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 1062, Wed Feb 13 21:15:14 2002 UTC revision 1288, Tue Oct 29 14:48:54 2002 UTC
# Line 14  Line 14 
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16  Name: Matthias Blume  Name: Matthias Blume
17    Date: 2002/10/28 09:50:00 EST
18    Tag: blume-20021028-typecheck
19    Description:
20    
21    Exported structure Typecheck from $smlnj/viscomp/core.cm.
22    
23    ----------------------------------------------------------------------
24    Name: Matthias Blume
25    Date: 2002/10/17 09:10:00 EDT
26    Tag: Release_110_42
27    Description:
28    
29    In good old tradition, there has been a slight hiccup so that we have
30    to patch 110.42 after the fact.   The old release tag has been replaced
31    (see below).
32    
33    The change solves a problem with two competing approaches the
34    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
35    each other's way.
36    
37    This change only affects the runtime system code and the installer script.
38    (No new bootfiles.)
39    
40    ----------------------------------------------------------------------
41    Name: Matthias Blume
42    Date: 2002/10/16 12:00:00 EDT
43    Tag: Release_110_42_removed
44    Description:
45    
46    New working release.  New bootfiles.
47    
48    ----------------------------------------------------------------------
49    Name: Matthias Blume
50    Date: 2002/10/10 13:10:00 EDT
51    Tag: blume-20021010-ppc-divs
52    Description:
53    
54    The mltree operator DIVS must be implemented with an overflow check on
55    the PPC because the hardware indicates divide-by-zero using "overflow" as
56    well.
57    
58    ----------------------------------------------------------------------
59    Name: Matthias Blume
60    Date: 2002/07/23 11:45:00
61    Tag: blume-20020723-smlnj-home
62    Description:
63    
64    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
65    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
66    CM_PATHCONFIG is also set), the path configuration file is assumed
67    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
68    the entire tree to some other place and everything will "just work".
69    
70    (Companion commands such as ml-build and ml-makedepend also sense this
71    variable.)
72    
73    ----------------------------------------------------------------------
74    Name: Matthias Blume
75    Date: 2002/07/12 21:19:00 EDT
76    Tag: blume-20020712-liveness
77    Description:
78    
79    Exported two useful "step" functions from liveness module (MLRISC).
80    
81    ----------------------------------------------------------------------
82    Name: Matthias Blume
83    Date: 2002/07/05 16:00 EDT
84    Tag: Release_110_41
85    Description:
86    
87    New working release.  New bootfiles.
88    
89    ----------------------------------------------------------------------
90    Name: Matthias Blume
91    Date: 2002/07/05 10:25:00 EDT
92    Tag: blume-20020705-btimp
93    Description:
94    
95    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
96    other clients can set up backtracing support.
97    
98    ----------------------------------------------------------------------
99    Name: Matthias Blume
100    Date: 2002/06/25 17:23:00 EDT
101    Tag: blume-20020625-fpmax
102    Description:
103    
104    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
105    case.  (The sense of the isNaN test was reversed -- which made min and
106    max always return their first argument.)
107    
108    ----------------------------------------------------------------------
109    Name: Matthias Blume
110    Date: 2002/06/11
111    Tag: blume-20020611-unixpath
112    Description:
113    
114    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
115    
116    ----------------------------------------------------------------------
117    Name: Matthias Blume
118    Date: 2002/06/10 16:35:00 EDT
119    Tag: blume-20020610-ieeereal
120    Description:
121    
122    I back-ported my implementation of IEEEReal.fromString from the
123    idlbasis-devel branch so that we can test it.
124    
125    Another small change is that ppDec tries to give more information
126    than just "<sig>" in the case of functors.  However, this code is
127    broken in some mysterious way if the functor's body's signature
128    has not been declared by ascription but gets inferred from the
129    implementation.  This needs fixing...
130    
131    ----------------------------------------------------------------------
132    Name: Matthias Blume
133    Date: 2002/05/31
134    Tag: blume-20020531-btrace-mode
135    Description:
136    
137    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
138    the wayside when I switched over to using Controls everywhere.)
139    
140    ----------------------------------------------------------------------
141    Name: Lal George
142    Date: 2002/05/23 12:21:40 EDT
143    Tag: george-20020523-visual-labels
144    Description:
145    
146    Labels are now displayed in the graphical output to make
147    the fall-through and target blocks obvious.
148    
149    ----------------------------------------------------------------------
150    Name: Matthias Blume
151    Date: 2002/05/22 11:03:00 EDT
152    Tag: blume-20020522-shrink
153    Description:
154    
155    John tweaked yesterday's fix for 1131 to handle an out-of-memory
156    situation that comes up when allocating huge arrays.
157    
158    ----------------------------------------------------------------------
159    Name: Matthias Blume
160    Date: 2002/05/21 16:00:00 EDT
161    Tag: Release_110_40
162    Description:
163    
164    New working release (110.40).  New bootfiles.
165    
166    [Also: John Reppy fixed GC bug 1131.]
167    
168    ----------------------------------------------------------------------
169    Name: Matthias Blume
170    Date: 2002/05/21 12:35:00 EDT
171    Tag: blume-20020521-cmdoc
172    Description:
173    
174    CM documentation update.
175    
176    ----------------------------------------------------------------------
177    Name: Matthias Blume
178    Date: 2002/05/21 10:55:00 EDT
179    Tag: blume-20020521-misc
180    Description:
181    
182    - John tweaked runtime to be silent on heap export (except when
183      GC messages are on).
184    - I added a few more things (cross-compiling versions of CMB) to
185      config/preloads (as suggestions).
186    
187    ----------------------------------------------------------------------
188    Name: Matthias Blume
189    Date: 2002/05/20 22:25:00 EDT
190    Tag: blume-20020520-controls
191    Description:
192    
193    - Added ControlUtil structure to control-lib.cm.
194    - Use it throughout.
195    - Used Controls facility to define MLRISC controls (as opposed to
196      registering MLRISC control ref cells with Controls after the fact)
197    - Fixed messed-up controls priorities.
198    
199    * Removed again all the stuff from config/preloads that one wouldn't
200      be able to preload at the time the initial heap image is built.
201      (Many libraries, e.g., CML, do not exist yet at this time.  The
202       only libraries that can be preloaded via config/preloads are those
203       that come bundled with the bootfiles.)
204    
205    ----------------------------------------------------------------------
206    Name: Matthias Blume
207    Date: 2002/05/20 10:59:00 EDT
208    Tag: blume-20020520-preloads
209    Description:
210    
211    Added a lot of commented-out suggestions for things to be included
212    in config/preloads.
