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 652, Tue Jun 6 02:14:56 2000 UTC revision 1291, Sat Dec 7 03:41:00 2002 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
17    Date: 2002/12/06 22:40:00 EST
18    Tag: blume-20021206-cm-fileid
19    Description:
20    
21    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
22    This fixes the problem of CM getting confused by files that suddenly
23    change their identity (e.g., by getting unlinked and recreated by some
24    text editor such as vi).  There might be a better/cheaper/cleaner way
25    of doing this, but for now this will have to do.
26    
27    ----------------------------------------------------------------------
28    Name: Matthias Blume
29    Date: 2002/10/28 09:50:00 EST
30    Tag: blume-20021028-typecheck
31    Description:
32    
33    Exported structure Typecheck from $smlnj/viscomp/core.cm.
34    
35    ----------------------------------------------------------------------
36    Name: Matthias Blume
37    Date: 2002/10/17 09:10:00 EDT
38    Tag: Release_110_42
39    Description:
40    
41    In good old tradition, there has been a slight hiccup so that we have
42    to patch 110.42 after the fact.   The old release tag has been replaced
43    (see below).
44    
45    The change solves a problem with two competing approaches the
46    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
47    each other's way.
48    
49    This change only affects the runtime system code and the installer script.
50    (No new bootfiles.)
51    
52    ----------------------------------------------------------------------
53    Name: Matthias Blume
54    Date: 2002/10/16 12:00:00 EDT
55    Tag: Release_110_42_removed
56    Description:
57    
58    New working release.  New bootfiles.
59    
60    ----------------------------------------------------------------------
61    Name: Matthias Blume
62    Date: 2002/10/10 13:10:00 EDT
63    Tag: blume-20021010-ppc-divs
64    Description:
65    
66    The mltree operator DIVS must be implemented with an overflow check on
67    the PPC because the hardware indicates divide-by-zero using "overflow" as
68    well.
69    
70    ----------------------------------------------------------------------
71    Name: Matthias Blume
72    Date: 2002/07/23 11:45:00
73    Tag: blume-20020723-smlnj-home
74    Description:
75    
76    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
77    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
78    CM_PATHCONFIG is also set), the path configuration file is assumed
79    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
80    the entire tree to some other place and everything will "just work".
81    
82    (Companion commands such as ml-build and ml-makedepend also sense this
83    variable.)
84    
85    ----------------------------------------------------------------------
86    Name: Matthias Blume
87    Date: 2002/07/12 21:19:00 EDT
88    Tag: blume-20020712-liveness
89    Description:
90    
91    Exported two useful "step" functions from liveness module (MLRISC).
92    
93    ----------------------------------------------------------------------
94    Name: Matthias Blume
95    Date: 2002/07/05 16:00 EDT
96    Tag: Release_110_41
97    Description:
98    
99    New working release.  New bootfiles.
100    
101    ----------------------------------------------------------------------
102    Name: Matthias Blume
103    Date: 2002/07/05 10:25:00 EDT
104    Tag: blume-20020705-btimp
105    Description:
106    
107    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
108    other clients can set up backtracing support.
109    
110    ----------------------------------------------------------------------
111    Name: Matthias Blume
112    Date: 2002/06/25 17:23:00 EDT
113    Tag: blume-20020625-fpmax
114    Description:
115    
116    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
117    case.  (The sense of the isNaN test was reversed -- which made min and
118    max always return their first argument.)
119    
120    ----------------------------------------------------------------------
121    Name: Matthias Blume
122    Date: 2002/06/11
123    Tag: blume-20020611-unixpath
124    Description:
125    
126    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
127    
128    ----------------------------------------------------------------------
129    Name: Matthias Blume
130    Date: 2002/06/10 16:35:00 EDT
131    Tag: blume-20020610-ieeereal
132    Description:
133    
134    I back-ported my implementation of IEEEReal.fromString from the
135    idlbasis-devel branch so that we can test it.
136    
137    Another small change is that ppDec tries to give more information
138    than just "<sig>" in the case of functors.  However, this code is
139    broken in some mysterious way if the functor's body's signature
140    has not been declared by ascription but gets inferred from the
141    implementation.  This needs fixing...
142    
143    ----------------------------------------------------------------------
144    Name: Matthias Blume
145    Date: 2002/05/31
146    Tag: blume-20020531-btrace-mode
147    Description:
148    
149    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
150    the wayside when I switched over to using Controls everywhere.)
151    
152    ----------------------------------------------------------------------
153    Name: Lal George
154    Date: 2002/05/23 12:21:40 EDT
155    Tag: george-20020523-visual-labels
156    Description:
157    
158    Labels are now displayed in the graphical output to make
159    the fall-through and target blocks obvious.
160    
161    ----------------------------------------------------------------------
162    Name: Matthias Blume
163    Date: 2002/05/22 11:03:00 EDT
164    Tag: blume-20020522-shrink
165    Description:
166    
167    John tweaked yesterday's fix for 1131 to handle an out-of-memory
168    situation that comes up when allocating huge arrays.
169    
170    ----------------------------------------------------------------------
171    Name: Matthias Blume
172    Date: 2002/05/21 16:00:00 EDT
173    Tag: Release_110_40
174    Description:
175    
176    New working release (110.40).  New bootfiles.
177    
178    [Also: John Reppy fixed GC bug 1131.]
179    
180    ----------------------------------------------------------------------
181    Name: Matthias Blume
182    Date: 2002/05/21 12:35:00 EDT
183    Tag: blume-20020521-cmdoc
184    Description:
185    
186    CM documentation update.
187    
188    ----------------------------------------------------------------------
189    Name: Matthias Blume
190    Date: 2002/05/21 10:55:00 EDT
191    Tag: blume-20020521-misc
192    Description:
193    
194    - John tweaked runtime to be silent on heap export (except when
195      GC messages are on).
196    - I added a few more things (cross-compiling versions of CMB) to
197      config/preloads (as suggestions).
198    
199    ----------------------------------------------------------------------
200    Name: Matthias Blume
201    Date: 2002/05/20 22:25:00 EDT
202    Tag: blume-20020520-controls
203    Description:
204    
205    - Added ControlUtil structure to control-lib.cm.
206    - Use it throughout.
207    - Used Controls facility to define MLRISC controls (as opposed to
208      registering MLRISC control ref cells with Controls after the fact)
209    - Fixed messed-up controls priorities.
210    
211    * Removed again all the stuff from config/preloads that one wouldn't
212      be able to preload at the time the initial heap image is built.
213      (Many libraries, e.g., CML, do not exist yet at this time.  The
214       only libraries that can be preloaded via config/preloads are those
215       that come bundled with the bootfiles.)
216    
217    ----------------------------------------------------------------------
218    Name: Matthias Blume
219    Date: 2002/05/20 10:59:00 EDT
220    Tag: blume-20020520-preloads
221    Description:
222    
223    Added a lot of commented-out suggestions for things to be included
224    in config/preloads.
225    
226    ----------------------------------------------------------------------
227    Name: Allen Leung
228    Date: 2002/05/18 14:20:00 EDT
229    Tag: leunga-20020518-mdl
230    Description:
231    
232    o Made the mdl tool stuff compile and run again.
233    o I've disabled all the stuff that depends on RTL specifications; they
234      are all badly broken anyway.
235    
236    ----------------------------------------------------------------------
237    Name: Matthias Blume
238    Date: 2002/05/17 16:49:00 EDT
239    Tag: blume-20020517-controls
240    Description:
241    
242    0. John Reppy made several modifications to the SML/NJ library.
243       In particular, there is a shiny new controls-lib.cm.
244    
245    1. Pushed new controls interface through compiler so that everything
246       compiles again.
247    
248    2. Added FormatComb and FORMAT_COMB to the CML version of the
249       SML/NJ library (so that CML compiles again).
250    
251    3. Modified init scripts because XXX_DEFAULT environment variables
252       are no longer with us.  (Boot-time initialization is now done
253       using the same environment variables that are also used for
254       startup-time initialization of controls.)
255    
256    ----------------------------------------------------------------------
257    Name: Lal George
258    Date: 2002/05/15 09:20:10 EDT
259    Tag: george-20020515-pseudo-op-decls
260    Description:
261    
262            All pseudo-ops emitted before the first segment declaration
263    such as TEXT, DATA, and BSS directives are assumed to be global
264    declarations and are emitted first in the assembly file. This is
265    useful in a number of situations where one has pseudo-ops that are not
266    specific to any segment, and also works around the constraint that one
267    cannot have client pseudo-ops in the TEXT segment.
268    
269            Because no segment is associated with these declarations it is
270    an error to allocate any space or objects before the first segment
271    directive and an exception will be raised. However, we cannot make
272    this check for client pseudo-ops.
273    
274            These top level declarations are a field in the CFG graph_info.
275    In theory you can continue to add to this field after the CFG has been
276    built -- provided you know what you are doing;-)
277    
278    ----------------------------------------------------------------------
279    Name: Matthias Blume
280    Date: 2002/05/13 16:40:00 EDT
281    Tag: blume-20020513-pp-etc
282    Description:
283    
284    A few minor bugfixes:
285    
286      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
287        (Bogus printouts for redefined bindings still occur.  Compiler
288        bug should no longer occur now.  We need to redo the prettyprinter
289        from scratch.)
290    
291      - CM pathname printer now also adds escape sequences for ( and )
292    
293      - commend and docu fixes for ml-nlffi
294    
295    ----------------------------------------------------------------------
296    Name: Matthias Blume
297    Date: 2002/05/10 16:40:00 EDT
298    Tag: blume-20020510-erg-textio
299    Description:
300    
301    Applied the following bugfix provided by Emden Gansner:
302    
303        Output is corrupted when outputSubstr is used rather than output.
304    
305        The problem occurs when a substring
306    
307            ss = (s, dataStart, dataLen)
308    
309        where dataStart > 0, fills a stream buffer with avail bytes left.
310        avail bytes of s, starting at index dataStart, are copied into the
311        buffer, the buffer is flushed, and then the remaining dataLen-avail
312        bytes of ss are copied into the beginning of the buffer. Instead of
313        starting this copy at index dataStart+avail in s, the current code
314        starts the copy at index avail.
315    
316        Fix:
317        In text-io-fn.sml, change line 695 from
318             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
319        to
320             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
321    
322    ----------------------------------------------------------------------
323    Name: Matthias Blume
324    Date: 2002/04/12 13:55:00 EDT
325    Tag: blume-20020412-assyntax
326    Description:
327    
328    1. Grabbed newer assyntax.h from the XFree86 project.
329    2. Fiddled with how to compile X86.prim.asm without warnings.
330    3. (Very) Minor cleanup in CM.
331    
332    ----------------------------------------------------------------------
333    Name: Matthias Blume
334    Date: 2002/04/01 (no joke!) 17:07:00 EST
335    Tag: blume-20020401-x86div
336    Description:
337    
338    Added full support for div/mod/rem/quot on the x86, using the machine
339    instruction's two results (without clumsily recomputing the remainder)
340    directly where appropriate.
341    
342    Some more extensive power-of-two support was added to the x86 instruction
343    selector (avoiding expensive divs, mods, and muls where they can be
344    replaced with cheaper shifts and masks).  However, this sort of thing
345    ought to be done earlier, e.g., within the CPS optimizer so that
346    all architectures benefit from it.
347    
348    The compiler compiles to a fixed point, but changes might be somewhat
349    fragile nevertheless.  Please, report any strange things that you might
350    see wrt. div/mod/quot/rem...
351    
352    ----------------------------------------------------------------------
353    Name: Matthias Blume
354    Date: 2002/03/29 17:22:00
355    Tag: blume-20020329-div
356    Description:
357    
358    Fixed my broken div/mod logic.  Unfortunately, this means that the
359    inline code for div/mod now has one more comparison than before.
360    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
361    The problem was with quotient = 0, because that alone does not tell
362    us which way the rounding went.  One then has to look at whether
363    remainder and divisor have the same sign...  :(
364    
365    Anyway, I replaced the bootfiles with fresh ones...
366    
367    ----------------------------------------------------------------------
368    Name: Matthias Blume
369    Date: 2002/03/29 14:10:00 EST
370    Tag: blume-20020329-inlprims
371    Description:
372    
373    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
374    
375    Primops have changed. This means that the bin/boot-file formats have
376    changed as well.
377    
378    To make sure that there is no confusion, I made a new version.
379    
380    
381    CHANGES:
382    
383    * removed REMT from mltree (remainder should never overflow).
384    
385    * added primops to deal with divisions of all flavors to the frontend
386    
387    * handled these primops all the way through so they map to their respective
388      MLRISC support
389    
390    * used these primops in the implementation of Int, Int32, Word, Word32
391    
392    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
393    
394    * parameterized INLMIN, INLMAX, and INLABS by a numkind
395    
396    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
397      floating point
398    
399    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
400      and Real (but Real.abs maps to a separate floating-point-only primop)
401    
402    
403    TODO items:
404    
405    * Hacked Alpha32 instruction selection, disabling the selection of REMx
406      instructions because the machine instruction encoder cannot handle
407      them.  (Hppa, PPC, and Sparc instruction selection did not handle
408      REM in the first place, and REM is supported by the x86 machine coder.)
409    
410    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
411      selection phase.  (The two can be streamlined because the hardware
412      delivers both quotient and remainder at the same time anyway.)
413    
414    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
415      (Currently the behavior is inconsistent both across architectures and
416      wrt. the draft Basis spec.)
417    
418    * Word8 should eventually be handled natively, too.
419    
420    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
421      as if there currently is no execution path that could trigger it in
422      SML/NJ.  (The assumptions underlying functions arith and promotable do not
423      hold for things like multiplication and division.)
424    
425    ----------------------------------------------------------------------
426    Name: Matthias Blume
427    Date: 2002/03/27 16:27:00 EST
428    Tag: blume-20020327-mlrisc-divisions
429    Description:
430    
431    Added support for all four division operations (ML's div, mod, quot,
432    and rem) to MLRISC.  In the course of doing so, I also rationalized
433    the naming (no more annoying switch-around of DIV and QUOT), by
434    parameterizing the operation by div_rounding_mode (which can be either
435    DIV_TO_ZERO or DIV_TO_NEGINF).
436    
437    The generic MLTreeGen functor takes care of compiling all four
438    operations down to only round-to-zero div.
439    
440    Missing pieces:
441    
442      * Doing something smarter than relying on MLTreeGen on architectures
443        like, e.g., the x86 where hardware division delivers both quotient and
444        remainder at the same time.  With this, the implementation of the
445        round-to-neginf operations could be further streamlined.
446    
447      * Remove inlining support for div/mod/rem from the frontend and replace it
448        with primops that get carried through to the backend.  Do this for all
449        int and word types.
450    
451    ----------------------------------------------------------------------
452    Name: Matthias Blume
453    Date: 2002/03/25 17:25:00 EST
454    Tag: blume-20020325-divmod
455    Description:
456    
457    I improved (hopefully without breaking them) the implementation of Int.div,
458    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
459    advantage of the following observations:
460    
461      Let  q = x quot y      r = x rem y
462           d = x div  y      m = x mod y
463    
464    where "quot" is the round-to-zero version of integer division that
465    hardware usually provides.  Then we have:
466    
467         r = x - q * y        where neither the * nor the - will overflow
468         d = if q >= 0 orelse x = q * y then q else q - 1
469                              where neither the * nor the - will overflow
470         m = if q >= 0 orelse r = 0 then r else r + y
471                              where the + will not overflow
472    
473    This results in substantial simplification of the generated code.
474    The following table shows the number of CFG nodes and edges generated
475    for
476            fun f (x, y) = x OPER y
477            (* with OPER \in div, mod, quot, rem *)
478    
479    
480        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
481        --------------------------------------------------------
482         div |         24 |         39 |         12 |         16
483         mod |         41 |         71 |         12 |         16
484        quot |          8 |         10 |          8 |         10
485         rem |         10 |         14 |          8 |         10
486    
487    
488    ----------------------------------------------------------------------
489    Name: Matthias Blume
490    Date: 2002/03/25 22:06:00 EST
491    Tag: blume-20020325-cprotobug
492    Description:
493    
494    Fixed a bug in cproto (c prototype decoder).
495    
496    ----------------------------------------------------------------------
497    Name: Matthias Blume
498    Date: 2002/03/25 16:00:00 EST
499    Tag: blume-20020325-raw-primops
500    Description:
501    
502    I did some cleanup to Allen's new primop code and
503    replaced yesterday's bootfiles with new ones.
