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 1009, Wed Jan 9 19:44:22 2002 UTC revision 1279, Wed Oct 16 14:21:07 2002 UTC
# Line 11  Line 11 
11  Date: yyyy/mm/dd  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: 2002/10/16 12:00:00 EDT
18    Tag: Release_110_42
19    Description:
20    
21    New working release.  New bootfiles.
22    
23    ----------------------------------------------------------------------
24    Name: Matthias Blume
25    Date: 2002/10/10 13:10:00 EDT
26    Tag: blume-20021010-ppc-divs
27    Description:
28    
29    The mltree operator DIVS must be implemented with an overflow check on
30    the PPC because the hardware indicates divide-by-zero using "overflow" as
31    well.
32    
33    ----------------------------------------------------------------------
34    Name: Matthias Blume
35    Date: 2002/07/23 11:45:00
36    Tag: blume-20020723-smlnj-home
37    Description:
38    
39    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
40    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
41    CM_PATHCONFIG is also set), the path configuration file is assumed
42    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
43    the entire tree to some other place and everything will "just work".
44    
45    (Companion commands such as ml-build and ml-makedepend also sense this
46    variable.)
47    
48    ----------------------------------------------------------------------
49    Name: Matthias Blume
50    Date: 2002/07/12 21:19:00 EDT
51    Tag: blume-20020712-liveness
52    Description:
53    
54    Exported two useful "step" functions from liveness module (MLRISC).
55    
56    ----------------------------------------------------------------------
57    Name: Matthias Blume
58    Date: 2002/07/05 16:00 EDT
59    Tag: Release_110_41
60    Description:
61    
62    New working release.  New bootfiles.
63    
64    ----------------------------------------------------------------------
65    Name: Matthias Blume
66    Date: 2002/07/05 10:25:00 EDT
67    Tag: blume-20020705-btimp
68    Description:
69    
70    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
71    other clients can set up backtracing support.
72    
73    ----------------------------------------------------------------------
74    Name: Matthias Blume
75    Date: 2002/06/25 17:23:00 EDT
76    Tag: blume-20020625-fpmax
77    Description:
78    
79    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
80    case.  (The sense of the isNaN test was reversed -- which made min and
81    max always return their first argument.)
82    
83    ----------------------------------------------------------------------
84    Name: Matthias Blume
85    Date: 2002/06/11
86    Tag: blume-20020611-unixpath
87    Description:
88    
89    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
90    
91    ----------------------------------------------------------------------
92    Name: Matthias Blume
93    Date: 2002/06/10 16:35:00 EDT
94    Tag: blume-20020610-ieeereal
95    Description:
96    
97    I back-ported my implementation of IEEEReal.fromString from the
98    idlbasis-devel branch so that we can test it.
99    
100    Another small change is that ppDec tries to give more information
101    than just "<sig>" in the case of functors.  However, this code is
102    broken in some mysterious way if the functor's body's signature
103    has not been declared by ascription but gets inferred from the
104    implementation.  This needs fixing...
105    
106    ----------------------------------------------------------------------
107    Name: Matthias Blume
108    Date: 2002/05/31
109    Tag: blume-20020531-btrace-mode
110    Description:
111    
112    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
113    the wayside when I switched over to using Controls everywhere.)
114    
115    ----------------------------------------------------------------------
116    Name: Lal George
117    Date: 2002/05/23 12:21:40 EDT
118    Tag: george-20020523-visual-labels
119    Description:
120    
121    Labels are now displayed in the graphical output to make
122    the fall-through and target blocks obvious.
123    
124    ----------------------------------------------------------------------
125    Name: Matthias Blume
126    Date: 2002/05/22 11:03:00 EDT
127    Tag: blume-20020522-shrink
128    Description:
129    
130    John tweaked yesterday's fix for 1131 to handle an out-of-memory
131    situation that comes up when allocating huge arrays.
132    
133    ----------------------------------------------------------------------
134    Name: Matthias Blume
135    Date: 2002/05/21 16:00:00 EDT
136    Tag: Release_110_40
137    Description:
138    
139    New working release (110.40).  New bootfiles.
140    
141    [Also: John Reppy fixed GC bug 1131.]
142    
143    ----------------------------------------------------------------------
144    Name: Matthias Blume
145    Date: 2002/05/21 12:35:00 EDT
146    Tag: blume-20020521-cmdoc
147    Description:
148    
149    CM documentation update.
150    
151    ----------------------------------------------------------------------
152    Name: Matthias Blume
153    Date: 2002/05/21 10:55:00 EDT
154    Tag: blume-20020521-misc
155    Description:
156    
157    - John tweaked runtime to be silent on heap export (except when
158      GC messages are on).
159    - I added a few more things (cross-compiling versions of CMB) to
160      config/preloads (as suggestions).
161    
162    ----------------------------------------------------------------------
163    Name: Matthias Blume
164    Date: 2002/05/20 22:25:00 EDT
165    Tag: blume-20020520-controls
166    Description:
167    
168    - Added ControlUtil structure to control-lib.cm.
169    - Use it throughout.
170    - Used Controls facility to define MLRISC controls (as opposed to
171      registering MLRISC control ref cells with Controls after the fact)
172    - Fixed messed-up controls priorities.
173    
174    * Removed again all the stuff from config/preloads that one wouldn't
175      be able to preload at the time the initial heap image is built.
176      (Many libraries, e.g., CML, do not exist yet at this time.  The
177       only libraries that can be preloaded via config/preloads are those
178       that come bundled with the bootfiles.)
179    
180    ----------------------------------------------------------------------
181    Name: Matthias Blume
182    Date: 2002/05/20 10:59:00 EDT
183    Tag: blume-20020520-preloads
184    Description:
185    
186    Added a lot of commented-out suggestions for things to be included
187    in config/preloads.
188    
189    ----------------------------------------------------------------------
190    Name: Allen Leung
191    Date: 2002/05/18 14:20:00 EDT
192    Tag: leunga-20020518-mdl
193    Description:
194    
195    o Made the mdl tool stuff compile and run again.
196    o I've disabled all the stuff that depends on RTL specifications; they
197      are all badly broken anyway.
198    
199    ----------------------------------------------------------------------
200    Name: Matthias Blume
201    Date: 2002/05/17 16:49:00 EDT
202    Tag: blume-20020517-controls
203    Description:
204    
205    0. John Reppy made several modifications to the SML/NJ library.
206       In particular, there is a shiny new controls-lib.cm.
207    
208    1. Pushed new controls interface through compiler so that everything
209       compiles again.
210    
211    2. Added FormatComb and FORMAT_COMB to the CML version of the
212       SML/NJ library (so that CML compiles again).
213    
214    3. Modified init scripts because XXX_DEFAULT environment variables
215       are no longer with us.  (Boot-time initialization is now done
216       using the same environment variables that are also used for
217       startup-time initialization of controls.)
218    
219    ----------------------------------------------------------------------
220    Name: Lal George
221    Date: 2002/05/15 09:20:10 EDT
222    Tag: george-20020515-pseudo-op-decls
223    Description:
224    
225            All pseudo-ops emitted before the first segment declaration
226    such as TEXT, DATA, and BSS directives are assumed to be global
227    declarations and are emitted first in the assembly file. This is
228    useful in a number of situations where one has pseudo-ops that are not
229    specific to any segment, and also works around the constraint that one
230    cannot have client pseudo-ops in the TEXT segment.
231    
232            Because no segment is associated with these declarations it is
233    an error to allocate any space or objects before the first segment
234    directive and an exception will be raised. However, we cannot make
235    this check for client pseudo-ops.
236    
237            These top level declarations are a field in the CFG graph_info.
238    In theory you can continue to add to this field after the CFG has been
239    built -- provided you know what you are doing;-)
240    
241    ----------------------------------------------------------------------
242    Name: Matthias Blume
243    Date: 2002/05/13 16:40:00 EDT
244    Tag: blume-20020513-pp-etc
245    Description:
246    
247    A few minor bugfixes:
248    
249      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
250        (Bogus printouts for redefined bindings still occur.  Compiler
251        bug should no longer occur now.  We need to redo the prettyprinter
252        from scratch.)