213    
214    ----------------------------------------------------------------------
215    Name: Allen Leung
216    Date: 2002/05/18 14:20:00 EDT
217    Tag: leunga-20020518-mdl
218    Description:
219    
220    o Made the mdl tool stuff compile and run again.
221    o I've disabled all the stuff that depends on RTL specifications; they
222      are all badly broken anyway.
223    
224    ----------------------------------------------------------------------
225    Name: Matthias Blume
226    Date: 2002/05/17 16:49:00 EDT
227    Tag: blume-20020517-controls
228    Description:
229    
230    0. John Reppy made several modifications to the SML/NJ library.
231       In particular, there is a shiny new controls-lib.cm.
232    
233    1. Pushed new controls interface through compiler so that everything
234       compiles again.
235    
236    2. Added FormatComb and FORMAT_COMB to the CML version of the
237       SML/NJ library (so that CML compiles again).
238    
239    3. Modified init scripts because XXX_DEFAULT environment variables
240       are no longer with us.  (Boot-time initialization is now done
241       using the same environment variables that are also used for
242       startup-time initialization of controls.)
243    
244    ----------------------------------------------------------------------
245    Name: Lal George
246    Date: 2002/05/15 09:20:10 EDT
247    Tag: george-20020515-pseudo-op-decls
248    Description:
249    
250            All pseudo-ops emitted before the first segment declaration
251    such as TEXT, DATA, and BSS directives are assumed to be global
252    declarations and are emitted first in the assembly file. This is
253    useful in a number of situations where one has pseudo-ops that are not
254    specific to any segment, and also works around the constraint that one
255    cannot have client pseudo-ops in the TEXT segment.
256    
257            Because no segment is associated with these declarations it is
258    an error to allocate any space or objects before the first segment
259    directive and an exception will be raised. However, we cannot make
260    this check for client pseudo-ops.
261    
262            These top level declarations are a field in the CFG graph_info.
263    In theory you can continue to add to this field after the CFG has been
264    built -- provided you know what you are doing;-)
265    
266    ----------------------------------------------------------------------
267    Name: Matthias Blume
268    Date: 2002/05/13 16:40:00 EDT
269    Tag: blume-20020513-pp-etc
270    Description:
271    
272    A few minor bugfixes:
273    
274      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
275        (Bogus printouts for redefined bindings still occur.  Compiler
276        bug should no longer occur now.  We need to redo the prettyprinter
277        from scratch.)
278    
279      - CM pathname printer now also adds escape sequences for ( and )
280    
281      - commend and docu fixes for ml-nlffi
282    
283    ----------------------------------------------------------------------
284    Name: Matthias Blume
285    Date: 2002/05/10 16:40:00 EDT
286    Tag: blume-20020510-erg-textio
287    Description:
288    
289    Applied the following bugfix provided by Emden Gansner:
290    
291        Output is corrupted when outputSubstr is used rather than output.
292    
293        The problem occurs when a substring
294    
295            ss = (s, dataStart, dataLen)
296    
297        where dataStart > 0, fills a stream buffer with avail bytes left.
298        avail bytes of s, starting at index dataStart, are copied into the
299        buffer, the buffer is flushed, and then the remaining dataLen-avail
300        bytes of ss are copied into the beginning of the buffer. Instead of
301        starting this copy at index dataStart+avail in s, the current code
302        starts the copy at index avail.
303    
304        Fix:
305        In text-io-fn.sml, change line 695 from
306             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
307        to
308             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
309    
310    ----------------------------------------------------------------------
311    Name: Matthias Blume
312    Date: 2002/04/12 13:55:00 EDT
313    Tag: blume-20020412-assyntax
314    Description:
315    
316    1. Grabbed newer assyntax.h from the XFree86 project.
317    2. Fiddled with how to compile X86.prim.asm without warnings.
318    3. (Very) Minor cleanup in CM.
319    
320    ----------------------------------------------------------------------
321    Name: Matthias Blume
322    Date: 2002/04/01 (no joke!) 17:07:00 EST
323    Tag: blume-20020401-x86div
324    Description:
325    
326    Added full support for div/mod/rem/quot on the x86, using the machine
327    instruction's two results (without clumsily recomputing the remainder)
328    directly where appropriate.
329    
330    Some more extensive power-of-two support was added to the x86 instruction
331    selector (avoiding expensive divs, mods, and muls where they can be
332    replaced with cheaper shifts and masks).  However, this sort of thing
333    ought to be done earlier, e.g., within the CPS optimizer so that
334    all architectures benefit from it.
335    
336    The compiler compiles to a fixed point, but changes might be somewhat
337    fragile nevertheless.  Please, report any strange things that you might
338    see wrt. div/mod/quot/rem...
339    
340    ----------------------------------------------------------------------
341    Name: Matthias Blume
342    Date: 2002/03/29 17:22:00
343    Tag: blume-20020329-div
344    Description:
345    
346    Fixed my broken div/mod logic.  Unfortunately, this means that the
347    inline code for div/mod now has one more comparison than before.
348    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
349    The problem was with quotient = 0, because that alone does not tell
350    us which way the rounding went.  One then has to look at whether
351    remainder and divisor have the same sign...  :(
352    
353    Anyway, I replaced the bootfiles with fresh ones...
354    
355    ----------------------------------------------------------------------
356    Name: Matthias Blume
357    Date: 2002/03/29 14:10:00 EST
358    Tag: blume-20020329-inlprims
359    Description:
360    
361    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
362    
363    Primops have changed. This means that the bin/boot-file formats have
364    changed as well.
365    
366    To make sure that there is no confusion, I made a new version.
367    
368    
369    CHANGES:
370    
371    * removed REMT from mltree (remainder should never overflow).
372    
373    * added primops to deal with divisions of all flavors to the frontend
374    
375    * handled these primops all the way through so they map to their respective
376      MLRISC support
377    
378    * used these primops in the implementation of Int, Int32, Word, Word32
379    
380    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
381    
382    * parameterized INLMIN, INLMAX, and INLABS by a numkind
383    
384    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
385      floating point
386    
387    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
388      and Real (but Real.abs maps to a separate floating-point-only primop)
389    
390    
391    TODO items:
392    
393    * Hacked Alpha32 instruction selection, disabling the selection of REMx
394      instructions because the machine instruction encoder cannot handle
395      them.  (Hppa, PPC, and Sparc instruction selection did not handle
396      REM in the first place, and REM is supported by the x86 machine coder.)
397    
398    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
399      selection phase.  (The two can be streamlined because the hardware
400      delivers both quotient and remainder at the same time anyway.)
