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 1053, Wed Feb 6 19:11:13 2002 UTC revision 1296, Tue Jan 7 15:40:55 2003 UTC
# Line 13  Line 13 
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Allen Leung
17    Date: 2003/01/07 10:40:00 EST
18    Tag: leunga-20030107-int-rem
19    Description:
20    
21    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
22    The arguments to the SUBL instruction were swapped.
23    
24    ----------------------------------------------------------------------
25    Name: Matthias Blume
26    Date: 2002/12/12 16:25:00 EST
27    Tag: blume-20021212-risc-ra
28    Description:
29    
30    Fixed a serious bug in the rewrite code for FP spilling/reloading that
31    sent the RA into an infinite loop when floating point registers get
32    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
33    110.37 and 110.38.)
34    There was another set of potential problems related to the handling of
35    MLRISC annotations (but those did not yet cause real problems, apparently).
36    
37    ----------------------------------------------------------------------
38    Name: Matthias Blume
39    Date: 2002/12/06 22:40:00 EST
40    Tag: blume-20021206-cm-fileid
41    Description:
42    
43    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
44    This fixes the problem of CM getting confused by files that suddenly
45    change their identity (e.g., by getting unlinked and recreated by some
46    text editor such as vi).  There might be a better/cheaper/cleaner way
47    of doing this, but for now this will have to do.
48    
49    ----------------------------------------------------------------------
50    Name: Matthias Blume
51    Date: 2002/10/28 09:50:00 EST
52    Tag: blume-20021028-typecheck
53    Description:
54    
55    Exported structure Typecheck from $smlnj/viscomp/core.cm.
56    
57    ----------------------------------------------------------------------
58    Name: Matthias Blume
59    Date: 2002/10/17 09:10:00 EDT
60    Tag: Release_110_42
61    Description:
62    
63    In good old tradition, there has been a slight hiccup so that we have
64    to patch 110.42 after the fact.   The old release tag has been replaced
65    (see below).
66    
67    The change solves a problem with two competing approaches the
68    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
69    each other's way.
70    
71    This change only affects the runtime system code and the installer script.
72    (No new bootfiles.)
73    
74    ----------------------------------------------------------------------
75    Name: Matthias Blume
76    Date: 2002/10/16 12:00:00 EDT
77    Tag: Release_110_42_removed
78    Description:
79    
80    New working release.  New bootfiles.
81    
82    ----------------------------------------------------------------------
83    Name: Matthias Blume
84    Date: 2002/10/10 13:10:00 EDT
85    Tag: blume-20021010-ppc-divs
86    Description:
87    
88    The mltree operator DIVS must be implemented with an overflow check on
89    the PPC because the hardware indicates divide-by-zero using "overflow" as
90    well.
91    
92    ----------------------------------------------------------------------
93    Name: Matthias Blume
94    Date: 2002/07/23 11:45:00
95    Tag: blume-20020723-smlnj-home
96    Description:
97    
98    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
99    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
100    CM_PATHCONFIG is also set), the path configuration file is assumed
101    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
102    the entire tree to some other place and everything will "just work".
103    
104    (Companion commands such as ml-build and ml-makedepend also sense this
105    variable.)
106    
107    ----------------------------------------------------------------------
108    Name: Matthias Blume
109    Date: 2002/07/12 21:19:00 EDT
110    Tag: blume-20020712-liveness
111    Description:
112    
113    Exported two useful "step" functions from liveness module (MLRISC).
114    
115    ----------------------------------------------------------------------
116    Name: Matthias Blume
117    Date: 2002/07/05 16:00 EDT
118    Tag: Release_110_41
119    Description:
120    
121    New working release.  New bootfiles.
122    
123    ----------------------------------------------------------------------
124    Name: Matthias Blume
125    Date: 2002/07/05 10:25:00 EDT
126    Tag: blume-20020705-btimp
127    Description:
128    
129    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
130    other clients can set up backtracing support.
131    
132    ----------------------------------------------------------------------
133    Name: Matthias Blume
134    Date: 2002/06/25 17:23:00 EDT
135    Tag: blume-20020625-fpmax
136    Description:
137    
138    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
139    case.  (The sense of the isNaN test was reversed -- which made min and
140    max always return their first argument.)
141    
142    ----------------------------------------------------------------------
143    Name: Matthias Blume
144    Date: 2002/06/11
145    Tag: blume-20020611-unixpath
146    Description:
147    
148    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
149    
150    ----------------------------------------------------------------------
151    Name: Matthias Blume
152    Date: 2002/06/10 16:35:00 EDT
153    Tag: blume-20020610-ieeereal
154    Description:
155    
156    I back-ported my implementation of IEEEReal.fromString from the
157    idlbasis-devel branch so that we can test it.
158    
159    Another small change is that ppDec tries to give more information
160    than just "<sig>" in the case of functors.  However, this code is
161    broken in some mysterious way if the functor's body's signature
162    has not been declared by ascription but gets inferred from the
163    implementation.  This needs fixing...
164    
165    ----------------------------------------------------------------------
166    Name: Matthias Blume
167    Date: 2002/05/31
168    Tag: blume-20020531-btrace-mode
169    Description:
170    
171    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
172    the wayside when I switched over to using Controls everywhere.)
173    
174    ----------------------------------------------------------------------
175    Name: Lal George
176    Date: 2002/05/23 12:21:40 EDT
177    Tag: george-20020523-visual-labels
178    Description:
179    
180    Labels are now displayed in the graphical output to make
181    the fall-through and target blocks obvious.
182    
183    ----------------------------------------------------------------------
184    Name: Matthias Blume
185    Date: 2002/05/22 11:03:00 EDT
186    Tag: blume-20020522-shrink
187    Description:
188    
189    John tweaked yesterday's fix for 1131 to handle an out-of-memory
190    situation that comes up when allocating huge arrays.
191    
192    ----------------------------------------------------------------------
193    Name: Matthias Blume
194    Date: 2002/05/21 16:00:00 EDT
195    Tag: Release_110_40
196    Description:
197    
198    New working release (110.40).  New bootfiles.
199    
200    [Also: John Reppy fixed GC bug 1131.]
201    
202    ----------------------------------------------------------------------
203    Name: Matthias Blume
204    Date: 2002/05/21 12:35:00 EDT
205    Tag: blume-20020521-cmdoc
206    Description:
207    
208    CM documentation update.
209    
210    ----------------------------------------------------------------------
211    Name: Matthias Blume
212    Date: 2002/05/21 10:55:00 EDT
213    Tag: blume-20020521-misc
214    Description:
215    
216    - John tweaked runtime to be silent on heap export (except when
217      GC messages are on).
218    - I added a few more things (cross-compiling versions of CMB) to
219      config/preloads (as suggestions).
220    
221    ----------------------------------------------------------------------
222    Name: Matthias Blume
223    Date: 2002/05/20 22:25:00 EDT
224    Tag: blume-20020520-controls
225    Description:
226    
227    - Added ControlUtil structure to control-lib.cm.
228    - Use it throughout.
229    - Used Controls facility to define MLRISC controls (as opposed to
230      registering MLRISC control ref cells with Controls after the fact)
231    - Fixed messed-up controls priorities.
232    
233    * Removed again all the stuff from config/preloads that one wouldn't
234      be able to preload at the time the initial heap image is built.