253    
254      - CM pathname printer now also adds escape sequences for ( and )
255    
256      - commend and docu fixes for ml-nlffi
257    
258    ----------------------------------------------------------------------
259    Name: Matthias Blume
260    Date: 2002/05/10 16:40:00 EDT
261    Tag: blume-20020510-erg-textio
262    Description:
263    
264    Applied the following bugfix provided by Emden Gansner:
265    
266        Output is corrupted when outputSubstr is used rather than output.
267    
268        The problem occurs when a substring
269    
270            ss = (s, dataStart, dataLen)
271    
272        where dataStart > 0, fills a stream buffer with avail bytes left.
273        avail bytes of s, starting at index dataStart, are copied into the
274        buffer, the buffer is flushed, and then the remaining dataLen-avail
275        bytes of ss are copied into the beginning of the buffer. Instead of
276        starting this copy at index dataStart+avail in s, the current code
277        starts the copy at index avail.
278    
279        Fix:
280        In text-io-fn.sml, change line 695 from
281             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
282        to
283             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
284    
285    ----------------------------------------------------------------------
286    Name: Matthias Blume
287    Date: 2002/04/12 13:55:00 EDT
288    Tag: blume-20020412-assyntax
289    Description:
290    
291    1. Grabbed newer assyntax.h from the XFree86 project.
292    2. Fiddled with how to compile X86.prim.asm without warnings.
293    3. (Very) Minor cleanup in CM.
294    
295    ----------------------------------------------------------------------
296    Name: Matthias Blume
297    Date: 2002/04/01 (no joke!) 17:07:00 EST
298    Tag: blume-20020401-x86div
299    Description:
300    
301    Added full support for div/mod/rem/quot on the x86, using the machine
302    instruction's two results (without clumsily recomputing the remainder)
303    directly where appropriate.
304    
305    Some more extensive power-of-two support was added to the x86 instruction
306    selector (avoiding expensive divs, mods, and muls where they can be
307    replaced with cheaper shifts and masks).  However, this sort of thing
308    ought to be done earlier, e.g., within the CPS optimizer so that
309    all architectures benefit from it.
310    
311    The compiler compiles to a fixed point, but changes might be somewhat
312    fragile nevertheless.  Please, report any strange things that you might
313    see wrt. div/mod/quot/rem...
314    
315    ----------------------------------------------------------------------
316    Name: Matthias Blume
317    Date: 2002/03/29 17:22:00
318    Tag: blume-20020329-div
319    Description:
320    
321    Fixed my broken div/mod logic.  Unfortunately, this means that the
322    inline code for div/mod now has one more comparison than before.
323    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
324    The problem was with quotient = 0, because that alone does not tell
325    us which way the rounding went.  One then has to look at whether
326    remainder and divisor have the same sign...  :(
327    
328    Anyway, I replaced the bootfiles with fresh ones...
329    
330    ----------------------------------------------------------------------
331    Name: Matthias Blume
332    Date: 2002/03/29 14:10:00 EST
333    Tag: blume-20020329-inlprims
334    Description:
335    
336    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
337    
338    Primops have changed. This means that the bin/boot-file formats have
339    changed as well.
340    
341    To make sure that there is no confusion, I made a new version.
342    
343    
344    CHANGES:
345    
346    * removed REMT from mltree (remainder should never overflow).
347    
348    * added primops to deal with divisions of all flavors to the frontend
349    
350    * handled these primops all the way through so they map to their respective
351      MLRISC support
352    
353    * used these primops in the implementation of Int, Int32, Word, Word32
354    
355    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
356    
357    * parameterized INLMIN, INLMAX, and INLABS by a numkind
358    
359    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
360      floating point
361    
362    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
363      and Real (but Real.abs maps to a separate floating-point-only primop)
364    
365    
366    TODO items:
367    
368    * Hacked Alpha32 instruction selection, disabling the selection of REMx
369      instructions because the machine instruction encoder cannot handle
370      them.  (Hppa, PPC, and Sparc instruction selection did not handle
371      REM in the first place, and REM is supported by the x86 machine coder.)
372    
373    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
374      selection phase.  (The two can be streamlined because the hardware
375      delivers both quotient and remainder at the same time anyway.)
376    
377    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
378      (Currently the behavior is inconsistent both across architectures and
379      wrt. the draft Basis spec.)
380    
381    * Word8 should eventually be handled natively, too.
382    
383    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
384      as if there currently is no execution path that could trigger it in
385      SML/NJ.  (The assumptions underlying functions arith and promotable do not
386      hold for things like multiplication and division.)
387    
388    ----------------------------------------------------------------------
389    Name: Matthias Blume
390    Date: 2002/03/27 16:27:00 EST
391    Tag: blume-20020327-mlrisc-divisions
392    Description:
393    
394    Added support for all four division operations (ML's div, mod, quot,
395    and rem) to MLRISC.  In the course of doing so, I also rationalized
396    the naming (no more annoying switch-around of DIV and QUOT), by
397    parameterizing the operation by div_rounding_mode (which can be either
398    DIV_TO_ZERO or DIV_TO_NEGINF).
399    
400    The generic MLTreeGen functor takes care of compiling all four
401    operations down to only round-to-zero div.
402    
403    Missing pieces:
404    
405      * Doing something smarter than relying on MLTreeGen on architectures
406        like, e.g., the x86 where hardware division delivers both quotient and
407        remainder at the same time.  With this, the implementation of the
408        round-to-neginf operations could be further streamlined.
409    
410      * Remove inlining support for div/mod/rem from the frontend and replace it
411        with primops that get carried through to the backend.  Do this for all
412        int and word types.
413    
414    ----------------------------------------------------------------------
415    Name: Matthias Blume
416    Date: 2002/03/25 17:25:00 EST
417    Tag: blume-20020325-divmod
418    Description:
419    
420    I improved (hopefully without breaking them) the implementation of Int.div,
421    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
422    advantage of the following observations:
423    
424      Let  q = x quot y      r = x rem y
425           d = x div  y      m = x mod y
426    
427    where "quot" is the round-to-zero version of integer division that
428    hardware usually provides.  Then we have:
429    
430         r = x - q * y        where neither the * nor the - will overflow
431         d = if q >= 0 orelse x = q * y then q else q - 1
432                              where neither the * nor the - will overflow
433         m = if q >= 0 orelse r = 0 then r else r + y
434                              where the + will not overflow
435    
436    This results in substantial simplification of the generated code.
437    The following table shows the number of CFG nodes and edges generated
438    for
439            fun f (x, y) = x OPER y
440            (* with OPER \in div, mod, quot, rem *)
441    
442    
443        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
444        --------------------------------------------------------
445         div |         24 |         39 |         12 |         16
446         mod |         41 |         71 |         12 |         16
447        quot |          8 |         10 |          8 |         10
448         rem |         10 |         14 |          8 |         10
449    
450    
451    ----------------------------------------------------------------------
452    Name: Matthias Blume
453    Date: 2002/03/25 22:06:00 EST
454    Tag: blume-20020325-cprotobug
455    Description:
456    
457    Fixed a bug in cproto (c prototype decoder).
458    
459    ----------------------------------------------------------------------
460    Name: Matthias Blume
461    Date: 2002/03/25 16:00:00 EST
462    Tag: blume-20020325-raw-primops
463    Description:
464    
465    I did some cleanup to Allen's new primop code and
466    replaced yesterday's bootfiles with new ones.
467    (But they are stored in the same place.)
468    
469    ----------------------------------------------------------------------
470    Name: Matthias Blume
471    Date: 2002/03/24 22:40:00 EST
472    Tag: blume-20020324-bootfiles
473    Description:
474    
475    Made the bootfiles that Allen asked for.
476    
477    ----------------------------------------------------------------------
478    Name: Allen Leung
479    Date: 2002/03/23 15:50:00 EST
480    Tag: leunga-20020323-flint-cps-rcc-primops
481    Description:
482    
483      1. Changes to FLINT primops:
484    
485        (* make a call to a C-function;
486         * The primop carries C function prototype information and specifies
487         * which of its (ML-) arguments are floating point. C prototype
488         * information is for use by the backend, ML information is for
489         * use by the CPS converter. *)
490      | RAW_CCALL of { c_proto: CTypes.c_proto,
491                       ml_args: ccall_type list,
492                       ml_res_opt: ccall_type option,
493                       reentrant : bool
494                     } option
495       (* Allocate uninitialized storage on the heap.