401    
402    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
403      (Currently the behavior is inconsistent both across architectures and
404      wrt. the draft Basis spec.)
405    
406    * Word8 should eventually be handled natively, too.
407    
408    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
409      as if there currently is no execution path that could trigger it in
410      SML/NJ.  (The assumptions underlying functions arith and promotable do not
411      hold for things like multiplication and division.)
412    
413    ----------------------------------------------------------------------
414    Name: Matthias Blume
415    Date: 2002/03/27 16:27:00 EST
416    Tag: blume-20020327-mlrisc-divisions
417    Description:
418    
419    Added support for all four division operations (ML's div, mod, quot,
420    and rem) to MLRISC.  In the course of doing so, I also rationalized
421    the naming (no more annoying switch-around of DIV and QUOT), by
422    parameterizing the operation by div_rounding_mode (which can be either
423    DIV_TO_ZERO or DIV_TO_NEGINF).
424    
425    The generic MLTreeGen functor takes care of compiling all four
426    operations down to only round-to-zero div.
427    
428    Missing pieces:
429    
430      * Doing something smarter than relying on MLTreeGen on architectures
431        like, e.g., the x86 where hardware division delivers both quotient and
432        remainder at the same time.  With this, the implementation of the
433        round-to-neginf operations could be further streamlined.
434    
435      * Remove inlining support for div/mod/rem from the frontend and replace it
436        with primops that get carried through to the backend.  Do this for all
437        int and word types.
438    
439    ----------------------------------------------------------------------
440    Name: Matthias Blume
441    Date: 2002/03/25 17:25:00 EST
442    Tag: blume-20020325-divmod
443    Description:
444    
445    I improved (hopefully without breaking them) the implementation of Int.div,
446    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
447    advantage of the following observations:
448    
449      Let  q = x quot y      r = x rem y
450           d = x div  y      m = x mod y
451    
452    where "quot" is the round-to-zero version of integer division that
453    hardware usually provides.  Then we have:
454    
455         r = x - q * y        where neither the * nor the - will overflow
456         d = if q >= 0 orelse x = q * y then q else q - 1
457                              where neither the * nor the - will overflow
458         m = if q >= 0 orelse r = 0 then r else r + y
459                              where the + will not overflow
460    
461    This results in substantial simplification of the generated code.
462    The following table shows the number of CFG nodes and edges generated
463    for
464            fun f (x, y) = x OPER y
465            (* with OPER \in div, mod, quot, rem *)
466    
467    
468        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
469        --------------------------------------------------------
470         div |         24 |         39 |         12 |         16
471         mod |         41 |         71 |         12 |         16
472        quot |          8 |         10 |          8 |         10
473         rem |         10 |         14 |          8 |         10
474    
475    
476    ----------------------------------------------------------------------
477    Name: Matthias Blume
478    Date: 2002/03/25 22:06:00 EST
479    Tag: blume-20020325-cprotobug
480    Description:
481    
482    Fixed a bug in cproto (c prototype decoder).
483    
484    ----------------------------------------------------------------------
485    Name: Matthias Blume
486    Date: 2002/03/25 16:00:00 EST
487    Tag: blume-20020325-raw-primops
488    Description:
489    
490    I did some cleanup to Allen's new primop code and
491    replaced yesterday's bootfiles with new ones.
492    (But they are stored in the same place.)
493    
494    ----------------------------------------------------------------------
495    Name: Matthias Blume
496    Date: 2002/03/24 22:40:00 EST
497    Tag: blume-20020324-bootfiles
498    Description:
499    
500    Made the bootfiles that Allen asked for.
501    
502    ----------------------------------------------------------------------
503    Name: Allen Leung
504    Date: 2002/03/23 15:50:00 EST
505    Tag: leunga-20020323-flint-cps-rcc-primops
506    Description:
507    
508      1. Changes to FLINT primops:
509    
510        (* make a call to a C-function;
511         * The primop carries C function prototype information and specifies
512         * which of its (ML-) arguments are floating point. C prototype
513         * information is for use by the backend, ML information is for
514         * use by the CPS converter. *)
515      | RAW_CCALL of { c_proto: CTypes.c_proto,
516                       ml_args: ccall_type list,
517                       ml_res_opt: ccall_type option,
518                       reentrant : bool
519                     } option
520       (* Allocate uninitialized storage on the heap.
521        * The record is meant to hold short-lived C objects, i.e., they
522        * are not ML pointers.  With the tag, the representation is
523        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
524        *)
525      | RAW_RECORD of {tag:bool,sz:int}
526      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
527    
528      2.  These CPS primops are now overloaded:
529    
530           rawload of {kind:numkind}
531           rawstore of {kind:numkind}
532    
533          The one argument form is:
534    
535             rawload {kind} address
536    
537          The two argument form is:
538    
539             rawload {kind} [ml object, byte-offset]
540    
541      3. RAW_CCALL/RCC now takes two extra arguments:
542    
543         a. The first is whether the C call is reentrant, i.e., whether
544            ML state should be saved and restored.
545         b. The second argument is a string argument specifying the name of
546            library and the C function.
547    
548         These things are currently not handled in the code generator, yet.
549    
550      4. In CProto,
551    
552         An encoding type of "bool" means "ml object" and is mapped into
553         C prototype of PTR.  Note that "bool" is different than "string",
554         even though "string" is also mapped into PTR, because "bool"
555         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
556    
557      5. Pickler/unpicker
558    
559         Changed to handle RAW_RECORD and newest RAW_CCALL
560    
561      6. MLRiscGen,
562    
563         1. Changed to handle the new rawload/rawstore/rawrecord operators.
564         2. Code for handling C Calls has been moved to a new module CPSCCalls,
565            in the file CodeGen/cpscompile/cps-c-calls.sml
566    
567      7. Added the conditional move operator
568    
569             condmove of branch
570    
571         to cps.  Generation of this is still buggy so it is currently
572         disabled.
573    
574    ----------------------------------------------------------------------
575    Name: Lal George
576    Date: 2002/03/22 14:18:25 EST
577    Tag: george-20020322-cps-branch-prob
578    Description:
579    
580    Implemented the Ball-Larus branch prediction-heuristics, and
581    incorporated graphical viewers for control flow graphs.
582    
583    Ball-Larus Heuristics:
584    ---------------------
585    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
586    
587    By design it uses the Dempster-Shafer theory for combining
588    probabilities.  For example, in the function:
589    
590        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
591    
592    the ball-larus heuristics predicts that the n=0 is unlikely
593    (OH-heuristic), and the 'then' branch is unlikely because of the
594    RH-heuristic -- giving the 'then' branch an even lower combined
595    probability using the Dempster-Shafer theory.