235      (Many libraries, e.g., CML, do not exist yet at this time.  The
236       only libraries that can be preloaded via config/preloads are those
237       that come bundled with the bootfiles.)
238    
239    ----------------------------------------------------------------------
240    Name: Matthias Blume
241    Date: 2002/05/20 10:59:00 EDT
242    Tag: blume-20020520-preloads
243    Description:
244    
245    Added a lot of commented-out suggestions for things to be included
246    in config/preloads.
247    
248    ----------------------------------------------------------------------
249    Name: Allen Leung
250    Date: 2002/05/18 14:20:00 EDT
251    Tag: leunga-20020518-mdl
252    Description:
253    
254    o Made the mdl tool stuff compile and run again.
255    o I've disabled all the stuff that depends on RTL specifications; they
256      are all badly broken anyway.
257    
258    ----------------------------------------------------------------------
259    Name: Matthias Blume
260    Date: 2002/05/17 16:49:00 EDT
261    Tag: blume-20020517-controls
262    Description:
263    
264    0. John Reppy made several modifications to the SML/NJ library.
265       In particular, there is a shiny new controls-lib.cm.
266    
267    1. Pushed new controls interface through compiler so that everything
268       compiles again.
269    
270    2. Added FormatComb and FORMAT_COMB to the CML version of the
271       SML/NJ library (so that CML compiles again).
272    
273    3. Modified init scripts because XXX_DEFAULT environment variables
274       are no longer with us.  (Boot-time initialization is now done
275       using the same environment variables that are also used for
276       startup-time initialization of controls.)
277    
278    ----------------------------------------------------------------------
279    Name: Lal George
280    Date: 2002/05/15 09:20:10 EDT
281    Tag: george-20020515-pseudo-op-decls
282    Description:
283    
284            All pseudo-ops emitted before the first segment declaration
285    such as TEXT, DATA, and BSS directives are assumed to be global
286    declarations and are emitted first in the assembly file. This is
287    useful in a number of situations where one has pseudo-ops that are not
288    specific to any segment, and also works around the constraint that one
289    cannot have client pseudo-ops in the TEXT segment.
290    
291            Because no segment is associated with these declarations it is
292    an error to allocate any space or objects before the first segment
293    directive and an exception will be raised. However, we cannot make
294    this check for client pseudo-ops.
295    
296            These top level declarations are a field in the CFG graph_info.
297    In theory you can continue to add to this field after the CFG has been
298    built -- provided you know what you are doing;-)
299    
300    ----------------------------------------------------------------------
301    Name: Matthias Blume
302    Date: 2002/05/13 16:40:00 EDT
303    Tag: blume-20020513-pp-etc
304    Description:
305    
306    A few minor bugfixes:
307    
308      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
309        (Bogus printouts for redefined bindings still occur.  Compiler
310        bug should no longer occur now.  We need to redo the prettyprinter
311        from scratch.)
312    
313      - CM pathname printer now also adds escape sequences for ( and )
314    
315      - commend and docu fixes for ml-nlffi
316    
317    ----------------------------------------------------------------------
318    Name: Matthias Blume
319    Date: 2002/05/10 16:40:00 EDT
320    Tag: blume-20020510-erg-textio
321    Description:
322    
323    Applied the following bugfix provided by Emden Gansner:
324    
325        Output is corrupted when outputSubstr is used rather than output.
326    
327        The problem occurs when a substring
328    
329            ss = (s, dataStart, dataLen)
330    
331        where dataStart > 0, fills a stream buffer with avail bytes left.
332        avail bytes of s, starting at index dataStart, are copied into the
333        buffer, the buffer is flushed, and then the remaining dataLen-avail
334        bytes of ss are copied into the beginning of the buffer. Instead of
335        starting this copy at index dataStart+avail in s, the current code
336        starts the copy at index avail.
337    
338        Fix:
339        In text-io-fn.sml, change line 695 from
340             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
341        to
342             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
343    
344    ----------------------------------------------------------------------
345    Name: Matthias Blume
346    Date: 2002/04/12 13:55:00 EDT
347    Tag: blume-20020412-assyntax
348    Description:
349    
350    1. Grabbed newer assyntax.h from the XFree86 project.
351    2. Fiddled with how to compile X86.prim.asm without warnings.
352    3. (Very) Minor cleanup in CM.
353    
354    ----------------------------------------------------------------------
355    Name: Matthias Blume
356    Date: 2002/04/01 (no joke!) 17:07:00 EST
357    Tag: blume-20020401-x86div
358    Description:
359    
360    Added full support for div/mod/rem/quot on the x86, using the machine
361    instruction's two results (without clumsily recomputing the remainder)
362    directly where appropriate.
363    
364    Some more extensive power-of-two support was added to the x86 instruction
365    selector (avoiding expensive divs, mods, and muls where they can be
366    replaced with cheaper shifts and masks).  However, this sort of thing
367    ought to be done earlier, e.g., within the CPS optimizer so that
368    all architectures benefit from it.
369    
370    The compiler compiles to a fixed point, but changes might be somewhat
371    fragile nevertheless.  Please, report any strange things that you might
372    see wrt. div/mod/quot/rem...
373    
374    ----------------------------------------------------------------------
375    Name: Matthias Blume
376    Date: 2002/03/29 17:22:00
377    Tag: blume-20020329-div
378    Description:
379    
380    Fixed my broken div/mod logic.  Unfortunately, this means that the
381    inline code for div/mod now has one more comparison than before.
382    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
383    The problem was with quotient = 0, because that alone does not tell
384    us which way the rounding went.  One then has to look at whether
385    remainder and divisor have the same sign...  :(
386    
387    Anyway, I replaced the bootfiles with fresh ones...
388    
389    ----------------------------------------------------------------------
390    Name: Matthias Blume
391    Date: 2002/03/29 14:10:00 EST
392    Tag: blume-20020329-inlprims
393    Description:
394    
395    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
396    
397    Primops have changed. This means that the bin/boot-file formats have
398    changed as well.
399    
400    To make sure that there is no confusion, I made a new version.
401    
402    
403    CHANGES:
404    
405    * removed REMT from mltree (remainder should never overflow).
406    
407    * added primops to deal with divisions of all flavors to the frontend
408    
409    * handled these primops all the way through so they map to their respective
410      MLRISC support
411    
412    * used these primops in the implementation of Int, Int32, Word, Word32
413    
414    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
415    
416    * parameterized INLMIN, INLMAX, and INLABS by a numkind
417    
418    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
419      floating point
420    
421    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
422      and Real (but Real.abs maps to a separate floating-point-only primop)
423    
424    
425    TODO items:
426    
427    * Hacked Alpha32 instruction selection, disabling the selection of REMx
428      instructions because the machine instruction encoder cannot handle
429      them.  (Hppa, PPC, and Sparc instruction selection did not handle
430      REM in the first place, and REM is supported by the x86 machine coder.)
431    
432    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
433      selection phase.  (The two can be streamlined because the hardware
434      delivers both quotient and remainder at the same time anyway.)
435    
436    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
437      (Currently the behavior is inconsistent both across architectures and
438      wrt. the draft Basis spec.)
439    
440    * Word8 should eventually be handled natively, too.