496        * The record is meant to hold short-lived C objects, i.e., they
497        * are not ML pointers.  With the tag, the representation is
498        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
499        *)
500      | RAW_RECORD of {tag:bool,sz:int}
501      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
502    
503      2.  These CPS primops are now overloaded:
504    
505           rawload of {kind:numkind}
506           rawstore of {kind:numkind}
507    
508          The one argument form is:
509    
510             rawload {kind} address
511    
512          The two argument form is:
513    
514             rawload {kind} [ml object, byte-offset]
515    
516      3. RAW_CCALL/RCC now takes two extra arguments:
517    
518         a. The first is whether the C call is reentrant, i.e., whether
519            ML state should be saved and restored.
520         b. The second argument is a string argument specifying the name of
521            library and the C function.
522    
523         These things are currently not handled in the code generator, yet.
524    
525      4. In CProto,
526    
527         An encoding type of "bool" means "ml object" and is mapped into
528         C prototype of PTR.  Note that "bool" is different than "string",
529         even though "string" is also mapped into PTR, because "bool"
530         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
531    
532      5. Pickler/unpicker
533    
534         Changed to handle RAW_RECORD and newest RAW_CCALL
535    
536      6. MLRiscGen,
537    
538         1. Changed to handle the new rawload/rawstore/rawrecord operators.
539         2. Code for handling C Calls has been moved to a new module CPSCCalls,
540            in the file CodeGen/cpscompile/cps-c-calls.sml
541    
542      7. Added the conditional move operator
543    
544             condmove of branch
545    
546         to cps.  Generation of this is still buggy so it is currently
547         disabled.
548    
549    ----------------------------------------------------------------------
550    Name: Lal George
551    Date: 2002/03/22 14:18:25 EST
552    Tag: george-20020322-cps-branch-prob
553    Description:
554    
555    Implemented the Ball-Larus branch prediction-heuristics, and
556    incorporated graphical viewers for control flow graphs.
557    
558    Ball-Larus Heuristics:
559    ---------------------
560    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
561    
562    By design it uses the Dempster-Shafer theory for combining
563    probabilities.  For example, in the function:
564    
565        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
566    
567    the ball-larus heuristics predicts that the n=0 is unlikely
568    (OH-heuristic), and the 'then' branch is unlikely because of the
569    RH-heuristic -- giving the 'then' branch an even lower combined
570    probability using the Dempster-Shafer theory.
571    
572    Finally, John Reppy's loop analysis in MLRISC, further lowers the
573    probability of the 'then' branch because of the loop in the else
574    branch.
575    
576    
577    Graphical Viewing:
578    ------------------
579    I merely plugged in Allen's graphical viewers into the compiler. The
580    additional code is not much. At the top level, saying:
581    
582            Control.MLRISC.getFlag "cfg-graphical-view" := true;
583    
584    will display the graphical view of the control flow graph just before
585    back-patching.  daVinci must be in your path for this to work. If
586    daVinci is not available, then the default viewer can be changed
587    using:
588    
589            Control.MLRISC.getString "viewer"
590    
591    which can be set to "dot" or "vcg" for the corresponding viewers. Of
592    course, these viewers must be in your path.
593    
594    The above will display the compilation unit at the level of clusters,
595    many of which are small, boring, and un-interesting. Also setting:
596    
597            Control.MLRISC.getInt "cfg-graphical-view_size"
598    
599    will display clusters that are larger than the value set by the above.
600    
601    
602    ----------------------------------------------------------------------
603    Name: Matthias Blume
604    Date: 2002/03/21 22:20:00 EST
605    Tag: blume-20020321-kmp-bugfix
606    Description:
607    
608    Changed the interface to the KMP routine in PreString and fixed
609    a minor bug in one place where it was used.
610    
611    ----------------------------------------------------------------------
612    Name: Allen Leung
613    Date: 2002/03/21 20:30:00 EST
614    Tag: leunga-20020321-cfg
615    Description:
616    
617      Fixed a potential problem in cfg edge splitting.
618    
619    ----------------------------------------------------------------------
620    Name: Allen Leung
621    Date: 2002/03/21 17:15:00 EST
622    Tag: leunga-20020321-x86-fp-cfg
623    Description:
624    
625      1. Recoded the buggy parts of x86-fp.
626    
627         a. All the block reordering code has been removed.
628            We now depend on the block placement phases to do this work.
629    
630         b. Critical edge splitting code has been simplified and moved into the
631            CFG modules, as where they belong.
632    
633         Both of these were quite buggy and complex.  The code is now much, much
634         simpler.
635    
636      2. X86 backend.
637    
638         a. Added instructions for 64-bit support.  Instruction selection for
639            64-bit has not been committed, however, since that
640            requires changes to MLTREE which haven't been approved by
641            Lal and John.
642    
643         b. Added support for FUCOMI and FUCOMIP when generating code for
644            PentiumPro and above.  We only generate these instructions in
645            the fast-fp mode.
646    
647         c. Added cases for JP and JNP in X86FreqProps.
648    
649      3. CFG
650    
651         CFG now has a bunch of methods for edge splitting and merging.
652    
653      4. Machine description.
654    
655         John's simplification of MLTREE_BASIS.fcond broke a few machine
656         description things:
657    
658         rtl-build.{sig,sml} and hppa.mdl fixed.
659    
660         NOTE: the machine description stuff in the repository is still broken.
661               Again, I can't put my fixes in because that involves
662               changes to MLTREE.
663    
664    ----------------------------------------------------------------------
665    Name: Matthias Blume
666    Date: 2002/03/20 15:55:00 EST
667    Tag: blume-20020320-kmp
668    Description:
669    
670    Implemented Knuth-Morris-Pratt string matching in PreString and used
671    it for String.isSubstring, Substring.isSubstring, and
672    Substring.position.
673    
674    (Might need some stress-testing.  Simple examples worked fine.)
675    
676    ----------------------------------------------------------------------
677    Name: Matthias Blume
678    Date: 2002/03/19 16:37:00 EST
679    Tag: blume-20020319-witnesses
680    Description:
681    
682    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
683    
684    This implements a generic mechanism for changing constness qualifiers
685    anywhere within big C types without resorting to outright "casts".
686    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
687    modify the constness at the outermost level.)
688    The implementation of "convert" is based on the idea of "witness"
689    values -- values that are not used by the operation but whose types
690    "testify" to their applicability.  On the implementation side, "convert"
691    is simply a projection (returning its second curried argument).  With
692    cross-module inlining, it should not result in any machine code being
693    generated.
694    
695    ----------------------------------------------------------------------
696    Name: Matthias Blume
697    Date: 2002/03/15 16:40:00 EST
698    Tag: blume-20020315-basis
699    Description:
700    
701    Provided (preliminary?) implementations for
702    
703      {String,Substring}.{concatWith,isSuffix,isSubstring}
704    
705    and
706    
707      Substring.full
708    
709    Those are in the Basis spec but they were missing in SML/NJ.
710    
711    ----------------------------------------------------------------------
712    Name: Matthias Blume
713    Date: 2002/03/14 21:30:00 EST
714    Tag: blume-20020314-controls
715    Description:
716    
717    Controls:
718    ---------
719    
720    1. Factored out the recently-added Controls : CONTROLS stuff and put
721       it into its own library $/controls-lib.cm.  The source tree for
722       this is under src/smlnj-lib/Controls.
723    
724    2. Changed the names of types and functions in this interface, so they
725       make a bit more "sense":
726    
727          module -> registry
728          'a registry -> 'a group
729    
730    3. The interface now deals in ref cells only.  The getter/setter interface
731       is (mostly) gone.
732    
733    4. Added a function that lets one register an already-existing ref cell.
734    
735    5. Made the corresponding modifications to the rest of the code so that
736       everything compiles again.
737    
738    6. Changed the implementation of Controls.MLRISC back to something closer
739       to the original.  In particular, this module (and therefore MLRISC)
740       does not depend on Controls.  There now is some link-time code in
741       int-sys.sml that registers the MLRISC controls with the Controls
742       module.