596    
597    Finally, John Reppy's loop analysis in MLRISC, further lowers the
598    probability of the 'then' branch because of the loop in the else
599    branch.
600    
601    
602    Graphical Viewing:
603    ------------------
604    I merely plugged in Allen's graphical viewers into the compiler. The
605    additional code is not much. At the top level, saying:
606    
607            Control.MLRISC.getFlag "cfg-graphical-view" := true;
608    
609    will display the graphical view of the control flow graph just before
610    back-patching.  daVinci must be in your path for this to work. If
611    daVinci is not available, then the default viewer can be changed
612    using:
613    
614            Control.MLRISC.getString "viewer"
615    
616    which can be set to "dot" or "vcg" for the corresponding viewers. Of
617    course, these viewers must be in your path.
618    
619    The above will display the compilation unit at the level of clusters,
620    many of which are small, boring, and un-interesting. Also setting:
621    
622            Control.MLRISC.getInt "cfg-graphical-view_size"
623    
624    will display clusters that are larger than the value set by the above.
625    
626    
627    ----------------------------------------------------------------------
628    Name: Matthias Blume
629    Date: 2002/03/21 22:20:00 EST
630    Tag: blume-20020321-kmp-bugfix
631    Description:
632    
633    Changed the interface to the KMP routine in PreString and fixed
634    a minor bug in one place where it was used.
635    
636    ----------------------------------------------------------------------
637    Name: Allen Leung
638    Date: 2002/03/21 20:30:00 EST
639    Tag: leunga-20020321-cfg
640    Description:
641    
642      Fixed a potential problem in cfg edge splitting.
643    
644    ----------------------------------------------------------------------
645    Name: Allen Leung
646    Date: 2002/03/21 17:15:00 EST
647    Tag: leunga-20020321-x86-fp-cfg
648    Description:
649    
650      1. Recoded the buggy parts of x86-fp.
651    
652         a. All the block reordering code has been removed.
653            We now depend on the block placement phases to do this work.
654    
655         b. Critical edge splitting code has been simplified and moved into the
656            CFG modules, as where they belong.
657    
658         Both of these were quite buggy and complex.  The code is now much, much
659         simpler.
660    
661      2. X86 backend.
662    
663         a. Added instructions for 64-bit support.  Instruction selection for
664            64-bit has not been committed, however, since that
665            requires changes to MLTREE which haven't been approved by
666            Lal and John.
667    
668         b. Added support for FUCOMI and FUCOMIP when generating code for
669            PentiumPro and above.  We only generate these instructions in
670            the fast-fp mode.
671    
672         c. Added cases for JP and JNP in X86FreqProps.
673    
674      3. CFG
675    
676         CFG now has a bunch of methods for edge splitting and merging.
677    
678      4. Machine description.
679    
680         John's simplification of MLTREE_BASIS.fcond broke a few machine
681         description things:
682    
683         rtl-build.{sig,sml} and hppa.mdl fixed.
684    
685         NOTE: the machine description stuff in the repository is still broken.
686               Again, I can't put my fixes in because that involves
687               changes to MLTREE.
688    
689    ----------------------------------------------------------------------
690    Name: Matthias Blume
691    Date: 2002/03/20 15:55:00 EST
692    Tag: blume-20020320-kmp
693    Description:
694    
695    Implemented Knuth-Morris-Pratt string matching in PreString and used
696    it for String.isSubstring, Substring.isSubstring, and
697    Substring.position.
698    
699    (Might need some stress-testing.  Simple examples worked fine.)
700    
701    ----------------------------------------------------------------------
702    Name: Matthias Blume
703    Date: 2002/03/19 16:37:00 EST
704    Tag: blume-20020319-witnesses
705    Description:
706    
707    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
708    
709    This implements a generic mechanism for changing constness qualifiers
710    anywhere within big C types without resorting to outright "casts".
711    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
712    modify the constness at the outermost level.)
713    The implementation of "convert" is based on the idea of "witness"
714    values -- values that are not used by the operation but whose types
715    "testify" to their applicability.  On the implementation side, "convert"
716    is simply a projection (returning its second curried argument).  With
717    cross-module inlining, it should not result in any machine code being
718    generated.
719    
720    ----------------------------------------------------------------------
721    Name: Matthias Blume
722    Date: 2002/03/15 16:40:00 EST
723    Tag: blume-20020315-basis
724    Description:
725    
726    Provided (preliminary?) implementations for
727    
728      {String,Substring}.{concatWith,isSuffix,isSubstring}
729    
730    and
731    
732      Substring.full
733    
734    Those are in the Basis spec but they were missing in SML/NJ.
735    
736    ----------------------------------------------------------------------
737    Name: Matthias Blume
738    Date: 2002/03/14 21:30:00 EST
739    Tag: blume-20020314-controls
740    Description:
741    
742    Controls:
743    ---------
744    
745    1. Factored out the recently-added Controls : CONTROLS stuff and put
746       it into its own library $/controls-lib.cm.  The source tree for
747       this is under src/smlnj-lib/Controls.
748    
749    2. Changed the names of types and functions in this interface, so they
750       make a bit more "sense":
751    
752          module -> registry
753          'a registry -> 'a group
754    
755    3. The interface now deals in ref cells only.  The getter/setter interface
756       is (mostly) gone.
757    
758    4. Added a function that lets one register an already-existing ref cell.
759    
760    5. Made the corresponding modifications to the rest of the code so that
761       everything compiles again.
762    
763    6. Changed the implementation of Controls.MLRISC back to something closer
764       to the original.  In particular, this module (and therefore MLRISC)
765       does not depend on Controls.  There now is some link-time code in
766       int-sys.sml that registers the MLRISC controls with the Controls
767       module.
768    
769    CM:
770    ---
771    
772      * One can now specify the lambda-split aggressiveness in init.cmi.
773    
774    ----------------------------------------------------------------------
775    Name: Allen Leung
776    Date: 2002/03/13 17:30:00 EST
777    Tag: leunga-20020313-x86-fp-unary
778    Description:
779    
780    Bug fix for:
781    
782    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
783    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
784    > - fun f(x,(y,z)) = Real.~ y;
785    > [autoloading]
786    > [autoloading done]
787    >       fchsl   (%eax), 184(%esp)
788    > Error: MLRisc bug: X86MCEmitter.emitInstr
789    >
790    > uncaught exception Error
791    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
792    
793    The problem was that the code generator did not generate any fp registers
794    in this case, and the ra didn't know that it needed to run the X86FP phase to
795    translate the pseudo fp instruction.   This only happened with unary fp
796    operators in certain situations.