441    
442    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
443      as if there currently is no execution path that could trigger it in
444      SML/NJ.  (The assumptions underlying functions arith and promotable do not
445      hold for things like multiplication and division.)
446    
447    ----------------------------------------------------------------------
448    Name: Matthias Blume
449    Date: 2002/03/27 16:27:00 EST
450    Tag: blume-20020327-mlrisc-divisions
451    Description:
452    
453    Added support for all four division operations (ML's div, mod, quot,
454    and rem) to MLRISC.  In the course of doing so, I also rationalized
455    the naming (no more annoying switch-around of DIV and QUOT), by
456    parameterizing the operation by div_rounding_mode (which can be either
457    DIV_TO_ZERO or DIV_TO_NEGINF).
458    
459    The generic MLTreeGen functor takes care of compiling all four
460    operations down to only round-to-zero div.
461    
462    Missing pieces:
463    
464      * Doing something smarter than relying on MLTreeGen on architectures
465        like, e.g., the x86 where hardware division delivers both quotient and
466        remainder at the same time.  With this, the implementation of the
467        round-to-neginf operations could be further streamlined.
468    
469      * Remove inlining support for div/mod/rem from the frontend and replace it
470        with primops that get carried through to the backend.  Do this for all
471        int and word types.
472    
473    ----------------------------------------------------------------------
474    Name: Matthias Blume
475    Date: 2002/03/25 17:25:00 EST
476    Tag: blume-20020325-divmod
477    Description:
478    
479    I improved (hopefully without breaking them) the implementation of Int.div,
480    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
481    advantage of the following observations:
482    
483      Let  q = x quot y      r = x rem y
484           d = x div  y      m = x mod y
485    
486    where "quot" is the round-to-zero version of integer division that
487    hardware usually provides.  Then we have:
488    
489         r = x - q * y        where neither the * nor the - will overflow
490         d = if q >= 0 orelse x = q * y then q else q - 1
491                              where neither the * nor the - will overflow
492         m = if q >= 0 orelse r = 0 then r else r + y
493                              where the + will not overflow
494    
495    This results in substantial simplification of the generated code.
496    The following table shows the number of CFG nodes and edges generated
497    for
498            fun f (x, y) = x OPER y
499            (* with OPER \in div, mod, quot, rem *)
500    
501    
502        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
503        --------------------------------------------------------
504         div |         24 |         39 |         12 |         16
505         mod |         41 |         71 |         12 |         16
506        quot |          8 |         10 |          8 |         10
507         rem |         10 |         14 |          8 |         10
508    
509    
510    ----------------------------------------------------------------------
511    Name: Matthias Blume
512    Date: 2002/03/25 22:06:00 EST
513    Tag: blume-20020325-cprotobug
514    Description:
515    
516    Fixed a bug in cproto (c prototype decoder).
517    
518    ----------------------------------------------------------------------
519    Name: Matthias Blume
520    Date: 2002/03/25 16:00:00 EST
521    Tag: blume-20020325-raw-primops
522    Description:
523    
524    I did some cleanup to Allen's new primop code and
525    replaced yesterday's bootfiles with new ones.
526    (But they are stored in the same place.)
527    
528    ----------------------------------------------------------------------
529    Name: Matthias Blume
530    Date: 2002/03/24 22:40:00 EST
531    Tag: blume-20020324-bootfiles
532    Description:
533    
534    Made the bootfiles that Allen asked for.
535    
536    ----------------------------------------------------------------------
537    Name: Allen Leung
538    Date: 2002/03/23 15:50:00 EST
539    Tag: leunga-20020323-flint-cps-rcc-primops
540    Description:
541    
542      1. Changes to FLINT primops:
543    
544        (* make a call to a C-function;
545         * The primop carries C function prototype information and specifies
546         * which of its (ML-) arguments are floating point. C prototype
547         * information is for use by the backend, ML information is for
548         * use by the CPS converter. *)
549      | RAW_CCALL of { c_proto: CTypes.c_proto,
550                       ml_args: ccall_type list,
551                       ml_res_opt: ccall_type option,
552                       reentrant : bool
553                     } option
554       (* Allocate uninitialized storage on the heap.
555        * The record is meant to hold short-lived C objects, i.e., they
556        * are not ML pointers.  With the tag, the representation is
557        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
558        *)
559      | RAW_RECORD of {tag:bool,sz:int}
560      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
561    
562      2.  These CPS primops are now overloaded:
563    
564           rawload of {kind:numkind}
565           rawstore of {kind:numkind}
566    
567          The one argument form is:
568    
569             rawload {kind} address
570    
571          The two argument form is:
572    
573             rawload {kind} [ml object, byte-offset]
574    
575      3. RAW_CCALL/RCC now takes two extra arguments:
576    
577         a. The first is whether the C call is reentrant, i.e., whether
578            ML state should be saved and restored.
579         b. The second argument is a string argument specifying the name of
580            library and the C function.
581    
582         These things are currently not handled in the code generator, yet.
583    
584      4. In CProto,
585    
586         An encoding type of "bool" means "ml object" and is mapped into
587         C prototype of PTR.  Note that "bool" is different than "string",
588         even though "string" is also mapped into PTR, because "bool"
589         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
590    
591      5. Pickler/unpicker
592    
593         Changed to handle RAW_RECORD and newest RAW_CCALL
594    
595      6. MLRiscGen,
596    
597         1. Changed to handle the new rawload/rawstore/rawrecord operators.
598         2. Code for handling C Calls has been moved to a new module CPSCCalls,
599            in the file CodeGen/cpscompile/cps-c-calls.sml
600    
601      7. Added the conditional move operator
602    
603             condmove of branch
604    
605         to cps.  Generation of this is still buggy so it is currently
606         disabled.
607    
608    ----------------------------------------------------------------------
609    Name: Lal George
610    Date: 2002/03/22 14:18:25 EST
611    Tag: george-20020322-cps-branch-prob
612    Description:
613    
614    Implemented the Ball-Larus branch prediction-heuristics, and
615    incorporated graphical viewers for control flow graphs.
616    
617    Ball-Larus Heuristics:
618    ---------------------
619    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
620    
621    By design it uses the Dempster-Shafer theory for combining
622    probabilities.  For example, in the function:
623    
624        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
625    
626    the ball-larus heuristics predicts that the n=0 is unlikely
627    (OH-heuristic), and the 'then' branch is unlikely because of the
628    RH-heuristic -- giving the 'then' branch an even lower combined
629    probability using the Dempster-Shafer theory.
630    
631    Finally, John Reppy's loop analysis in MLRISC, further lowers the
632    probability of the 'then' branch because of the loop in the else
633    branch.
634    
635    
636    Graphical Viewing:
637    ------------------
638    I merely plugged in Allen's graphical viewers into the compiler. The
639    additional code is not much. At the top level, saying:
640    
641            Control.MLRISC.getFlag "cfg-graphical-view" := true;
642    
643    will display the graphical view of the control flow graph just before
644    back-patching.  daVinci must be in your path for this to work. If
645    daVinci is not available, then the default viewer can be changed
646    using:
647    
648            Control.MLRISC.getString "viewer"
649    
650    which can be set to "dot" or "vcg" for the corresponding viewers. Of
651    course, these viewers must be in your path.
