death at       of doing any "global reformatting" to avoid an untimely death at
5314       Dave's hands. :)       Dave's hands. :)
5315    

Legend:
Removed from v.916  
changed lines
  Added in v.1335

root@smlnj-gforge.cs.uchicago.edu
ViewVC Help
Powered by ViewVC 1.0.0