797    
798    ----------------------------------------------------------------------
799    Name: Matthias Blume
800    Date: 2002/03/13 14:00:00 EST
801    Tag: blume-20020313-overload-etc
802    Description:
803    
804    1. Added _overload as a synonym for overload for backward compatibility.
805       (Control.overloadKW must be true for either version to be accepted.)
806    
807    2. Fixed bug in install script that caused more things to be installed
808       than what was requested in config/targets.
809    
810    3. Made CM aware of the (_)overload construct so that autoloading
811       works.
812    
813    ----------------------------------------------------------------------
814    Name: Matthias Blume
815    Date: 2002/03/12 22:03:00 EST
816    Tag: blume-20020312-url
817    Description:
818    
819    Forgot to update BOOT and srcarchiveurl.
820    
821    ----------------------------------------------------------------------
822    Name: Matthias Blume
823    Date: 2002/03/12 17:30:00 EST
824    Tag: blume-20020312-version110392
825    Description:
826    
827    Yet another version number bump (because of small changes to the
828    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
829    
830    Changes:
831    
832      The new pid generation scheme described a few weeks ago was overly
833      complicated.  I implemented a new mechanism that is simpler and
834      provides a bit more "stability":  Once CM has seen a compilation
835      unit, it keeps its identity constant (as long as you do not delete
836      those crucial CM/GUID/* files).  This means that when you change
837      an interface, compile, then go back to the old interface, and
838      compile again, you arrive at the original pid.
839    
840      There now also is a mechanism that instructs CM to use the plain
841      environment hash as a module's pid (effectively making its GUID
842      the empty string).  For this, "noguid" must be specified as an
843      option to the .sml file in question within its .cm file.
844      This is most useful for code that is being generated by tools such
845      as ml-nlffigen (because during development programmers tend to
846      erase the tool's entire output directory tree including CM's cached
847      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
848      revert to the old, broken behavior of SML/NJ, but in specific cases
849      where there is no danger of interface confusion, its use is ok
850      (I think).
851    
852      ml-nlffigen by default generates "noguid" annotations.  They can be
853      turned off by specifying -guid in its command line.
854    
855    ----------------------------------------------------------------------
856    Name: Lal George
857    Date: 2002/03/12 12 14:42:36 EST
858    Tag: george-20020312-frequency-computation
859    Description:
860    
861    Integrated jump chaining and static block frequency into the
862    compiler. More details and numbers later.
863    
864    ----------------------------------------------------------------------
865    Name: Lal George
866    Date: 2002/03/11 11 22:38:53 EST
867    Tag: george-20020311-jump-chain-elim
868    Description:
869    
870    Tested the jump chain elimination on all architectures (except the
871    hppa).  This is on by default right now and is profitable for the
872    alpha and x86, however, it may not be profitable for the sparc and ppc
873    when compiling the compiler.
874    
875    The gc test will typically jump to a label at the end of the cluster,
876    where there is another jump to an external cluster containing the actual
877    code to invoke gc. This is to allow factoring of common gc invocation
878    sequences. That is to say, we generate:
879    
880            f:
881               testgc
882               ja   L1      % jump if above to L1
883    
884            L1:
885               jmp L2
886    
887    
888    After jump chain elimination the 'ja L1' instructions is converted to
889    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
890    up being implemented in their long form (if L2 is far away) using:
891    
892            jbe     L3      % jump if below or equal to L3
893            jmp     L2
894         L3:
895            ...
896    
897    
898    For large compilation units L2  may be far away.
899    
900    
901    ----------------------------------------------------------------------
902    Name: Matthias Blume
903    Date: 2002/03/11 13:30:00 EST
904    Tag: blume-20020311-mltreeeval
905    Description:
906    
907    A functor parameter was missing.
908    
909    ----------------------------------------------------------------------
910    Name: Allen Leung
911    Date: 2002/03/11 10:30:00 EST
912    Tag: leunga-20020311-runtime-string0
913    Description:
914    
915       The representation of the empty string now points to a
916    legal null terminated C string instead of unit.  It is now possible
917    to convert an ML string into C string with InlineT.CharVector.getData.
918    This compiles into one single machine instruction.
919    
920    ----------------------------------------------------------------------
921    Name: Allen Leung
922    Date: 2002/03/10 23:55:00 EST
923    Tag: leunga-20020310-x86-call
924    Description:
925    
926       Added machine generation for CALL instruction (relative displacement mode)
927    
928    ----------------------------------------------------------------------
929    Name: Matthias Blume
930    Date: 2002/03/08 16:05:00
931    Tag: blume-20020308-entrypoints
932    Description:
933    
934    Version number bumped to 110.39.1.  NEW BOOTFILES!
935    
936    Entrypoints: non-zero offset into a code object where execution should begin.
937    
938    - Added the notion of an entrypoint to CodeObj.
939    - Added reading/writing of entrypoint info to Binfile.
940    - Made runtime system bootloader aware of entrypoints.
941    - Use the address of the label of the first function given to mlriscGen
942      as the entrypoint.  This address is currently always 0, but it will
943      not be 0 once we turn on block placement.
944    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
945      with entry points) from mlriscGen.
946    
947    ----------------------------------------------------------------------
948    Name: Allen Leung
949    Date: 2002/03/07 20:45:00 EST
950    Tag: leunga-20020307-x86-cmov
951    Description:
952    
953       Bug fixes for CMOVcc on x86.
954    
955       1. Added machine code generation for CMOVcc
956       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
957       3. CMOVcc cannot have an immediate operand as argument.
958    
959    ----------------------------------------------------------------------
960    Name: Matthias Blume
961    Date: 2002/03/07 16:15:00 EST
962    Tag: blume-20020307-controls
963    Description:
964    
965    This is a very large but mostly boring patch which makes (almost)
966    every tuneable compiler knob (i.e., pretty much everything under
967    Control.* plus a few other things) configurable via both the command
968    line and environment variables in the style CM did its configuration
969    until now.
970    
971    Try starting sml with '-h' (or, if you are brave, '-H')
972    
973    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
974    implements the underlying generic mechanism.
975    
976    The interface to some of the existing such facilities has changed somewhat.
977    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
978    (The getFoo interface is still there for backward-compatibility, but its
979    use is deprecated.)
980    
981    The ml-build script passes -Cxxx=yyy command-line arguments through so
982    that one can now twiddle the compiler settings when using this "batch"
983    compiler.