652    
653    The above will display the compilation unit at the level of clusters,
654    many of which are small, boring, and un-interesting. Also setting:
655    
656            Control.MLRISC.getInt "cfg-graphical-view_size"
657    
658    will display clusters that are larger than the value set by the above.
659    
660    
661    ----------------------------------------------------------------------
662    Name: Matthias Blume
663    Date: 2002/03/21 22:20:00 EST
664    Tag: blume-20020321-kmp-bugfix
665    Description:
666    
667    Changed the interface to the KMP routine in PreString and fixed
668    a minor bug in one place where it was used.
669    
670    ----------------------------------------------------------------------
671    Name: Allen Leung
672    Date: 2002/03/21 20:30:00 EST
673    Tag: leunga-20020321-cfg
674    Description:
675    
676      Fixed a potential problem in cfg edge splitting.
677    
678    ----------------------------------------------------------------------
679    Name: Allen Leung
680    Date: 2002/03/21 17:15:00 EST
681    Tag: leunga-20020321-x86-fp-cfg
682    Description:
683    
684      1. Recoded the buggy parts of x86-fp.
685    
686         a. All the block reordering code has been removed.
687            We now depend on the block placement phases to do this work.
688    
689         b. Critical edge splitting code has been simplified and moved into the
690            CFG modules, as where they belong.
691    
692         Both of these were quite buggy and complex.  The code is now much, much
693         simpler.
694    
695      2. X86 backend.
696    
697         a. Added instructions for 64-bit support.  Instruction selection for
698            64-bit has not been committed, however, since that
699            requires changes to MLTREE which haven't been approved by
700            Lal and John.
701    
702         b. Added support for FUCOMI and FUCOMIP when generating code for
703            PentiumPro and above.  We only generate these instructions in
704            the fast-fp mode.
705    
706         c. Added cases for JP and JNP in X86FreqProps.
707    
708      3. CFG
709    
710         CFG now has a bunch of methods for edge splitting and merging.
711    
712      4. Machine description.
713    
714         John's simplification of MLTREE_BASIS.fcond broke a few machine
715         description things:
716    
717         rtl-build.{sig,sml} and hppa.mdl fixed.
718    
719         NOTE: the machine description stuff in the repository is still broken.
720               Again, I can't put my fixes in because that involves
721               changes to MLTREE.
722    
723    ----------------------------------------------------------------------
724    Name: Matthias Blume
725    Date: 2002/03/20 15:55:00 EST
726    Tag: blume-20020320-kmp
727    Description:
728    
729    Implemented Knuth-Morris-Pratt string matching in PreString and used
730    it for String.isSubstring, Substring.isSubstring, and
731    Substring.position.
732    
733    (Might need some stress-testing.  Simple examples worked fine.)
734    
735    ----------------------------------------------------------------------
736    Name: Matthias Blume
737    Date: 2002/03/19 16:37:00 EST
738    Tag: blume-20020319-witnesses
739    Description:
740    
741    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
742    
743    This implements a generic mechanism for changing constness qualifiers
744    anywhere within big C types without resorting to outright "casts".
745    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
746    modify the constness at the outermost level.)
747    The implementation of "convert" is based on the idea of "witness"
748    values -- values that are not used by the operation but whose types
749    "testify" to their applicability.  On the implementation side, "convert"
750    is simply a projection (returning its second curried argument).  With
751    cross-module inlining, it should not result in any machine code being
752    generated.
753    
754    ----------------------------------------------------------------------
755    Name: Matthias Blume
756    Date: 2002/03/15 16:40:00 EST
757    Tag: blume-20020315-basis
758    Description:
759    
760    Provided (preliminary?) implementations for
761    
762      {String,Substring}.{concatWith,isSuffix,isSubstring}
763    
764    and
765    
766      Substring.full
767    
768    Those are in the Basis spec but they were missing in SML/NJ.
769    
770    ----------------------------------------------------------------------
771    Name: Matthias Blume
772    Date: 2002/03/14 21:30:00 EST
773    Tag: blume-20020314-controls
774    Description:
775    
776    Controls:
777    ---------
778    
779    1. Factored out the recently-added Controls : CONTROLS stuff and put
780       it into its own library $/controls-lib.cm.  The source tree for
781       this is under src/smlnj-lib/Controls.
782    
783    2. Changed the names of types and functions in this interface, so they
784       make a bit more "sense":
785    
786          module -> registry
787          'a registry -> 'a group
788    
789    3. The interface now deals in ref cells only.  The getter/setter interface
790       is (mostly) gone.
791    
792    4. Added a function that lets one register an already-existing ref cell.
793    
794    5. Made the corresponding modifications to the rest of the code so that
795       everything compiles again.
796    
797    6. Changed the implementation of Controls.MLRISC back to something closer
798       to the original.  In particular, this module (and therefore MLRISC)
799       does not depend on Controls.  There now is some link-time code in
800       int-sys.sml that registers the MLRISC controls with the Controls
801       module.
802    
803    CM:
804    ---
805    
806      * One can now specify the lambda-split aggressiveness in init.cmi.
807    
808    ----------------------------------------------------------------------
809    Name: Allen Leung
810    Date: 2002/03/13 17:30:00 EST
811    Tag: leunga-20020313-x86-fp-unary
812    Description:
813    
814    Bug fix for:
815    
816    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
817    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
818    > - fun f(x,(y,z)) = Real.~ y;
819    > [autoloading]
820    > [autoloading done]
821    >       fchsl   (%eax), 184(%esp)
822    > Error: MLRisc bug: X86MCEmitter.emitInstr
823    >
824    > uncaught exception Error
825    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
826    
827    The problem was that the code generator did not generate any fp registers
828    in this case, and the ra didn't know that it needed to run the X86FP phase to
829    translate the pseudo fp instruction.   This only happened with unary fp
830    operators in certain situations.
831    
832    ----------------------------------------------------------------------
833    Name: Matthias Blume
834    Date: 2002/03/13 14:00:00 EST
835    Tag: blume-20020313-overload-etc
836    Description:
837    
838    1. Added _overload as a synonym for overload for backward compatibility.
839       (Control.overloadKW must be true for either version to be accepted.)
840    
841    2. Fixed bug in install script that caused more things to be installed
842       than what was requested in config/targets.
843    
844    3. Made CM aware of the (_)overload construct so that autoloading
845       works.
846    
847    ----------------------------------------------------------------------
848    Name: Matthias Blume
849    Date: 2002/03/12 22:03:00 EST
850    Tag: blume-20020312-url
851    Description:
852    
853    Forgot to update BOOT and srcarchiveurl.
854    
855    ----------------------------------------------------------------------
856    Name: Matthias Blume
857    Date: 2002/03/12 17:30:00 EST
858    Tag: blume-20020312-version110392
859    Description:
860    
861    Yet another version number bump (because of small changes to the
862    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
863    
864    Changes:
865    
866      The new pid generation scheme described a few weeks ago was overly
867      complicated.  I implemented a new mechanism that is simpler and
868      provides a bit more "stability":  Once CM has seen a compilation
869      unit, it keeps its identity constant (as long as you do not delete
870      those crucial CM/GUID/* files).  This means that when you change
871      an interface, compile, then go back to the old interface, and
872      compile again, you arrive at the original pid.