504    (But they are stored in the same place.)
505    
506    ----------------------------------------------------------------------
507    Name: Matthias Blume
508    Date: 2002/03/24 22:40:00 EST
509    Tag: blume-20020324-bootfiles
510    Description:
511    
512    Made the bootfiles that Allen asked for.
513    
514    ----------------------------------------------------------------------
515    Name: Allen Leung
516    Date: 2002/03/23 15:50:00 EST
517    Tag: leunga-20020323-flint-cps-rcc-primops
518    Description:
519    
520      1. Changes to FLINT primops:
521    
522        (* make a call to a C-function;
523         * The primop carries C function prototype information and specifies
524         * which of its (ML-) arguments are floating point. C prototype
525         * information is for use by the backend, ML information is for
526         * use by the CPS converter. *)
527      | RAW_CCALL of { c_proto: CTypes.c_proto,
528                       ml_args: ccall_type list,
529                       ml_res_opt: ccall_type option,
530                       reentrant : bool
531                     } option
532       (* Allocate uninitialized storage on the heap.
533        * The record is meant to hold short-lived C objects, i.e., they
534        * are not ML pointers.  With the tag, the representation is
535        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
536        *)
537      | RAW_RECORD of {tag:bool,sz:int}
538      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
539    
540      2.  These CPS primops are now overloaded:
541    
542           rawload of {kind:numkind}
543           rawstore of {kind:numkind}
544    
545          The one argument form is:
546    
547             rawload {kind} address
548    
549          The two argument form is:
550    
551             rawload {kind} [ml object, byte-offset]
552    
553      3. RAW_CCALL/RCC now takes two extra arguments:
554    
555         a. The first is whether the C call is reentrant, i.e., whether
556            ML state should be saved and restored.
557         b. The second argument is a string argument specifying the name of
558            library and the C function.
559    
560         These things are currently not handled in the code generator, yet.
561    
562      4. In CProto,
563    
564         An encoding type of "bool" means "ml object" and is mapped into
565         C prototype of PTR.  Note that "bool" is different than "string",
566         even though "string" is also mapped into PTR, because "bool"
567         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
568    
569      5. Pickler/unpicker
570    
571         Changed to handle RAW_RECORD and newest RAW_CCALL
572    
573      6. MLRiscGen,
574    
575         1. Changed to handle the new rawload/rawstore/rawrecord operators.
576         2. Code for handling C Calls has been moved to a new module CPSCCalls,
577            in the file CodeGen/cpscompile/cps-c-calls.sml
578    
579      7. Added the conditional move operator
580    
581             condmove of branch
582    
583         to cps.  Generation of this is still buggy so it is currently
584         disabled.
585    
586    ----------------------------------------------------------------------
587    Name: Lal George
588    Date: 2002/03/22 14:18:25 EST
589    Tag: george-20020322-cps-branch-prob
590    Description:
591    
592    Implemented the Ball-Larus branch prediction-heuristics, and
593    incorporated graphical viewers for control flow graphs.
594    
595    Ball-Larus Heuristics:
596    ---------------------
597    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
598    
599    By design it uses the Dempster-Shafer theory for combining
600    probabilities.  For example, in the function:
601    
602        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
603    
604    the ball-larus heuristics predicts that the n=0 is unlikely
605    (OH-heuristic), and the 'then' branch is unlikely because of the
606    RH-heuristic -- giving the 'then' branch an even lower combined
607    probability using the Dempster-Shafer theory.
608    
609    Finally, John Reppy's loop analysis in MLRISC, further lowers the
610    probability of the 'then' branch because of the loop in the else
611    branch.
612    
613    
614    Graphical Viewing:
615    ------------------
616    I merely plugged in Allen's graphical viewers into the compiler. The
617    additional code is not much. At the top level, saying:
618    
619            Control.MLRISC.getFlag "cfg-graphical-view" := true;
620    
621    will display the graphical view of the control flow graph just before
622    back-patching.  daVinci must be in your path for this to work. If
623    daVinci is not available, then the default viewer can be changed
624    using:
625    
626            Control.MLRISC.getString "viewer"
627    
628    which can be set to "dot" or "vcg" for the corresponding viewers. Of
629    course, these viewers must be in your path.
630    
631    The above will display the compilation unit at the level of clusters,
632    many of which are small, boring, and un-interesting. Also setting:
633    
634            Control.MLRISC.getInt "cfg-graphical-view_size"
635    
636    will display clusters that are larger than the value set by the above.
637    
638    
639    ----------------------------------------------------------------------
640    Name: Matthias Blume
641    Date: 2002/03/21 22:20:00 EST
642    Tag: blume-20020321-kmp-bugfix
643    Description:
644    
645    Changed the interface to the KMP routine in PreString and fixed
646    a minor bug in one place where it was used.
647    
648    ----------------------------------------------------------------------
649    Name: Allen Leung
650    Date: 2002/03/21 20:30:00 EST
651    Tag: leunga-20020321-cfg
652    Description:
653    
654      Fixed a potential problem in cfg edge splitting.
655    
656    ----------------------------------------------------------------------
657    Name: Allen Leung
658    Date: 2002/03/21 17:15:00 EST
659    Tag: leunga-20020321-x86-fp-cfg
660    Description:
661    
662      1. Recoded the buggy parts of x86-fp.
663    
664         a. All the block reordering code has been removed.
665            We now depend on the block placement phases to do this work.
666    
667         b. Critical edge splitting code has been simplified and moved into the
668            CFG modules, as where they belong.
669    
670         Both of these were quite buggy and complex.  The code is now much, much
671         simpler.
672    
673      2. X86 backend.
674    
675         a. Added instructions for 64-bit support.  Instruction selection for
676            64-bit has not been committed, however, since that
677            requires changes to MLTREE which haven't been approved by
678            Lal and John.
679    
680         b. Added support for FUCOMI and FUCOMIP when generating code for
681            PentiumPro and above.  We only generate these instructions in
682            the fast-fp mode.
683    
684         c. Added cases for JP and JNP in X86FreqProps.
685    
686      3. CFG
687    
688         CFG now has a bunch of methods for edge splitting and merging.
689    
690      4. Machine description.
691    
692         John's simplification of MLTREE_BASIS.fcond broke a few machine
693         description things:
694    
695         rtl-build.{sig,sml} and hppa.mdl fixed.
696    
697         NOTE: the machine description stuff in the repository is still broken.
698               Again, I can't put my fixes in because that involves
699               changes to MLTREE.
700    
701    ----------------------------------------------------------------------
702    Name: Matthias Blume
703    Date: 2002/03/20 15:55:00 EST
704    Tag: blume-20020320-kmp
705    Description:
706    
707    Implemented Knuth-Morris-Pratt string matching in PreString and used
708    it for String.isSubstring, Substring.isSubstring, and
709    Substring.position.
710    
711    (Might need some stress-testing.  Simple examples worked fine.)
712    
713    ----------------------------------------------------------------------
714    Name: Matthias Blume
715    Date: 2002/03/19 16:37:00 EST
716    Tag: blume-20020319-witnesses
717    Description:
718    
719    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
720    
721    This implements a generic mechanism for changing constness qualifiers
722    anywhere within big C types without resorting to outright "casts".
723    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
724    modify the constness at the outermost level.)
725    The implementation of "convert" is based on the idea of "witness"
726    values -- values that are not used by the operation but whose types
727    "testify" to their applicability.  On the implementation side, "convert"
728    is simply a projection (returning its second curried argument).  With
729    cross-module inlining, it should not result in any machine code being
730    generated.
731    
732    ----------------------------------------------------------------------
733    Name: Matthias Blume
734    Date: 2002/03/15 16:40:00 EST
735    Tag: blume-20020315-basis
736    Description:
737    
738    Provided (preliminary?) implementations for
739    
740      {String,Substring}.{concatWith,isSuffix,isSubstring}
741    
742    and
743    
744      Substring.full
745    
746    Those are in the Basis spec but they were missing in SML/NJ.
747    
748    ----------------------------------------------------------------------
749    Name: Matthias Blume
750    Date: 2002/03/14 21:30:00 EST
751    Tag: blume-20020314-controls
752    Description:
753    
754    Controls:
755    ---------
756    
757    1. Factored out the recently-added Controls : CONTROLS stuff and put
758       it into its own library $/controls-lib.cm.  The source tree for
759       this is under src/smlnj-lib/Controls.
760    
761    2. Changed the names of types and functions in this interface, so they
762       make a bit more "sense":
763    
764          module -> registry
765          'a registry -> 'a group
766    
767    3. The interface now deals in ref cells only.  The getter/setter interface
768       is (mostly) gone.
769    
770    4. Added a function that lets one register an already-existing ref cell.
771    
772    5. Made the corresponding modifications to the rest of the code so that
773       everything compiles again.
774    
775    6. Changed the implementation of Controls.MLRISC back to something closer
776       to the original.  In particular, this module (and therefore MLRISC)
777       does not depend on Controls.  There now is some link-time code in
778       int-sys.sml that registers the MLRISC controls with the Controls
779       module.
780    
781    CM:
782    ---
783    
784      * One can now specify the lambda-split aggressiveness in init.cmi.
785    
786    ----------------------------------------------------------------------
787    Name: Allen Leung
788    Date: 2002/03/13 17:30:00 EST
789    Tag: leunga-20020313-x86-fp-unary
790    Description:
791    
792    Bug fix for:
793    
794    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
795    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
796    > - fun f(x,(y,z)) = Real.~ y;
797    > [autoloading]
798    > [autoloading done]
799    >       fchsl   (%eax), 184(%esp)
800    > Error: MLRisc bug: X86MCEmitter.emitInstr
801    >
802    > uncaught exception Error
803    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
804    
805    The problem was that the code generator did not generate any fp registers
806    in this case, and the ra didn't know that it needed to run the X86FP phase to
807    translate the pseudo fp instruction.   This only happened with unary fp
808    operators in certain situations.
809    
810    ----------------------------------------------------------------------
811    Name: Matthias Blume
812    Date: 2002/03/13 14:00:00 EST
813    Tag: blume-20020313-overload-etc
814    Description:
815    
816    1. Added _overload as a synonym for overload for backward compatibility.
817       (Control.overloadKW must be true for either version to be accepted.)
818    
819    2. Fixed bug in install script that caused more things to be installed
820       than what was requested in config/targets.
821    
822    3. Made CM aware of the (_)overload construct so that autoloading
823       works.
824    
825    ----------------------------------------------------------------------
826    Name: Matthias Blume
827    Date: 2002/03/12 22:03:00 EST
828    Tag: blume-20020312-url
829    Description:
830    
831    Forgot to update BOOT and srcarchiveurl.
832    
833    ----------------------------------------------------------------------
834    Name: Matthias Blume
835    Date: 2002/03/12 17:30:00 EST
836    Tag: blume-20020312-version110392
837    Description:
838    
839    Yet another version number bump (because of small changes to the
840    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
841    
842    Changes:
843    
844      The new pid generation scheme described a few weeks ago was overly
845      complicated.  I implemented a new mechanism that is simpler and
846      provides a bit more "stability":  Once CM has seen a compilation
847      unit, it keeps its identity constant (as long as you do not delete
848      those crucial CM/GUID/* files).  This means that when you change
849      an interface, compile, then go back to the old interface, and
850      compile again, you arrive at the original pid.
851    
852      There now also is a mechanism that instructs CM to use the plain
853      environment hash as a module's pid (effectively making its GUID
854      the empty string).  For this, "noguid" must be specified as an
855      option to the .sml file in question within its .cm file.
856      This is most useful for code that is being generated by tools such
857      as ml-nlffigen (because during development programmers tend to
858      erase the tool's entire output directory tree including CM's cached
859      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
860      revert to the old, broken behavior of SML/NJ, but in specific cases
861      where there is no danger of interface confusion, its use is ok
862      (I think).
863    
864      ml-nlffigen by default generates "noguid" annotations.  They can be
865      turned off by specifying -guid in its command line.
866    
867    ----------------------------------------------------------------------
868    Name: Lal George
869    Date: 2002/03/12 12 14:42:36 EST
870    Tag: george-20020312-frequency-computation
871    Description:
872    
873    Integrated jump chaining and static block frequency into the
874    compiler. More details and numbers later.
875    
876    ----------------------------------------------------------------------
877    Name: Lal George
878    Date: 2002/03/11 11 22:38:53 EST
879    Tag: george-20020311-jump-chain-elim
880    Description:
881    
882    Tested the jump chain elimination on all architectures (except the
883    hppa).  This is on by default right now and is profitable for the
884    alpha and x86, however, it may not be profitable for the sparc and ppc
885    when compiling the compiler.
886    
887    The gc test will typically jump to a label at the end of the cluster,
888    where there is another jump to an external cluster containing the actual
889    code to invoke gc. This is to allow factoring of common gc invocation
890    sequences. That is to say, we generate:
891    
892            f:
893               testgc
894               ja   L1      % jump if above to L1
895    
896            L1:
897               jmp L2
898    
899    
900    After jump chain elimination the 'ja L1' instructions is converted to
901    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
902    up being implemented in their long form (if L2 is far away) using:
903    
904            jbe     L3      % jump if below or equal to L3
905            jmp     L2
906         L3:
907            ...
908    
909    
910    For large compilation units L2  may be far away.
911    
912    
913    ----------------------------------------------------------------------
914    Name: Matthias Blume
915    Date: 2002/03/11 13:30:00 EST
916    Tag: blume-20020311-mltreeeval
917    Description:
918    
919    A functor parameter was missing.
920    
921    ----------------------------------------------------------------------
922    Name: Allen Leung
923    Date: 2002/03/11 10:30:00 EST
924    Tag: leunga-20020311-runtime-string0
925    Description:
926    
927       The representation of the empty string now points to a
928    legal null terminated C string instead of unit.  It is now possible
929    to convert an ML string into C string with InlineT.CharVector.getData.
930    This compiles into one single machine instruction.
931    
932    ----------------------------------------------------------------------
933    Name: Allen Leung
934    Date: 2002/03/10 23:55:00 EST
935    Tag: leunga-20020310-x86-call
936    Description:
937    
938       Added machine generation for CALL instruction (relative displacement mode)
939    
940    ----------------------------------------------------------------------
941    Name: Matthias Blume
942    Date: 2002/03/08 16:05:00
943    Tag: blume-20020308-entrypoints
944    Description:
945    
946    Version number bumped to 110.39.1.  NEW BOOTFILES!
947    
948    Entrypoints: non-zero offset into a code object where execution should begin.
949    
950    - Added the notion of an entrypoint to CodeObj.
951    - Added reading/writing of entrypoint info to Binfile.
952    - Made runtime system bootloader aware of entrypoints.
953    - Use the address of the label of the first function given to mlriscGen
954      as the entrypoint.  This address is currently always 0, but it will
955      not be 0 once we turn on block placement.
956    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
957      with entry points) from mlriscGen.
958    
959    ----------------------------------------------------------------------
960    Name: Allen Leung
961    Date: 2002/03/07 20:45:00 EST
962    Tag: leunga-20020307-x86-cmov
963    Description:
964    
965       Bug fixes for CMOVcc on x86.
966    
967       1. Added machine code generation for CMOVcc
968       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
969       3. CMOVcc cannot have an immediate operand as argument.
970    
971    ----------------------------------------------------------------------
972    Name: Matthias Blume
973    Date: 2002/03/07 16:15:00 EST
974    Tag: blume-20020307-controls
975    Description:
976    
977    This is a very large but mostly boring patch which makes (almost)
978    every tuneable compiler knob (i.e., pretty much everything under
979    Control.* plus a few other things) configurable via both the command
980    line and environment variables in the style CM did its configuration
981    until now.
982    
983    Try starting sml with '-h' (or, if you are brave, '-H')
984    
985    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
986    implements the underlying generic mechanism.
987    
988    The interface to some of the existing such facilities has changed somewhat.
989    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
990    (The getFoo interface is still there for backward-compatibility, but its
991    use is deprecated.)
992    
993    The ml-build script passes -Cxxx=yyy command-line arguments through so
994    that one can now twiddle the compiler settings when using this "batch"
995    compiler.
