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 1044, Mon Jan 28 21:36:08 2002 UTC revision 1229, Fri May 31 21:10:34 2002 UTC
# Line 14  Line 14 
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16  Name: Matthias Blume  Name: Matthias Blume
17    Date: 2002/05/31
18    Tag: blume-20020531-btrace-mode
19    Description:
20    
21    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
22    the wayside when I switched over to using Controls everywhere.)
23    
24    ----------------------------------------------------------------------
25    Name: Lal George
26    Date: 2002/05/23 12:21:40 EDT
27    Tag: george-20020523-visual-labels
28    Description:
29    
30    Labels are now displayed in the graphical output to make
31    the fall-through and target blocks obvious.
32    
33    ----------------------------------------------------------------------
34    Name: Matthias Blume
35    Date: 2002/05/22 11:03:00 EDT
36    Tag: blume-20020522-shrink
37    Description:
38    
39    John tweaked yesterday's fix for 1131 to handle an out-of-memory
40    situation that comes up when allocating huge arrays.
41    
42    ----------------------------------------------------------------------
43    Name: Matthias Blume
44    Date: 2002/05/21 16:00:00 EDT
45    Tag: Release_110_40
46    Description:
47    
48    New working release (110.40).  New bootfiles.
49    
50    [Also: John Reppy fixed GC bug 1131.]
51    
52    ----------------------------------------------------------------------
53    Name: Matthias Blume
54    Date: 2002/05/21 12:35:00 EDT
55    Tag: blume-20020521-cmdoc
56    Description:
57    
58    CM documentation update.
59    
60    ----------------------------------------------------------------------
61    Name: Matthias Blume
62    Date: 2002/05/21 10:55:00 EDT
63    Tag: blume-20020521-misc
64    Description:
65    
66    - John tweaked runtime to be silent on heap export (except when
67      GC messages are on).
68    - I added a few more things (cross-compiling versions of CMB) to
69      config/preloads (as suggestions).
70    
71    ----------------------------------------------------------------------
72    Name: Matthias Blume
73    Date: 2002/05/20 22:25:00 EDT
74    Tag: blume-20020520-controls
75    Description:
76    
77    - Added ControlUtil structure to control-lib.cm.
78    - Use it throughout.
79    - Used Controls facility to define MLRISC controls (as opposed to
80      registering MLRISC control ref cells with Controls after the fact)
81    - Fixed messed-up controls priorities.
82    
83    * Removed again all the stuff from config/preloads that one wouldn't
84      be able to preload at the time the initial heap image is built.
85      (Many libraries, e.g., CML, do not exist yet at this time.  The
86       only libraries that can be preloaded via config/preloads are those
87       that come bundled with the bootfiles.)
88    
89    ----------------------------------------------------------------------
90    Name: Matthias Blume
91    Date: 2002/05/20 10:59:00 EDT
92    Tag: blume-20020520-preloads
93    Description:
94    
95    Added a lot of commented-out suggestions for things to be included
96    in config/preloads.
97    
98    ----------------------------------------------------------------------
99    Name: Allen Leung
100    Date: 2002/05/18 14:20:00 EDT
101    Tag: leunga-20020518-mdl
102    Description:
103    
104    o Made the mdl tool stuff compile and run again.
105    o I've disabled all the stuff that depends on RTL specifications; they
106      are all badly broken anyway.
107    
108    ----------------------------------------------------------------------
109    Name: Matthias Blume
110    Date: 2002/05/17 16:49:00 EDT
111    Tag: blume-20020517-controls
112    Description:
113    
114    0. John Reppy made several modifications to the SML/NJ library.
115       In particular, there is a shiny new controls-lib.cm.
116    
117    1. Pushed new controls interface through compiler so that everything
118       compiles again.
119    
120    2. Added FormatComb and FORMAT_COMB to the CML version of the
121       SML/NJ library (so that CML compiles again).
122    
123    3. Modified init scripts because XXX_DEFAULT environment variables
124       are no longer with us.  (Boot-time initialization is now done
125       using the same environment variables that are also used for
126       startup-time initialization of controls.)
127    
128    ----------------------------------------------------------------------
129    Name: Lal George
130    Date: 2002/05/15 09:20:10 EDT
131    Tag: george-20020515-pseudo-op-decls
132    Description:
133    
134            All pseudo-ops emitted before the first segment declaration
135    such as TEXT, DATA, and BSS directives are assumed to be global
136    declarations and are emitted first in the assembly file. This is
137    useful in a number of situations where one has pseudo-ops that are not
138    specific to any segment, and also works around the constraint that one
139    cannot have client pseudo-ops in the TEXT segment.
140    
141            Because no segment is associated with these declarations it is
142    an error to allocate any space or objects before the first segment
143    directive and an exception will be raised. However, we cannot make
144    this check for client pseudo-ops.
145    
146            These top level declarations are a field in the CFG graph_info.
147    In theory you can continue to add to this field after the CFG has been
148    built -- provided you know what you are doing;-)
149    
150    ----------------------------------------------------------------------
151    Name: Matthias Blume
152    Date: 2002/05/13 16:40:00 EDT
153    Tag: blume-20020513-pp-etc
154    Description:
155    
156    A few minor bugfixes:
157    
158      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
159        (Bogus printouts for redefined bindings still occur.  Compiler
160        bug should no longer occur now.  We need to redo the prettyprinter
161        from scratch.)
162    
163      - CM pathname printer now also adds escape sequences for ( and )
164    
165      - commend and docu fixes for ml-nlffi
166    
167    ----------------------------------------------------------------------
168    Name: Matthias Blume
169    Date: 2002/05/10 16:40:00 EDT
170    Tag: blume-20020510-erg-textio
171    Description:
172    
173    Applied the following bugfix provided by Emden Gansner:
174    
175        Output is corrupted when outputSubstr is used rather than output.
176    
177        The problem occurs when a substring
178    
179            ss = (s, dataStart, dataLen)
180    
181        where dataStart > 0, fills a stream buffer with avail bytes left.
182        avail bytes of s, starting at index dataStart, are copied into the
183        buffer, the buffer is flushed, and then the remaining dataLen-avail
184        bytes of ss are copied into the beginning of the buffer. Instead of
185        starting this copy at index dataStart+avail in s, the current code
186        starts the copy at index avail.
187    
188        Fix:
189        In text-io-fn.sml, change line 695 from
190             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
191        to
192             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
193    
194    ----------------------------------------------------------------------
195    Name: Matthias Blume
196    Date: 2002/04/12 13:55:00 EDT
197    Tag: blume-20020412-assyntax
198    Description:
199    
200    1. Grabbed newer assyntax.h from the XFree86 project.
201    2. Fiddled with how to compile X86.prim.asm without warnings.
202    3. (Very) Minor cleanup in CM.