984    
985    TODO items:
986    
987    We should go through and throw out all controls that are no longer
988    connected to anything.  Moreover, we should go through and provide
989    meaningful (and correct!) documentation strings for those controls
990    that still are connected.
991    
992    Currently, multiple calls to Controls.new are accepted (only the first
993    has any effect).  Eventually we should make sure that every control
994    is being made (via Controls.new) exactly once.  Future access can then
995    be done using Controls.acc.
996    
997    Finally, it would probably be a good idea to use the getter-setter
998    interface to controls rather than ref cells.  For the time being, both
999    styles are provided by the Controls module, but getter-setter pairs are
1000    better if thread-safety is of any concern because they can be wrapped.
1001    
1002    *****************************************
1003    
1004    One bug fix: The function blockPlacement in three of the MLRISC
1005    backpatch files used to be hard-wired to one of two possibilities at
1006    link time (according to the value of the placementFlag).  But (I
1007    think) it should rather sense the flag every time.
1008    
1009    *****************************************
1010    
1011    Other assorted changes (by other people who did not supply a HISTORY entry):
1012    
1013    1. the cross-module inliner now works much better (Monnier)
1014    2. representation of weights, frequencies, and probabilities in MLRISC
1015       changed in preparation of using those for weighted block placement
1016       (Reppy, George)
1017    
1018    ----------------------------------------------------------------------
1019    Name: Lal George
1020    Date: 2002/03/07 14:44:24 EST 2002
1021    Tag: george-20020307-weighted-block-placement
1022    
1023    Tested the weighted block placement optimization on all architectures
1024    (except the hppa) using AMPL to generate the block and edge frequencies.
1025    Changes were required in the machine properties to correctly
1026    categorize trap instructions. There is an MLRISC flag
1027    "weighted-block-placement" that can be used to enable weighted block
1028    placement, but this will be ineffective without block/edge
1029    frequencies (coming soon).
1030    
1031    
1032    ----------------------------------------------------------------------
1033    Name: Lal George
1034    Date: 2002/03/05 17:24:48 EST
1035    Tag: george-20020305-linkage-cluster
1036    
1037    In order to support the block placement optimization, a new cluster
1038    is generated as the very first cluster (called the linkage cluster).
1039    It contains a single jump to the 'real' entry point for the compilation
1040    unit. Block placement has no effect on the linkage cluster itself, but
1041    all the other clusters  have full freedom in the manner in which they
1042    reorder blocks or functions.
1043    
1044    On the x86 the typical linkage code that is generated is:
1045       ----------------------
1046            .align 2
1047       L0:
1048            addl    $L1-L0, 72(%esp)
1049            jmp     L1
1050    
1051    
1052            .align  2
1053       L1:
1054       ----------------------
1055    
1056    72(%esp) is the memory location for the stdlink register. This
1057    must contain the address of the CPS function being called. In the
1058    above example, it contains the address of  L0; before
1059    calling L1 (the real entry point for the compilation unit), it
1060    must contain the address for L1, and hence
1061    
1062            addl $L1-L0, 72(%esp)
1063    
1064    I have tested this on all architectures except the hppa.The increase
1065    in code size is of course negligible
1066    
1067    ----------------------------------------------------------------------
1068    Name: Allen Leung
1069    Date: 2002/03/03 13:20:00 EST
1070    Tag: leunga-20020303-mlrisc-tools
1071    
1072      Added #[ ... ] expressions to mlrisc tools
1073    
1074    ----------------------------------------------------------------------
1075    Name: Matthias Blume
1076    Date: 2002/02/27 12:29:00 EST
1077    Tag: blume-20020227-cdebug
1078    Description:
1079    
1080    - made types in structure C and C_Debug to be equal
1081    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1082    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1083    
1084    ----------------------------------------------------------------------
1085    Name: Matthias Blume
1086    Date: 2002/02/26 12:00:00 EST
1087    Tag: blume-20020226-ffi
1088    Description:
1089    
1090    1. Fixed a minor bug in CM's "noweb" tool:
1091       If numbering is turned off, then truly don't number (i.e., do not
1092       supply the -L option to noweb).  The previous behavior was to supply
1093       -L'' -- which caused noweb to use the "default" line numbering scheme.
1094       Thanks to Chris Richards for pointing this out (and supplying the fix).
1095    
1096    2. Once again, I reworked some aspects of the FFI:
1097    
1098       A. The incomplete/complete type business:
1099    
1100       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1101         gone!
1102       - ML types representing an incomplete type are now *equal* to
1103         ML types representing their corresponding complete types (just like
1104         in C).  This is still safe because ml-nlffigen will not generate
1105         RTTI for incomplete types, nor will it generate functions that
1106         require access to such RTTI.   But when ML code generated from both
1107         incomplete and complete versions of the C type meet, the ML types
1108         are trivially interoperable.
1109    
1110         NOTE:  These changes restore the full generality of the translation
1111         (which was previously lost when I eliminated functorization)!
1112    
1113       B. Enum types:
1114    
1115       - Structure C now has a type constructor "enum" that is similar to
1116         how the "su" constructor works.  However, "enum" is not a phantom
1117         type because each "T enum" has values (and is isomorphic to
1118         MLRep.Signed.int).
1119       - There are generic access operations for enum objects (using
1120         MLRep.Signed.int).
1121       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1122         * The structure contains the definition of type "mlrep" (the ML-side
1123         representation type of the enum).  Normally, mlrep is the same
1124         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1125         then mlrep will be defined as a datatype -- thus facilitating
1126         pattern matching on mlrep values.
1127         ("-ec" will be suppressed if there are duplicate values in an
1128          enumeration.)
1129         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1130         will be generated for each C enum constant xxx.
1131         * Conversion functions m2i and i2m convert between mlrep and
1132         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1133         * Coversion functions c and ml convert between mlrep and "tag enum".
1134         * Access functions (get/set) fetch and store mlrep values.
1135       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1136         enumerations are merged into one single enumeration represented by
1137         structure E_'.
1138    
1139    ----------------------------------------------------------------------
1140    Name: Allen Leung
1141    Date: 2002/02/25 04:45:00 EST
1142    Tag: leunga-20020225-cps-spill
1143    
1144    This is a new implementation of the CPS spill phase.
1145    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1146    In case of problems, replace it with the old file spill.sml
1147    
1148    The current compiler runs into some serious performance problems when
1149    constructing a large record.  This can happen when we try to compile a
1150    structure with many items.  Even a very simple structure like the following
1151    makes the compiler slow down.