996    
997    TODO items:
998    
999    We should go through and throw out all controls that are no longer
1000    connected to anything.  Moreover, we should go through and provide
1001    meaningful (and correct!) documentation strings for those controls
1002    that still are connected.
1003    
1004    Currently, multiple calls to Controls.new are accepted (only the first
1005    has any effect).  Eventually we should make sure that every control
1006    is being made (via Controls.new) exactly once.  Future access can then
1007    be done using Controls.acc.
1008    
1009    Finally, it would probably be a good idea to use the getter-setter
1010    interface to controls rather than ref cells.  For the time being, both
1011    styles are provided by the Controls module, but getter-setter pairs are
1012    better if thread-safety is of any concern because they can be wrapped.
1013    
1014    *****************************************
1015    
1016    One bug fix: The function blockPlacement in three of the MLRISC
1017    backpatch files used to be hard-wired to one of two possibilities at
1018    link time (according to the value of the placementFlag).  But (I
1019    think) it should rather sense the flag every time.
1020    
1021    *****************************************
1022    
1023    Other assorted changes (by other people who did not supply a HISTORY entry):
1024    
1025    1. the cross-module inliner now works much better (Monnier)
1026    2. representation of weights, frequencies, and probabilities in MLRISC
1027       changed in preparation of using those for weighted block placement
1028       (Reppy, George)
1029    
1030    ----------------------------------------------------------------------
1031    Name: Lal George
1032    Date: 2002/03/07 14:44:24 EST 2002
1033    Tag: george-20020307-weighted-block-placement
1034    
1035    Tested the weighted block placement optimization on all architectures
1036    (except the hppa) using AMPL to generate the block and edge frequencies.
1037    Changes were required in the machine properties to correctly
1038    categorize trap instructions. There is an MLRISC flag
1039    "weighted-block-placement" that can be used to enable weighted block
1040    placement, but this will be ineffective without block/edge
1041    frequencies (coming soon).
1042    
1043    
1044    ----------------------------------------------------------------------
1045    Name: Lal George
1046    Date: 2002/03/05 17:24:48 EST
1047    Tag: george-20020305-linkage-cluster
1048    
1049    In order to support the block placement optimization, a new cluster
1050    is generated as the very first cluster (called the linkage cluster).
1051    It contains a single jump to the 'real' entry point for the compilation
1052    unit. Block placement has no effect on the linkage cluster itself, but
1053    all the other clusters  have full freedom in the manner in which they
1054    reorder blocks or functions.
1055    
1056    On the x86 the typical linkage code that is generated is:
1057       ----------------------
1058            .align 2
1059       L0:
1060            addl    $L1-L0, 72(%esp)
1061            jmp     L1
1062    
1063    
1064            .align  2
1065       L1:
1066       ----------------------
1067    
1068    72(%esp) is the memory location for the stdlink register. This
1069    must contain the address of the CPS function being called. In the
1070    above example, it contains the address of  L0; before
1071    calling L1 (the real entry point for the compilation unit), it
1072    must contain the address for L1, and hence
1073    
1074            addl $L1-L0, 72(%esp)
1075    
1076    I have tested this on all architectures except the hppa.The increase
1077    in code size is of course negligible
1078    
1079    ----------------------------------------------------------------------
1080    Name: Allen Leung
1081    Date: 2002/03/03 13:20:00 EST
1082    Tag: leunga-20020303-mlrisc-tools
1083    
1084      Added #[ ... ] expressions to mlrisc tools
1085    
1086    ----------------------------------------------------------------------
1087    Name: Matthias Blume
1088    Date: 2002/02/27 12:29:00 EST
1089    Tag: blume-20020227-cdebug
1090    Description:
1091    
1092    - made types in structure C and C_Debug to be equal
1093    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1094    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1095    
1096    ----------------------------------------------------------------------
1097    Name: Matthias Blume
1098    Date: 2002/02/26 12:00:00 EST
1099    Tag: blume-20020226-ffi
1100    Description:
1101    
1102    1. Fixed a minor bug in CM's "noweb" tool:
1103       If numbering is turned off, then truly don't number (i.e., do not
1104       supply the -L option to noweb).  The previous behavior was to supply
1105       -L'' -- which caused noweb to use the "default" line numbering scheme.
1106       Thanks to Chris Richards for pointing this out (and supplying the fix).
1107    
1108    2. Once again, I reworked some aspects of the FFI:
1109    
1110       A. The incomplete/complete type business:
1111    
1112       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1113         gone!
1114       - ML types representing an incomplete type are now *equal* to
1115         ML types representing their corresponding complete types (just like
1116         in C).  This is still safe because ml-nlffigen will not generate
1117         RTTI for incomplete types, nor will it generate functions that
1118         require access to such RTTI.   But when ML code generated from both
1119         incomplete and complete versions of the C type meet, the ML types
1120         are trivially interoperable.
1121    
1122         NOTE:  These changes restore the full generality of the translation
1123         (which was previously lost when I eliminated functorization)!
1124    
1125       B. Enum types:
1126    
1127       - Structure C now has a type constructor "enum" that is similar to
1128         how the "su" constructor works.  However, "enum" is not a phantom
1129         type because each "T enum" has values (and is isomorphic to
1130         MLRep.Signed.int).
1131       - There are generic access operations for enum objects (using
1132         MLRep.Signed.int).
1133       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1134         * The structure contains the definition of type "mlrep" (the ML-side
1135         representation type of the enum).  Normally, mlrep is the same
1136         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1137         then mlrep will be defined as a datatype -- thus facilitating
1138         pattern matching on mlrep values.
1139         ("-ec" will be suppressed if there are duplicate values in an
1140          enumeration.)
1141         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1142         will be generated for each C enum constant xxx.
1143         * Conversion functions m2i and i2m convert between mlrep and
1144         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1145         * Coversion functions c and ml convert between mlrep and "tag enum".
1146         * Access functions (get/set) fetch and store mlrep values.
1147       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1148         enumerations are merged into one single enumeration represented by
1149         structure E_'.
1150    
1151    ----------------------------------------------------------------------
1152    Name: Allen Leung
1153    Date: 2002/02/25 04:45:00 EST
1154    Tag: leunga-20020225-cps-spill
1155    
1156    This is a new implementation of the CPS spill phase.
1157    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1158    In case of problems, replace it with the old file spill.sml
1159    
1160    The current compiler runs into some serious performance problems when
1161    constructing a large record.  This can happen when we try to compile a
1162    structure with many items.  Even a very simple structure like the following
1163    makes the compiler slow down.
1164    
1165        structure Foo = struct
1166           val x_1 = 0w1 : Word32.int
1167           val x_2 = 0w2 : Word32.int
1168           val x_3 = 0w3 : Word32.int
1169           ...
1170           val x_N = 0wN : Word32.int
1171        end
1172    
1173    The following table shows the compile time, from N=1000 to N=4000,
1174    with the old compiler:
1175    
1176    N
1177    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1178           MLRISC ra                               0.06u  0.00s  0.05g
1179              (spills = 0 reloads = 0)
1180           TOTAL                                   0.63u  0.07s  0.21g
1181    
1182    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1183           MLRISC ra                               5.68u  0.59s  3.93g
1184              (spills = 0 reloads = 0)
1185           TOTAL                                   14.71u  0.99s  4.81g
1186    
1187    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1188           MLRISC ra                               5.54u  0.65s  3.91g
1189              (spills = 543 reloads = 1082)
1190           TOTAL                                   65.40u  3.13s  6.00g
1191    
1192    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1193           MLRISC ra                               0.80u  0.10s  0.55g
1194              (spills = 42 reloads = 84)
1195           TOTAL                                   129.42u  5.10s  4.13g
1196    
1197    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1198           MLRISC ra                               2.69u  0.27s  1.38g
1199              (spills = 62 reloads = 124)
1200           TOTAL                                   682.48u  19.61s  13.99g
1201    
1202    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1203           MLRISC ra                               4.96u  0.27s  2.72g
1204              (spills = 85 reloads = 170)
1205           TOTAL                                   2375.26u  57.21s  48.00g
1206    
1207    As you can see the old cps spill module suffers from some serious
1208    performance problem.  But since I cannot decipher the old code fully,
1209    instead of patching the problems up, I'm reimplementing it
1210    with a different algorithm.  The new code is more modular,
1211    smaller when compiled, and substantially faster
1212    (O(n log n) time and O(n) space).  Timing of the new spill module:
1213    
1214    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1215          MLRISC ra                               0.25u  0.02s  0.15g
1216             (spills=1 reloads=3)
1217          TOTAL                                   7.74u  0.34s  1.62g
1218    
1219    Implementation details:
1220    
1221    As far as I can tell, the purpose of the CPS spill module is to make sure the
1222    number of live variables at any program point (the bandwidth)
1223    does not exceed a certain limit, which is determined by the
1224    size of the spill area.
1225    
1226    When the bandwidth is too large, we decrease the register pressure by
1227    packing live variables into spill records.  How we achieve this is
1228    completely different than what we did in the old code.
1229    
1230    First, there is something about the MLRiscGen code generator
1231    that we should be aware of:
1232    
1233    o MLRiscGen performs code motion!
1234    
1235       In particular, it will move floating point computations and
1236       address computations involving only the heap pointer to
1237       their use sites (if there is only a single use).
1238       What this means is that if we have a CPS record construction
1239       statement
1240    
1241           RECORD(k,vl,w,e)
1242    
1243       we should never count the new record address w as live if w
1244       has only one use (which is often the case).
1245    
1246       We should do something similar to floating point, but the transformation
1247       there is much more complex, so I won't deal with that.
1248    
1249    Secondly, there are now two new cps primops at our disposal:
1250    
1251     1. rawrecord of record_kind option
1252        This pure operator allocates some uninitialized storage from the heap.
1253        There are two forms:
1254    
1255         rawrecord NONE [INT n]  allocates a tagless record of length n
1256         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1257                                     and initializes the tag.
1258    
1259     2. rawupdate of cty
1260          rawupdate cty (v,i,x)
1261          Assigns to x to the ith component of record v.
1262          The storelist is not updated.
1263    
1264    We use these new primops for both spilling and increment record construction.
1265    
1266     1. Spilling.
1267    
1268        This is implemented with a linear scan algorithm (but generalized
1269        to trees).  The algorithm will create a single spill record at the
1270        beginning of the cps function and use rawupdate to spill to it,
1271        and SELECT or SELp to reload from it.  So both spills and reloads
1272        are fine-grain operations.  In contrast, in the old algorithm
1273        "spills" have to be bundled together in records.
1274    
1275        Ideally, we should sink the spill record construction to where
1276        it is needed.  We can even split the spill record into multiple ones
1277        at the places where they are needed.  But CPS is not a good
1278        representation for global code motion, so I'll keep it simple and
1279        am not attempting this.
1280    
1281     2. Incremental record construction (aka record splitting).
1282    
1283        Long records with many component values which are simulatenously live
1284        (recall that single use record addresses are not considered to
1285         be live) are constructed with rawrecord and rawupdate.
1286        We allocate space on the heap with rawrecord first, then gradually
1287        fill it in with rawupdate.  This is the technique suggested to me
1288        by Matthias.
1289    
1290        Some restrictions on when this is applicable:
1291        1. It is not a VECTOR record.  The code generator currently does not handle
1292           this case. VECTOR record uses double indirection like arrays.
1293        2. All the record component values are defined in the same "basic block"
1294           as the record constructor.  This is to prevent speculative
1295           record construction.
1296    
1297    ----------------------------------------------------------------------
1298    Name: Allen Leung
1299    Date: 2002/02/22 01:02:00 EST
1300    Tag: leunga-20020222-mlrisc-tools
1301    
1302    Minor bug fixes in the parser and rewriter
1303    
1304    ----------------------------------------------------------------------
1305    Name: Allen Leung
1306    Date: 2002/02/21 20:20:00 EST
1307    Tag: leunga-20020221-peephole
1308    
1309    Regenerated the peephole files.  Some contained typos in the specification
1310    and some didn't compile because of pretty printing bugs in the old version
1311    of 'nowhere'.
1312    
1313    ----------------------------------------------------------------------
1314    Name: Allen Leung
1315    Date: 2002/02/19 20:20:00 EST
1316    Tag: leunga-20020219-mlrisc-tools
1317    Description:
1318    
1319       Minor bug fixes to the mlrisc-tools library:
1320    
1321       1.  Fixed up parsing colon suffixed keywords
1322       2.  Added the ability to shut the error messages up
1323       3.  Reimplemented the pretty printer and fixed up/improved
1324           the pretty printing of handle and -> types.
1325       4.  Fixed up generation of literal symbols in the nowhere tool.
1326       5.  Added some SML keywords to to sml.sty
1327    
1328    ----------------------------------------------------------------------
1329    Name: Matthias Blume
1330    Date: 2002/02/19 16:20:00 EST
1331    Tag: blume-20020219-cmffi
1332    Description:
1333    
1334    A wild mix of changes, some minor, some major:
1335    
1336    * All C FFI-related libraries are now anchored under $c:
1337        $/c.cm      --> $c/c.cm
1338        $/c-int.cm  --> $c/internals/c-int.cm
1339        $/memory.cm --> $c/memory/memory.cm
1340    
1341    * "make" tool (in CM) now treats its argument pathname slightly
1342      differently:
1343        1. If the native expansion is an absolute name, then before invoking
1344           the "make" command on it, CM will apply OS.Path.mkRelative
1345           (with relativeTo = OS.FileSys.getDir()) to it.
1346        2. The argument will be passed through to subsequent phases of CM
1347           processing without "going native".  In particular, if the argument
1348           was an anchored path, then "make" will not lose track of that anchor.
1349    
1350    * Compiler backends now "know" their respective C calling conventions
1351      instead of having to be told about it by ml-nlffigen.  This relieves
1352      ml-nlffigen from one of its burdens.
1353    
1354    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1355    
1356    * Export C_DEBUG and C_Debug from $c/c.cm.
1357    
1358    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1359      subtyping relationship between incomplete pointers and their complete
1360      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1361      with the convention of instantiating 'o with ('t, 'c) obj whenever
1362      the pointer target type is complete.  In the incomplete case, 'o
1363      will be instantiated with some "'c iobj" -- a type obtained by
1364      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1365    
1366      Operations that work on both incomplete and complete pointer types are
1367      typed as taking an 'o ptr while operations that require the target to
1368      be known are typed as taking some ('t, 'c) obj ptr.
1369    
1370      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1371      where void is an eqtype without any values.  This makes it possible
1372      to work on voidptr values using functions meant to operate on light
1373      incomplete pointers.
1374    
1375    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1376      been vastly simplified.
1377    
1378    ----------------------------------------------------------------------
1379    Name: Matthias Blume
1380    Date: 2002/02/19 10:48:00 EST
1381    Tag: blume-20020219-pqfix
1382    Description:
1383    
1384    Applied Chris Okasaki's bug fix for priority queues.
1385    
1386    ----------------------------------------------------------------------
1387    Name: Matthias Blume
1388    Date: 2002/02/15 17:05:00
1389    Tag: Release_110_39
1390    Description:
1391    
1392    Last-minute retagging is becoming a tradition... :-(
1393    
1394    This is the working release 110.39.
1395    
1396    ----------------------------------------------------------------------
1397    Name: Matthias Blume
1398    Date: 2002/02/15 16:00:00 EST
1399    Tag: Release_110_39-orig
1400    Description:
1401    
1402    Working release 110.39.  New bootfiles.
1403    
1404    (Update: There was a small bug in the installer so it wouldn't work
1405    with all shells.  So I retagged. -Matthias)
1406    
1407    ----------------------------------------------------------------------
1408    Name: Matthias Blume
1409    Date: 2002/02/15 14:17:00 EST
1410    Tag: blume-20020215-showbindings
1411    Description:
1412    
1413    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1414    the latter can be useful for exploring what bindings are available at
1415    the interactive prompt.  (The first function returns only the list
1416    of symbols that are really bound, the second prints those but also the
1417    ones that CM's autoloading mechanism knows about.)
1418    
1419    ----------------------------------------------------------------------
1420    Name: Matthias Blume
1421    Date: 2002/02/15 12:08:00 EST
1422    Tag: blume-20020215-iptrs
1423    Description:
1424    
1425    Two improvements to ml-nlffigen:
1426    
1427      1. Write files only if they do not exist or if their current contents
1428         do not coincide with what's being written.  (That is, avoid messing
1429         with the time stamps unless absolutely necessary.)