203    
204    ----------------------------------------------------------------------
205    Name: Matthias Blume
206    Date: 2002/04/01 (no joke!) 17:07:00 EST
207    Tag: blume-20020401-x86div
208    Description:
209    
210    Added full support for div/mod/rem/quot on the x86, using the machine
211    instruction's two results (without clumsily recomputing the remainder)
212    directly where appropriate.
213    
214    Some more extensive power-of-two support was added to the x86 instruction
215    selector (avoiding expensive divs, mods, and muls where they can be
216    replaced with cheaper shifts and masks).  However, this sort of thing
217    ought to be done earlier, e.g., within the CPS optimizer so that
218    all architectures benefit from it.
219    
220    The compiler compiles to a fixed point, but changes might be somewhat
221    fragile nevertheless.  Please, report any strange things that you might
222    see wrt. div/mod/quot/rem...
223    
224    ----------------------------------------------------------------------
225    Name: Matthias Blume
226    Date: 2002/03/29 17:22:00
227    Tag: blume-20020329-div
228    Description:
229    
230    Fixed my broken div/mod logic.  Unfortunately, this means that the
231    inline code for div/mod now has one more comparison than before.
232    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
233    The problem was with quotient = 0, because that alone does not tell
234    us which way the rounding went.  One then has to look at whether
235    remainder and divisor have the same sign...  :(
236    
237    Anyway, I replaced the bootfiles with fresh ones...
238    
239    ----------------------------------------------------------------------
240    Name: Matthias Blume
241    Date: 2002/03/29 14:10:00 EST
242    Tag: blume-20020329-inlprims
243    Description:
244    
245    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
246    
247    Primops have changed. This means that the bin/boot-file formats have
248    changed as well.
249    
250    To make sure that there is no confusion, I made a new version.
251    
252    
253    CHANGES:
254    
255    * removed REMT from mltree (remainder should never overflow).
256    
257    * added primops to deal with divisions of all flavors to the frontend
258    
259    * handled these primops all the way through so they map to their respective
260      MLRISC support
261    
262    * used these primops in the implementation of Int, Int32, Word, Word32
263    
264    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
265    
266    * parameterized INLMIN, INLMAX, and INLABS by a numkind
267    
268    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
269      floating point
270    
271    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
272      and Real (but Real.abs maps to a separate floating-point-only primop)
273    
274    
275    TODO items:
276    
277    * Hacked Alpha32 instruction selection, disabling the selection of REMx
278      instructions because the machine instruction encoder cannot handle
279      them.  (Hppa, PPC, and Sparc instruction selection did not handle
280      REM in the first place, and REM is supported by the x86 machine coder.)
281    
282    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
283      selection phase.  (The two can be streamlined because the hardware
284      delivers both quotient and remainder at the same time anyway.)
285    
286    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
287      (Currently the behavior is inconsistent both across architectures and
288      wrt. the draft Basis spec.)
289    
290    * Word8 should eventually be handled natively, too.
291    
292    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
293      as if there currently is no execution path that could trigger it in
294      SML/NJ.  (The assumptions underlying functions arith and promotable do not
295      hold for things like multiplication and division.)
296    
297    ----------------------------------------------------------------------
298    Name: Matthias Blume
299    Date: 2002/03/27 16:27:00 EST
300    Tag: blume-20020327-mlrisc-divisions
301    Description:
302    
303    Added support for all four division operations (ML's div, mod, quot,
304    and rem) to MLRISC.  In the course of doing so, I also rationalized
305    the naming (no more annoying switch-around of DIV and QUOT), by
306    parameterizing the operation by div_rounding_mode (which can be either
307    DIV_TO_ZERO or DIV_TO_NEGINF).
308    
309    The generic MLTreeGen functor takes care of compiling all four
310    operations down to only round-to-zero div.
311    
312    Missing pieces:
313    
314      * Doing something smarter than relying on MLTreeGen on architectures
315        like, e.g., the x86 where hardware division delivers both quotient and
316        remainder at the same time.  With this, the implementation of the
317        round-to-neginf operations could be further streamlined.
318    
319      * Remove inlining support for div/mod/rem from the frontend and replace it
320        with primops that get carried through to the backend.  Do this for all
321        int and word types.
322    
323    ----------------------------------------------------------------------
324    Name: Matthias Blume
325    Date: 2002/03/25 17:25:00 EST
326    Tag: blume-20020325-divmod
327    Description:
328    
329    I improved (hopefully without breaking them) the implementation of Int.div,
330    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
331    advantage of the following observations:
332    
333      Let  q = x quot y      r = x rem y
334           d = x div  y      m = x mod y
335    
336    where "quot" is the round-to-zero version of integer division that
337    hardware usually provides.  Then we have:
338    
339         r = x - q * y        where neither the * nor the - will overflow
340         d = if q >= 0 orelse x = q * y then q else q - 1
341                              where neither the * nor the - will overflow
342         m = if q >= 0 orelse r = 0 then r else r + y
343                              where the + will not overflow
344    
345    This results in substantial simplification of the generated code.
346    The following table shows the number of CFG nodes and edges generated
347    for
348            fun f (x, y) = x OPER y
349            (* with OPER \in div, mod, quot, rem *)
350    
351    
352        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
353        --------------------------------------------------------
354         div |         24 |         39 |         12 |         16
355         mod |         41 |         71 |         12 |         16
356        quot |          8 |         10 |          8 |         10
357         rem |         10 |         14 |          8 |         10
358    
359    
360    ----------------------------------------------------------------------
361    Name: Matthias Blume
362    Date: 2002/03/25 22:06:00 EST
363    Tag: blume-20020325-cprotobug
364    Description:
365    
366    Fixed a bug in cproto (c prototype decoder).
367    
368    ----------------------------------------------------------------------
369    Name: Matthias Blume
370    Date: 2002/03/25 16:00:00 EST
371    Tag: blume-20020325-raw-primops
372    Description:
373    
374    I did some cleanup to Allen's new primop code and
375    replaced yesterday's bootfiles with new ones.
376    (But they are stored in the same place.)
377    
378    ----------------------------------------------------------------------
379    Name: Matthias Blume
380    Date: 2002/03/24 22:40:00 EST
381    Tag: blume-20020324-bootfiles
382    Description:
383    
384    Made the bootfiles that Allen asked for.
385    
386    ----------------------------------------------------------------------
387    Name: Allen Leung
388    Date: 2002/03/23 15:50:00 EST
389    Tag: leunga-20020323-flint-cps-rcc-primops
390    Description:
391    
392      1. Changes to FLINT primops:
393    
394        (* make a call to a C-function;
395         * The primop carries C function prototype information and specifies
396         * which of its (ML-) arguments are floating point. C prototype
397         * information is for use by the backend, ML information is for
398         * use by the CPS converter. *)
399      | RAW_CCALL of { c_proto: CTypes.c_proto,
400                       ml_args: ccall_type list,
401                       ml_res_opt: ccall_type option,
402                       reentrant : bool
403                     } option
404       (* Allocate uninitialized storage on the heap.