743    
744    CM:
745    ---
746    
747      * One can now specify the lambda-split aggressiveness in init.cmi.
748    
749    ----------------------------------------------------------------------
750    Name: Allen Leung
751    Date: 2002/03/13 17:30:00 EST
752    Tag: leunga-20020313-x86-fp-unary
753    Description:
754    
755    Bug fix for:
756    
757    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
758    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
759    > - fun f(x,(y,z)) = Real.~ y;
760    > [autoloading]
761    > [autoloading done]
762    >       fchsl   (%eax), 184(%esp)
763    > Error: MLRisc bug: X86MCEmitter.emitInstr
764    >
765    > uncaught exception Error
766    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
767    
768    The problem was that the code generator did not generate any fp registers
769    in this case, and the ra didn't know that it needed to run the X86FP phase to
770    translate the pseudo fp instruction.   This only happened with unary fp
771    operators in certain situations.
772    
773    ----------------------------------------------------------------------
774    Name: Matthias Blume
775    Date: 2002/03/13 14:00:00 EST
776    Tag: blume-20020313-overload-etc
777    Description:
778    
779    1. Added _overload as a synonym for overload for backward compatibility.
780       (Control.overloadKW must be true for either version to be accepted.)
781    
782    2. Fixed bug in install script that caused more things to be installed
783       than what was requested in config/targets.
784    
785    3. Made CM aware of the (_)overload construct so that autoloading
786       works.
787    
788    ----------------------------------------------------------------------
789    Name: Matthias Blume
790    Date: 2002/03/12 22:03:00 EST
791    Tag: blume-20020312-url
792    Description:
793    
794    Forgot to update BOOT and srcarchiveurl.
795    
796    ----------------------------------------------------------------------
797    Name: Matthias Blume
798    Date: 2002/03/12 17:30:00 EST
799    Tag: blume-20020312-version110392
800    Description:
801    
802    Yet another version number bump (because of small changes to the
803    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
804    
805    Changes:
806    
807      The new pid generation scheme described a few weeks ago was overly
808      complicated.  I implemented a new mechanism that is simpler and
809      provides a bit more "stability":  Once CM has seen a compilation
810      unit, it keeps its identity constant (as long as you do not delete
811      those crucial CM/GUID/* files).  This means that when you change
812      an interface, compile, then go back to the old interface, and
813      compile again, you arrive at the original pid.
814    
815      There now also is a mechanism that instructs CM to use the plain
816      environment hash as a module's pid (effectively making its GUID
817      the empty string).  For this, "noguid" must be specified as an
818      option to the .sml file in question within its .cm file.
819      This is most useful for code that is being generated by tools such
820      as ml-nlffigen (because during development programmers tend to
821      erase the tool's entire output directory tree including CM's cached
822      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
823      revert to the old, broken behavior of SML/NJ, but in specific cases
824      where there is no danger of interface confusion, its use is ok
825      (I think).
826    
827      ml-nlffigen by default generates "noguid" annotations.  They can be
828      turned off by specifying -guid in its command line.
829    
830    ----------------------------------------------------------------------
831    Name: Lal George
832    Date: 2002/03/12 12 14:42:36 EST
833    Tag: george-20020312-frequency-computation
834    Description:
835    
836    Integrated jump chaining and static block frequency into the
837    compiler. More details and numbers later.
838    
839    ----------------------------------------------------------------------
840    Name: Lal George
841    Date: 2002/03/11 11 22:38:53 EST
842    Tag: george-20020311-jump-chain-elim
843    Description:
844    
845    Tested the jump chain elimination on all architectures (except the
846    hppa).  This is on by default right now and is profitable for the
847    alpha and x86, however, it may not be profitable for the sparc and ppc
848    when compiling the compiler.
849    
850    The gc test will typically jump to a label at the end of the cluster,
851    where there is another jump to an external cluster containing the actual
852    code to invoke gc. This is to allow factoring of common gc invocation
853    sequences. That is to say, we generate:
854    
855            f:
856               testgc
857               ja   L1      % jump if above to L1
858    
859            L1:
860               jmp L2
861    
862    
863    After jump chain elimination the 'ja L1' instructions is converted to
864    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
865    up being implemented in their long form (if L2 is far away) using:
866    
867            jbe     L3      % jump if below or equal to L3
868            jmp     L2
869         L3:
870            ...
871    
872    
873    For large compilation units L2  may be far away.
874    
875    
876    ----------------------------------------------------------------------
877    Name: Matthias Blume
878    Date: 2002/03/11 13:30:00 EST
879    Tag: blume-20020311-mltreeeval
880    Description:
881    
882    A functor parameter was missing.
883    
884    ----------------------------------------------------------------------
885    Name: Allen Leung
886    Date: 2002/03/11 10:30:00 EST
887    Tag: leunga-20020311-runtime-string0
888    Description:
889    
890       The representation of the empty string now points to a
891    legal null terminated C string instead of unit.  It is now possible
892    to convert an ML string into C string with InlineT.CharVector.getData.
893    This compiles into one single machine instruction.
894    
895    ----------------------------------------------------------------------
896    Name: Allen Leung
897    Date: 2002/03/10 23:55:00 EST
898    Tag: leunga-20020310-x86-call
899    Description:
900    
901       Added machine generation for CALL instruction (relative displacement mode)
902    
903    ----------------------------------------------------------------------
904    Name: Matthias Blume
905    Date: 2002/03/08 16:05:00
906    Tag: blume-20020308-entrypoints
907    Description:
908    
909    Version number bumped to 110.39.1.  NEW BOOTFILES!
910    
911    Entrypoints: non-zero offset into a code object where execution should begin.
912    
913    - Added the notion of an entrypoint to CodeObj.
914    - Added reading/writing of entrypoint info to Binfile.
915    - Made runtime system bootloader aware of entrypoints.
916    - Use the address of the label of the first function given to mlriscGen
917      as the entrypoint.  This address is currently always 0, but it will
918      not be 0 once we turn on block placement.
919    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
920      with entry points) from mlriscGen.
921    
922    ----------------------------------------------------------------------
923    Name: Allen Leung
924    Date: 2002/03/07 20:45:00 EST
925    Tag: leunga-20020307-x86-cmov
926    Description:
927    
928       Bug fixes for CMOVcc on x86.
929    
930       1. Added machine code generation for CMOVcc
931       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
932       3. CMOVcc cannot have an immediate operand as argument.
933    
934    ----------------------------------------------------------------------
935    Name: Matthias Blume
936    Date: 2002/03/07 16:15:00 EST
937    Tag: blume-20020307-controls
938    Description:
939    
940    This is a very large but mostly boring patch which makes (almost)
941    every tuneable compiler knob (i.e., pretty much everything under
942    Control.* plus a few other things) configurable via both the command
943    line and environment variables in the style CM did its configuration
944    until now.
945    
946    Try starting sml with '-h' (or, if you are brave, '-H')
947    
948    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
949    implements the underlying generic mechanism.
950    
951    The interface to some of the existing such facilities has changed somewhat.
952    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
953    (The getFoo interface is still there for backward-compatibility, but its
954    use is deprecated.)
955    
956    The ml-build script passes -Cxxx=yyy command-line arguments through so
957    that one can now twiddle the compiler settings when using this "batch"
958    compiler.
959    
960    TODO items:
961    
962    We should go through and throw out all controls that are no longer
963    connected to anything.  Moreover, we should go through and provide
964    meaningful (and correct!) documentation strings for those controls
965    that still are connected.
966    
967    Currently, multiple calls to Controls.new are accepted (only the first
968    has any effect).  Eventually we should make sure that every control
969    is being made (via Controls.new) exactly once.  Future access can then
970    be done using Controls.acc.
971    
972    Finally, it would probably be a good idea to use the getter-setter
973    interface to controls rather than ref cells.  For the time being, both
974    styles are provided by the Controls module, but getter-setter pairs are
975    better if thread-safety is of any concern because they can be wrapped.
976    
977    *****************************************
978    
979    One bug fix: The function blockPlacement in three of the MLRISC
980    backpatch files used to be hard-wired to one of two possibilities at
981    link time (according to the value of the placementFlag).  But (I
982    think) it should rather sense the flag every time.