1430    
1431      2. Implement a "repository" mechanism for generated files related
1432         to "incomplete pointer types".   See the README file for details.
1433    
1434    ----------------------------------------------------------------------
1435    Name: Matthias Blume
1436    Date: 2002/02/14 11:50:00 EST
1437    Tag: blume-20020214-quote
1438    Description:
1439    
1440    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1441    because of the new and improved tag generation scheme.  (Thanks to Allen
1442    Leung for pointing it out.)
1443    
1444    ----------------------------------------------------------------------
1445    Name: Lal George
1446    Date: 2002/02/14 09:55:27 EST 2002
1447    Tag: george-20020214-isabelle-bug
1448    Description:
1449    
1450    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1451    of Isabelle on the x86.
1452    
1453    From Allen:
1454    -----------
1455     I've found the problem:
1456    
1457         in ra-core.sml, I use the counter "blocked" to keep track of the
1458         true number of elements in the freeze queue.  When the counter goes
1459         to zero, I skip examining the queue.  But I've messed up the
1460         bookkeeping in combine():
1461    
1462             else ();
1463             case !ucol of
1464               PSEUDO => (if !cntv > 0 then
1465                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1466                                        ^^^^^^^^^^^^^^^^^^^^^^^
1467                      moveu := mergeMoveList(!movev, !moveu)
1468                     )
1469                  else ();
1470    
1471         combine() is called to coalesce two nodes u and v.
1472         I think I was thinking that if the move counts of u and v are both
1473         greater than zero then after they are coalesced then one node is
1474         removed from the freeze queue.  Apparently I was thinking that
1475         both u and v are of low degree, but that's clearly not necessarily true.
1476    
1477    
1478    02/12/2002:
1479        Here's the patch.  HOL now compiles.
1480    
1481        I don't know how this impact on performance (compile
1482        time or runtime).  This bug caused the RA (especially on the x86)
1483        to go thru the potential spill phase when there are still nodes on the
1484        freeze queue.
1485    
1486    
1487    
1488    
1489    ----------------------------------------------------------------------
1490    Name: Matthias Blume
1491    Date: 2002/02/13 22:40:00 EST
1492    Tag: blume-20020213-fptr-rtti
1493    Description:
1494    
1495    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1496    updates.
1497    
1498    ----------------------------------------------------------------------
1499    Name: Matthias Blume
1500    Date: 2002/02/13 16:41:00 EST
1501    Tag: blume-20020213-cmlpq
1502    Description:
1503    
1504    Added new priority queue export symbols (which have just been added to
1505    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1506    not compile and the installer would choke.)
1507    
1508    ----------------------------------------------------------------------
1509    Name: Matthias Blume
1510    Date: 2002/02/13 16:15:00 EST
1511    Tag: blume-20020213-various
1512    Description:
1513    
1514    1. More tweaks to ml-nlffigen:
1515    
1516       - better internal datastructures (resulting in slight speedup)
1517       - "-match" option requires exact match
1518       - "localized" gensym counters (untagged structs/unions nested within
1519         other structs/unions or within typedefs get a fresh counter; their
1520         tag will be prefixed by a concatenation of their parents' tags)
1521       - bug fixes (related to calculation of transitive closure of types
1522         to be included in the output)
1523    
1524    2. Minor Basis updates:
1525    
1526       - added implementations for List.collate and Option.app
1527    
1528    ----------------------------------------------------------------------
1529    Name: Matthias Blume
1530    Date: 2002/02/11 15:55:00 EST
1531    Tag: blume-20020211-gensym
1532    Description:
1533    
1534    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1535    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1536    names (ML structure names that correspond to unnamed C structs, unions,
1537    and enums), so that separate runs of ml-nlffigen do not clash.
1538    
1539    ----------------------------------------------------------------------
1540    Name: Matthias Blume
1541    Date: 2002/02/11 12:05:00 EST
1542    Tag: blume-20020211-gensml
1543    Description:
1544    
1545    A quick fix for a problem with GenSML (in the pgraph-util library):
1546    Make generation of toplevel "local" optional.  (Strictly speaking,
1547    signature definitions within "local" are not legal SML.)
1548    
1549    Other than that: updates to INSTALL and cm/TODO.
1550    
1551    ----------------------------------------------------------------------
1552    Name: Matthias Blume
1553    Date: 2002/02/08 15:00:00 EST
1554    Tag: blume-20020208-uniquepid
1555    Description:
1556    
1557    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1558    
1559    1. The installer (config/install.sh) has gotten smarter:
1560    
1561         - Configuration options are a bit easier to specify now
1562           (in config/targets).
1563         - Bug in recognizing .tar.bz2 files fixed.
1564         - Installer automatically resolves dependencies between
1565           configuration options (e.g., if you ask for eXene, you will
1566           also get cml -- regardless whether you asked for it or not).
1567         - Installer can run in "quieter mode" by setting the environment
1568           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1569           "completely silent", though.
1570         - Build HashCons library as part of smlnj-lib.
1571    
1572    2. A new scheme for assigning persistent identifiers to compilation
1573       units (and, by extension, to types etc.) has been put into place.
1574       This fixes a long-standing bug where types and even dynamic values
1575       can get internally confused, thereby compromising type safety
1576       (abstraction) and dynamic correctness.  See
1577    
1578         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1579    
1580       for an example of how things could go wrong until now.
1581    
1582       The downside of the new scheme is that pids are not quite as
1583       persistent as they used to be: CM will generate a fresh pid
1584       for every compilation unit that it thinks it sees for the first
1585       time.  That means that if you compile starting from a clean, fresh
1586       source tree at two different times, you end up with different
1587       binaries.
1588    
1589       Cutoff recompilation, however, has not been compromised because
1590       CM keeps pid information in special caches between runs.
1591    
1592    ----------------------------------------------------------------------
1593    Name: Lal George
1594    Date: 2002/02/07 15:34:13 EST 2002
1595    Tag: <none>
1596    Description:
1597    
1598    Compilers that generate assembly code may produce  global labels
1599    whose value is resolved at link time. The various peephole optimization
1600    modules did not take this in account.
1601    
1602    TODO. The Labels.addrOf function should really return an option
1603    type so that clients are forced to deal with this issue, rather
1604    than an exception being raised.
1605    
1606    ----------------------------------------------------------------------
1607    Name: Lal George
1608    Date: 2002/02/06 13:55:02 EST
1609    Tag: george-20020206-ra-breakup
1610    Description:
1611    
1612    1. A bug fix from Allen.
1613    
1614        A typo causes extra fstp %st(0)'s to be generated at compensation
1615        edges, which might cause stack underflow traps at runtime.  This
1616        occurs in fft where there are extraneous fstps right before the 'into'
1617        trap instruction (in this case they are harmless since none of the
1618        integers overflow.)
1619    
1620    2. Pulled out various utility modules that were embedded in the modules
1621       of the register allocator. I need these modules for other purposes, but
1622       they are not complete enough to put into a library (just yet).
1623    ----------------------------------------------------------------------
1624    Name: Matthias Blume
1625    Date: 2002/01/31 16:05:00 EST
1626    Tag: blume-20020131-sparc-ccalls
1627    Description:
1628    
1629    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1630       of extra stack space by mistake.  Fixed.
1631    
1632    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1633    
1634    ----------------------------------------------------------------------
1635    Name: Allen Leung
1636    Date: 2002/01/30
1637    Tag: leunga-20020130-nowhere-bug-fix
1638    Description:
1639    
1640       MLRISC bug fixes:
1641       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1642       2. MachineInt.fromString was negating its value.
1643    
1644    ----------------------------------------------------------------------
1645    Name: Matthias Blume
1646    Date: 2002/01/29
1647    Tag: blume-20020129-INSTALL
1648    Description:
1649    
1650    - Added somewhat detailed installation instructions (file INSTALL).
1651    - Fixed curl-detection bug in config/install.sh.
1652    - It is now possible to select the URL getter using the URLGETTER
1653      environment variable:
1654    
1655          not set / "unknown"      --> automatic detection (script tries wget,
1656                                       curl, and lynx)
1657          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1658                                       how to properly invoke them)
1659          other                    --> use $URLGETTER directly, it must take
1660                                       precisely two command-line arguments
1661                                       (source URL and destination file name)
1662    
1663    ----------------------------------------------------------------------
1664    Name: Matthias Blume
1665    Date: 2002/01/28
1666    Tag: blume-20020128-sparc-ccalls
1667    Description:
1668    
1669    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1670    - Make use of the allocParam argument in sparc-c-calls.
1671    
1672    ----------------------------------------------------------------------
1673    Name: Matthias Blume
1674    Date: 2002/01/28
1675    Tag: blume-20020128-allocParam
1676    Description:
1677    
1678    John Reppy:  Changes c-calls API to accept client-callback for
1679    allocating extra stack space.
1680    me: Corresponding changes to mlriscGen (using a dummy argument that
1681        does not change the current behavior).
1682    
1683    ----------------------------------------------------------------------
1684    Name: Matthias Blume
1685    Date: 2002/01/28 12:00:00
1686    Tag: Release_110_38
1687    Description:
1688    
1689    This time for real!!!
1690    
1691    ----------------------------------------------------------------------
1692    Name: Matthias Blume
1693    Date: 2002/01/28 10:56:00 EST
1694    Tag: blume-20020128-retraction
1695    Description:
1696    
1697    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1698       with blume-Release_110_38-retracted.)
1699    
1700    1. Fixed a problem with incorrect rounding modes in real64.sml.
1701       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1702    
1703    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1704       and enums fixed.  The naming of corresponding ML identifiers should
1705       now be consistent again.
1706    
1707    ----------------------------------------------------------------------
1708    Name: Allen Leung
1709    Date: 2002/01/27
1710    Tag: leunga-20020127-nowhere
1711    Description:
1712    
1713       Added a target called nowhere in the configuration scripts.
1714       Enabling this will build the MLRISC 'nowhere' tool (for translating
1715       programs with where-clauses into legal SML code) during installation.
1716    
1717    ----------------------------------------------------------------------
1718    Name: Matthias Blume
1719    Date: 2002/01/25 21:27:00 EST
1720    Tag: blume-Release_110_38-retracted
1721    Description:
1722    
1723    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1724    
1725    README will be added later.
1726    
1727    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1728    (Release_110_38) removed.  Reason: Last-minute bug fixes.
1729    
1730    ----------------------------------------------------------------------
1731    Name: Matthias Blume
1732    Date: 2002/01/25
1733    Tag: blume-20020125-ffi
1734    Description:
1735    
1736    A large number of tweaks and improvements to ml-nlffi-lib and
1737    ml-nlffigen:
1738    
1739       - ML represenation types have been streamlined
1740       - getter and setter functions work with concrete values, not abstract
1741         ones where possible
1742       - ml-nlffigen command line more flexible (see README file there)
1743       - some bugs have been fixed (hopefully)
1744    
1745    ----------------------------------------------------------------------
1746    Name: Lal George
1747    Date: 2002/01/24
1748    Tag: george-20020124-risc-ra-interface
1749    Description:
1750    
1751       There is a dramatic simplification in the interface to the
1752       register allocator for RISC architectures as a result of making
1753       parallel copy instructions explicit.
1754    
1755    ----------------------------------------------------------------------
1756    Name: Matthias Blume
1757    Date: 2002/01/22
1758    Tag: blume-20020122-x86-ccalls
1759    Description:
1760    
1761    Bug fix for c-calls on x86 (having to do with how char- and
1762    short-arguments are being handled).
1763    
1764    ----------------------------------------------------------------------
1765    Name: Matthias Blume
1766    Date: 2002/01/21
1767    Tag: blume-20020121-ff
1768    Description:
1769    
1770    Another day of fiddling with the FFI...
1771    
1772    1. Bug fix/workaround:  CKIT does not complain about negative array
1773       dimensions, so ml-nlffigen has to guard itself against this possibility.
1774       (Otherwise a negative dimension would send it into an infinite loop.)
1775    
1776    2. Some of the abstract types (light objects, light pointers, most "base"
1777       types) in structure C are now eqtypes.
1778    
1779    3. Added constructors and test functions for NULL function pointers.
1780    
1781    ----------------------------------------------------------------------
1782    Name: Matthias Blume
1783    Date: 2002/01/18
1784    Tag: blume-20020118-ready-for-new-release
1785    Description:
1786    
1787    Made config/srcarchiveurl point to a new place.  (Will provide boot
1788    files shortly.)
1789    
1790    Maybe we christen this to be 110.38?
1791    
1792    ----------------------------------------------------------------------
1793    Name: Matthias Blume
1794    Date: 2002/01/18
1795    Tag: blume-20020118-more-ffifiddle
1796    Description:
1797    
1798    Today's FFI fiddling:
1799    
1800      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
1801        C.Set.* that use concrete (MLRep.*) arguments and results instead
1802        of abstract ones.
1803    
1804      - Provided word-style bit operations etc. for "int" representation
1805        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1806        and Long.
1807    
1808    ----------------------------------------------------------------------
1809    Name: Matthias Blume
1810    Date: 2002/01/18
1811    Tag: blume-20020118-use-x86-fp
1812    Description:
1813    
1814    Now that x86-fast-fp seems to be working, I turned it back on again
1815    by default.  (Seems to work fine now, even with the FFI.)
1816    
1817    Other than that, I added some documentation about the FFI to
1818    src/ml-nlffigen/README and updated the FFI test examples in
1819    src/ml-nlffi-lib/Tests/*.
1820    
1821    ----------------------------------------------------------------------
1822    Name: Allen Leung
1823    Date: 2002/01/17
1824    Tag: leunga-20020117-x86-fast-fp-call
1825    Description:
1826    
1827       1. Fixed a problem with handling return fp values when x86's fast fp
1828          mode is turned on.
1829    
1830       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
1831          of %f32.
1832    
1833       3. Added a constructor INT32lit to the ast of MLRISC tools.
1834    
1835    ----------------------------------------------------------------------
1836    Name: Matthias Blume
1837    Date: 2002/01/16
1838    Tag: blume-20020116-ffifiddle
1839    Description:
1840    
1841    More fiddling with the FFI interface:
1842    
1843     - Make constness 'c instead of rw wherever possible.  This eliminates
1844       the need for certain explicit coercions.  (However, due to ML's
1845       value polymorphism, there will still be many cases where explicit
1846       coercions are necessary.  Phantom types are not the whole answer
1847       to modeling a subtyping relationship in ML.)
1848    
1849     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
1850    
1851     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
1852       clumsy workaround for varargs problem.)
1853    
1854    ----------------------------------------------------------------------
1855    Name: Lal George
1856    Date: 2002/01/15
1857    Tag: <none>
1858    Description:
1859    
1860    1. Since COPY instructions are no longer native to the architecture,
1861       a generic functor can be used to implement the expandCopies function.
1862    
1863    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1864       TEXT segment.
1865    
1866    ----------------------------------------------------------------------
1867    Name: Matthias Blume
1868    Date: 2002/01/15
1869    Tag: blume-20020115-ffiupdates
1870    Description:
1871    
1872    1. Fix for bug resulting in single-precision float values being returned
1873       incorrectly from FFI calls.
1874    
1875    2. Small modifications to C FFI API:
1876    
1877        - memory-allocation routines return straight objects (no options)
1878          and raise an exception in out-of-memory situations
1879        - unsafe extensions to cast between function pointers and pointers
1880          from/to ints
1881        - added structure C_Debug as an alternative to structure C where
1882          pointer-dereferencing (|*| and |*!) always check for null-pointers
1883        - added open_lib' to DynLinkage;  open_lib' works like open_lib
1884          but also takes a (possibly empty) list of existing library handles
1885          that the current library depends on
1886    
1887    ----------------------------------------------------------------------
1888    Name: Matthias Blume
1889    Date: 2002/01/10
1890    Tag: blume-20020110-newffigen
1891    Description:
1892    
1893    1. Updates to portable graph code.
1894    
1895    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
1896       more scalable now so that even huge interfaces such as the one
1897       for GTK compile in finite time and space. :-)
1898       See src/ml-nlffigen/README for details on what's new.
1899    
1900    ----------------------------------------------------------------------
1901    Name: Lal George
1902    Date: 2001/01/09 14:31:35 EST 2002
1903    Tag: george-20011206-rm-native-copy
1904    Description:
1905    
1906            Removed the native COPY and FCOPY instructions
1907            from all the architectures and replaced it with the
1908            explicit COPY instruction from the previous commit.