405        * The record is meant to hold short-lived C objects, i.e., they
406        * are not ML pointers.  With the tag, the representation is
407        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
408        *)
409      | RAW_RECORD of {tag:bool,sz:int}
410      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
411    
412      2.  These CPS primops are now overloaded:
413    
414           rawload of {kind:numkind}
415           rawstore of {kind:numkind}
416    
417          The one argument form is:
418    
419             rawload {kind} address
420    
421          The two argument form is:
422    
423             rawload {kind} [ml object, byte-offset]
424    
425      3. RAW_CCALL/RCC now takes two extra arguments:
426    
427         a. The first is whether the C call is reentrant, i.e., whether
428            ML state should be saved and restored.
429         b. The second argument is a string argument specifying the name of
430            library and the C function.
431    
432         These things are currently not handled in the code generator, yet.
433    
434      4. In CProto,
435    
436         An encoding type of "bool" means "ml object" and is mapped into
437         C prototype of PTR.  Note that "bool" is different than "string",
438         even though "string" is also mapped into PTR, because "bool"
439         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
440    
441      5. Pickler/unpicker
442    
443         Changed to handle RAW_RECORD and newest RAW_CCALL
444    
445      6. MLRiscGen,
446    
447         1. Changed to handle the new rawload/rawstore/rawrecord operators.
448         2. Code for handling C Calls has been moved to a new module CPSCCalls,
449            in the file CodeGen/cpscompile/cps-c-calls.sml
450    
451      7. Added the conditional move operator
452    
453             condmove of branch
454    
455         to cps.  Generation of this is still buggy so it is currently
456         disabled.
457    
458    ----------------------------------------------------------------------
459    Name: Lal George
460    Date: 2002/03/22 14:18:25 EST
461    Tag: george-20020322-cps-branch-prob
462    Description:
463    
464    Implemented the Ball-Larus branch prediction-heuristics, and
465    incorporated graphical viewers for control flow graphs.
466    
467    Ball-Larus Heuristics:
468    ---------------------
469    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
470    
471    By design it uses the Dempster-Shafer theory for combining
472    probabilities.  For example, in the function:
473    
474        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
475    
476    the ball-larus heuristics predicts that the n=0 is unlikely
477    (OH-heuristic), and the 'then' branch is unlikely because of the
478    RH-heuristic -- giving the 'then' branch an even lower combined
479    probability using the Dempster-Shafer theory.
480    
481    Finally, John Reppy's loop analysis in MLRISC, further lowers the
482    probability of the 'then' branch because of the loop in the else
483    branch.
484    
485    
486    Graphical Viewing:
487    ------------------
488    I merely plugged in Allen's graphical viewers into the compiler. The
489    additional code is not much. At the top level, saying:
490    
491            Control.MLRISC.getFlag "cfg-graphical-view" := true;
492    
493    will display the graphical view of the control flow graph just before
494    back-patching.  daVinci must be in your path for this to work. If
495    daVinci is not available, then the default viewer can be changed
496    using:
497    
498            Control.MLRISC.getString "viewer"
499    
500    which can be set to "dot" or "vcg" for the corresponding viewers. Of
501    course, these viewers must be in your path.
502    
503    The above will display the compilation unit at the level of clusters,
504    many of which are small, boring, and un-interesting. Also setting:
505    
506            Control.MLRISC.getInt "cfg-graphical-view_size"
507    
508    will display clusters that are larger than the value set by the above.
509    
510    
511    ----------------------------------------------------------------------
512    Name: Matthias Blume
513    Date: 2002/03/21 22:20:00 EST
514    Tag: blume-20020321-kmp-bugfix
515    Description:
516    
517    Changed the interface to the KMP routine in PreString and fixed
518    a minor bug in one place where it was used.
519    
520    ----------------------------------------------------------------------
521    Name: Allen Leung
522    Date: 2002/03/21 20:30:00 EST
523    Tag: leunga-20020321-cfg
524    Description:
525    
526      Fixed a potential problem in cfg edge splitting.
527    
528    ----------------------------------------------------------------------
529    Name: Allen Leung
530    Date: 2002/03/21 17:15:00 EST
531    Tag: leunga-20020321-x86-fp-cfg
532    Description:
533    
534      1. Recoded the buggy parts of x86-fp.
535    
536         a. All the block reordering code has been removed.
537            We now depend on the block placement phases to do this work.
538    
539         b. Critical edge splitting code has been simplified and moved into the
540            CFG modules, as where they belong.
541    
542         Both of these were quite buggy and complex.  The code is now much, much
543         simpler.
544    
545      2. X86 backend.
546    
547         a. Added instructions for 64-bit support.  Instruction selection for
548            64-bit has not been committed, however, since that
549            requires changes to MLTREE which haven't been approved by
550            Lal and John.
551    
552         b. Added support for FUCOMI and FUCOMIP when generating code for
553            PentiumPro and above.  We only generate these instructions in
554            the fast-fp mode.
555    
556         c. Added cases for JP and JNP in X86FreqProps.
557    
558      3. CFG
559    
560         CFG now has a bunch of methods for edge splitting and merging.
561    
562      4. Machine description.
563    
564         John's simplification of MLTREE_BASIS.fcond broke a few machine
565         description things:
566    
567         rtl-build.{sig,sml} and hppa.mdl fixed.
568    
569         NOTE: the machine description stuff in the repository is still broken.
570               Again, I can't put my fixes in because that involves
571               changes to MLTREE.
572    
573    ----------------------------------------------------------------------
574    Name: Matthias Blume
575    Date: 2002/03/20 15:55:00 EST
576    Tag: blume-20020320-kmp
577    Description:
578    
579    Implemented Knuth-Morris-Pratt string matching in PreString and used
580    it for String.isSubstring, Substring.isSubstring, and
581    Substring.position.
582    
583    (Might need some stress-testing.  Simple examples worked fine.)
584    
585    ----------------------------------------------------------------------
586    Name: Matthias Blume
587    Date: 2002/03/19 16:37:00 EST
588    Tag: blume-20020319-witnesses
589    Description:
590    
591    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
592    
593    This implements a generic mechanism for changing constness qualifiers
594    anywhere within big C types without resorting to outright "casts".
595    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
596    modify the constness at the outermost level.)
597    The implementation of "convert" is based on the idea of "witness"
598    values -- values that are not used by the operation but whose types
599    "testify" to their applicability.  On the implementation side, "convert"
600    is simply a projection (returning its second curried argument).  With
601    cross-module inlining, it should not result in any machine code being
602    generated.