1152    
1153        structure Foo = struct
1154           val x_1 = 0w1 : Word32.int
1155           val x_2 = 0w2 : Word32.int
1156           val x_3 = 0w3 : Word32.int
1157           ...
1158           val x_N = 0wN : Word32.int
1159        end
1160    
1161    The following table shows the compile time, from N=1000 to N=4000,
1162    with the old compiler:
1163    
1164    N
1165    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1166           MLRISC ra                               0.06u  0.00s  0.05g
1167              (spills = 0 reloads = 0)
1168           TOTAL                                   0.63u  0.07s  0.21g
1169    
1170    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1171           MLRISC ra                               5.68u  0.59s  3.93g
1172              (spills = 0 reloads = 0)
1173           TOTAL                                   14.71u  0.99s  4.81g
1174    
1175    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1176           MLRISC ra                               5.54u  0.65s  3.91g
1177              (spills = 543 reloads = 1082)
1178           TOTAL                                   65.40u  3.13s  6.00g
1179    
1180    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1181           MLRISC ra                               0.80u  0.10s  0.55g
1182              (spills = 42 reloads = 84)
1183           TOTAL                                   129.42u  5.10s  4.13g
1184    
1185    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1186           MLRISC ra                               2.69u  0.27s  1.38g
1187              (spills = 62 reloads = 124)
1188           TOTAL                                   682.48u  19.61s  13.99g
1189    
1190    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1191           MLRISC ra                               4.96u  0.27s  2.72g
1192              (spills = 85 reloads = 170)
1193           TOTAL                                   2375.26u  57.21s  48.00g
1194    
1195    As you can see the old cps spill module suffers from some serious
1196    performance problem.  But since I cannot decipher the old code fully,
1197    instead of patching the problems up, I'm reimplementing it
1198    with a different algorithm.  The new code is more modular,
1199    smaller when compiled, and substantially faster
1200    (O(n log n) time and O(n) space).  Timing of the new spill module:
1201    
1202    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1203          MLRISC ra                               0.25u  0.02s  0.15g
1204             (spills=1 reloads=3)
1205          TOTAL                                   7.74u  0.34s  1.62g
1206    
1207    Implementation details:
1208    
1209    As far as I can tell, the purpose of the CPS spill module is to make sure the
1210    number of live variables at any program point (the bandwidth)
1211    does not exceed a certain limit, which is determined by the
1212    size of the spill area.
1213    
1214    When the bandwidth is too large, we decrease the register pressure by
1215    packing live variables into spill records.  How we achieve this is
1216    completely different than what we did in the old code.
1217    
1218    First, there is something about the MLRiscGen code generator
1219    that we should be aware of:
1220    
1221    o MLRiscGen performs code motion!
1222    
1223       In particular, it will move floating point computations and
1224       address computations involving only the heap pointer to
1225       their use sites (if there is only a single use).
1226       What this means is that if we have a CPS record construction
1227       statement
1228    
1229           RECORD(k,vl,w,e)
1230    
1231       we should never count the new record address w as live if w
1232       has only one use (which is often the case).
1233    
1234       We should do something similar to floating point, but the transformation
1235       there is much more complex, so I won't deal with that.
1236    
1237    Secondly, there are now two new cps primops at our disposal:
1238    
1239     1. rawrecord of record_kind option
1240        This pure operator allocates some uninitialized storage from the heap.
1241        There are two forms:
1242    
1243         rawrecord NONE [INT n]  allocates a tagless record of length n
1244         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1245                                     and initializes the tag.
1246    
1247     2. rawupdate of cty
1248          rawupdate cty (v,i,x)
1249          Assigns to x to the ith component of record v.
1250          The storelist is not updated.
1251    
1252    We use these new primops for both spilling and increment record construction.
1253    
1254     1. Spilling.
1255    
1256        This is implemented with a linear scan algorithm (but generalized
1257        to trees).  The algorithm will create a single spill record at the
1258        beginning of the cps function and use rawupdate to spill to it,
1259        and SELECT or SELp to reload from it.  So both spills and reloads
1260        are fine-grain operations.  In contrast, in the old algorithm
1261        "spills" have to be bundled together in records.
1262    
1263        Ideally, we should sink the spill record construction to where
1264        it is needed.  We can even split the spill record into multiple ones
1265        at the places where they are needed.  But CPS is not a good
1266        representation for global code motion, so I'll keep it simple and
1267        am not attempting this.
1268    
1269     2. Incremental record construction (aka record splitting).
1270    
1271        Long records with many component values which are simulatenously live
1272        (recall that single use record addresses are not considered to
1273         be live) are constructed with rawrecord and rawupdate.
1274        We allocate space on the heap with rawrecord first, then gradually
1275        fill it in with rawupdate.  This is the technique suggested to me
1276        by Matthias.
1277    
1278        Some restrictions on when this is applicable:
1279        1. It is not a VECTOR record.  The code generator currently does not handle
1280           this case. VECTOR record uses double indirection like arrays.
1281        2. All the record component values are defined in the same "basic block"
1282           as the record constructor.  This is to prevent speculative
1283           record construction.
1284    
1285    ----------------------------------------------------------------------
1286    Name: Allen Leung
1287    Date: 2002/02/22 01:02:00 EST
1288    Tag: leunga-20020222-mlrisc-tools
1289    
1290    Minor bug fixes in the parser and rewriter
1291    
1292    ----------------------------------------------------------------------
1293    Name: Allen Leung
1294    Date: 2002/02/21 20:20:00 EST
1295    Tag: leunga-20020221-peephole
1296    
1297    Regenerated the peephole files.  Some contained typos in the specification
1298    and some didn't compile because of pretty printing bugs in the old version
1299    of 'nowhere'.
1300    
1301    ----------------------------------------------------------------------
1302    Name: Allen Leung
1303    Date: 2002/02/19 20:20:00 EST
1304    Tag: leunga-20020219-mlrisc-tools
1305    Description:
1306    
1307       Minor bug fixes to the mlrisc-tools library:
1308    
1309       1.  Fixed up parsing colon suffixed keywords
1310       2.  Added the ability to shut the error messages up
1311       3.  Reimplemented the pretty printer and fixed up/improved
1312           the pretty printing of handle and -> types.
1313       4.  Fixed up generation of literal symbols in the nowhere tool.