873    
874      There now also is a mechanism that instructs CM to use the plain
875      environment hash as a module's pid (effectively making its GUID
876      the empty string).  For this, "noguid" must be specified as an
877      option to the .sml file in question within its .cm file.
878      This is most useful for code that is being generated by tools such
879      as ml-nlffigen (because during development programmers tend to
880      erase the tool's entire output directory tree including CM's cached
881      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
882      revert to the old, broken behavior of SML/NJ, but in specific cases
883      where there is no danger of interface confusion, its use is ok
884      (I think).
885    
886      ml-nlffigen by default generates "noguid" annotations.  They can be
887      turned off by specifying -guid in its command line.
888    
889    ----------------------------------------------------------------------
890    Name: Lal George
891    Date: 2002/03/12 12 14:42:36 EST
892    Tag: george-20020312-frequency-computation
893    Description:
894    
895    Integrated jump chaining and static block frequency into the
896    compiler. More details and numbers later.
897    
898    ----------------------------------------------------------------------
899    Name: Lal George
900    Date: 2002/03/11 11 22:38:53 EST
901    Tag: george-20020311-jump-chain-elim
902    Description:
903    
904    Tested the jump chain elimination on all architectures (except the
905    hppa).  This is on by default right now and is profitable for the
906    alpha and x86, however, it may not be profitable for the sparc and ppc
907    when compiling the compiler.
908    
909    The gc test will typically jump to a label at the end of the cluster,
910    where there is another jump to an external cluster containing the actual
911    code to invoke gc. This is to allow factoring of common gc invocation
912    sequences. That is to say, we generate:
913    
914            f:
915               testgc
916               ja   L1      % jump if above to L1
917    
918            L1:
919               jmp L2
920    
921    
922    After jump chain elimination the 'ja L1' instructions is converted to
923    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
924    up being implemented in their long form (if L2 is far away) using:
925    
926            jbe     L3      % jump if below or equal to L3
927            jmp     L2
928         L3:
929            ...
930    
931    
932    For large compilation units L2  may be far away.
933    
934    
935    ----------------------------------------------------------------------
936    Name: Matthias Blume
937    Date: 2002/03/11 13:30:00 EST
938    Tag: blume-20020311-mltreeeval
939    Description:
940    
941    A functor parameter was missing.
942    
943    ----------------------------------------------------------------------
944    Name: Allen Leung
945    Date: 2002/03/11 10:30:00 EST
946    Tag: leunga-20020311-runtime-string0
947    Description:
948    
949       The representation of the empty string now points to a
950    legal null terminated C string instead of unit.  It is now possible
951    to convert an ML string into C string with InlineT.CharVector.getData.
952    This compiles into one single machine instruction.
953    
954    ----------------------------------------------------------------------
955    Name: Allen Leung
956    Date: 2002/03/10 23:55:00 EST
957    Tag: leunga-20020310-x86-call
958    Description:
959    
960       Added machine generation for CALL instruction (relative displacement mode)
961    
962    ----------------------------------------------------------------------
963    Name: Matthias Blume
964    Date: 2002/03/08 16:05:00
965    Tag: blume-20020308-entrypoints
966    Description:
967    
968    Version number bumped to 110.39.1.  NEW BOOTFILES!
969    
970    Entrypoints: non-zero offset into a code object where execution should begin.
971    
972    - Added the notion of an entrypoint to CodeObj.
973    - Added reading/writing of entrypoint info to Binfile.
974    - Made runtime system bootloader aware of entrypoints.
975    - Use the address of the label of the first function given to mlriscGen
976      as the entrypoint.  This address is currently always 0, but it will
977      not be 0 once we turn on block placement.
978    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
979      with entry points) from mlriscGen.
980    
981    ----------------------------------------------------------------------
982    Name: Allen Leung
983    Date: 2002/03/07 20:45:00 EST
984    Tag: leunga-20020307-x86-cmov
985    Description:
986    
987       Bug fixes for CMOVcc on x86.
988    
989       1. Added machine code generation for CMOVcc
990       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
991       3. CMOVcc cannot have an immediate operand as argument.
992    
993    ----------------------------------------------------------------------
994    Name: Matthias Blume
995    Date: 2002/03/07 16:15:00 EST
996    Tag: blume-20020307-controls
997    Description:
998    
999    This is a very large but mostly boring patch which makes (almost)
1000    every tuneable compiler knob (i.e., pretty much everything under
1001    Control.* plus a few other things) configurable via both the command
1002    line and environment variables in the style CM did its configuration
1003    until now.
1004    
1005    Try starting sml with '-h' (or, if you are brave, '-H')
1006    
1007    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1008    implements the underlying generic mechanism.
1009    
1010    The interface to some of the existing such facilities has changed somewhat.
1011    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1012    (The getFoo interface is still there for backward-compatibility, but its
1013    use is deprecated.)
1014    
1015    The ml-build script passes -Cxxx=yyy command-line arguments through so
1016    that one can now twiddle the compiler settings when using this "batch"
1017    compiler.
1018    
1019    TODO items:
1020    
1021    We should go through and throw out all controls that are no longer
1022    connected to anything.  Moreover, we should go through and provide
1023    meaningful (and correct!) documentation strings for those controls
1024    that still are connected.
1025    
1026    Currently, multiple calls to Controls.new are accepted (only the first
1027    has any effect).  Eventually we should make sure that every control
1028    is being made (via Controls.new) exactly once.  Future access can then
1029    be done using Controls.acc.
1030    
1031    Finally, it would probably be a good idea to use the getter-setter
1032    interface to controls rather than ref cells.  For the time being, both
1033    styles are provided by the Controls module, but getter-setter pairs are
1034    better if thread-safety is of any concern because they can be wrapped.
1035    
1036    *****************************************
1037    
1038    One bug fix: The function blockPlacement in three of the MLRISC
1039    backpatch files used to be hard-wired to one of two possibilities at
1040    link time (according to the value of the placementFlag).  But (I
1041    think) it should rather sense the flag every time.
1042    
1043    *****************************************
1044    
1045    Other assorted changes (by other people who did not supply a HISTORY entry):
1046    
1047    1. the cross-module inliner now works much better (Monnier)
1048    2. representation of weights, frequencies, and probabilities in MLRISC
1049       changed in preparation of using those for weighted block placement
1050       (Reppy, George)
1051    
1052    ----------------------------------------------------------------------
1053    Name: Lal George
1054    Date: 2002/03/07 14:44:24 EST 2002
1055    Tag: george-20020307-weighted-block-placement
1056    
1057    Tested the weighted block placement optimization on all architectures
1058    (except the hppa) using AMPL to generate the block and edge frequencies.
1059    Changes were required in the machine properties to correctly
1060    categorize trap instructions. There is an MLRISC flag
1061    "weighted-block-placement" that can be used to enable weighted block
1062    placement, but this will be ineffective without block/edge
1063    frequencies (coming soon).
1064    
1065    
1066    ----------------------------------------------------------------------
1067    Name: Lal George
1068    Date: 2002/03/05 17:24:48 EST
1069    Tag: george-20020305-linkage-cluster
1070    
1071    In order to support the block placement optimization, a new cluster
1072    is generated as the very first cluster (called the linkage cluster).