603    
604    ----------------------------------------------------------------------
605    Name: Matthias Blume
606    Date: 2002/03/15 16:40:00 EST
607    Tag: blume-20020315-basis
608    Description:
609    
610    Provided (preliminary?) implementations for
611    
612      {String,Substring}.{concatWith,isSuffix,isSubstring}
613    
614    and
615    
616      Substring.full
617    
618    Those are in the Basis spec but they were missing in SML/NJ.
619    
620    ----------------------------------------------------------------------
621    Name: Matthias Blume
622    Date: 2002/03/14 21:30:00 EST
623    Tag: blume-20020314-controls
624    Description:
625    
626    Controls:
627    ---------
628    
629    1. Factored out the recently-added Controls : CONTROLS stuff and put
630       it into its own library $/controls-lib.cm.  The source tree for
631       this is under src/smlnj-lib/Controls.
632    
633    2. Changed the names of types and functions in this interface, so they
634       make a bit more "sense":
635    
636          module -> registry
637          'a registry -> 'a group
638    
639    3. The interface now deals in ref cells only.  The getter/setter interface
640       is (mostly) gone.
641    
642    4. Added a function that lets one register an already-existing ref cell.
643    
644    5. Made the corresponding modifications to the rest of the code so that
645       everything compiles again.
646    
647    6. Changed the implementation of Controls.MLRISC back to something closer
648       to the original.  In particular, this module (and therefore MLRISC)
649       does not depend on Controls.  There now is some link-time code in
650       int-sys.sml that registers the MLRISC controls with the Controls
651       module.
652    
653    CM:
654    ---
655    
656      * One can now specify the lambda-split aggressiveness in init.cmi.
657    
658    ----------------------------------------------------------------------
659    Name: Allen Leung
660    Date: 2002/03/13 17:30:00 EST
661    Tag: leunga-20020313-x86-fp-unary
662    Description:
663    
664    Bug fix for:
665    
666    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
667    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
668    > - fun f(x,(y,z)) = Real.~ y;
669    > [autoloading]
670    > [autoloading done]
671    >       fchsl   (%eax), 184(%esp)
672    > Error: MLRisc bug: X86MCEmitter.emitInstr
673    >
674    > uncaught exception Error
675    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
676    
677    The problem was that the code generator did not generate any fp registers
678    in this case, and the ra didn't know that it needed to run the X86FP phase to
679    translate the pseudo fp instruction.   This only happened with unary fp
680    operators in certain situations.
681    
682    ----------------------------------------------------------------------
683    Name: Matthias Blume
684    Date: 2002/03/13 14:00:00 EST
685    Tag: blume-20020313-overload-etc
686    Description:
687    
688    1. Added _overload as a synonym for overload for backward compatibility.
689       (Control.overloadKW must be true for either version to be accepted.)
690    
691    2. Fixed bug in install script that caused more things to be installed
692       than what was requested in config/targets.
693    
694    3. Made CM aware of the (_)overload construct so that autoloading
695       works.
696    
697    ----------------------------------------------------------------------
698    Name: Matthias Blume
699    Date: 2002/03/12 22:03:00 EST
700    Tag: blume-20020312-url
701    Description:
702    
703    Forgot to update BOOT and srcarchiveurl.
704    
705    ----------------------------------------------------------------------
706    Name: Matthias Blume
707    Date: 2002/03/12 17:30:00 EST
708    Tag: blume-20020312-version110392
709    Description:
710    
711    Yet another version number bump (because of small changes to the
712    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
713    
714    Changes:
715    
716      The new pid generation scheme described a few weeks ago was overly
717      complicated.  I implemented a new mechanism that is simpler and
718      provides a bit more "stability":  Once CM has seen a compilation
719      unit, it keeps its identity constant (as long as you do not delete
720      those crucial CM/GUID/* files).  This means that when you change
721      an interface, compile, then go back to the old interface, and
722      compile again, you arrive at the original pid.
723    
724      There now also is a mechanism that instructs CM to use the plain
725      environment hash as a module's pid (effectively making its GUID
726      the empty string).  For this, "noguid" must be specified as an
727      option to the .sml file in question within its .cm file.
728      This is most useful for code that is being generated by tools such
729      as ml-nlffigen (because during development programmers tend to
730      erase the tool's entire output directory tree including CM's cached
731      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
732      revert to the old, broken behavior of SML/NJ, but in specific cases
733      where there is no danger of interface confusion, its use is ok
734      (I think).
735    
736      ml-nlffigen by default generates "noguid" annotations.  They can be
737      turned off by specifying -guid in its command line.
738    
739    ----------------------------------------------------------------------
740    Name: Lal George
741    Date: 2002/03/12 12 14:42:36 EST
742    Tag: george-20020312-frequency-computation
743    Description:
744    
745    Integrated jump chaining and static block frequency into the
746    compiler. More details and numbers later.
747    
748    ----------------------------------------------------------------------
749    Name: Lal George
750    Date: 2002/03/11 11 22:38:53 EST
751    Tag: george-20020311-jump-chain-elim
752    Description:
753    
754    Tested the jump chain elimination on all architectures (except the
755    hppa).  This is on by default right now and is profitable for the
756    alpha and x86, however, it may not be profitable for the sparc and ppc
757    when compiling the compiler.
758    
759    The gc test will typically jump to a label at the end of the cluster,
760    where there is another jump to an external cluster containing the actual
761    code to invoke gc. This is to allow factoring of common gc invocation
762    sequences. That is to say, we generate:
763    
764            f:
765               testgc
766               ja   L1      % jump if above to L1
767    
768            L1:
769               jmp L2
770    
771    
772    After jump chain elimination the 'ja L1' instructions is converted to
773    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
774    up being implemented in their long form (if L2 is far away) using:
775    
776            jbe     L3      % jump if below or equal to L3
777            jmp     L2
778         L3:
779            ...
780    
781    
782    For large compilation units L2  may be far away.
783    
784    
785    ----------------------------------------------------------------------
786    Name: Matthias Blume
787    Date: 2002/03/11 13:30:00 EST
788    Tag: blume-20020311-mltreeeval
789    Description:
790    
791    A functor parameter was missing.
792    
793    ----------------------------------------------------------------------
794    Name: Allen Leung
795    Date: 2002/03/11 10:30:00 EST
796    Tag: leunga-20020311-runtime-string0
797    Description:
798    
799       The representation of the empty string now points to a
800    legal null terminated C string instead of unit.  It is now possible
801    to convert an ML string into C string with InlineT.CharVector.getData.
802    This compiles into one single machine instruction.
803    
804    ----------------------------------------------------------------------
805    Name: Allen Leung
806    Date: 2002/03/10 23:55:00 EST
807    Tag: leunga-20020310-x86-call
808    Description:
809    
810       Added machine generation for CALL instruction (relative displacement mode)
811    
812    ----------------------------------------------------------------------
813    Name: Matthias Blume
814    Date: 2002/03/08 16:05:00
815    Tag: blume-20020308-entrypoints
816    Description:
817    
818    Version number bumped to 110.39.1.  NEW BOOTFILES!
819    
820    Entrypoints: non-zero offset into a code object where execution should begin.
821    
822    - Added the notion of an entrypoint to CodeObj.