983    
984    *****************************************
985    
986    Other assorted changes (by other people who did not supply a HISTORY entry):
987    
988    1. the cross-module inliner now works much better (Monnier)
989    2. representation of weights, frequencies, and probabilities in MLRISC
990       changed in preparation of using those for weighted block placement
991       (Reppy, George)
992    
993    ----------------------------------------------------------------------
994    Name: Lal George
995    Date: 2002/03/07 14:44:24 EST 2002
996    Tag: george-20020307-weighted-block-placement
997    
998    Tested the weighted block placement optimization on all architectures
999    (except the hppa) using AMPL to generate the block and edge frequencies.
1000    Changes were required in the machine properties to correctly
1001    categorize trap instructions. There is an MLRISC flag
1002    "weighted-block-placement" that can be used to enable weighted block
1003    placement, but this will be ineffective without block/edge
1004    frequencies (coming soon).
1005    
1006    
1007    ----------------------------------------------------------------------
1008    Name: Lal George
1009    Date: 2002/03/05 17:24:48 EST
1010    Tag: george-20020305-linkage-cluster
1011    
1012    In order to support the block placement optimization, a new cluster
1013    is generated as the very first cluster (called the linkage cluster).
1014    It contains a single jump to the 'real' entry point for the compilation
1015    unit. Block placement has no effect on the linkage cluster itself, but
1016    all the other clusters  have full freedom in the manner in which they
1017    reorder blocks or functions.
1018    
1019    On the x86 the typical linkage code that is generated is:
1020       ----------------------
1021            .align 2
1022       L0:
1023            addl    $L1-L0, 72(%esp)
1024            jmp     L1
1025    
1026    
1027            .align  2
1028       L1:
1029       ----------------------
1030    
1031    72(%esp) is the memory location for the stdlink register. This
1032    must contain the address of the CPS function being called. In the
1033    above example, it contains the address of  L0; before
1034    calling L1 (the real entry point for the compilation unit), it
1035    must contain the address for L1, and hence
1036    
1037            addl $L1-L0, 72(%esp)
1038    
1039    I have tested this on all architectures except the hppa.The increase
1040    in code size is of course negligible
1041    
1042    ----------------------------------------------------------------------
1043    Name: Allen Leung
1044    Date: 2002/03/03 13:20:00 EST
1045    Tag: leunga-20020303-mlrisc-tools
1046    
1047      Added #[ ... ] expressions to mlrisc tools
1048    
1049    ----------------------------------------------------------------------
1050    Name: Matthias Blume
1051    Date: 2002/02/27 12:29:00 EST
1052    Tag: blume-20020227-cdebug
1053    Description:
1054    
1055    - made types in structure C and C_Debug to be equal
1056    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1057    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1058    
1059    ----------------------------------------------------------------------
1060    Name: Matthias Blume
1061    Date: 2002/02/26 12:00:00 EST
1062    Tag: blume-20020226-ffi
1063    Description:
1064    
1065    1. Fixed a minor bug in CM's "noweb" tool:
1066       If numbering is turned off, then truly don't number (i.e., do not
1067       supply the -L option to noweb).  The previous behavior was to supply
1068       -L'' -- which caused noweb to use the "default" line numbering scheme.
1069       Thanks to Chris Richards for pointing this out (and supplying the fix).
1070    
1071    2. Once again, I reworked some aspects of the FFI:
1072    
1073       A. The incomplete/complete type business:
1074    
1075       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1076         gone!
1077       - ML types representing an incomplete type are now *equal* to
1078         ML types representing their corresponding complete types (just like
1079         in C).  This is still safe because ml-nlffigen will not generate
1080         RTTI for incomplete types, nor will it generate functions that
1081         require access to such RTTI.   But when ML code generated from both
1082         incomplete and complete versions of the C type meet, the ML types
1083         are trivially interoperable.
1084    
1085         NOTE:  These changes restore the full generality of the translation
1086         (which was previously lost when I eliminated functorization)!
1087    
1088       B. Enum types:
1089    
1090       - Structure C now has a type constructor "enum" that is similar to
1091         how the "su" constructor works.  However, "enum" is not a phantom
1092         type because each "T enum" has values (and is isomorphic to
1093         MLRep.Signed.int).
1094       - There are generic access operations for enum objects (using
1095         MLRep.Signed.int).
1096       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1097         * The structure contains the definition of type "mlrep" (the ML-side
1098         representation type of the enum).  Normally, mlrep is the same
1099         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1100         then mlrep will be defined as a datatype -- thus facilitating
1101         pattern matching on mlrep values.
1102         ("-ec" will be suppressed if there are duplicate values in an
1103          enumeration.)
1104         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1105         will be generated for each C enum constant xxx.
1106         * Conversion functions m2i and i2m convert between mlrep and
1107         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1108         * Coversion functions c and ml convert between mlrep and "tag enum".
1109         * Access functions (get/set) fetch and store mlrep values.
1110       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1111         enumerations are merged into one single enumeration represented by
1112         structure E_'.
1113    
1114    ----------------------------------------------------------------------
1115    Name: Allen Leung
1116    Date: 2002/02/25 04:45:00 EST
1117    Tag: leunga-20020225-cps-spill
1118    
1119    This is a new implementation of the CPS spill phase.
1120    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1121    In case of problems, replace it with the old file spill.sml
1122    
1123    The current compiler runs into some serious performance problems when
1124    constructing a large record.  This can happen when we try to compile a
1125    structure with many items.  Even a very simple structure like the following
1126    makes the compiler slow down.
1127    
1128        structure Foo = struct
1129           val x_1 = 0w1 : Word32.int
1130           val x_2 = 0w2 : Word32.int
1131           val x_3 = 0w3 : Word32.int
1132           ...
1133           val x_N = 0wN : Word32.int
1134        end
1135    
1136    The following table shows the compile time, from N=1000 to N=4000,
1137    with the old compiler:
1138    
1139    N
1140    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1141           MLRISC ra                               0.06u  0.00s  0.05g
1142              (spills = 0 reloads = 0)
1143           TOTAL                                   0.63u  0.07s  0.21g
1144    
1145    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1146           MLRISC ra                               5.68u  0.59s  3.93g
1147              (spills = 0 reloads = 0)
1148           TOTAL                                   14.71u  0.99s  4.81g
1149    
1150    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1151           MLRISC ra                               5.54u  0.65s  3.91g
1152              (spills = 543 reloads = 1082)
1153           TOTAL                                   65.40u  3.13s  6.00g
1154    
1155    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1156           MLRISC ra                               0.80u  0.10s  0.55g
1157              (spills = 42 reloads = 84)
1158           TOTAL                                   129.42u  5.10s  4.13g
1159    
1160    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1161           MLRISC ra                               2.69u  0.27s  1.38g
1162              (spills = 62 reloads = 124)
1163           TOTAL                                   682.48u  19.61s  13.99g
1164    
1165    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1166           MLRISC ra                               4.96u  0.27s  2.72g
1167              (spills = 85 reloads = 170)
1168           TOTAL                                   2375.26u  57.21s  48.00g
1169    
1170    As you can see the old cps spill module suffers from some serious
1171    performance problem.  But since I cannot decipher the old code fully,
1172    instead of patching the problems up, I'm reimplementing it
1173    with a different algorithm.  The new code is more modular,
1174    smaller when compiled, and substantially faster
1175    (O(n log n) time and O(n) space).  Timing of the new spill module:
1176    
1177    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1178          MLRISC ra                               0.25u  0.02s  0.15g
1179             (spills=1 reloads=3)
1180          TOTAL                                   7.74u  0.34s  1.62g
1181    
1182    Implementation details:
1183    
1184    As far as I can tell, the purpose of the CPS spill module is to make sure the
1185    number of live variables at any program point (the bandwidth)
1186    does not exceed a certain limit, which is determined by the
1187    size of the spill area.
1188    
1189    When the bandwidth is too large, we decrease the register pressure by
1190    packing live variables into spill records.  How we achieve this is
1191    completely different than what we did in the old code.
1192    
1193    First, there is something about the MLRiscGen code generator
1194    that we should be aware of:
1195    
1196    o MLRiscGen performs code motion!