1073    It contains a single jump to the 'real' entry point for the compilation
1074    unit. Block placement has no effect on the linkage cluster itself, but
1075    all the other clusters  have full freedom in the manner in which they
1076    reorder blocks or functions.
1077    
1078    On the x86 the typical linkage code that is generated is:
1079       ----------------------
1080            .align 2
1081       L0:
1082            addl    $L1-L0, 72(%esp)
1083            jmp     L1
1084    
1085    
1086            .align  2
1087       L1:
1088       ----------------------
1089    
1090    72(%esp) is the memory location for the stdlink register. This
1091    must contain the address of the CPS function being called. In the
1092    above example, it contains the address of  L0; before
1093    calling L1 (the real entry point for the compilation unit), it
1094    must contain the address for L1, and hence
1095    
1096            addl $L1-L0, 72(%esp)
1097    
1098    I have tested this on all architectures except the hppa.The increase
1099    in code size is of course negligible
1100    
1101    ----------------------------------------------------------------------
1102    Name: Allen Leung
1103    Date: 2002/03/03 13:20:00 EST
1104    Tag: leunga-20020303-mlrisc-tools
1105    
1106      Added #[ ... ] expressions to mlrisc tools
1107    
1108    ----------------------------------------------------------------------
1109    Name: Matthias Blume
1110    Date: 2002/02/27 12:29:00 EST
1111    Tag: blume-20020227-cdebug
1112    Description:
1113    
1114    - made types in structure C and C_Debug to be equal
1115    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1116    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1117    
1118    ----------------------------------------------------------------------
1119    Name: Matthias Blume
1120    Date: 2002/02/26 12:00:00 EST
1121    Tag: blume-20020226-ffi
1122    Description:
1123    
1124    1. Fixed a minor bug in CM's "noweb" tool:
1125       If numbering is turned off, then truly don't number (i.e., do not
1126       supply the -L option to noweb).  The previous behavior was to supply
1127       -L'' -- which caused noweb to use the "default" line numbering scheme.
1128       Thanks to Chris Richards for pointing this out (and supplying the fix).
1129    
1130    2. Once again, I reworked some aspects of the FFI:
1131    
1132       A. The incomplete/complete type business:
1133    
1134       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1135         gone!
1136       - ML types representing an incomplete type are now *equal* to
1137         ML types representing their corresponding complete types (just like
1138         in C).  This is still safe because ml-nlffigen will not generate
1139         RTTI for incomplete types, nor will it generate functions that
1140         require access to such RTTI.   But when ML code generated from both
1141         incomplete and complete versions of the C type meet, the ML types
1142         are trivially interoperable.
1143    
1144         NOTE:  These changes restore the full generality of the translation
1145         (which was previously lost when I eliminated functorization)!
1146    
1147       B. Enum types:
1148    
1149       - Structure C now has a type constructor "enum" that is similar to
1150         how the "su" constructor works.  However, "enum" is not a phantom
1151         type because each "T enum" has values (and is isomorphic to
1152         MLRep.Signed.int).
1153       - There are generic access operations for enum objects (using
1154         MLRep.Signed.int).
1155       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1156         * The structure contains the definition of type "mlrep" (the ML-side
1157         representation type of the enum).  Normally, mlrep is the same
1158         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1159         then mlrep will be defined as a datatype -- thus facilitating
1160         pattern matching on mlrep values.
1161         ("-ec" will be suppressed if there are duplicate values in an
1162          enumeration.)
1163         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1164         will be generated for each C enum constant xxx.
1165         * Conversion functions m2i and i2m convert between mlrep and
1166         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1167         * Coversion functions c and ml convert between mlrep and "tag enum".
1168         * Access functions (get/set) fetch and store mlrep values.
1169       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1170         enumerations are merged into one single enumeration represented by
1171         structure E_'.
1172    
1173    ----------------------------------------------------------------------
1174    Name: Allen Leung
1175    Date: 2002/02/25 04:45:00 EST
1176    Tag: leunga-20020225-cps-spill
1177    
1178    This is a new implementation of the CPS spill phase.
1179    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1180    In case of problems, replace it with the old file spill.sml
1181    
1182    The current compiler runs into some serious performance problems when
1183    constructing a large record.  This can happen when we try to compile a
1184    structure with many items.  Even a very simple structure like the following
1185    makes the compiler slow down.
1186    
1187        structure Foo = struct
1188           val x_1 = 0w1 : Word32.int
1189           val x_2 = 0w2 : Word32.int
1190           val x_3 = 0w3 : Word32.int
1191           ...
1192           val x_N = 0wN : Word32.int
1193        end
1194    
1195    The following table shows the compile time, from N=1000 to N=4000,
1196    with the old compiler:
1197    
1198    N
1199    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1200           MLRISC ra                               0.06u  0.00s  0.05g
1201              (spills = 0 reloads = 0)
1202           TOTAL                                   0.63u  0.07s  0.21g
1203    
1204    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1205           MLRISC ra                               5.68u  0.59s  3.93g
1206              (spills = 0 reloads = 0)
1207           TOTAL                                   14.71u  0.99s  4.81g
1208    
1209    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1210           MLRISC ra                               5.54u  0.65s  3.91g
1211              (spills = 543 reloads = 1082)
1212           TOTAL                                   65.40u  3.13s  6.00g
1213    
1214    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1215           MLRISC ra                               0.80u  0.10s  0.55g
1216              (spills = 42 reloads = 84)
1217           TOTAL                                   129.42u  5.10s  4.13g
1218    
1219    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1220           MLRISC ra                               2.69u  0.27s  1.38g
1221              (spills = 62 reloads = 124)
1222           TOTAL                                   682.48u  19.61s  13.99g
1223    
1224    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1225           MLRISC ra                               4.96u  0.27s  2.72g
1226              (spills = 85 reloads = 170)
1227           TOTAL                                   2375.26u  57.21s  48.00g
1228    
1229    As you can see the old cps spill module suffers from some serious
1230    performance problem.  But since I cannot decipher the old code fully,
1231    instead of patching the problems up, I'm reimplementing it
1232    with a different algorithm.  The new code is more modular,
1233    smaller when compiled, and substantially faster
1234    (O(n log n) time and O(n) space).  Timing of the new spill module:
1235    
1236    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1237          MLRISC ra                               0.25u  0.02s  0.15g
1238             (spills=1 reloads=3)
1239          TOTAL                                   7.74u  0.34s  1.62g
1240    
1241    Implementation details:
1242    
1243    As far as I can tell, the purpose of the CPS spill module is to make sure the
1244    number of live variables at any program point (the bandwidth)
1245    does not exceed a certain limit, which is determined by the
1246    size of the spill area.
1247    
1248    When the bandwidth is too large, we decrease the register pressure by
1249    packing live variables into spill records.  How we achieve this is
1250    completely different than what we did in the old code.
1251    
1252    First, there is something about the MLRiscGen code generator
1253    that we should be aware of:
1254    
1255    o MLRiscGen performs code motion!
1256    
1257       In particular, it will move floating point computations and
1258       address computations involving only the heap pointer to
1259       their use sites (if there is only a single use).
1260       What this means is that if we have a CPS record construction
1261       statement
1262    
1263           RECORD(k,vl,w,e)
1264    
1265       we should never count the new record address w as live if w
1266       has only one use (which is often the case).