1909    
1910            It is now possible to simplify many of the optimizations
1911            modules that manipulate copies. This has not been
1912            done in this change.
1913    
1914    ----------------------------------------------------------------------
1915    Name: Lal George
1916    Date: 2001/12/06 16:50:13 EST 2001
1917    Tag: george-20011206-mlrisc-instruction
1918    Description:
1919    
1920    Changed the representation of instructions from being fully abstract
1921    to being partially concrete. That is to say:
1922    
1923      from
1924            type instruction
1925    
1926      to
1927            type instr                              (* machine instruction *)
1928    
1929            datatype instruction =
1930                LIVE of {regs: C.cellset, spilled: C.cellset}
1931              | KILL of {regs: C.cellset, spilled: C.cellset}
1932              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
1933              | ANNOTATION of {i: instruction, a: Annotations.annotation}
1934              | INSTR of instr
1935    
1936    This makes the handling of certain special instructions that appear on
1937    all architectures easier and uniform.
1938    
1939    LIVE and KILL say that a list of registers are live or killed at the
1940    program point where they appear. No spill code is generated when an
1941    element of the 'regs' field is spilled, but the register is moved to
1942    the 'spilled' (which is present, more for debugging than anything else).
1943    
1944    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
1945    We used to generate:
1946    
1947            DEFFREG f1
1948            f1 := f2 + f3
1949            trapb
1950    
1951    but now generate:
1952    
1953            f1 := f2 + f3
1954            trapb
1955            LIVE {regs=[f1,f2,f3], spilled=[]}
1956    
1957    Furthermore, the DEFFREG (hack) required that all floating point instruction
1958    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
1959    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
1960    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
1961    this architecture).
1962    
1963    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
1964    all the architectures. This will result in further simplification of the
1965    register allocator that must be aware of them for coalescing purposes, and
1966    will also simplify certain aspects of the machine description that provides
1967    callbacks related to parallel copies.
1968    
1969    ANNOTATION should be obvious, and now INSTR represents the honest to God
1970    machine instruction set!
1971    
1972    The <arch>/instructions/<arch>Instr.sml files define certain utility
1973    functions for making porting easier -- essentially converting upper case
1974    to lower case. All machine instructions (of type instr) are in upper case,
1975    and the lower case form generates an MLRISC instruction. For example on
1976    the alpha we have:
1977    
1978      datatype instr =
1979         LDA of {r:cell, b:cell, d:operand}
1980       | ...
1981    
1982      val lda : {r:cell, b:cell, d:operand} -> instruction
1983        ...
1984    
1985    where lda is just (INSTR o LDA), etc.
1986    
1987    ----------------------------------------------------------------------
1988    Name: Matthias Blume
1989    Date: 2001/11/22 21:40:00 EST
1990    Tag: Release_110_37
1991    Description:
1992    
1993    Release 110.37.  This time for real.
1994    
1995    ----------------------------------------------------------------------
1996    Name: Matthias Blume
1997    Date: 2001/11/21 16:35:00 EST
1998    Tag: blume-20011121-foot-in-mouth
1999    Description:
2000    
2001    Removed the "Release_110_37" tag because of a serious bug.
2002    This will be re-tagged once the bug is fixed.
2003    
2004    ----------------------------------------------------------------------
2005    Name: Matthias Blume
2006    Date: 2001/11/21 16:14:00 EST
2007    Tag: blume-20011121-forgottenfile
2008    Description:
2009    
2010    Forgot to add a file.  (Just a .tex-file -- part of
2011    the CM manual source.)
2012    
2013    ----------------------------------------------------------------------
2014    Name: Matthias Blume
2015    Date: 2001/11/21 16:10:00 EST
2016    Tag: blume-20011121-invalid_110_37
2017    Description:
2018    
2019    Note: I removed the original tag "Release_110_37" from this commit
2020          because we found a serious bug in all non-x86 backends.
2021          - Matthias
2022    
2023    1. Modifications to the SML/NJ code generator and to the runtime system
2024       so that code object name strings are directly inserted into code
2025       objects at code generation time.  The only business the runtime system
2026       has with this is now to read the name strings on occasions.
2027       (The encoding of the name string has also changed somewhat.)
2028    
2029    2. CM now implements a simple "set calculus" for specifying export lists.
2030       In particular, it is now possible to refer to the export lists of
2031       other libraries/groups/sources and form unions as well as differences.
2032       See the latest CM manual for details.
2033    
2034    3. An separate notion of "proxy" libraries has again be eliminated from
2035       CM's model.  (Proxy libraries are now simply a special case of using
2036       the export list calculus.)
2037    
2038    4. Some of the existing libraries now take advantage of the new set
2039       calculus.
2040       (Notice that not all libraries have been converted because some
2041       of the existing .cm-files are supposed to be backward compatible
2042       with 110.0.x.)
2043    
2044    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2045       "General.exnMessage"!  The former relies on a certain hook to be
2046       initialized, and that often does not happen in the stand-alone case.)
2047    
2048    ----------------------------------------------------------------------
2049    Name: Lal George
2050    Date: 2001/11/21  13:56:18 EST
2051    Tag: george-2001121-pseudo-ops
2052    Description:
2053    
2054      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2055      ought to never be any question of incompatabilities with
2056      pseudo-op syntax expected by host assemblers.
2057    
2058      For now, only modules supporting GAS syntax are implemented
2059      but more should follow, such as MASM, and vendor assembler
2060      syntax, e.g. IBM as, Sun as, etc.
2061    
2062    ----------------------------------------------------------------------
2063    Name: Matthias Blume
2064    Date: 2001/11/14 11:52:00 EST
2065    Tag: blume-20011114-srcname
2066    Description:
2067    
2068    1. Routed the name of the current source file to mlriscgen where it
2069       should be directly emitted into the code object.  (This last part
2070       is yet to be done.)
2071    
2072    2. Some cleanup of the pgraph code to make it match the proposal that
2073       I put out the other day.  (The proposal notwithstanding, things are
2074       still in flux here.)
2075    
2076    ----------------------------------------------------------------------
2077    Name: Lal George
2078    Date: 2001/11/14 09:44:04 EST
2079    Tag:
2080    Description:
2081    
2082      Fix for a backpatching bug reported by Allen.
2083    
2084      Because the boundary between short and long span-dependent
2085      instructions is +/- 128, there are an astounding number of
2086      span-dependent instructions whose size is over estimated.
2087    
2088      Allen came up with the idea of letting the size of span
2089      dependent instructions be non-monotonic, for a maxIter
2090      number of times, after which the size must be monotonically
2091      increasing.
2092    
2093      This table shows the number of span-dependent instructions
2094      whose size was over-estimated as a function of maxIter, for the
2095      file Parse/parse/ml.grm.sml:
2096    
2097         maxIter            # of instructions:
2098            10                      687
2099            20                      438
2100            30                      198
2101            40                        0
2102    
2103      In compiling the compiler, there is no significant difference in
2104      compilation speed between maxIter=10 and maxIter=40. Actually,
2105      my measurements showed that maxIter=40 was a tad faster than
2106      maxIter=10! Also 96% of the  files in the compiler reach a fix
2107      point within 13 iterations, so fixing maxIter at 40, while high,
2108      is okay.
2109    
2110    ----------------------------------------------------------------------
2111    Name: Matthias Blume
2112    Date: 2001/10/31 15:25:00 EST
2113    Tag: blume-20011031-pgraph
2114    Description:
2115    
2116    CKIT:
2117    * Changed the "Function" constructor of type Ast.ctype to carry optional
2118      argument identifiers.
2119    * Changed the return type of TypeUtil.getFunction accordingly.
2120    * Type equality ignores the argument names.
2121    * TypeUtil.composite tries to preserve argument names but gives up quickly
2122      if there is a mismatch.
2123    
2124    installation script:
2125    * attempts to use "curl" if available (unless "wget" is available as well)
2126    
2127    CM:
2128    * has an experimental implementation of "portable graphs" which I will
2129      soon propose as an implementation-independent library format
2130    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2131    
2132    NLFFI-LIB:
2133    * some cleanup (all cosmetic)
2134    
2135    NLFFIGEN:
2136    * temporarily disabled the mechanism that suppresses ML output for
2137      C definitions whose identifiers start with an underscore character
2138    * generate val bindings for enum constants
2139    * user can request that only one style (light or heavy) is being used;
2140      default is to use both (command-line arguments: -heavy and -light)
2141    * fixed bug in handling of function types involving incomplete pointers
2142    * generate ML entry points that take record arguments (i.e., using
2143      named arguments) for C functions that have a prototype with named
2144      arguments
2145      (see changes to CKIT)
2146    
2147    ----------------------------------------------------------------------
2148    Name: Allen Leung
2149    Date: 2001/10/27 20:34:00 EDT
2150    Tag: leunga-20011027-x86-fast-fp-call
2151    Description:
2152    
2153       Fixed the bug described in blume-20010920-slowfp.
2154    
2155       The fix involves
2156          1. generating FCOPYs in FSTP in ia32-svid
2157          2. marking a CALL with the appropriate annotation
2158    
2159    ----------------------------------------------------------------------
2160    Name: Matthias Blume
2161    Date: 2001/10/16 11:32:00 EDT
2162    Tag: blume-20011016-netbsd
2163    Description:
2164    
2165    Underscore patch from Chris Richards (fixing problem with compiling
2166    runtime system under recent NetBSD).
2167    
2168    ----------------------------------------------------------------------
2169    Name: Allen Leung
2170    Date: 2001/10/12 17:18:32 EDT 2001
2171    Tag: leung-20011012-x86-printflowgraph
2172    Description:
2173    
2174    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2175    
2176    ----------------------------------------------------------------------
2177    Name: Lal George
2178    Date: 2001/10/11 23:51:34 EDT
2179    Tag: george-20011011-too-many-instrs
2180    Description:
2181    
2182    The representation of a program point never expected to see more
2183    than 65536 instructions in a basic block!
2184    
2185    ----------------------------------------------------------------------
2186    Name: Lal George
2187    Date: 2001/10/09 09:41:37 EDT
2188    Tag: george-20011008-mlrisc-labels
2189    Description:
2190    
2191    Changed the machine description files to support printing of
2192    local and global labels in assembly code, based on host assembler
2193    conventions.
2194    
2195    ----------------------------------------------------------------------
2196    Name: Matthias Blume
2197    Date: 2001/09/25 15:25:00 EDT
2198    Tag: blume-20010925-exninfo
2199    Description:
2200    
2201    I provided a non-hook implementation of exnName (at the toplevel) and
2202    made the "dummy" implementation of exnMessage (at the toplevel) more
2203    useful: if nothing gets "hooked in", then at least you are going to
2204    see the exception name and a message indicating why you don't see more.
2205    
2206    [For the time being, programs that need exnMessage and want to use
2207    ml-build should either use General.exnMessage (strongly recommended) or
2208    refer to structure General at some other point so that CM sees a
2209    static dependency.]
2210    
2211    [Similar remarks go for "print" and "use":  If you want to use their
2212    functionality in stand-alone programs generated by ml-build, then use
2213    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2214    
2215    ----------------------------------------------------------------------
2216    Name: Matthias Blume
2217    Date: 2001/09/20 17:28:00 EDT
2218    Tag: blume-20010920-slowfp
2219    Description:
2220    
2221    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2222    
2223    ----------------------------------------------------------------------
2224    Name: Matthias Blume
2225    Date: 2001/09/20 17:20:00 EDT
2226    Tag: blume-20010920-canonicalpaths
2227    Description:
2228    
2229    0. Updated the BOOT file (something that I forgot to do earlier).
2230    
2231    1. Small internal change to CM so that it avoids "/../" in filenames
2232       as much as possible (but only where it is safe).
2233    
2234    2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2235       that arguments that contain delimiters are passed through correctly.
2236       This change also means that all "special" arguments of the form
2237       @SMLxxx... must come first.
2238    
2239    3. Changed install script to put relative anchor names for tool commands
2240       into pathconfig.
2241    
2242    ----------------------------------------------------------------------
2243    Name: Matthias Blume
2244    Date: 2001/09/18 15:35:00 EDT
2245    Tag: blume-20010918-readme11036
2246    Description:
2247    
2248    Added README files.
2249    
2250    ----------------------------------------------------------------------
2251    Name: Matthias Blume
2252    Date: 2001/09/18 11:45:00 EDT
2253    Tag: Release_110_36 (retag)
2254    Description:
2255    
2256    Fixed mistake in config/preloads. Retagged as 110.36.
2257    
2258    ----------------------------------------------------------------------
2259    Name: Matthias Blume
2260    Date: 2001/09/18 09:40:00 EDT
2261    Tag: Release_110_36_orig (tag changed)
2262    Description:
2263    
2264    New version (110.36).  New bootfiles.
2265    
2266    ----------------------------------------------------------------------
2267    Name: Matthias Blume
2268    Date: 2001/09/14 16:15:00 EDT
2269    Tag: blume-20010914-x86fastfp
2270    Description:
2271    
2272    John committed some changes that Allen made, in particular a (hopefully)
2273    correctly working version of the x86-fp module.
2274    
2275    I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
2276    flag to "true".  Everything seems to compile to a fixpoint ok, and
2277    "mandelbrot" speeds up by about 15%.
2278    
2279    ----------------------------------------------------------------------
2280    Name: Matthias Blume
2281    Date: 2001/09/13 11:20:00 EDT
2282    Tag: blume-20010913-minimal
2283    Description:
2284    
2285    1. Stefan Monnier's patch to fix a miscompilation problem that
2286       was brought to light by John Reppy's work on Moby.
2287    
2288    2. Implemented a minimal "structure Compiler" that contains just
2289       "version" and "architecture".  The minimal version will be
2290       available when the full version is not.  This is for backward-
2291       compatibility with code that wants to test Compiler.version.
2292    
2293    ----------------------------------------------------------------------
2294    Name: Matthias Blume
2295    Date: 2001/08/28 14:03:00 EDT
2296    Tag: blume-20010828-ml-lex
2297    Description:
2298    
2299    Fix for bug 1581, received from Neophytos Michael.
2300    
2301    ----------------------------------------------------------------------
2302    Name: Matthias Blume
2303    Date: 2001/08/27 11:20:00 EDT
2304    Tag: blume-20010827-readme11035
2305    Description:
2306    
2307    Fleshed out the README file for 110.35.
2308    
2309    ----------------------------------------------------------------------
2310    Name: Matthias Blume
2311    Date: 2001/08/24 17:10:00 EDT
2312    Tag: Release_110_35
2313    Description:
2314    
2315    New version number (110.35).  New bootfiles.
2316    
2317    ----------------------------------------------------------------------
2318    Name: Lal George
2319    Date: 2001/08/24 13:47:18 EDT 2001
2320    Tag: george-20010824-MLRISC-graphs
2321    Description:
2322    
2323     removed clusters from MLRISC completely and replaced with graphs.
2324    
2325    ----------------------------------------------------------------------
2326    Name: Matthias Blume
2327    Date: 2001/08/23 17:50:00 EDT
2328    Tag: blume-20010823-toplevel
2329    Description:
2330    
2331    - some reorganization of the code that implements various kinds of
2332      environments in the compiler (static, dynamic, symbolic, combined)
2333    - re-implemented the EnvRef module so that evalStream works properly
2334      (if the stream contains references to "use", "CM.make", etc.)
2335    - cleaned up evalloop.sml and interact.sml (but they need more cleaning)
2336    
2337    ----------------------------------------------------------------------
2338    Name: Matthias Blume
2339    Date: 2001/08/20 15:50 EDT
2340    Tag: blume20010820-slipup
2341    Description:
2342    
2343    I forgot to commit a few files.  Here they are...
2344    
2345    ----------------------------------------------------------------------
2346    Name: Matthias Blume
2347    Date: 2001/08/20 15:35:00 EDT
2348    Tag: blume-20010820-debugprof
2349    Description:
2350    
2351    !!!! NEW BOOTFILES !!!!
2352    
2353    This is another round of reorganizing the compiler sources.  This
2354    time the main goal was to factor out all the "instrumentation"
2355    passes (for profiling and backtracing) into their own library.
2356    The difficulty was to do it in such a way that it does not depend
2357    on elaborate.cm but only on elabdata.cm.