1197    
1198       In particular, it will move floating point computations and
1199       address computations involving only the heap pointer to
1200       their use sites (if there is only a single use).
1201       What this means is that if we have a CPS record construction
1202       statement
1203    
1204           RECORD(k,vl,w,e)
1205    
1206       we should never count the new record address w as live if w
1207       has only one use (which is often the case).
1208    
1209       We should do something similar to floating point, but the transformation
1210       there is much more complex, so I won't deal with that.
1211    
1212    Secondly, there are now two new cps primops at our disposal:
1213    
1214     1. rawrecord of record_kind option
1215        This pure operator allocates some uninitialized storage from the heap.
1216        There are two forms:
1217    
1218         rawrecord NONE [INT n]  allocates a tagless record of length n
1219         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1220                                     and initializes the tag.
1221    
1222     2. rawupdate of cty
1223          rawupdate cty (v,i,x)
1224          Assigns to x to the ith component of record v.
1225          The storelist is not updated.
1226    
1227    We use these new primops for both spilling and increment record construction.
1228    
1229     1. Spilling.
1230    
1231        This is implemented with a linear scan algorithm (but generalized
1232        to trees).  The algorithm will create a single spill record at the
1233        beginning of the cps function and use rawupdate to spill to it,
1234        and SELECT or SELp to reload from it.  So both spills and reloads
1235        are fine-grain operations.  In contrast, in the old algorithm
1236        "spills" have to be bundled together in records.
1237    
1238        Ideally, we should sink the spill record construction to where
1239        it is needed.  We can even split the spill record into multiple ones
1240        at the places where they are needed.  But CPS is not a good
1241        representation for global code motion, so I'll keep it simple and
1242        am not attempting this.
1243    
1244     2. Incremental record construction (aka record splitting).
1245    
1246        Long records with many component values which are simulatenously live
1247        (recall that single use record addresses are not considered to
1248         be live) are constructed with rawrecord and rawupdate.
1249        We allocate space on the heap with rawrecord first, then gradually
1250        fill it in with rawupdate.  This is the technique suggested to me
1251        by Matthias.
1252    
1253        Some restrictions on when this is applicable:
1254        1. It is not a VECTOR record.  The code generator currently does not handle
1255           this case. VECTOR record uses double indirection like arrays.
1256        2. All the record component values are defined in the same "basic block"
1257           as the record constructor.  This is to prevent speculative
1258           record construction.
1259    
1260    ----------------------------------------------------------------------
1261    Name: Allen Leung
1262    Date: 2002/02/22 01:02:00 EST
1263    Tag: leunga-20020222-mlrisc-tools
1264    
1265    Minor bug fixes in the parser and rewriter
1266    
1267    ----------------------------------------------------------------------
1268    Name: Allen Leung
1269    Date: 2002/02/21 20:20:00 EST
1270    Tag: leunga-20020221-peephole
1271    
1272    Regenerated the peephole files.  Some contained typos in the specification
1273    and some didn't compile because of pretty printing bugs in the old version
1274    of 'nowhere'.
1275    
1276    ----------------------------------------------------------------------
1277    Name: Allen Leung
1278    Date: 2002/02/19 20:20:00 EST
1279    Tag: leunga-20020219-mlrisc-tools
1280    Description:
1281    
1282       Minor bug fixes to the mlrisc-tools library:
1283    
1284       1.  Fixed up parsing colon suffixed keywords
1285       2.  Added the ability to shut the error messages up
1286       3.  Reimplemented the pretty printer and fixed up/improved
1287           the pretty printing of handle and -> types.
1288       4.  Fixed up generation of literal symbols in the nowhere tool.
1289       5.  Added some SML keywords to to sml.sty
1290    
1291    ----------------------------------------------------------------------
1292    Name: Matthias Blume
1293    Date: 2002/02/19 16:20:00 EST
1294    Tag: blume-20020219-cmffi
1295    Description:
1296    
1297    A wild mix of changes, some minor, some major:
1298    
1299    * All C FFI-related libraries are now anchored under $c:
1300        $/c.cm      --> $c/c.cm
1301        $/c-int.cm  --> $c/internals/c-int.cm
1302        $/memory.cm --> $c/memory/memory.cm
1303    
1304    * "make" tool (in CM) now treats its argument pathname slightly
1305      differently:
1306        1. If the native expansion is an absolute name, then before invoking
1307           the "make" command on it, CM will apply OS.Path.mkRelative
1308           (with relativeTo = OS.FileSys.getDir()) to it.
1309        2. The argument will be passed through to subsequent phases of CM
1310           processing without "going native".  In particular, if the argument
1311           was an anchored path, then "make" will not lose track of that anchor.
1312    
1313    * Compiler backends now "know" their respective C calling conventions
1314      instead of having to be told about it by ml-nlffigen.  This relieves
1315      ml-nlffigen from one of its burdens.
1316    
1317    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1318    
1319    * Export C_DEBUG and C_Debug from $c/c.cm.
1320    
1321    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1322      subtyping relationship between incomplete pointers and their complete
1323      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1324      with the convention of instantiating 'o with ('t, 'c) obj whenever
1325      the pointer target type is complete.  In the incomplete case, 'o
1326      will be instantiated with some "'c iobj" -- a type obtained by
1327      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1328    
1329      Operations that work on both incomplete and complete pointer types are
1330      typed as taking an 'o ptr while operations that require the target to
1331      be known are typed as taking some ('t, 'c) obj ptr.
1332    
1333      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1334      where void is an eqtype without any values.  This makes it possible
1335      to work on voidptr values using functions meant to operate on light
1336      incomplete pointers.
1337    
1338    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1339      been vastly simplified.
1340    
1341    ----------------------------------------------------------------------
1342    Name: Matthias Blume
1343    Date: 2002/02/19 10:48:00 EST
1344    Tag: blume-20020219-pqfix
1345    Description:
1346    
1347    Applied Chris Okasaki's bug fix for priority queues.
1348    
1349    ----------------------------------------------------------------------
1350    Name: Matthias Blume
1351    Date: 2002/02/15 17:05:00
1352    Tag: Release_110_39
1353    Description:
1354    
1355    Last-minute retagging is becoming a tradition... :-(
1356    
1357    This is the working release 110.39.
1358    
1359    ----------------------------------------------------------------------
1360    Name: Matthias Blume
1361    Date: 2002/02/15 16:00:00 EST
1362    Tag: Release_110_39-orig
1363    Description:
1364    
1365    Working release 110.39.  New bootfiles.
1366    
1367    (Update: There was a small bug in the installer so it wouldn't work
1368    with all shells.  So I retagged. -Matthias)
1369    
1370    ----------------------------------------------------------------------
1371    Name: Matthias Blume
1372    Date: 2002/02/15 14:17:00 EST
1373    Tag: blume-20020215-showbindings
1374    Description:
1375    
1376    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1377    the latter can be useful for exploring what bindings are available at
1378    the interactive prompt.  (The first function returns only the list
1379    of symbols that are really bound, the second prints those but also the
1380    ones that CM's autoloading mechanism knows about.)
1381    
1382    ----------------------------------------------------------------------
1383    Name: Matthias Blume
1384    Date: 2002/02/15 12:08:00 EST
1385    Tag: blume-20020215-iptrs
1386    Description:
1387    
1388    Two improvements to ml-nlffigen:
1389    
1390      1. Write files only if they do not exist or if their current contents
1391         do not coincide with what's being written.  (That is, avoid messing
1392         with the time stamps unless absolutely necessary.)
1393    
1394      2. Implement a "repository" mechanism for generated files related
1395         to "incomplete pointer types".   See the README file for details.
1396    
1397    ----------------------------------------------------------------------
1398    Name: Matthias Blume
1399    Date: 2002/02/14 11:50:00 EST
1400    Tag: blume-20020214-quote
1401    Description:
1402    
1403    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1404    because of the new and improved tag generation scheme.  (Thanks to Allen
1405    Leung for pointing it out.)
1406    
1407    ----------------------------------------------------------------------
1408    Name: Lal George
1409    Date: 2002/02/14 09:55:27 EST 2002
1410    Tag: george-20020214-isabelle-bug
1411    Description:
1412    
1413    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1414    of Isabelle on the x86.