823    - Added reading/writing of entrypoint info to Binfile.
824    - Made runtime system bootloader aware of entrypoints.
825    - Use the address of the label of the first function given to mlriscGen
826      as the entrypoint.  This address is currently always 0, but it will
827      not be 0 once we turn on block placement.
828    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
829      with entry points) from mlriscGen.
830    
831    ----------------------------------------------------------------------
832    Name: Allen Leung
833    Date: 2002/03/07 20:45:00 EST
834    Tag: leunga-20020307-x86-cmov
835    Description:
836    
837       Bug fixes for CMOVcc on x86.
838    
839       1. Added machine code generation for CMOVcc
840       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
841       3. CMOVcc cannot have an immediate operand as argument.
842    
843    ----------------------------------------------------------------------
844    Name: Matthias Blume
845    Date: 2002/03/07 16:15:00 EST
846    Tag: blume-20020307-controls
847    Description:
848    
849    This is a very large but mostly boring patch which makes (almost)
850    every tuneable compiler knob (i.e., pretty much everything under
851    Control.* plus a few other things) configurable via both the command
852    line and environment variables in the style CM did its configuration
853    until now.
854    
855    Try starting sml with '-h' (or, if you are brave, '-H')
856    
857    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
858    implements the underlying generic mechanism.
859    
860    The interface to some of the existing such facilities has changed somewhat.
861    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
862    (The getFoo interface is still there for backward-compatibility, but its
863    use is deprecated.)
864    
865    The ml-build script passes -Cxxx=yyy command-line arguments through so
866    that one can now twiddle the compiler settings when using this "batch"
867    compiler.
868    
869    TODO items:
870    
871    We should go through and throw out all controls that are no longer
872    connected to anything.  Moreover, we should go through and provide
873    meaningful (and correct!) documentation strings for those controls
874    that still are connected.
875    
876    Currently, multiple calls to Controls.new are accepted (only the first
877    has any effect).  Eventually we should make sure that every control
878    is being made (via Controls.new) exactly once.  Future access can then
879    be done using Controls.acc.
880    
881    Finally, it would probably be a good idea to use the getter-setter
882    interface to controls rather than ref cells.  For the time being, both
883    styles are provided by the Controls module, but getter-setter pairs are
884    better if thread-safety is of any concern because they can be wrapped.
885    
886    *****************************************
887    
888    One bug fix: The function blockPlacement in three of the MLRISC
889    backpatch files used to be hard-wired to one of two possibilities at
890    link time (according to the value of the placementFlag).  But (I
891    think) it should rather sense the flag every time.
892    
893    *****************************************
894    
895    Other assorted changes (by other people who did not supply a HISTORY entry):
896    
897    1. the cross-module inliner now works much better (Monnier)
898    2. representation of weights, frequencies, and probabilities in MLRISC
899       changed in preparation of using those for weighted block placement
900       (Reppy, George)
901    
902    ----------------------------------------------------------------------
903    Name: Lal George
904    Date: 2002/03/07 14:44:24 EST 2002
905    Tag: george-20020307-weighted-block-placement
906    
907    Tested the weighted block placement optimization on all architectures
908    (except the hppa) using AMPL to generate the block and edge frequencies.
909    Changes were required in the machine properties to correctly
910    categorize trap instructions. There is an MLRISC flag
911    "weighted-block-placement" that can be used to enable weighted block
912    placement, but this will be ineffective without block/edge
913    frequencies (coming soon).
914    
915    
916    ----------------------------------------------------------------------
917    Name: Lal George
918    Date: 2002/03/05 17:24:48 EST
919    Tag: george-20020305-linkage-cluster
920    
921    In order to support the block placement optimization, a new cluster
922    is generated as the very first cluster (called the linkage cluster).
923    It contains a single jump to the 'real' entry point for the compilation
924    unit. Block placement has no effect on the linkage cluster itself, but
925    all the other clusters  have full freedom in the manner in which they
926    reorder blocks or functions.
927    
928    On the x86 the typical linkage code that is generated is:
929       ----------------------
930            .align 2
931       L0:
932            addl    $L1-L0, 72(%esp)
933            jmp     L1
934    
935    
936            .align  2
937       L1:
938       ----------------------
939    
940    72(%esp) is the memory location for the stdlink register. This
941    must contain the address of the CPS function being called. In the
942    above example, it contains the address of  L0; before
943    calling L1 (the real entry point for the compilation unit), it
944    must contain the address for L1, and hence
945    
946            addl $L1-L0, 72(%esp)
947    
948    I have tested this on all architectures except the hppa.The increase
949    in code size is of course negligible
950    
951    ----------------------------------------------------------------------
952    Name: Allen Leung
953    Date: 2002/03/03 13:20:00 EST
954    Tag: leunga-20020303-mlrisc-tools
955    
956      Added #[ ... ] expressions to mlrisc tools
957    
958    ----------------------------------------------------------------------
959    Name: Matthias Blume
960    Date: 2002/02/27 12:29:00 EST
961    Tag: blume-20020227-cdebug
962    Description:
963    
964    - made types in structure C and C_Debug to be equal
965    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
966    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
967    
968    ----------------------------------------------------------------------
969    Name: Matthias Blume
970    Date: 2002/02/26 12:00:00 EST
971    Tag: blume-20020226-ffi
972    Description:
973    
974    1. Fixed a minor bug in CM's "noweb" tool:
975       If numbering is turned off, then truly don't number (i.e., do not
976       supply the -L option to noweb).  The previous behavior was to supply
977       -L'' -- which caused noweb to use the "default" line numbering scheme.
978       Thanks to Chris Richards for pointing this out (and supplying the fix).
979    
980    2. Once again, I reworked some aspects of the FFI:
981    
982       A. The incomplete/complete type business:
983    
984       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
985         gone!
986       - ML types representing an incomplete type are now *equal* to
987         ML types representing their corresponding complete types (just like
988         in C).  This is still safe because ml-nlffigen will not generate
989         RTTI for incomplete types, nor will it generate functions that
990         require access to such RTTI.   But when ML code generated from both
991         incomplete and complete versions of the C type meet, the ML types
992         are trivially interoperable.
993    
994         NOTE:  These changes restore the full generality of the translation
995         (which was previously lost when I eliminated functorization)!
996    
997       B. Enum types:
998    
999       - Structure C now has a type constructor "enum" that is similar to
1000         how the "su" constructor works.  However, "enum" is not a phantom
1001         type because each "T enum" has values (and is isomorphic to
1002         MLRep.Signed.int).
1003       - There are generic access operations for enum objects (using
1004         MLRep.Signed.int).
1005       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1006         * The structure contains the definition of type "mlrep" (the ML-side
1007         representation type of the enum).  Normally, mlrep is the same
1008         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1009         then mlrep will be defined as a datatype -- thus facilitating
1010         pattern matching on mlrep values.