2358    
2359    Therefore there have been further changes to both elaborate.cm and
2360    elabdata.cm -- more "generic" things have been moved from the former
2361    to the latter.  As a result, I was forced to split the assignment
2362    of numbers indicating "primtyc"s into two portions: SML-generic and
2363    SML/NJ-specific.  Since it would have been awkward to maintain,
2364    I bit the bullet and actually _changed_ the mapping between these
2365    numbers and primtycs.  The bottom line of this is that you need
2366    a new set of bin- and bootfiles.
2367    
2368    I have built new bootfiles for all architectures, so doing a fresh
2369    checkout and config/install.sh should be all you need.
2370    
2371    The newly created library's name is
2372    
2373        $smlnj/viscomp/debugprof.cm
2374    
2375    and its sources live under
2376    
2377        src/compiler/DebugProf
2378    
2379    ----------------------------------------------------------------------
2380    Name: Matthias Blume
2381    Date: 2001/08/15 17:15:00 EDT
2382    Tag: blume-20010815-compreorg
2383    Description:
2384    
2385    This is a first cut at reorganizing the CM libraries that make up the
2386    core of the compiler.  The idea is to separate out pieces that could
2387    be used independently by tools, e.g., the parser, the typechecker, etc.
2388    
2389    The current status is a step in this direction, but it is not quite
2390    satisfactory yet.  Expect more changes in the future.
2391    
2392    Here is the current (new) organization...
2393    
2394        What used to be $smlnj/viscomp/core.cm is now divided into
2395        six CM libraries:
2396    
2397             $smlnj/viscomp/basics.cm
2398                           /parser.cm
2399                           /elabdata.cm
2400                           /elaborate.cm
2401                           /execute.cm
2402                           /core.cm
2403    
2404        The CM files for these libraries live under src/system/smlnj/viscomp.
2405        All these libraries are proxy libraries that contain precisely
2406        one CM library component.  Here are the locations of the components
2407        (all within the src/compiler tree):
2408    
2409             Basics/basics.cm
2410             Parse/parser.cm
2411             ElabData/elabdata.cm
2412             Elaborator/elaborate.cm
2413             Execution/execute.cm
2414             core.cm
2415    
2416         [This organization is the same that has been used already
2417         for a while for the architecture-specific parts of the visible
2418         compiler and for the old version of core.cm.]
2419    
2420         As you will notice, many source files have been moved from their
2421         respective original locations to a new home in one of the above
2422         subtrees.
2423    
2424         The division of labor between the new libraries is the following:
2425    
2426             basics.cm:
2427                - Simple, basic definitions that pertain to many (or all) of
2428                  the other libraries.
2429             parser.cm:
2430                - The SML parser, producing output of type Ast.dec.
2431                - The type family for Ast is also defined and exported here.
2432             elabdata.cm:
2433                - The datatypes that describe input and output of the elaborator.
2434                  This includes types, absyn, and static environments.
2435             elaborator.cm:
2436                - The SML/NJ type checker and elaborator.
2437                  This maps an Ast.dec (with a given static environment) to
2438                  an Absyn.dec (with a new static environment).
2439                - This libraries implements certain modules that used to be
2440                  structures as functors (to remove dependencies on FLINT).
2441             execute.cm:
2442                - Everything having to do with executing binary code objects.
2443                - Dynamic environments.
2444             core.cm:
2445                - SML/NJ-specific instantiations of the elaborator and MLRISC.
2446                - Top-level modules.
2447                - FLINT (this should eventually become its own library)
2448    
2449    Notes:
2450    
2451    I am not 100% happy with the way I separated the elaborator (and its
2452    data structures) from FLINT.  Two instances of the same problem:
2453    
2454        1. Data structures contain certain fields that carry FLINT-specific
2455           information.  I hacked around this using exn and the property list
2456           module from smlnj-lib.  But the fact that there are middle-end
2457           specific fields around at all is a bit annoying.
2458    
2459        2. The elaborator calculates certain FLINT-related information.  I tried
2460           to make this as abstract as I could using functorization, but, again,
2461           the fact that the elaborator has to perform calculations on behalf
2462           of the middle-end at all is not nice.
2463    
2464        3. Having to used exn and property lists is unfortunate because it
2465           weakens type checking.  The other alternative (parameterizing
2466           nearly *everything*) is not appealing, though.
2467    
2468    I removed the "rebinding =" warning hack because due to the new organization
2469    it was awkward to maintain it.  As a result, the compiler now issues some of
2470    these warnings when compiling init.cmi during bootstrap compilation. On
2471    the plus side, you also get a warning when you do, for example:
2472       val op = = Int32.+
2473    which was not the case up to now.
2474    
2475    I placed "assign" and "deref" into the _Core structure so that the
2476    code that deals with the "lazy" keyword can find them there.  This
2477    removes the need for having access to the primitive environment
2478    during elaboration.
2479    
2480    ----------------------------------------------------------------------
2481    Name: Matthias Blume
2482    Date: 2001/08/13
2483    Tag: blume-20010813-closures
2484    Description:
2485    
2486    This fix was sent to us by Zhong Shao.  It is supposed to improve the
2487    performance of certain loops by avoiding needless closure allocation.
2488    
2489    ----------------------------------------------------------------------
2490    Name: Lal George
2491    Date: 2001/07/31 10:03:23 EDT 2001
2492    Tag: george-20010731-x86-fmalloc
2493    Description: Fixed bug in x86 calls
2494    
2495        There was a bug where call instructions would mysteriously
2496        vanish. The call instruction had to be one that returned
2497        a floating point value.
2498    
2499    ----------------------------------------------------------------------
2500    Name: Lal George
2501    Date: 2001/07/19 16:36:29 EDT 2001
2502    Tag: george-20010719-simple-cells
2503    Description:
2504    
2505    I have dramatically simplified the interface for CELLS in MLRISC.
2506    
2507    In summary, the cells interface is broken up into three parts:
2508    
2509      1. CellsBasis : CELLS_BASIS
2510    
2511            CellsBasis is a top level structure and common for all
2512            architectures.  it contains the definitions of basic datatypes
2513            and utility  functions over these types.
2514    
2515      2. functor Cells() : CELLS
2516    
2517            Cells generates an interface for CELLS that incorporates the
2518            specific resources on the target architecture, such as the
2519            presence of special register classes, their number and size,
2520            and various useful substructures.
2521    
2522      3. <ARCH>CELLS
2523    
2524            e.g. SparcCells: SPARCCELLS
2525    
2526            <ARCH>CELLS usually contains additional bindings for special
2527            registers  on the architecture, such as:
2528    
2529                    val r0 : cell           (* register zero *)
2530                    val y : cell            (* Y register *)
2531                    val psr : cell          (* processor status register *)
2532                    ...
2533    
2534            The structure returned by applying the Cells functor is opened
2535            in this interface.
2536    
2537    The main implication of all this is that the datatypes for cells is
2538    split between CellsBasis and CELLS -- a fairly simple change for user
2539    code.
2540    
2541    In the old scheme the CELLS interface had a definitional binding of
2542    the form:
2543    
2544            signature CELLS = sig
2545    
2546               structure CellsBasis = CellsBasis
2547    
2548               ...
2549    
2550            end
2551    
2552    With all the sharing constraints that goes on in MLRISC, this old
2553    design  quickly leads to errors such as:
2554    
2555            "structure definition spec inside of sharing ... "
2556    
2557    
2558    and appears to require an unacceptable amount of sharing and where
2559    constraint hackery.
2560    
2561    I think this error message (the interaction of definitional specs and
2562    sharing) requires more explanation on our web page.
2563    
2564    ----------------------------------------------------------------------
2565    Name: Matthias Blume
2566    Date: 2001/07/19 15:00:00 EDT
2567    Tag: blume-20010719-libreorg
2568    Description:
2569    
2570    This update puts together a fairly extensive but straightforward change
2571    to the way the libraries that implement the interactive system are
2572    organized:
2573    
2574       The biggest change is the elimination of structure Compiler.  As a
2575       replacement for this structure, there is now a CM library
2576       (known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
2577       that exports all the substructures of the original structure Compiler
2578       directly.  So instead of saying Compiler.Foo.bar one now simply
2579       says Foo.bar.  (The CM libraries actually export a collection of
2580       structures that is richer than the collection of substructures of
2581       structure Compiler.)
2582    
2583       To make the transition smooth, there is a separate library called
2584       $smlnj/compiler/compiler.cm which puts together and exports the
2585       original structure Compiler (or at least something very close to it).
2586    
2587       There are five members of the original structure Compiler
2588       that are not exported directly but which instead became members
2589       of a new structure Backend (described by signature BACKEND).  These are:
2590       structure Profile (: PROFILE), structure Compile (: COMPILE), structure
2591       Interact (: INTERACT), structure Machine (: MACHINE), and val
2592       architecture (: string).
2593    
2594       Structure Compiler.Version has become structure CompilerVersion.
2595    
2596       Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
2597       by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
2598       or x86, respectively.
2599       Each of these exports the same frontend structures that
2600       $smlnj/compiler.cm exports.  But they do not have a structure Backend
2601       and instead export some structure <Arch>Backend where <Arch> is Alpha32,
2602       Hppa, PPC, Sparc, or X86, respectively.
2603    
2604       Library $smlnj/compiler/all.cm exports the union of the exports of
2605       $smlnj/compiler/<arch>.cm
2606    
2607       There are no structures <Arch>Compiler anymore, use
2608       $smlnj/compiler/<arch>.cm instead.
2609    
2610       Library host-compiler-0.cm is gone.  Instead, the internal library
2611       that instantiates CM is now called cm0.cm.  Selection of the host
2612       compiler (backend) is no longer done here but. (Responsibility for it
2613       now lies with $smlnj/compiler/current.cm.  This seems to be more
2614       logical.)
2615    
2616       Many individual files have been moved or renamed.  Some files have
2617       been split into multiple files, and some "dead" files have been deleted.
2618    
2619    Aside from these changes to library organization, there are also changes
2620    to the way the code itself is organized:
2621    
2622       Structure Binfile has been re-implemented in such a way that it no
2623       longer needs any knowledge of the compiler.  It exclusively deals
2624       with the details of binfile layout.  It no longer invokes the
2625       compiler (for the purpose of creating new prospective binfile
2626       content), and it no longer has any knowledge of how to interpret
2627       pickles.
2628    
2629       Structure Compile (: COMPILE) has been stripped down to the bare
2630       essentials of compilation.  It no longer deals with linking/execution.
2631       The interface has been cleaned up considerably.
2632    
2633       Utility routines for dealing with linking and execution have been
2634       moved into their own substructures.
2635    
2636       (The ultimate goal of these changes is to provide a light-weight
2637       binfile loader/linker (at least for, e.g., stable libraries) that
2638       does not require CM or the compiler to be present.)
2639    
2640    CM documentation has been updated to reflect the changes to library
2641    organization.
2642    
2643    ----------------------------------------------------------------------
2644    Name: Matthias Blume
2645    Date: 2001/07/10 17:30:00 EDT
2646    Tag: Release_110_34
2647    Description:
2648    
2649    Minor tweak to 110.34 (re-tagged):
2650    
2651      - README.html file added to CVS repository
2652      - runtime compiles properly under FreeBSD 3.X and 4.X
2653    
2654    ----------------------------------------------------------------------
2655    Name: Matthias Blume
2656    Date: 2001/07/10 17:30:00 EDT
2657    Tag: Release_110_34
2658    Description:
2659    
2660    New version number (110.34). New bootfiles.
2661    
2662    ----------------------------------------------------------------------
2663    Name: Matthias Blume
2664    Date: 2001/07/09 16:00:00 EDT
2665    Tag: blume-20010709-more-varargs
2666    Description:
2667    
2668    I changed the handling of varargs in ml-nlffigen again:
2669    The ellipsis ... will now simply be ignored (with an accompanying warning).
2670    
2671    The immediate effect is that you can actually call a varargs function
2672    from ML -- but you can't actually supply any arguments beyond the ones
2673    specified explicitly.  (For example, you can call printf with its format
2674    string, but you cannot pass additional arguments.)
2675    
2676    This behavior is only marginally more useful than the one before, but
2677    it has the advantage that a function or, more importantly, a function
2678    type never gets dropped on the floor, thus avoiding follow-up problems with
2679    other types that refer to the offending one.
2680    
2681    ----------------------------------------------------------------------
2682    Name: Matthias Blume
2683    Date: 2001/07/09 11:25:00 EDT
2684    Tag: blume-20010709-varargs
2685    Description:
2686    
2687    1. ckit-lib.cm now exports structure Error
2688    2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
2689       with a warning accompanied by a source location.  Moreover, it
2690       merely skips the offending function or type and proceeds with the
2691       rest of its work.u  As a result, one can safely feed C code containing
2692       "..." to ml-nlffigen.
2693    3. There are some internal improvements to CM, providing slightly
2694       more general string substitutions in the tools subsystem.
2695    
2696    ----------------------------------------------------------------------
2697    Name: Matthias Blume
2698    Date: 2001/06/27 15:10:00 EDT
2699    Tag: blume-20010627-concur
2700    Description:
2701    
2702    Fixed a small bug in CM's handling of parallel compilation.
2703    (You could observe the bug by Control-C-interrupting an ordinary
2704    CMB.make or CM.stabilize and then attaching some compile servers.
2705    The result was that all of a sudden the previously interrupted
2706    compilation would continue on its own.  This was because of
2707    an over-optimization: CM did not bother to clean out certain queues
2708    when no servers were attached "anyway", resulting in the contents
2709    of these queues to grab control when new servers did get attached.)
2710    
2711    There is also another minor update to the CM manual.
2712    
2713    ----------------------------------------------------------------------
2714    Name: Matthias Blume
2715    Date: 2001/06/26 16:15:00 EDT
2716    Tag: blume-20010626-cmdoc
2717    Description:
2718    
2719    Minor typo fixed in CM manual (syntax diagram for libraries).
2720    
2721    ----------------------------------------------------------------------
2722    Name: Matthias Blume
2723    Date: 2001/06/25 22:55:00 EDT
2724    Tag: blume-20010625-x86pc
2725    Description:
2726    
2727    Fixed a nasty bug in the X86 assembly code that caused signal
2728    handlers to fail (crash) randomly.
2729    
2730    ----------------------------------------------------------------------
2731    Name: Matthias Blume
2732    Date: 2001/06/25 12:05:00 EDT
2733    Tag: blume-20010625-nlffigen
2734    Description:
2735    
2736    This update fixes a number of minor bugs in ml-nlffigen as reported by
2737    Nick Carter <nbc@andrew.cmu.edu>.
2738    
2739      1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
2740      2. Default names for generated files are now derived from the name of
2741         the C file *without its directory*.  In particular, this causes generated
2742         files to be placed locally even if the C file is in some system directory.
2743      3. Default names for generated signatures and structures are also derived
2744         from the C file name without its directory.  This avoids silly things
2745         like "structure GL/GL".
2746         (Other silly names are still possible because ml-nlffigen does not do
2747          a thorough check of whether generated names are legal ML identifiers.
2748          When in doubt, use command line arguments to force particular names.)
2749    
2750    ----------------------------------------------------------------------
2751    Name: Matthias Blume
2752    Date: 2001/06/21 12:25:00 EDT
2753    Tag: blume-20010621-eXene
2754    Description:
2755    
2756    eXene now compiles and (sort of) works again.
2757    
2758    The library name (for version > 110.33) is $/eXene.cm.
2759    
2760    I also added an new example in src/eXene/examples/nbody.  See the
2761    README file there for details.
2762    
2763    ----------------------------------------------------------------------
2764    Name: Matthias Blume
2765    Date: 2001/06/20 16:40:00 EDT
2766    Tag: blume-20010620-cml
2767    Description:
2768    
2769    CML now compiles and works again.
2770    
2771    Libraries (for version > 110.33):
2772    
2773      $cml/cml.cm            Main CML library.
2774      $cml/basis.cm          CML's version of $/basis.cm.
2775      $cml/cml-internal.cm   Internal helper library.
2776      $cml/core-cml.cm       Internal helper library.
2777      $cml-lib/trace-cml.cm  Tracing facility.
2778      $cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
2779    
2780    The installer (config/install.sh) has been taught how to properly
2781    install this stuff.