1415    
1416    From Allen:
1417    -----------
1418     I've found the problem:
1419    
1420         in ra-core.sml, I use the counter "blocked" to keep track of the
1421         true number of elements in the freeze queue.  When the counter goes
1422         to zero, I skip examining the queue.  But I've messed up the
1423         bookkeeping in combine():
1424    
1425             else ();
1426             case !ucol of
1427               PSEUDO => (if !cntv > 0 then
1428                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1429                                        ^^^^^^^^^^^^^^^^^^^^^^^
1430                      moveu := mergeMoveList(!movev, !moveu)
1431                     )
1432                  else ();
1433    
1434         combine() is called to coalesce two nodes u and v.
1435         I think I was thinking that if the move counts of u and v are both
1436         greater than zero then after they are coalesced then one node is
1437         removed from the freeze queue.  Apparently I was thinking that
1438         both u and v are of low degree, but that's clearly not necessarily true.
1439    
1440    
1441    02/12/2002:
1442        Here's the patch.  HOL now compiles.
1443    
1444        I don't know how this impact on performance (compile
1445        time or runtime).  This bug caused the RA (especially on the x86)
1446        to go thru the potential spill phase when there are still nodes on the
1447        freeze queue.
1448    
1449    
1450    
1451    
1452    ----------------------------------------------------------------------
1453    Name: Matthias Blume
1454    Date: 2002/02/13 22:40:00 EST
1455    Tag: blume-20020213-fptr-rtti
1456    Description:
1457    
1458    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1459    updates.
1460    
1461    ----------------------------------------------------------------------
1462    Name: Matthias Blume
1463    Date: 2002/02/13 16:41:00 EST
1464    Tag: blume-20020213-cmlpq
1465    Description:
1466    
1467    Added new priority queue export symbols (which have just been added to
1468    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1469    not compile and the installer would choke.)
1470    
1471    ----------------------------------------------------------------------
1472    Name: Matthias Blume
1473    Date: 2002/02/13 16:15:00 EST
1474    Tag: blume-20020213-various
1475    Description:
1476    
1477    1. More tweaks to ml-nlffigen:
1478    
1479       - better internal datastructures (resulting in slight speedup)
1480       - "-match" option requires exact match
1481       - "localized" gensym counters (untagged structs/unions nested within
1482         other structs/unions or within typedefs get a fresh counter; their
1483         tag will be prefixed by a concatenation of their parents' tags)
1484       - bug fixes (related to calculation of transitive closure of types
1485         to be included in the output)
1486    
1487    2. Minor Basis updates:
1488    
1489       - added implementations for List.collate and Option.app
1490    
1491    ----------------------------------------------------------------------
1492    Name: Matthias Blume
1493    Date: 2002/02/11 15:55:00 EST
1494    Tag: blume-20020211-gensym
1495    Description:
1496    
1497    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1498    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1499    names (ML structure names that correspond to unnamed C structs, unions,
1500    and enums), so that separate runs of ml-nlffigen do not clash.
1501    
1502    ----------------------------------------------------------------------
1503    Name: Matthias Blume
1504    Date: 2002/02/11 12:05:00 EST
1505    Tag: blume-20020211-gensml
1506    Description:
1507    
1508    A quick fix for a problem with GenSML (in the pgraph-util library):
1509    Make generation of toplevel "local" optional.  (Strictly speaking,
1510    signature definitions within "local" are not legal SML.)
1511    
1512    Other than that: updates to INSTALL and cm/TODO.
1513    
1514    ----------------------------------------------------------------------
1515    Name: Matthias Blume
1516    Date: 2002/02/08 15:00:00 EST
1517    Tag: blume-20020208-uniquepid
1518    Description:
1519    
1520    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1521    
1522    1. The installer (config/install.sh) has gotten smarter:
1523    
1524         - Configuration options are a bit easier to specify now
1525           (in config/targets).
1526         - Bug in recognizing .tar.bz2 files fixed.
1527         - Installer automatically resolves dependencies between
1528           configuration options (e.g., if you ask for eXene, you will
1529           also get cml -- regardless whether you asked for it or not).
1530         - Installer can run in "quieter mode" by setting the environment
1531           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1532           "completely silent", though.
1533         - Build HashCons library as part of smlnj-lib.
1534    
1535    2. A new scheme for assigning persistent identifiers to compilation
1536       units (and, by extension, to types etc.) has been put into place.
1537       This fixes a long-standing bug where types and even dynamic values
1538       can get internally confused, thereby compromising type safety
1539       (abstraction) and dynamic correctness.  See
1540    
1541         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1542    
1543       for an example of how things could go wrong until now.
1544    
1545       The downside of the new scheme is that pids are not quite as
1546       persistent as they used to be: CM will generate a fresh pid
1547       for every compilation unit that it thinks it sees for the first
1548       time.  That means that if you compile starting from a clean, fresh
1549       source tree at two different times, you end up with different
1550       binaries.
1551    
1552       Cutoff recompilation, however, has not been compromised because
1553       CM keeps pid information in special caches between runs.
1554    
1555    ----------------------------------------------------------------------
1556    Name: Lal George
1557    Date: 2002/02/07 15:34:13 EST 2002
1558    Tag: <none>
1559    Description:
1560    
1561    Compilers that generate assembly code may produce  global labels
1562    whose value is resolved at link time. The various peephole optimization
1563    modules did not take this in account.
1564    
1565    TODO. The Labels.addrOf function should really return an option
1566    type so that clients are forced to deal with this issue, rather
1567    than an exception being raised.
1568    
1569    ----------------------------------------------------------------------
1570    Name: Lal George
1571    Date: 2002/02/06 13:55:02 EST
1572    Tag: george-20020206-ra-breakup
1573    Description:
1574    
1575    1. A bug fix from Allen.
1576    
1577        A typo causes extra fstp %st(0)'s to be generated at compensation
1578        edges, which might cause stack underflow traps at runtime.  This
1579        occurs in fft where there are extraneous fstps right before the 'into'
1580        trap instruction (in this case they are harmless since none of the
1581        integers overflow.)
1582    
1583    2. Pulled out various utility modules that were embedded in the modules
1584       of the register allocator. I need these modules for other purposes, but
1585       they are not complete enough to put into a library (just yet).
1586    ----------------------------------------------------------------------
1587    Name: Matthias Blume
1588    Date: 2002/01/31 16:05:00 EST
1589    Tag: blume-20020131-sparc-ccalls
1590    Description:
1591    
1592    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1593       of extra stack space by mistake.  Fixed.
1594    
1595    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1596    
1597    ----------------------------------------------------------------------
1598    Name: Allen Leung
1599    Date: 2002/01/30
1600    Tag: leunga-20020130-nowhere-bug-fix
1601    Description:
1602    
1603       MLRISC bug fixes:
1604       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1605       2. MachineInt.fromString was negating its value.
1606    
1607    ----------------------------------------------------------------------
1608    Name: Matthias Blume
1609    Date: 2002/01/29
1610    Tag: blume-20020129-INSTALL
1611    Description:
1612    
1613    - Added somewhat detailed installation instructions (file INSTALL).
1614    - Fixed curl-detection bug in config/install.sh.
1615    - It is now possible to select the URL getter using the URLGETTER
1616      environment variable:
1617    
1618          not set / "unknown"      --> automatic detection (script tries wget,
1619                                       curl, and lynx)
1620          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1621                                       how to properly invoke them)
1622          other                    --> use $URLGETTER directly, it must take
1623                                       precisely two command-line arguments
1624                                       (source URL and destination file name)
1625    
1626    ----------------------------------------------------------------------
1627    Name: Matthias Blume
1628    Date: 2002/01/28
1629    Tag: blume-20020128-sparc-ccalls
1630    Description:
1631    
1632    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1633    - Make use of the allocParam argument in sparc-c-calls.
1634    
1635    ----------------------------------------------------------------------
1636    Name: Matthias Blume
1637    Date: 2002/01/28
1638    Tag: blume-20020128-allocParam
1639    Description:
1640    
1641    John Reppy:  Changes c-calls API to accept client-callback for
1642    allocating extra stack space.