1011         ("-ec" will be suppressed if there are duplicate values in an
1012          enumeration.)
1013         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1014         will be generated for each C enum constant xxx.
1015         * Conversion functions m2i and i2m convert between mlrep and
1016         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1017         * Coversion functions c and ml convert between mlrep and "tag enum".
1018         * Access functions (get/set) fetch and store mlrep values.
1019       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1020         enumerations are merged into one single enumeration represented by
1021         structure E_'.
1022    
1023    ----------------------------------------------------------------------
1024    Name: Allen Leung
1025    Date: 2002/02/25 04:45:00 EST
1026    Tag: leunga-20020225-cps-spill
1027    
1028    This is a new implementation of the CPS spill phase.
1029    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1030    In case of problems, replace it with the old file spill.sml
1031    
1032    The current compiler runs into some serious performance problems when
1033    constructing a large record.  This can happen when we try to compile a
1034    structure with many items.  Even a very simple structure like the following
1035    makes the compiler slow down.
1036    
1037        structure Foo = struct
1038           val x_1 = 0w1 : Word32.int
1039           val x_2 = 0w2 : Word32.int
1040           val x_3 = 0w3 : Word32.int
1041           ...
1042           val x_N = 0wN : Word32.int
1043        end
1044    
1045    The following table shows the compile time, from N=1000 to N=4000,
1046    with the old compiler:
1047    
1048    N
1049    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1050           MLRISC ra                               0.06u  0.00s  0.05g
1051              (spills = 0 reloads = 0)
1052           TOTAL                                   0.63u  0.07s  0.21g
1053    
1054    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1055           MLRISC ra                               5.68u  0.59s  3.93g
1056              (spills = 0 reloads = 0)
1057           TOTAL                                   14.71u  0.99s  4.81g
1058    
1059    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1060           MLRISC ra                               5.54u  0.65s  3.91g
1061              (spills = 543 reloads = 1082)
1062           TOTAL                                   65.40u  3.13s  6.00g
1063    
1064    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1065           MLRISC ra                               0.80u  0.10s  0.55g
1066              (spills = 42 reloads = 84)
1067           TOTAL                                   129.42u  5.10s  4.13g
1068    
1069    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1070           MLRISC ra                               2.69u  0.27s  1.38g
1071              (spills = 62 reloads = 124)
1072           TOTAL                                   682.48u  19.61s  13.99g
1073    
1074    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1075           MLRISC ra                               4.96u  0.27s  2.72g
1076              (spills = 85 reloads = 170)
1077           TOTAL                                   2375.26u  57.21s  48.00g
1078    
1079    As you can see the old cps spill module suffers from some serious
1080    performance problem.  But since I cannot decipher the old code fully,
1081    instead of patching the problems up, I'm reimplementing it
1082    with a different algorithm.  The new code is more modular,
1083    smaller when compiled, and substantially faster
1084    (O(n log n) time and O(n) space).  Timing of the new spill module:
1085    
1086    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1087          MLRISC ra                               0.25u  0.02s  0.15g
1088             (spills=1 reloads=3)
1089          TOTAL                                   7.74u  0.34s  1.62g
1090    
1091    Implementation details:
1092    
1093    As far as I can tell, the purpose of the CPS spill module is to make sure the
1094    number of live variables at any program point (the bandwidth)
1095    does not exceed a certain limit, which is determined by the
1096    size of the spill area.
1097    
1098    When the bandwidth is too large, we decrease the register pressure by
1099    packing live variables into spill records.  How we achieve this is
1100    completely different than what we did in the old code.
1101    
1102    First, there is something about the MLRiscGen code generator
1103    that we should be aware of:
1104    
1105    o MLRiscGen performs code motion!
1106    
1107       In particular, it will move floating point computations and
1108       address computations involving only the heap pointer to
1109       their use sites (if there is only a single use).
1110       What this means is that if we have a CPS record construction
1111       statement
1112    
1113           RECORD(k,vl,w,e)
1114    
1115       we should never count the new record address w as live if w
1116       has only one use (which is often the case).
1117    
1118       We should do something similar to floating point, but the transformation
1119       there is much more complex, so I won't deal with that.
1120    
1121    Secondly, there are now two new cps primops at our disposal:
1122    
1123     1. rawrecord of record_kind option
1124        This pure operator allocates some uninitialized storage from the heap.
1125        There are two forms:
1126    
1127         rawrecord NONE [INT n]  allocates a tagless record of length n
1128         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1129                                     and initializes the tag.
1130    
1131     2. rawupdate of cty
1132          rawupdate cty (v,i,x)
1133          Assigns to x to the ith component of record v.
1134          The storelist is not updated.
1135    
1136    We use these new primops for both spilling and increment record construction.
1137    
1138     1. Spilling.
1139    
1140        This is implemented with a linear scan algorithm (but generalized
1141        to trees).  The algorithm will create a single spill record at the
1142        beginning of the cps function and use rawupdate to spill to it,
1143        and SELECT or SELp to reload from it.  So both spills and reloads
1144        are fine-grain operations.  In contrast, in the old algorithm
1145        "spills" have to be bundled together in records.
1146    
1147        Ideally, we should sink the spill record construction to where
1148        it is needed.  We can even split the spill record into multiple ones
1149        at the places where they are needed.  But CPS is not a good
1150        representation for global code motion, so I'll keep it simple and
1151        am not attempting this.
1152    
1153     2. Incremental record construction (aka record splitting).
1154    
1155        Long records with many component values which are simulatenously live
1156        (recall that single use record addresses are not considered to
1157         be live) are constructed with rawrecord and rawupdate.
1158        We allocate space on the heap with rawrecord first, then gradually
1159        fill it in with rawupdate.  This is the technique suggested to me
1160        by Matthias.
1161    
1162        Some restrictions on when this is applicable:
1163        1. It is not a VECTOR record.  The code generator currently does not handle
1164           this case. VECTOR record uses double indirection like arrays.
1165        2. All the record component values are defined in the same "basic block"
1166           as the record constructor.  This is to prevent speculative
1167           record construction.
1168    
1169    ----------------------------------------------------------------------
1170    Name: Allen Leung
1171    Date: 2002/02/22 01:02:00 EST
1172    Tag: leunga-20020222-mlrisc-tools
1173    
1174    Minor bug fixes in the parser and rewriter
1175    
1176    ----------------------------------------------------------------------
1177    Name: Allen Leung
1178    Date: 2002/02/21 20:20:00 EST
1179    Tag: leunga-20020221-peephole
1180    
1181    Regenerated the peephole files.  Some contained typos in the specification
1182    and some didn't compile because of pretty printing bugs in the old version
1183    of 'nowhere'.