1267    
1268       We should do something similar to floating point, but the transformation
1269       there is much more complex, so I won't deal with that.
1270    
1271    Secondly, there are now two new cps primops at our disposal:
1272    
1273     1. rawrecord of record_kind option
1274        This pure operator allocates some uninitialized storage from the heap.
1275        There are two forms:
1276    
1277         rawrecord NONE [INT n]  allocates a tagless record of length n
1278         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1279                                     and initializes the tag.
1280    
1281     2. rawupdate of cty
1282          rawupdate cty (v,i,x)
1283          Assigns to x to the ith component of record v.
1284          The storelist is not updated.
1285    
1286    We use these new primops for both spilling and increment record construction.
1287    
1288     1. Spilling.
1289    
1290        This is implemented with a linear scan algorithm (but generalized
1291        to trees).  The algorithm will create a single spill record at the
1292        beginning of the cps function and use rawupdate to spill to it,
1293        and SELECT or SELp to reload from it.  So both spills and reloads
1294        are fine-grain operations.  In contrast, in the old algorithm
1295        "spills" have to be bundled together in records.
1296    
1297        Ideally, we should sink the spill record construction to where
1298        it is needed.  We can even split the spill record into multiple ones
1299        at the places where they are needed.  But CPS is not a good
1300        representation for global code motion, so I'll keep it simple and
1301        am not attempting this.
1302    
1303     2. Incremental record construction (aka record splitting).
1304    
1305        Long records with many component values which are simulatenously live
1306        (recall that single use record addresses are not considered to
1307         be live) are constructed with rawrecord and rawupdate.
1308        We allocate space on the heap with rawrecord first, then gradually
1309        fill it in with rawupdate.  This is the technique suggested to me
1310        by Matthias.
1311    
1312        Some restrictions on when this is applicable:
1313        1. It is not a VECTOR record.  The code generator currently does not handle
1314           this case. VECTOR record uses double indirection like arrays.
1315        2. All the record component values are defined in the same "basic block"
1316           as the record constructor.  This is to prevent speculative
1317           record construction.
1318    
1319    ----------------------------------------------------------------------
1320    Name: Allen Leung
1321    Date: 2002/02/22 01:02:00 EST
1322    Tag: leunga-20020222-mlrisc-tools
1323    
1324    Minor bug fixes in the parser and rewriter
1325    
1326    ----------------------------------------------------------------------
1327    Name: Allen Leung
1328    Date: 2002/02/21 20:20:00 EST
1329    Tag: leunga-20020221-peephole
1330    
1331    Regenerated the peephole files.  Some contained typos in the specification
1332    and some didn't compile because of pretty printing bugs in the old version
1333    of 'nowhere'.
1334    
1335    ----------------------------------------------------------------------
1336    Name: Allen Leung
1337    Date: 2002/02/19 20:20:00 EST
1338    Tag: leunga-20020219-mlrisc-tools
1339    Description:
1340    
1341       Minor bug fixes to the mlrisc-tools library:
1342    
1343       1.  Fixed up parsing colon suffixed keywords
1344       2.  Added the ability to shut the error messages up
1345       3.  Reimplemented the pretty printer and fixed up/improved
1346           the pretty printing of handle and -> types.
1347       4.  Fixed up generation of literal symbols in the nowhere tool.
1348       5.  Added some SML keywords to to sml.sty
1349    
1350    ----------------------------------------------------------------------
1351    Name: Matthias Blume
1352    Date: 2002/02/19 16:20:00 EST
1353    Tag: blume-20020219-cmffi
1354    Description:
1355    
1356    A wild mix of changes, some minor, some major:
1357    
1358    * All C FFI-related libraries are now anchored under $c:
1359        $/c.cm      --> $c/c.cm
1360        $/c-int.cm  --> $c/internals/c-int.cm
1361        $/memory.cm --> $c/memory/memory.cm
1362    
1363    * "make" tool (in CM) now treats its argument pathname slightly
1364      differently:
1365        1. If the native expansion is an absolute name, then before invoking
1366           the "make" command on it, CM will apply OS.Path.mkRelative
1367           (with relativeTo = OS.FileSys.getDir()) to it.
1368        2. The argument will be passed through to subsequent phases of CM
1369           processing without "going native".  In particular, if the argument
1370           was an anchored path, then "make" will not lose track of that anchor.
1371    
1372    * Compiler backends now "know" their respective C calling conventions
1373      instead of having to be told about it by ml-nlffigen.  This relieves
1374      ml-nlffigen from one of its burdens.
1375    
1376    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1377    
1378    * Export C_DEBUG and C_Debug from $c/c.cm.
1379    
1380    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1381      subtyping relationship between incomplete pointers and their complete
1382      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1383      with the convention of instantiating 'o with ('t, 'c) obj whenever
1384      the pointer target type is complete.  In the incomplete case, 'o
1385      will be instantiated with some "'c iobj" -- a type obtained by
1386      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1387    
1388      Operations that work on both incomplete and complete pointer types are
1389      typed as taking an 'o ptr while operations that require the target to
1390      be known are typed as taking some ('t, 'c) obj ptr.
1391    
1392      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1393      where void is an eqtype without any values.  This makes it possible
1394      to work on voidptr values using functions meant to operate on light
1395      incomplete pointers.
1396    
1397    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1398      been vastly simplified.
1399    
1400    ----------------------------------------------------------------------
1401    Name: Matthias Blume
1402    Date: 2002/02/19 10:48:00 EST
1403    Tag: blume-20020219-pqfix
1404    Description:
1405    
1406    Applied Chris Okasaki's bug fix for priority queues.
1407    
1408    ----------------------------------------------------------------------
1409    Name: Matthias Blume
1410    Date: 2002/02/15 17:05:00
1411    Tag: Release_110_39
1412    Description:
1413    
1414    Last-minute retagging is becoming a tradition... :-(
1415    
1416    This is the working release 110.39.
1417    
1418    ----------------------------------------------------------------------
1419    Name: Matthias Blume
1420    Date: 2002/02/15 16:00:00 EST
1421    Tag: Release_110_39-orig
1422    Description:
1423    
1424    Working release 110.39.  New bootfiles.
1425    
1426    (Update: There was a small bug in the installer so it wouldn't work
1427    with all shells.  So I retagged. -Matthias)
1428    
1429    ----------------------------------------------------------------------
1430    Name: Matthias Blume
1431    Date: 2002/02/15 14:17:00 EST
1432    Tag: blume-20020215-showbindings
1433    Description:
1434    
1435    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1436    the latter can be useful for exploring what bindings are available at
1437    the interactive prompt.  (The first function returns only the list
1438    of symbols that are really bound, the second prints those but also the
1439    ones that CM's autoloading mechanism knows about.)
1440    
1441    ----------------------------------------------------------------------
1442    Name: Matthias Blume
1443    Date: 2002/02/15 12:08:00 EST
1444    Tag: blume-20020215-iptrs
1445    Description:
1446    
1447    Two improvements to ml-nlffigen:
1448    
1449      1. Write files only if they do not exist or if their current contents
1450         do not coincide with what's being written.  (That is, avoid messing
1451         with the time stamps unless absolutely necessary.)
1452    
1453      2. Implement a "repository" mechanism for generated files related
1454         to "incomplete pointer types".   See the README file for details.