1314       5.  Added some SML keywords to to sml.sty
1315    
1316    ----------------------------------------------------------------------
1317    Name: Matthias Blume
1318    Date: 2002/02/19 16:20:00 EST
1319    Tag: blume-20020219-cmffi
1320    Description:
1321    
1322    A wild mix of changes, some minor, some major:
1323    
1324    * All C FFI-related libraries are now anchored under $c:
1325        $/c.cm      --> $c/c.cm
1326        $/c-int.cm  --> $c/internals/c-int.cm
1327        $/memory.cm --> $c/memory/memory.cm
1328    
1329    * "make" tool (in CM) now treats its argument pathname slightly
1330      differently:
1331        1. If the native expansion is an absolute name, then before invoking
1332           the "make" command on it, CM will apply OS.Path.mkRelative
1333           (with relativeTo = OS.FileSys.getDir()) to it.
1334        2. The argument will be passed through to subsequent phases of CM
1335           processing without "going native".  In particular, if the argument
1336           was an anchored path, then "make" will not lose track of that anchor.
1337    
1338    * Compiler backends now "know" their respective C calling conventions
1339      instead of having to be told about it by ml-nlffigen.  This relieves
1340      ml-nlffigen from one of its burdens.
1341    
1342    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1343    
1344    * Export C_DEBUG and C_Debug from $c/c.cm.
1345    
1346    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1347      subtyping relationship between incomplete pointers and their complete
1348      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1349      with the convention of instantiating 'o with ('t, 'c) obj whenever
1350      the pointer target type is complete.  In the incomplete case, 'o
1351      will be instantiated with some "'c iobj" -- a type obtained by
1352      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1353    
1354      Operations that work on both incomplete and complete pointer types are
1355      typed as taking an 'o ptr while operations that require the target to
1356      be known are typed as taking some ('t, 'c) obj ptr.
1357    
1358      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1359      where void is an eqtype without any values.  This makes it possible
1360      to work on voidptr values using functions meant to operate on light
1361      incomplete pointers.
1362    
1363    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1364      been vastly simplified.
1365    
1366    ----------------------------------------------------------------------
1367    Name: Matthias Blume
1368    Date: 2002/02/19 10:48:00 EST
1369    Tag: blume-20020219-pqfix
1370    Description:
1371    
1372    Applied Chris Okasaki's bug fix for priority queues.
1373    
1374    ----------------------------------------------------------------------
1375    Name: Matthias Blume
1376    Date: 2002/02/15 17:05:00
1377    Tag: Release_110_39
1378    Description:
1379    
1380    Last-minute retagging is becoming a tradition... :-(
1381    
1382    This is the working release 110.39.
1383    
1384    ----------------------------------------------------------------------
1385    Name: Matthias Blume
1386    Date: 2002/02/15 16:00:00 EST
1387    Tag: Release_110_39-orig
1388    Description:
1389    
1390    Working release 110.39.  New bootfiles.
1391    
1392    (Update: There was a small bug in the installer so it wouldn't work
1393    with all shells.  So I retagged. -Matthias)
1394    
1395    ----------------------------------------------------------------------
1396    Name: Matthias Blume
1397    Date: 2002/02/15 14:17:00 EST
1398    Tag: blume-20020215-showbindings
1399    Description:
1400    
1401    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1402    the latter can be useful for exploring what bindings are available at
1403    the interactive prompt.  (The first function returns only the list
1404    of symbols that are really bound, the second prints those but also the
1405    ones that CM's autoloading mechanism knows about.)
1406    
1407    ----------------------------------------------------------------------
1408    Name: Matthias Blume
1409    Date: 2002/02/15 12:08:00 EST
1410    Tag: blume-20020215-iptrs
1411    Description:
1412    
1413    Two improvements to ml-nlffigen:
1414    
1415      1. Write files only if they do not exist or if their current contents
1416         do not coincide with what's being written.  (That is, avoid messing
1417         with the time stamps unless absolutely necessary.)
1418    
1419      2. Implement a "repository" mechanism for generated files related
1420         to "incomplete pointer types".   See the README file for details.
1421    
1422    ----------------------------------------------------------------------
1423    Name: Matthias Blume
1424    Date: 2002/02/14 11:50:00 EST
1425    Tag: blume-20020214-quote
1426    Description:
1427    
1428    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1429    because of the new and improved tag generation scheme.  (Thanks to Allen
1430    Leung for pointing it out.)
1431    
1432    ----------------------------------------------------------------------
1433    Name: Lal George
1434    Date: 2002/02/14 09:55:27 EST 2002
1435    Tag: george-20020214-isabelle-bug
1436    Description:
1437    
1438    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1439    of Isabelle on the x86.
1440    
1441    From Allen:
1442    -----------
1443     I've found the problem:
1444    
1445         in ra-core.sml, I use the counter "blocked" to keep track of the
1446         true number of elements in the freeze queue.  When the counter goes
1447         to zero, I skip examining the queue.  But I've messed up the
1448         bookkeeping in combine():
1449    
1450             else ();
1451             case !ucol of
1452               PSEUDO => (if !cntv > 0 then
1453                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1454                                        ^^^^^^^^^^^^^^^^^^^^^^^
1455                      moveu := mergeMoveList(!movev, !moveu)
1456                     )
1457                  else ();
1458    
1459         combine() is called to coalesce two nodes u and v.
1460         I think I was thinking that if the move counts of u and v are both
1461         greater than zero then after they are coalesced then one node is
1462         removed from the freeze queue.  Apparently I was thinking that
1463         both u and v are of low degree, but that's clearly not necessarily true.
1464    
1465    
1466    02/12/2002:
1467        Here's the patch.  HOL now compiles.
1468    
1469        I don't know how this impact on performance (compile
1470        time or runtime).  This bug caused the RA (especially on the x86)
1471        to go thru the potential spill phase when there are still nodes on the
1472        freeze queue.
1473    
1474    
1475    
1476    
1477    ----------------------------------------------------------------------
1478    Name: Matthias Blume
1479    Date: 2002/02/13 22:40:00 EST
1480    Tag: blume-20020213-fptr-rtti
1481    Description:
1482    
1483    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1484    updates.
1485    
1486    ----------------------------------------------------------------------
1487    Name: Matthias Blume
1488    Date: 2002/02/13 16:41:00 EST
1489    Tag: blume-20020213-cmlpq
1490    Description:
1491    
1492    Added new priority queue export symbols (which have just been added to
1493    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1494    not compile and the installer would choke.)
1495    
1496    ----------------------------------------------------------------------
1497    Name: Matthias Blume
1498  Date: 2002/02/13 16:15:00 EST  Date: 2002/02/13 16:15:00 EST
1499  Tag: blume-20020213-various  Tag: blume-20020213-various
1500  Description:  Description:

Legend:
Removed from v.1062  
changed lines
  Added in v.1288

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