1184    
1185    ----------------------------------------------------------------------
1186    Name: Allen Leung
1187    Date: 2002/02/19 20:20:00 EST
1188    Tag: leunga-20020219-mlrisc-tools
1189    Description:
1190    
1191       Minor bug fixes to the mlrisc-tools library:
1192    
1193       1.  Fixed up parsing colon suffixed keywords
1194       2.  Added the ability to shut the error messages up
1195       3.  Reimplemented the pretty printer and fixed up/improved
1196           the pretty printing of handle and -> types.
1197       4.  Fixed up generation of literal symbols in the nowhere tool.
1198       5.  Added some SML keywords to to sml.sty
1199    
1200    ----------------------------------------------------------------------
1201    Name: Matthias Blume
1202    Date: 2002/02/19 16:20:00 EST
1203    Tag: blume-20020219-cmffi
1204    Description:
1205    
1206    A wild mix of changes, some minor, some major:
1207    
1208    * All C FFI-related libraries are now anchored under $c:
1209        $/c.cm      --> $c/c.cm
1210        $/c-int.cm  --> $c/internals/c-int.cm
1211        $/memory.cm --> $c/memory/memory.cm
1212    
1213    * "make" tool (in CM) now treats its argument pathname slightly
1214      differently:
1215        1. If the native expansion is an absolute name, then before invoking
1216           the "make" command on it, CM will apply OS.Path.mkRelative
1217           (with relativeTo = OS.FileSys.getDir()) to it.
1218        2. The argument will be passed through to subsequent phases of CM
1219           processing without "going native".  In particular, if the argument
1220           was an anchored path, then "make" will not lose track of that anchor.
1221    
1222    * Compiler backends now "know" their respective C calling conventions
1223      instead of having to be told about it by ml-nlffigen.  This relieves
1224      ml-nlffigen from one of its burdens.
1225    
1226    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1227    
1228    * Export C_DEBUG and C_Debug from $c/c.cm.
1229    
1230    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1231      subtyping relationship between incomplete pointers and their complete
1232      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1233      with the convention of instantiating 'o with ('t, 'c) obj whenever
1234      the pointer target type is complete.  In the incomplete case, 'o
1235      will be instantiated with some "'c iobj" -- a type obtained by
1236      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1237    
1238      Operations that work on both incomplete and complete pointer types are
1239      typed as taking an 'o ptr while operations that require the target to
1240      be known are typed as taking some ('t, 'c) obj ptr.
1241    
1242      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1243      where void is an eqtype without any values.  This makes it possible
1244      to work on voidptr values using functions meant to operate on light
1245      incomplete pointers.
1246    
1247    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1248      been vastly simplified.
1249    
1250    ----------------------------------------------------------------------
1251    Name: Matthias Blume
1252    Date: 2002/02/19 10:48:00 EST
1253    Tag: blume-20020219-pqfix
1254    Description:
1255    
1256    Applied Chris Okasaki's bug fix for priority queues.
1257    
1258    ----------------------------------------------------------------------
1259    Name: Matthias Blume
1260    Date: 2002/02/15 17:05:00
1261    Tag: Release_110_39
1262    Description:
1263    
1264    Last-minute retagging is becoming a tradition... :-(
1265    
1266    This is the working release 110.39.
1267    
1268    ----------------------------------------------------------------------
1269    Name: Matthias Blume
1270    Date: 2002/02/15 16:00:00 EST
1271    Tag: Release_110_39-orig
1272    Description:
1273    
1274    Working release 110.39.  New bootfiles.
1275    
1276    (Update: There was a small bug in the installer so it wouldn't work
1277    with all shells.  So I retagged. -Matthias)
1278    
1279    ----------------------------------------------------------------------
1280    Name: Matthias Blume
1281    Date: 2002/02/15 14:17:00 EST
1282    Tag: blume-20020215-showbindings
1283    Description:
1284    
1285    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1286    the latter can be useful for exploring what bindings are available at
1287    the interactive prompt.  (The first function returns only the list
1288    of symbols that are really bound, the second prints those but also the
1289    ones that CM's autoloading mechanism knows about.)
1290    
1291    ----------------------------------------------------------------------
1292    Name: Matthias Blume
1293    Date: 2002/02/15 12:08:00 EST
1294    Tag: blume-20020215-iptrs
1295    Description:
1296    
1297    Two improvements to ml-nlffigen:
1298    
1299      1. Write files only if they do not exist or if their current contents
1300         do not coincide with what's being written.  (That is, avoid messing
1301         with the time stamps unless absolutely necessary.)
1302    
1303      2. Implement a "repository" mechanism for generated files related
1304         to "incomplete pointer types".   See the README file for details.
1305    
1306    ----------------------------------------------------------------------
1307    Name: Matthias Blume
1308    Date: 2002/02/14 11:50:00 EST
1309    Tag: blume-20020214-quote
1310    Description:
1311    
1312    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1313    because of the new and improved tag generation scheme.  (Thanks to Allen
1314    Leung for pointing it out.)
1315    
1316    ----------------------------------------------------------------------
1317    Name: Lal George
1318    Date: 2002/02/14 09:55:27 EST 2002
1319    Tag: george-20020214-isabelle-bug
1320    Description:
1321    
1322    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1323    of Isabelle on the x86.
1324    
1325    From Allen:
1326    -----------
1327     I've found the problem:
1328    
1329         in ra-core.sml, I use the counter "blocked" to keep track of the
1330         true number of elements in the freeze queue.  When the counter goes
1331         to zero, I skip examining the queue.  But I've messed up the
1332         bookkeeping in combine():
1333    
1334             else ();
1335             case !ucol of
1336               PSEUDO => (if !cntv > 0 then
1337                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1338                                        ^^^^^^^^^^^^^^^^^^^^^^^
1339                      moveu := mergeMoveList(!movev, !moveu)
1340                     )
1341                  else ();
1342    
1343         combine() is called to coalesce two nodes u and v.
1344         I think I was thinking that if the move counts of u and v are both
1345         greater than zero then after they are coalesced then one node is
1346         removed from the freeze queue.  Apparently I was thinking that
1347         both u and v are of low degree, but that's clearly not necessarily true.
1348    
1349    
1350    02/12/2002:
1351        Here's the patch.  HOL now compiles.
1352    
1353        I don't know how this impact on performance (compile
1354        time or runtime).  This bug caused the RA (especially on the x86)
1355        to go thru the potential spill phase when there are still nodes on the
1356        freeze queue.
1357    
1358    
1359    
1360    
1361    ----------------------------------------------------------------------
1362    Name: Matthias Blume
1363    Date: 2002/02/13 22:40:00 EST
1364    Tag: blume-20020213-fptr-rtti
1365    Description:
1366    
1367    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1368    updates.
1369    
1370    ----------------------------------------------------------------------
1371    Name: Matthias Blume
1372    Date: 2002/02/13 16:41:00 EST
1373    Tag: blume-20020213-cmlpq
1374    Description:
1375    
1376    Added new priority queue export symbols (which have just been added to
1377    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1378    not compile and the installer would choke.)