1455    
1456    ----------------------------------------------------------------------
1457    Name: Matthias Blume
1458    Date: 2002/02/14 11:50:00 EST
1459    Tag: blume-20020214-quote
1460    Description:
1461    
1462    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1463    because of the new and improved tag generation scheme.  (Thanks to Allen
1464    Leung for pointing it out.)
1465    
1466    ----------------------------------------------------------------------
1467    Name: Lal George
1468    Date: 2002/02/14 09:55:27 EST 2002
1469    Tag: george-20020214-isabelle-bug
1470    Description:
1471    
1472    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1473    of Isabelle on the x86.
1474    
1475    From Allen:
1476    -----------
1477     I've found the problem:
1478    
1479         in ra-core.sml, I use the counter "blocked" to keep track of the
1480         true number of elements in the freeze queue.  When the counter goes
1481         to zero, I skip examining the queue.  But I've messed up the
1482         bookkeeping in combine():
1483    
1484             else ();
1485             case !ucol of
1486               PSEUDO => (if !cntv > 0 then
1487                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1488                                        ^^^^^^^^^^^^^^^^^^^^^^^
1489                      moveu := mergeMoveList(!movev, !moveu)
1490                     )
1491                  else ();
1492    
1493         combine() is called to coalesce two nodes u and v.
1494         I think I was thinking that if the move counts of u and v are both
1495         greater than zero then after they are coalesced then one node is
1496         removed from the freeze queue.  Apparently I was thinking that
1497         both u and v are of low degree, but that's clearly not necessarily true.
1498    
1499    
1500    02/12/2002:
1501        Here's the patch.  HOL now compiles.
1502    
1503        I don't know how this impact on performance (compile
1504        time or runtime).  This bug caused the RA (especially on the x86)
1505        to go thru the potential spill phase when there are still nodes on the
1506        freeze queue.
1507    
1508    
1509    
1510    
1511    ----------------------------------------------------------------------
1512    Name: Matthias Blume
1513    Date: 2002/02/13 22:40:00 EST
1514    Tag: blume-20020213-fptr-rtti
1515    Description:
1516    
1517    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1518    updates.
1519    
1520    ----------------------------------------------------------------------
1521    Name: Matthias Blume
1522    Date: 2002/02/13 16:41:00 EST
1523    Tag: blume-20020213-cmlpq
1524    Description:
1525    
1526    Added new priority queue export symbols (which have just been added to
1527    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1528    not compile and the installer would choke.)
1529    
1530    ----------------------------------------------------------------------
1531    Name: Matthias Blume
1532    Date: 2002/02/13 16:15:00 EST
1533    Tag: blume-20020213-various
1534    Description:
1535    
1536    1. More tweaks to ml-nlffigen:
1537    
1538       - better internal datastructures (resulting in slight speedup)
1539       - "-match" option requires exact match
1540       - "localized" gensym counters (untagged structs/unions nested within
1541         other structs/unions or within typedefs get a fresh counter; their
1542         tag will be prefixed by a concatenation of their parents' tags)
1543       - bug fixes (related to calculation of transitive closure of types
1544         to be included in the output)
1545    
1546    2. Minor Basis updates:
1547    
1548       - added implementations for List.collate and Option.app
1549    
1550    ----------------------------------------------------------------------
1551    Name: Matthias Blume
1552    Date: 2002/02/11 15:55:00 EST
1553    Tag: blume-20020211-gensym
1554    Description:
1555    
1556    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1557    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1558    names (ML structure names that correspond to unnamed C structs, unions,
1559    and enums), so that separate runs of ml-nlffigen do not clash.
1560    
1561    ----------------------------------------------------------------------
1562    Name: Matthias Blume
1563    Date: 2002/02/11 12:05:00 EST
1564    Tag: blume-20020211-gensml
1565    Description:
1566    
1567    A quick fix for a problem with GenSML (in the pgraph-util library):
1568    Make generation of toplevel "local" optional.  (Strictly speaking,
1569    signature definitions within "local" are not legal SML.)
1570    
1571    Other than that: updates to INSTALL and cm/TODO.
1572    
1573    ----------------------------------------------------------------------
1574    Name: Matthias Blume
1575    Date: 2002/02/08 15:00:00 EST
1576    Tag: blume-20020208-uniquepid
1577    Description:
1578    
1579    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1580    
1581    1. The installer (config/install.sh) has gotten smarter:
1582    
1583         - Configuration options are a bit easier to specify now
1584           (in config/targets).
1585         - Bug in recognizing .tar.bz2 files fixed.
1586         - Installer automatically resolves dependencies between
1587           configuration options (e.g., if you ask for eXene, you will
1588           also get cml -- regardless whether you asked for it or not).
1589         - Installer can run in "quieter mode" by setting the environment
1590           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1591           "completely silent", though.
1592         - Build HashCons library as part of smlnj-lib.
1593    
1594    2. A new scheme for assigning persistent identifiers to compilation
1595       units (and, by extension, to types etc.) has been put into place.
1596       This fixes a long-standing bug where types and even dynamic values
1597       can get internally confused, thereby compromising type safety
1598       (abstraction) and dynamic correctness.  See
1599    
1600         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1601    
1602       for an example of how things could go wrong until now.
1603    
1604       The downside of the new scheme is that pids are not quite as
1605       persistent as they used to be: CM will generate a fresh pid
1606       for every compilation unit that it thinks it sees for the first
1607       time.  That means that if you compile starting from a clean, fresh
1608       source tree at two different times, you end up with different
1609       binaries.
1610    
1611       Cutoff recompilation, however, has not been compromised because
1612       CM keeps pid information in special caches between runs.
1613    
1614    ----------------------------------------------------------------------
1615    Name: Lal George
1616    Date: 2002/02/07 15:34:13 EST 2002
1617    Tag: <none>
1618    Description:
1619    
1620    Compilers that generate assembly code may produce  global labels
1621    whose value is resolved at link time. The various peephole optimization
1622    modules did not take this in account.
1623    
1624    TODO. The Labels.addrOf function should really return an option
1625    type so that clients are forced to deal with this issue, rather
1626    than an exception being raised.
1627    
1628    ----------------------------------------------------------------------
1629  Name: Lal George  Name: Lal George
1630  Date: 2002/02/06 13:55:02 EST  Date: 2002/02/06 13:55:02 EST
1631  Tag: george-20020206-ra-breakup  Tag: george-20020206-ra-breakup
# Line 26  Line 1639 
1639      trap instruction (in this case they are harmless since none of the      trap instruction (in this case they are harmless since none of the
1640      integers overflow.)      integers overflow.)
1641    
1642  2. Plugged out various utility modules that were embedded in the modules  2. Pulled out various utility modules that were embedded in the modules
1643     of the register allocator. I need these modules for other purposes, but     of the register allocator. I need these modules for other purposes, but
1644     they are not complete enough to put into a library just yet.     they are not complete enough to put into a library (just yet).
1645  ----------------------------------------------------------------------  ----------------------------------------------------------------------
1646  Name: Matthias Blume  Name: Matthias Blume
1647  Date: 2002/01/31 16:05:00 EST  Date: 2002/01/31 16:05:00 EST

Legend:
Removed from v.1053  
changed lines
  Added in v.1296

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