2782    
2783    ----------------------------------------------------------------------
2784    Name: Matthias Blume
2785    Date: 2001/06/19 17:55:00 EDT
2786    Tag: blume-20010619-instantiate
2787    Description:
2788    
2789    This un-breaks the fix for bug 1432.
2790    (The bug was originally fixed in 110.9 but I broke it again some
2791    time after that.)
2792    
2793    ----------------------------------------------------------------------
2794    Name: Matthias Blume
2795    Date: 2001/06/19 17:25:00 EDT
2796    Tag: blume-20010619-signals
2797    Description:
2798    
2799    This should (hopefully) fix the long-standing signal handling bug.
2800    (The runtime system was constructing a continuation record with an
2801    incorrect descriptor which would cause the GC to drop data on the floor...)
2802    
2803    ----------------------------------------------------------------------
2804    Name: Matthias Blume
2805    Date: 2001/06/15 15:05:00 EDT
2806    Tag: blume-20010615-moresparc
2807    Description:
2808    
2809    Here is a short late-hour update related to Sparc c-calls:
2810    
2811     -- made handling of double-word arguments a bit smarter
2812    
2813     -- instruction selection phase tries to collapse certain clumsily
2814        constructed ML-Trees; typical example:
2815    
2816            ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
2817    
2818        This currently has no further impact on SML/NJ since mlriscGen does
2819        not seem to generate such patterns in the first place, and c-calls
2820        (which did generate them in the beginning) has meanwhile been fixed
2821        so as to avoid them as well.
2822    
2823    ----------------------------------------------------------------------
2824    Name: Matthias Blume
2825    Date: 2001/06/15 15:05:00 EDT
2826    Tag: blume-20010615-sparc
2827    Description:
2828    
2829    The purpose of this update is to provide an implementation of NLFFI
2830    on Sparc machines.
2831    
2832    Here are the changes in detail:
2833    
2834       * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
2835       the Sparc implementation of the c-calls API.
2836       * The Sparc backend of SML/NJ has been modified to uniformely use %fp
2837       for accessing the ML frame.  Thus, we have a real frame pointer and
2838       can freely modify %sp without need for an omit-frame-ptr phase.
2839       The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
2840       this case.
2841       * ml-nlffigen has been taught to produce code for different architectures
2842       and calling conventions.
2843       * In a way similar to what was done in the x86 case, the Sparc
2844       backend uses its own specific extension to mltree.  (For example,
2845       it needs to be able to generate UNIMP instructions which are part
2846       of the calling convention.)
2847       * ml-nlffi-lib was reorganized to make it more modular (in particular,
2848       to make it easier to plug in new machine- and os-dependent parts).
2849    
2850    There are some other fairly unrelated bug fixes and cleanups as well:
2851    
2852       * I further hacked the .cm files for MLRISC tools (like MDLGen) so
2853       that they properly share their libraries with existing SML/NJ libraries.
2854       * I fixed a minor cosmetic bug in CM, supressing certain spurious
2855       follow-up error messages.
2856       * Updates to CM/CMB documentation.
2857    
2858    TODO items:
2859    
2860       * MLRISC should use a different register as its asmTemp on the Sparc.
2861         (The current %o2 is a really bad choice because it is part of the
2862         calling conventions, so things might interfere in unexpected ways.)
2863    
2864    ----------------------------------------------------------------------
2865    Name: Matthias Blume
2866    Date: 2001/06/07
2867    Tag: blume-20010607-calls
2868    Description:
2869    
2870    A number of internal changes related to C calls and calling conventions:
2871    
2872    1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
2873       number of bytes popped implicitly (by the callee).  In most cases
2874       this field is 0 but on x86/win32 it is some non-zero value.  This
2875       is information provided for the benefit of the "omit-frameptr" pass.
2876    2. The CALL instruction on the x86 carries a similar "pops" field.
2877       The instruction selection phase copies its value from the ML-Tree
2878       CALL statement.
2879    3. On all other architectures, the instruction selection phase checks
2880       whether "pops=0" and complains if not.
2881    4. The c-calls implementation for x86 now accepts two calling conventions:
2882       "ccall" and "stdcall".  When "ccall" is selected, the caller cleans
2883       up after the call and pops is set to 0.  For "stdcall", the caller
2884       does nothing, leaving the cleanup to the callee; pops is set to
2885       the number of bytes that were pushed onto the stack.
2886    5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
2887       distinguish between "ccall" and "stdcall".
2888    6. The UNIMP instruction has been added to the supported Sparc instruction
2889       set. (This is needed for implementing the official C calling convention
2890       on this architecture.)
2891    7. I fixed some of the .cm files under src/MLRISC/Tools to make them
2892       work with the latest CM.
2893    
2894    ----------------------------------------------------------------------
2895    Name: Matthias Blume
2896    Date: 2001/06/05 15:10:00 EDT
2897    Tag: blume-20010605-cm-index
2898    Description:
2899    
2900    0. The "lambdasplit" parameter for class "sml" in CM has been documented.
2901    
2902    1. CM can now generate "index files".  These are human-readable files
2903       that list on a per-.cm-file basis each toplevel symbol defined or
2904       imported.  The location of the index file for
2905       <p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
2906       To enable index-file generation, set CM.Control.generate_index to true
2907       or export an environment-symbol: export CM_GENERATE_INDEX=true.
2908    
2909       The CM manual has been updated accordingly.
2910    
2911    2. I made some slight modifications to the c-calls API in MLRISC.
2912    
2913         a) There is now a callback to support saving/restoring of
2914            dedicated but caller-save registers around the actual call
2915            instruction.
2916         b) One can optionally specify a comment-annotation for the
2917            call instruction.
2918    
2919    3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
2920       (For example, the comment annotation shows the C prototype of
2921        the function being called.)
2922    
2923    ----------------------------------------------------------------------
2924    Name: Matthias Blume
2925    Date: 2001/06/01 13:30:00 EDT
2926    Tag: blume-20010601-nlffi-cleanup
2927    Description:
2928    
2929    This is mostly a cleanup of MLFFI stuff:
2930    
2931       - some signature files have been put into a more exposed place
2932       - the ugly 'f type parameter is gone (simplifies types tremendously!)
2933       - ml-nlffigen changed accordingly
2934       - tutorial updated
2935    
2936    Other changes:
2937    
2938       - author's affiliation in CM manual(s) updated
2939       - some more recognized keywords added to Allen's sml.sty
2940    
2941    ----------------------------------------------------------------------
2942    Name: Matthias Blume
2943    Date: 2001/05/25 15:30:00 EDT
2944    Tag: blume-20010525-iptr
2945    Description:
2946    
2947      - put the official 110.33-README (as it appears on the ftp server) under
2948        CVS
2949      - fixed a small bug related to incomplete pointer types in
2950        ml-nlffigen
2951      - small cosmetic change to the ml-nlffi-lib's "arr" type constructor
2952        (it does not need the 'f type parameter)
2953    
2954    ----------------------------------------------------------------------
2955    Name: Matthias Blume
2956    Date: 2001/05/23 14:30:00 EDT
2957    Tag: Release_110_33
2958    Description:
2959    
2960    New version number (110.33).  New bootfiles.
2961    
2962    ----------------------------------------------------------------------
2963    Name: Matthias Blume
2964    Date: 2001/05/22 18:06:00 EDT
2965    Tag: blume-20010522-targets
2966    Description:
2967    
2968    Made install.sh use file config/targets.customized if it exists, falling
2969    back to config/targets if it doesn't.  This way one can have a customized
2970    version of the targets file without touching the "real thing", thus
2971    eliminating the constant fear of accidentally checking something bogus
2972    back into the CVS repository...  (File config/targets.customized must
2973    not be added to the repository!)
2974    
2975    ----------------------------------------------------------------------
2976    Name: Matthias Blume
2977    Date: 2001/05/22 16:30:00 EDT
2978    Tag: blume-20010522-minitut
2979    Description:
2980    
2981    1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
2982       struct returns.
2983    2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
2984       incomplete, preliminary documentation for NLFFI.
2985    
2986    ----------------------------------------------------------------------
2987    Name: Matthias Blume
2988    Date: 2001/05/14 11:30:00 EDT
2989    Tag: blume-20010514-script
2990    Description:
2991    
2992    Some bugs in install script fixed.
2993    
2994    In addition to that I also made a slight change to the NLFFI API:
2995    Functors generated by ml-nlffigen now take the dynamic library as a
2996    straight functor argument, not as a suspended one.  (The original
2997    functor code used to force the suspension right away anyway, so there
2998    was nothing gained by this complication of the interface.)
2999    
3000    ----------------------------------------------------------------------
3001    Name: Matthias Blume
3002    Date: 2001/05/11 14:35:00 EDT
3003    Tag: blume-20010511-ml-nlffi
3004    Description:
3005    
3006    I finally took the plunge and added my new FFI code to the main
3007    repository.  For x86-linux it is now ready for prime-time.
3008    
3009    There are two new subdirectories of "src":
3010    
3011      - ml-nlffi-lib:
3012           The utility library for programs using the FFI interface.
3013           Here is the implementation of $/c.cm and its associated low-level
3014           partners $/c-int.cm and $/memory.cm.
3015      - ml-nlffigen:
3016           A stand-alone program for generating ML glue code from C source
3017           code.
3018    
3019    Building ml-nlffigen requires $/ckit-lib.cm.
3020    
3021    The config/install.sh script has been updates to do the Right Thing
3022    (hopefully).
3023    
3024    Notice that the source tree for the C-Kit will not be put under "src"
3025    but directly under the installation root directory.  (This is the
3026    structure that currently exists on the CVS server when you check out
3027    module "sml".)  Fortunately, config/install.sh knows about this oddity.
3028    
3029    Bugs: No documentation yet.
3030    
3031    ----------------------------------------------------------------------
3032    Name: Matthias Blume
3033    Date: 2001/05/09 16:35:00 EDT
3034    Tag: blume-20010509-cpscontract
3035    Description:
3036    
3037    Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
3038    wrapper/unwrapper elimination did not decrement usage counts and some
3039    dead variables got overlooked by the dead-up logic.)
3040    
3041    ----------------------------------------------------------------------
3042    Name: Lal George
3043    Date: 2001/05/08  17:26:09 EDT
3044    Tag: george-20010508-omit-frameptr
3045    Description:
3046    
3047    Changes to implement the omit-frame-pointer optimization to support
3048    raw C calls. For now, there is only support on the Intel x86, but
3049    other architectures will follow as more experience is gained with this.
3050    
3051    
3052    ----------------------------------------------------------------------
3053    Name: Matthias Blume
3054    Date: 2001/05/07 14:40:00 EDT
3055    Tag: blume-20010507-proxies
3056    Description:
3057    
3058    I made into "proxy libraries" all libraries that qualify for such a
3059    change.  (A qualifying library is a library that has another library or
3060    groups as its sole member and repeats that member's export list
3061    verbatim.  A proxy library avoids this repetition by omitting its export
3062    list, effectively inheriting the list that its (only) member exports.
3063    See the CM manual for more explanation.)
3064    The main effect is that explicit export lists for these libraries
3065    do not have to be kepts in sync, making maintenance a bit easier.
3066    
3067    I also added copyright notices to many .cm-files.
3068    
3069    Last but not least, I made a new set of bootfiles.
3070    
3071    ----------------------------------------------------------------------
3072    Name: Matthias Blume
3073    Date: 2001/05/04 17:00:00 EDT
3074    Tag: blume-20010504-cm-lsplit
3075    Description:
3076    
3077    0. John merged pending changes to $/smlnj-lib.cm
3078    
3079    1. Allen's previous change accidentally backed out of one of Lal's
3080       earlier changes.  I undid this mistake (re-introducing Lal's change).
3081    
3082    2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
3083       within the compiler where applicable.  There is some code simplification
3084       because of that.
3085    
3086    3. The "split" phase (in FLINT) is now part of the default list of phases.
3087       Compiler.Control.LambdaSplitting.* can be used to globally control the
3088       lambda-splitting (cross-module-inlining) engine.  In addition to that,
3089       it can now also be controlled on a per-source basis: CM has been taught
3090       a new tool parameter applicable to ML source files.
3091    
3092       - To turn lambda-splitting off completely:
3093            local open Compiler.Control.LambdaSplitting in
3094                val _ = set Off
3095            end
3096       - To make "no lambda-splitting" the global default (but allow per-source
3097         overriding); this is the initial setting:
3098            local open Compiler.Control.LambdaSplitting in
3099                val _ = set (Default NONE)
3100            end
3101       - To make "lambda-splitting with aggressiveness a" the global default
3102         (and allow per-source overriding):
3103            local open Compiler.Control.LambdaSplitting in
3104                val _ = set (Default (SOME a))
3105            end
3106    
3107       - To turn lambda-splitting off for a given ML souce file (say: a.sml)
3108         write (in the respective .cm-file):
3109            a.sml (lambdasplitting:off)
3110       - To turn lambda-splitting for a.sml on with minimal aggressiveness:
3111            a.sml (lambdasplitting:on)
3112       - To turn lambda-splitting for a.sml on with aggressiveness <a> (where
3113         <a> is a decimal non-negative integer):
3114            a.sml (lambdasplitting:<a>)
3115       - To turn lambda-splitting for a.sml on with maximal aggressiveness:
3116            a.sml (lambdasplitting:infinity)
3117       - To use the global default for a.sml:
3118            a.sml (lambdasplitting:default)
3119         or simply
3120            a.sml
3121    
3122    ----------------------------------------------------------------------
3123    Name: Allen Leung
3124    Date: 2001/05/04 01:57:00 EDT
3125    Tag: leunga-20010504-sync
3126    Description:
3127    
3128      MLRISC features.
3129    
3130      1. Fix to CMPXCHG instructions.
3131      2. Changed RA interface to allow annotations in callbacks.
3132      3. Added a new method to the stream interface to allow annotations updates.
3133    
3134    ----------------------------------------------------------------------
3135    Name: Matthias Blume
3136    Date: 2001/05/01 11:45:00 EDT
3137    Tag: blume-20010501-pcedittmp
3138    Description:
3139    
3140    Changed install.sh to use the current working directory instead of
3141    /usr/tmp for a temporary file (pcedittmp).  The previous choice
3142    of /usr/tmp caused trouble with MacOS X because of file premission
3143    problems.
3144    
3145    ----------------------------------------------------------------------
3146    Name: Matthias Blume
3147    Date: 2001/04/20 11:10:00 EDT
3148    Tag: blume-20010420-inMLflag
3149    Description:
3150    
3151     - added vp_limitPtrMask to vproc-state.h
3152       (for use by the raw-C-calls mechanism to implement proper interrupt
3153        handling)
3154     - made the ML compiler aware of various data-structure offsets so it
3155       can generate code for accessing the vp_inML flag and vp_limitPtrMask
3156     - tweaked mlriscGen.sml to have it emit interrupt-handling code for
3157       raw C-calls
3158    
3159    ----------------------------------------------------------------------
3160    Name: Lal George
3161    Date: 2001/04/20 09:15:28 EDT
3162    Tag: george-20010420-macosX
3163    Description:
3164    
3165     - Changes to port to Mac OS X; Darwin.
3166    
3167     - In the process I found that sqrt was broken on the PPC, because the
3168       fsqrt instruction is not implemented.
3169    
3170    ----------------------------------------------------------------------
3171    Name: Matthias Blume
3172    Date: 2001/04/18 12:45:00 EDT
3173    Tag: blume-20010418-ccalls
3174    Description:
3175    
3176     - fixed two off-by-4 errors in the x86-specific c-calls implementation
3177       (this bug prevented structure arguments containing pointers from being
3178        passed correctly)
3179     - changed the raw-C-call code in mlriscGen.sml in such a way that
3180       structure arguments are represented as a pointer to the beginning
3181       of the structure (instead of having a series of synthesized arguments,
3182       one for each structure member)
3183    
3184     - made makeml script's verbosity level configurable via environment
3185       variable (MAKEML_VERBOSITY)
3186    
3187     - eliminated placeholder implementations for f32l, w16s, i16s, and f32s
3188       in rawmem-x86.sml; we are now using the real thing
3189    
3190    ----------------------------------------------------------------------
3191    Name: Matthias Blume
3192    Date: 2001/03/22 16:25:00 EST
3193    Tag: blume-20010322-bootfiles
3194    Description:
3195    
3196    Created a new set of bootfiles (for your automatic installation convenience).
3197    
3198    ----------------------------------------------------------------------
3199    Name: Matthias Blume
3200    Date: 2001/03/22 15:10:00 EST
3201    Tag: blume-20010322-rawmem-parcm
3202    Description:
3203    
3204    1. All "raw memory access" primitives for the new FFI are implemented now
3205       (at least on the x86).