1379    
1380    ----------------------------------------------------------------------
1381    Name: Matthias Blume
1382    Date: 2002/02/13 16:15:00 EST
1383    Tag: blume-20020213-various
1384    Description:
1385    
1386    1. More tweaks to ml-nlffigen:
1387    
1388       - better internal datastructures (resulting in slight speedup)
1389       - "-match" option requires exact match
1390       - "localized" gensym counters (untagged structs/unions nested within
1391         other structs/unions or within typedefs get a fresh counter; their
1392         tag will be prefixed by a concatenation of their parents' tags)
1393       - bug fixes (related to calculation of transitive closure of types
1394         to be included in the output)
1395    
1396    2. Minor Basis updates:
1397    
1398       - added implementations for List.collate and Option.app
1399    
1400    ----------------------------------------------------------------------
1401    Name: Matthias Blume
1402    Date: 2002/02/11 15:55:00 EST
1403    Tag: blume-20020211-gensym
1404    Description:
1405    
1406    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1407    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1408    names (ML structure names that correspond to unnamed C structs, unions,
1409    and enums), so that separate runs of ml-nlffigen do not clash.
1410    
1411    ----------------------------------------------------------------------
1412    Name: Matthias Blume
1413    Date: 2002/02/11 12:05:00 EST
1414    Tag: blume-20020211-gensml
1415    Description:
1416    
1417    A quick fix for a problem with GenSML (in the pgraph-util library):
1418    Make generation of toplevel "local" optional.  (Strictly speaking,
1419    signature definitions within "local" are not legal SML.)
1420    
1421    Other than that: updates to INSTALL and cm/TODO.
1422    
1423    ----------------------------------------------------------------------
1424    Name: Matthias Blume
1425    Date: 2002/02/08 15:00:00 EST
1426    Tag: blume-20020208-uniquepid
1427    Description:
1428    
1429    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1430    
1431    1. The installer (config/install.sh) has gotten smarter:
1432    
1433         - Configuration options are a bit easier to specify now
1434           (in config/targets).
1435         - Bug in recognizing .tar.bz2 files fixed.
1436         - Installer automatically resolves dependencies between
1437           configuration options (e.g., if you ask for eXene, you will
1438           also get cml -- regardless whether you asked for it or not).
1439         - Installer can run in "quieter mode" by setting the environment
1440           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1441           "completely silent", though.
1442         - Build HashCons library as part of smlnj-lib.
1443    
1444    2. A new scheme for assigning persistent identifiers to compilation
1445       units (and, by extension, to types etc.) has been put into place.
1446       This fixes a long-standing bug where types and even dynamic values
1447       can get internally confused, thereby compromising type safety
1448       (abstraction) and dynamic correctness.  See
1449    
1450         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1451    
1452       for an example of how things could go wrong until now.
1453    
1454       The downside of the new scheme is that pids are not quite as
1455       persistent as they used to be: CM will generate a fresh pid
1456       for every compilation unit that it thinks it sees for the first
1457       time.  That means that if you compile starting from a clean, fresh
1458       source tree at two different times, you end up with different
1459       binaries.
1460    
1461       Cutoff recompilation, however, has not been compromised because
1462       CM keeps pid information in special caches between runs.
1463    
1464    ----------------------------------------------------------------------
1465    Name: Lal George
1466    Date: 2002/02/07 15:34:13 EST 2002
1467    Tag: <none>
1468    Description:
1469    
1470    Compilers that generate assembly code may produce  global labels
1471    whose value is resolved at link time. The various peephole optimization
1472    modules did not take this in account.
1473    
1474    TODO. The Labels.addrOf function should really return an option
1475    type so that clients are forced to deal with this issue, rather
1476    than an exception being raised.
1477    
1478    ----------------------------------------------------------------------
1479    Name: Lal George
1480    Date: 2002/02/06 13:55:02 EST
1481    Tag: george-20020206-ra-breakup
1482    Description:
1483    
1484    1. A bug fix from Allen.
1485    
1486        A typo causes extra fstp %st(0)'s to be generated at compensation
1487        edges, which might cause stack underflow traps at runtime.  This
1488        occurs in fft where there are extraneous fstps right before the 'into'
1489        trap instruction (in this case they are harmless since none of the
1490        integers overflow.)
1491    
1492    2. Pulled out various utility modules that were embedded in the modules
1493       of the register allocator. I need these modules for other purposes, but
1494       they are not complete enough to put into a library (just yet).
1495    ----------------------------------------------------------------------
1496    Name: Matthias Blume
1497    Date: 2002/01/31 16:05:00 EST
1498    Tag: blume-20020131-sparc-ccalls
1499    Description:
1500    
1501    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1502       of extra stack space by mistake.  Fixed.
1503    
1504    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1505    
1506    ----------------------------------------------------------------------
1507    Name: Allen Leung
1508    Date: 2002/01/30
1509    Tag: leunga-20020130-nowhere-bug-fix
1510    Description:
1511    
1512       MLRISC bug fixes:
1513       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1514       2. MachineInt.fromString was negating its value.
1515    
1516    ----------------------------------------------------------------------
1517    Name: Matthias Blume
1518    Date: 2002/01/29
1519    Tag: blume-20020129-INSTALL
1520    Description:
1521    
1522    - Added somewhat detailed installation instructions (file INSTALL).
1523    - Fixed curl-detection bug in config/install.sh.
1524    - It is now possible to select the URL getter using the URLGETTER
1525      environment variable:
1526    
1527          not set / "unknown"      --> automatic detection (script tries wget,
1528                                       curl, and lynx)
1529          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1530                                       how to properly invoke them)
1531          other                    --> use $URLGETTER directly, it must take
1532                                       precisely two command-line arguments
1533                                       (source URL and destination file name)
1534    
1535    ----------------------------------------------------------------------
1536    Name: Matthias Blume
1537    Date: 2002/01/28
1538    Tag: blume-20020128-sparc-ccalls
1539    Description:
1540    
1541    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1542    - Make use of the allocParam argument in sparc-c-calls.
1543    
1544    ----------------------------------------------------------------------
1545    Name: Matthias Blume
1546  Date: 2002/01/28  Date: 2002/01/28
1547  Tag: blume-20020128-allocParam  Tag: blume-20020128-allocParam
1548  Description:  Description:
# Line 584  Line 2113 
2113    
2114  ----------------------------------------------------------------------  ----------------------------------------------------------------------
2115  Name: Matthias Blume  Name: Matthias Blume
 >>>>>>> 1.169  
2116  Date: 2001/09/18 15:35:00 EDT  Date: 2001/09/18 15:35:00 EDT
2117  Tag: blume-20010918-readme11036  Tag: blume-20010918-readme11036
2118  Description:  Description:

Legend:
Removed from v.1044  
changed lines
  Added in v.1229

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