1643    me: Corresponding changes to mlriscGen (using a dummy argument that
1644        does not change the current behavior).
1645    
1646    ----------------------------------------------------------------------
1647    Name: Matthias Blume
1648    Date: 2002/01/28 12:00:00
1649    Tag: Release_110_38
1650    Description:
1651    
1652    This time for real!!!
1653    
1654    ----------------------------------------------------------------------
1655    Name: Matthias Blume
1656    Date: 2002/01/28 10:56:00 EST
1657    Tag: blume-20020128-retraction
1658    Description:
1659    
1660    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1661       with blume-Release_110_38-retracted.)
1662    
1663    1. Fixed a problem with incorrect rounding modes in real64.sml.
1664       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1665    
1666    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1667       and enums fixed.  The naming of corresponding ML identifiers should
1668       now be consistent again.
1669    
1670    ----------------------------------------------------------------------
1671    Name: Allen Leung
1672    Date: 2002/01/27
1673    Tag: leunga-20020127-nowhere
1674    Description:
1675    
1676       Added a target called nowhere in the configuration scripts.
1677       Enabling this will build the MLRISC 'nowhere' tool (for translating
1678       programs with where-clauses into legal SML code) during installation.
1679    
1680    ----------------------------------------------------------------------
1681    Name: Matthias Blume
1682    Date: 2002/01/25 21:27:00 EST
1683    Tag: blume-Release_110_38-retracted
1684    Description:
1685    
1686    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1687    
1688    README will be added later.
1689    
1690    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1691    (Release_110_38) removed.  Reason: Last-minute bug fixes.
1692    
1693    ----------------------------------------------------------------------
1694    Name: Matthias Blume
1695    Date: 2002/01/25
1696    Tag: blume-20020125-ffi
1697    Description:
1698    
1699    A large number of tweaks and improvements to ml-nlffi-lib and
1700    ml-nlffigen:
1701    
1702       - ML represenation types have been streamlined
1703       - getter and setter functions work with concrete values, not abstract
1704         ones where possible
1705       - ml-nlffigen command line more flexible (see README file there)
1706       - some bugs have been fixed (hopefully)
1707    
1708    ----------------------------------------------------------------------
1709    Name: Lal George
1710    Date: 2002/01/24
1711    Tag: george-20020124-risc-ra-interface
1712    Description:
1713    
1714       There is a dramatic simplification in the interface to the
1715       register allocator for RISC architectures as a result of making
1716       parallel copy instructions explicit.
1717    
1718    ----------------------------------------------------------------------
1719    Name: Matthias Blume
1720    Date: 2002/01/22
1721    Tag: blume-20020122-x86-ccalls
1722    Description:
1723    
1724    Bug fix for c-calls on x86 (having to do with how char- and
1725    short-arguments are being handled).
1726    
1727    ----------------------------------------------------------------------
1728    Name: Matthias Blume
1729    Date: 2002/01/21
1730    Tag: blume-20020121-ff
1731    Description:
1732    
1733    Another day of fiddling with the FFI...
1734    
1735    1. Bug fix/workaround:  CKIT does not complain about negative array
1736       dimensions, so ml-nlffigen has to guard itself against this possibility.
1737       (Otherwise a negative dimension would send it into an infinite loop.)
1738    
1739    2. Some of the abstract types (light objects, light pointers, most "base"
1740       types) in structure C are now eqtypes.
1741    
1742    3. Added constructors and test functions for NULL function pointers.
1743    
1744    ----------------------------------------------------------------------
1745    Name: Matthias Blume
1746    Date: 2002/01/18
1747    Tag: blume-20020118-ready-for-new-release
1748    Description:
1749    
1750    Made config/srcarchiveurl point to a new place.  (Will provide boot
1751    files shortly.)
1752    
1753    Maybe we christen this to be 110.38?
1754    
1755    ----------------------------------------------------------------------
1756    Name: Matthias Blume
1757    Date: 2002/01/18
1758    Tag: blume-20020118-more-ffifiddle
1759    Description:
1760    
1761    Today's FFI fiddling:
1762    
1763      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
1764        C.Set.* that use concrete (MLRep.*) arguments and results instead
1765        of abstract ones.
1766    
1767      - Provided word-style bit operations etc. for "int" representation
1768        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1769        and Long.
1770    
1771    ----------------------------------------------------------------------
1772    Name: Matthias Blume
1773    Date: 2002/01/18
1774    Tag: blume-20020118-use-x86-fp
1775    Description:
1776    
1777    Now that x86-fast-fp seems to be working, I turned it back on again
1778    by default.  (Seems to work fine now, even with the FFI.)
1779    
1780    Other than that, I added some documentation about the FFI to
1781    src/ml-nlffigen/README and updated the FFI test examples in
1782    src/ml-nlffi-lib/Tests/*.
1783    
1784    ----------------------------------------------------------------------
1785    Name: Allen Leung
1786    Date: 2002/01/17
1787    Tag: leunga-20020117-x86-fast-fp-call
1788    Description:
1789    
1790       1. Fixed a problem with handling return fp values when x86's fast fp
1791          mode is turned on.
1792    
1793       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
1794          of %f32.
1795    
1796       3. Added a constructor INT32lit to the ast of MLRISC tools.
1797    
1798    ----------------------------------------------------------------------
1799    Name: Matthias Blume
1800    Date: 2002/01/16
1801    Tag: blume-20020116-ffifiddle
1802    Description:
1803    
1804    More fiddling with the FFI interface:
1805    
1806     - Make constness 'c instead of rw wherever possible.  This eliminates
1807       the need for certain explicit coercions.  (However, due to ML's
1808       value polymorphism, there will still be many cases where explicit
1809       coercions are necessary.  Phantom types are not the whole answer
1810       to modeling a subtyping relationship in ML.)
1811    
1812     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
1813    
1814     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
1815       clumsy workaround for varargs problem.)
1816    
1817    ----------------------------------------------------------------------
1818    Name: Lal George
1819    Date: 2002/01/15
1820    Tag: <none>
1821    Description:
1822    
1823    1. Since COPY instructions are no longer native to the architecture,
1824       a generic functor can be used to implement the expandCopies function.
1825    
1826    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1827       TEXT segment.
1828    
1829    ----------------------------------------------------------------------
1830    Name: Matthias Blume
1831    Date: 2002/01/15
1832    Tag: blume-20020115-ffiupdates
1833    Description:
1834    
1835    1. Fix for bug resulting in single-precision float values being returned
1836       incorrectly from FFI calls.
1837    
1838    2. Small modifications to C FFI API:
1839    
1840        - memory-allocation routines return straight objects (no options)
1841          and raise an exception in out-of-memory situations
1842        - unsafe extensions to cast between function pointers and pointers
1843          from/to ints
1844        - added structure C_Debug as an alternative to structure C where
1845          pointer-dereferencing (|*| and |*!) always check for null-pointers
1846        - added open_lib' to DynLinkage;  open_lib' works like open_lib
1847          but also takes a (possibly empty) list of existing library handles
1848          that the current library depends on
1849    
1850    ----------------------------------------------------------------------
1851    Name: Matthias Blume
1852    Date: 2002/01/10
1853    Tag: blume-20020110-newffigen
1854    Description:
1855    
1856    1. Updates to portable graph code.
1857    
1858    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
1859       more scalable now so that even huge interfaces such as the one
1860       for GTK compile in finite time and space. :-)
1861       See src/ml-nlffigen/README for details on what's new.
1862    
1863  ----------------------------------------------------------------------  ----------------------------------------------------------------------
1864  Name: Lal George  Name: Lal George
1865  Date: 2001/01/09 14:31:35 EST 2002  Date: 2001/01/09 14:31:35 EST 2002
# Line 355  Line 2204 
2204    
2205  ----------------------------------------------------------------------  ----------------------------------------------------------------------
2206  Name: Matthias Blume  Name: Matthias Blume
 >>>>>>> 1.169  
2207  Date: 2001/09/18 15:35:00 EDT  Date: 2001/09/18 15:35:00 EDT
2208  Tag: blume-20010918-readme11036  Tag: blume-20010918-readme11036
2209  Description:  Description:

Legend:
Removed from v.1009  
changed lines
  Added in v.1279

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