3206    2. Some further cleanup of CM's parallel make mechanism.
3207    
3208    ----------------------------------------------------------------------
3209    Name: Matthias Blume
3210    Date: 2001/03/19 17:53:00 EST
3211    Tag: blume-20010319-parallel
3212    Description:
3213    
3214    Parallel make (using compile servers) now works again.
3215    
3216    To this end, CM.stabilize and CMB.make have been modified to work in
3217    two passes when compile servers are attached:
3218       1. Compile everything, do not perform stabilization; this pass
3219          uses compile servers
3220       2. Stabilize everything; this pass does not use compile servers
3221    If there are no compile servers, the two passes are combined into one
3222    (as before).  Splitting the passes increases the inherent parallelism
3223    in the dependency graph because the entire graph including all
3224    libraries is available at the same time.  This, in turn, improves
3225    server utilization.  The downside is that the master process will
3226    have to do some extra work after compilation is done (because for
3227    technical reasons it must re-read all the binfiles during stabilization).
3228    
3229    ----------------------------------------------------------------------
3230    Name: Matthias Blume
3231    Date: 2001/03/16 12:22:00 EST
3232    Tag: blume-20010316-bootfiles
3233    Description:
3234    
3235    Created a new set of bootfiles (for your automatic installation convenience).
3236    
3237    ----------------------------------------------------------------------
3238    Name: Matthias Blume
3239    Date: 2001/03/16 11:00:00 EST
3240    Tag: blume-20010316-MLTREE-fixup
3241    Description:
3242    
3243    This is a minor fixup for an (untagged) earlier commit by Allen.
3244    (A file was missing).
3245    
3246    ----------------------------------------------------------------------
3247    Name: Allen Leung
3248    Date: Mon Mar  5 18:54:57 EST 2001
3249    Tag: leunga-20010305-cut-support
3250    
3251    1. New support for alternative control-flow in MLTREE.
3252       Currently we support
3253    
3254          FLOW_TO(CALL ...., [k1,...,kn])
3255    
3256       This is needed for 'cuts to' in C-- and try/handle-like constructs
3257       in Moby
3258    
3259       New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
3260    
3261    2. Register Allocator
3262    
3263       Changes in interface [from Fermin, John]
3264    
3265    3. Alpha 8-bit SLL support [Fermin]
3266    
3267    4. All architectures
3268    
3269       A new module (ClusterExpandCopies) for expanding parallel copies.
3270    
3271    ----------------------------------------------------------------------
3272    Name: Allen Leung
3273    Date: 2001/02/27 23:07:00 EST
3274    Tag: leunga-20010227-minor-stuff
3275    
3276    1. Alpha bug fix for CMOVNE
3277    2. Handle mltree COND(..,FCMP ...,...)
3278    3. Bug fix in simplifier
3279    
3280    ----------------------------------------------------------------------
3281    Name: Matthias Blume
3282    Date: 2001/01/30 17:50:00 EST
3283    Tag: blume-20010130-sync
3284    Description:
3285    
3286    This is just a minor update to sync my devel branch with the main brach.
3287    The only visible change is the addition of some README files.
3288    
3289    ----------------------------------------------------------------------
3290    Name: Matthias Blume
3291    Date: 2001/01/12 23:30:00 JST
3292    Tag: blume-20010112-bootfiles
3293    Description:
3294    
3295    Made a new set of bootfiles that goes with the current state of the
3296    repository.
3297    
3298    ----------------------------------------------------------------------
3299    Name: Matthias Blume
3300    Date: 2001/01/12 21:20:00 JST
3301    Tag: blume-20010112-sync
3302    Description:
3303    
3304    I am just flushing out some minor changes that had accumulated in
3305    my private branch in order to sync with the main tree.  (This is
3306    mainly because I had CVS trouble when trying to merge _into_ my
3307    private branch.)
3308    
3309    Most people should be completely unaffected by this.
3310    
3311    ----------------------------------------------------------------------
3312    Name: Allen Leung
3313    Date: Thu Jan 11 21:03:00 EST 2001
3314    Tag: leunga-20010111-labexp=mltree
3315    Description:
3316    
3317    1.  Removed the type LabelExp and replace it by MLTree.
3318    2.  Rewritten mltree-simplify with the pattern matcher tool.
3319    3.  There were some bugs in alpha code generator which would break
3320        64-bit code generation.
3321    4.  Redo the tools to generate code with the
3322    5.  The CM files in MLRISC (and in src/system/smlnj/MLRISC)
3323        are now generated by perl scripts.
3324    
3325    ----------------------------------------------------------------------
3326    Name: Matthias Blume
3327    Date: 2001/01/10 21:55:00 JST
3328    Tag: blume-20010110-rcc
3329    Description:
3330    
3331    The RCC stuff now seems to work (but only on the x86).
3332    This required hacking of the c-calls interface (and -implementation) in
3333    MLRISC.
3334    
3335    Normal compiler users should be unaffected.
3336    
3337    ----------------------------------------------------------------------
3338    Name: Matthias Blume
3339    Date: 2001/01/09 01:20:00 JST
3340    Tag: blume-20010109-rcc
3341    Description:
3342    
3343    This is a fairly big patch, flushing out a large number of pending
3344    changes that I made to my development copy over the last couple of days.
3345    
3346    Of practical relevance at this moment is a workaround for a pickling
3347    bug that Allen ran into the other day.  The cause of the bug itself is
3348    still unknown and it might be hard to fix it properly, but the
3349    workaround has some merits of its own (namely somewhat reducing pickling
3350    overhead for certain libraries).  Therefore, I think this solution should
3351    be satisfactory at this time.
3352    
3353    The rest of the changes (i.e., the vast majority) has to do with my
3354    ongoing efforts of providing direct support for C function calls from
3355    ML.  At the moment there is a new primop "RAW_CCALL", typing magic
3356    in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new
3357    case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to
3358    translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to
3359    handle RCC.
3360    
3361    The last part (the changes to mlriscGen.sml) are still known to be
3362    wrong on the x86 and not implemented on all other architectures.  But
3363    the infrastructure is in place. I had to change a few functor
3364    signatures in the backend to be able to route the CCalls interface
3365    from MLRISC there, and I had to specialize the mltree type (on the
3366    x86) to include the necessary extensions. (The extensions themselves
3367    were already there and redy to go in MLRISC/x86).
3368    
3369    Everything should be very happy as soon as someone helps me with
3370    mlriscGen.sml...
3371    
3372    In any case, nothing of this should matter to anyone as long as the
3373    new primop is not being used (which is going to be the case unless you
3374    find it where I hid it :). The rest of the compiler is completely
3375    unaffected.
3376    
3377    ----------------------------------------------------------------------
3378    Name: Matthias Blume
3379    Date: 2001/01/05 00:30:00 JST
3380    Tag: blume-20010105-primops
3381    Description:
3382    
3383    Added some experimental support for work that I am doing right now.
3384    These changes mostly concern added primops, but there is also a new
3385    experimental C library in the runtime system (but currently not enabled
3386    anywhere except on Linux/X86).
3387    
3388    In the course of adding primops (and playing with them), I discovered that
3389    Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly
3390    broken.  (Zhong was very right he labeled this stuff as "major gross hack".)
3391    To recover, I made type information in INL_PRIM mandatory and changed
3392    prim.sml as well as built-in.sml accordingly.  The InLine structure now
3393    has complete, correct type information (i.e., no bottom types).
3394    
3395    Since all these changes mean that we need new binfiles, I also bumped the
3396    version number to 110.32.1.
3397    
3398    ----------------------------------------------------------------------
3399    Name: Matthias Blume
3400    Date: 2000/12/30 22:10:00 JST
3401    Tag: blume-20001230-various
3402    Description:
3403    
3404    Added proxy libraries for MLRISC and let MLRISC libraries refer
3405    to each other using path anchors.  (See CM manual for explanation.)
3406    
3407    Updated CM documentation.
3408    
3409    Fixed some bugs in CM.
3410    
3411    Implemented "proxy" libraries (= syntactic sugar for CM).
3412    
3413    Added "-quiet" option to makeml and changed runtime system accordingly.
3414    
3415    Added cleanup handler for exportML to reset timers and compiler stats.
3416    
3417    ----------------------------------------------------------------------
3418    Name: Lal George
3419    Date: 2000/12/22 22:22:58 EST 2000
3420    Tag: Release_110_32
3421    Description:
3422    
3423            Infinite precision used throughout MLRISC.
3424            see MLRISC/mltree/machine-int.sig
3425    
3426    ----------------------------------------------------------------------
3427    Name: Matthias Blume
3428    Date: 2000/12/22 23:16:00 JST
3429    Tag: blume-20001222-warn
3430    Description:
3431    
3432    Corrected wording and formatting of some CM warning message which I
3433    broke in my previous patch.
3434    
3435    ----------------------------------------------------------------------
3436    Name: Matthias Blume
3437    Date: 2000/12/22 21:20:00 JST
3438    Tag: blume-20001222-anchorenv
3439    Description:
3440    
3441    Fixed CM's handling of anchor environments in connection with CMB.make.
3442    
3443    ----------------------------------------------------------------------
3444    Name: Matthias Blume
3445    Date: 2000/12/22 13:15:00 JST
3446    Tag: blume-20001222-cleanup
3447    Description:
3448    
3449    Removed src/cm/ffi which does not (and did not) belong here.
3450    
3451    ----------------------------------------------------------------------
3452    Name: Matthias Blume
3453    Date: 2000/12/21 23:55:00 JST
3454    Tag: blume-20001221-exn
3455    Description:
3456    
3457    Probably most important: CM no longer silently swallows all exceptions
3458    in the compiler.
3459    Plus: some other minor CM changes.  For example, CM now reports some
3460    sizes for generated binfiles (code, data, envpickle, lambdapickle).
3461    
3462    ----------------------------------------------------------------------
3463    Name: Matthias Blume
3464    Date: 2000/12/15 00:01:05 JST
3465    Tag: blume-20001215-dirtool
3466    Description:
3467    
3468    - "dir" tool added.
3469    - improvements and cleanup to Tools structure
3470    - documentation updates
3471    
3472    ----------------------------------------------------------------------
3473    Name: Allen Leung
3474    Date: Thu Dec 14 03:45:24 EST 2000
3475    Description:
3476    Tag:  leunga-20001214-int-inf
3477    Description:
3478    
3479       In IntInf, added these standard functions, which are missing from our
3480    implementation:
3481    
3482        andb : int * int -> int
3483        xorb : int * int -> int
3484        orb  : int * int -> int
3485        notb : int -> int
3486         <<   : int * word -> int
3487        ~>>  : int * word -> int
3488    
3489       Not tested, I hope they are correct.
3490    
3491    ----------------------------------------------------------------------
3492    Name: Allen Leung
3493    Date: Fri Dec  8 19:23:26 EST 2000
3494    Description:
3495    Tag:  leunga-20001208-nowhere
3496    Description:
3497    
3498      Slight improvements to the 'nowhere' tool to handle OR-patterns,
3499    to generate better error messages etc.  Plus a brief manual.
3500    
3501    ----------------------------------------------------------------------
3502    Name: Lal George
3503    Date: 2000/12/08 09:54:02 EST 2000
3504    Tag: Release_110_31
3505    Description:
3506    
3507    - Version 110.31
3508    ----------------------------------------------------------------------
3509    Name: Allen Leung
3510    Date: Thu Dec  7 22:01:04 EST 2000
3511    Tag:  leunga-20001207-cell-monster-hack
3512    Description:
3513    
3514    Major MLRISC internal changes.  Affect all clients.
3515    Summary:
3516    
3517    1.  Type CELLS.cell = int is now replaced by a datatype.
3518        As a result, the old regmap is now gone.  Almost all interfaces
3519        in MLRISC change as a consequence.
3520    
3521    2.  A new brand version of machine description tool (v3.0) that generates
3522        modules expecting the new interface.  The old version is removed.
3523    
3524    3.  The RA interface has been further abstracted into two new functors.
3525        RISC_RA and X86RA.  These functors have much simpler interfaces.
3526        [See also directory MLRISC/demo.]
3527    
3528    4.  Some other new source->source code generation tools are available:
3529    
3530        a. MLRISC/Tools/RewriteGen -- generate rewriters from rules.
3531        b. MLRISC/Tools/WhereGen -- expands conditional pattern matching rules.
3532           I use this tool to generate the peephole optimizers---with the new
3533           cell type changes, peephole rules are becoming difficult to write
3534           without conditional pattern matching.
3535    
3536    5.  More Intmap -> IntHashTable change.  Previous changes by Matthias didn't
3537        cover the entire MLRISC source tree so many things broke.
3538    
3539    6.  CM files have been moved to the subdirectory MLRISC/cm.
3540        They are moved because there are a lot of them and they clutter up the
3541        root dir.
3542    
3543    7.  More detailed documentation to come...
3544    
3545        NOTE: To rebuild from 110.30 (ftp distribution), you'll have to do
3546        a makeml -rebuild first.  This is because of other other
3547        changes that Matthias has made (see below).
3548    
3549    
3550    ----------------------------------------------------------------------
3551    Name: Matthias Blume
3552    Date: 2000/11/30 23:12:00 JST
3553    Tag: blume-20001130-filereorg
3554    Description:
3555    
3556    Some manual updates and some file reorganizations in CM.
3557    
3558    ----------------------------------------------------------------------
3559    Name: Matthias Blume
3560    Date: 2000/11/24 17:45:00 JST
3561    Tag: blume-20001124-link
3562    Description:
3563    
3564    Drastically improved link traversal code for the case that the dynamic
3565    value was already loaded at bootstrap time.  As a result, CM and CMB
3566    now both load blazingly fast -- even on a very slow machine.  Also,
3567    memory consumption has been further reduced by this.
3568    
3569    Warning: The format of the PIDMAP file has changed.  THerefore, to
3570    bootstrap you have to do this:
3571    
3572    1. Run CMB.make
3573    2. Make a symbolic link for the boot directory:
3574         ln -s sml.boot.ARCH-OS xxx
3575    3. "Rebuild" the boot directory:
3576         ./makeml -boot xxx -rebuild sml ; rm xxx
3577    4. Boot normally:
3578          ./makeml
3579    
3580    ----------------------------------------------------------------------
3581    Name: Matthias Blume
3582    Date: 2000/11/21 21:20:00 JST
3583    Tag: blume-20001121-tools
3584    Description:
3585    
3586    Continued hacking on autoloading problem -- with success this time.
3587    Also changed tool-plugin mechanism.  See new CM manual.
3588    
3589    ----------------------------------------------------------------------
3590    Name: Matthias Blume
3591    Date: 2000/11/19 14:30:00 JST
3592    Tag:  blume-20001119-autoload
3593    Description:
3594    
3595    Some hacking to make autoloading faster.  Success for CMB, no success
3596    so far for CM.  There is a reduced structure CM' that autoloads faster.
3597    (This is a temporary, non-documented hack to be eliminated again when
3598    the general problem is solved.)
3599    
3600    ----------------------------------------------------------------------
3601    Name: Matthias Blume
3602    Date: 2000/11/17 14:10:00 JST
3603    Tag: blume-20001117-pickle-lib
3604    Description:
3605    
3606    1. Eliminated comp-lib.cm
3607    2. Made pickle-lib.cm
3608    3. Eliminated all uses of intset.sml (from comp-lib.cm)
3609    4. Replaced all uses of intmap.{sig,sml} (from comp-lib.cm) with
3610       equivalent constructs from smlnj-lib.cm (INtHashTable).
3611    5. Point 4. also goes for those uses of intmap.* in MLRISC.
3612       Duplicated intmap modules thrown out.
3613    6. Hunted down all duplicated SCC code and replaced it with
3614       equivalent stuff (GraphSCCFn from smlnj-lib.cm).
3615    7. Rewrote Feedback module.
3616    8. Moved sortedlist.sml into viscomp-lib.cm.  Eventually it
3617       should be thrown out and equivalent modules from smlnj-lib.cm
3618       should be used (IntRedBlackSet, IntListSet, ...).
3619    
3620    Confirmed that compiler compiles to fixpoint.
3621    
3622    ----------------------------------------------------------------------