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 1034, Thu Jan 24 05:53:53 2002 UTC revision 1240, Tue Jun 11 15:35:34 2002 UTC
# Line 13  Line 13 
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume
17    Date: 2002/06/11
18    Tag: blume-20020611-unixpath
19    Description:
20    
21    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
22    
23    ----------------------------------------------------------------------
24    Name: Matthias Blume
25    Date: 2002/06/10 16:35:00 EDT
26    Tag: blume-20020610-ieeereal
27    Description:
28    
29    I back-ported my implementation of IEEEReal.fromString from the
30    idlbasis-devel branch so that we can test it.
31    
32    Another small change is that ppDec tries to give more information
33    than just "<sig>" in the case of functors.  However, this code is
34    broken in some mysterious way if the functor's body's signature
35    has not been declared by ascription but gets inferred from the
36    implementation.  This needs fixing...
37    
38    ----------------------------------------------------------------------
39    Name: Matthias Blume
40    Date: 2002/05/31
41    Tag: blume-20020531-btrace-mode
42    Description:
43    
44    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
45    the wayside when I switched over to using Controls everywhere.)
46    
47    ----------------------------------------------------------------------
48    Name: Lal George
49    Date: 2002/05/23 12:21:40 EDT
50    Tag: george-20020523-visual-labels
51    Description:
52    
53    Labels are now displayed in the graphical output to make
54    the fall-through and target blocks obvious.
55    
56    ----------------------------------------------------------------------
57    Name: Matthias Blume
58    Date: 2002/05/22 11:03:00 EDT
59    Tag: blume-20020522-shrink
60    Description:
61    
62    John tweaked yesterday's fix for 1131 to handle an out-of-memory
63    situation that comes up when allocating huge arrays.
64    
65    ----------------------------------------------------------------------
66    Name: Matthias Blume
67    Date: 2002/05/21 16:00:00 EDT
68    Tag: Release_110_40
69    Description:
70    
71    New working release (110.40).  New bootfiles.
72    
73    [Also: John Reppy fixed GC bug 1131.]
74    
75    ----------------------------------------------------------------------
76    Name: Matthias Blume
77    Date: 2002/05/21 12:35:00 EDT
78    Tag: blume-20020521-cmdoc
79    Description:
80    
81    CM documentation update.
82    
83    ----------------------------------------------------------------------
84    Name: Matthias Blume
85    Date: 2002/05/21 10:55:00 EDT
86    Tag: blume-20020521-misc
87    Description:
88    
89    - John tweaked runtime to be silent on heap export (except when
90      GC messages are on).
91    - I added a few more things (cross-compiling versions of CMB) to
92      config/preloads (as suggestions).
93    
94    ----------------------------------------------------------------------
95    Name: Matthias Blume
96    Date: 2002/05/20 22:25:00 EDT
97    Tag: blume-20020520-controls
98    Description:
99    
100    - Added ControlUtil structure to control-lib.cm.
101    - Use it throughout.
102    - Used Controls facility to define MLRISC controls (as opposed to
103      registering MLRISC control ref cells with Controls after the fact)
104    - Fixed messed-up controls priorities.
105    
106    * Removed again all the stuff from config/preloads that one wouldn't
107      be able to preload at the time the initial heap image is built.
108      (Many libraries, e.g., CML, do not exist yet at this time.  The
109       only libraries that can be preloaded via config/preloads are those
110       that come bundled with the bootfiles.)
111    
112    ----------------------------------------------------------------------
113    Name: Matthias Blume
114    Date: 2002/05/20 10:59:00 EDT
115    Tag: blume-20020520-preloads
116    Description:
117    
118    Added a lot of commented-out suggestions for things to be included
119    in config/preloads.
120    
121    ----------------------------------------------------------------------
122    Name: Allen Leung
123    Date: 2002/05/18 14:20:00 EDT
124    Tag: leunga-20020518-mdl
125    Description:
126    
127    o Made the mdl tool stuff compile and run again.
128    o I've disabled all the stuff that depends on RTL specifications; they
129      are all badly broken anyway.
130    
131    ----------------------------------------------------------------------
132    Name: Matthias Blume
133    Date: 2002/05/17 16:49:00 EDT
134    Tag: blume-20020517-controls
135    Description:
136    
137    0. John Reppy made several modifications to the SML/NJ library.
138       In particular, there is a shiny new controls-lib.cm.
139    
140    1. Pushed new controls interface through compiler so that everything
141       compiles again.
142    
143    2. Added FormatComb and FORMAT_COMB to the CML version of the
144       SML/NJ library (so that CML compiles again).
145    
146    3. Modified init scripts because XXX_DEFAULT environment variables
147       are no longer with us.  (Boot-time initialization is now done
148       using the same environment variables that are also used for
149       startup-time initialization of controls.)
150    
151    ----------------------------------------------------------------------
152    Name: Lal George
153    Date: 2002/05/15 09:20:10 EDT
154    Tag: george-20020515-pseudo-op-decls
155    Description:
156    
157            All pseudo-ops emitted before the first segment declaration
158    such as TEXT, DATA, and BSS directives are assumed to be global
159    declarations and are emitted first in the assembly file. This is
160    useful in a number of situations where one has pseudo-ops that are not
161    specific to any segment, and also works around the constraint that one
162    cannot have client pseudo-ops in the TEXT segment.
163    
164            Because no segment is associated with these declarations it is
165    an error to allocate any space or objects before the first segment
166    directive and an exception will be raised. However, we cannot make
167    this check for client pseudo-ops.
168    
169            These top level declarations are a field in the CFG graph_info.
170    In theory you can continue to add to this field after the CFG has been
171    built -- provided you know what you are doing;-)
172    
173    ----------------------------------------------------------------------
174    Name: Matthias Blume
175    Date: 2002/05/13 16:40:00 EDT
176    Tag: blume-20020513-pp-etc
177    Description:
178    
179    A few minor bugfixes:
180    
181      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
182        (Bogus printouts for redefined bindings still occur.  Compiler
183        bug should no longer occur now.  We need to redo the prettyprinter
184        from scratch.)
185    
186      - CM pathname printer now also adds escape sequences for ( and )
187    
188      - commend and docu fixes for ml-nlffi
189    
190    ----------------------------------------------------------------------
191    Name: Matthias Blume
192    Date: 2002/05/10 16:40:00 EDT
193    Tag: blume-20020510-erg-textio
194    Description:
195    
196    Applied the following bugfix provided by Emden Gansner:
197    
198        Output is corrupted when outputSubstr is used rather than output.
199    
200        The problem occurs when a substring
201    
202            ss = (s, dataStart, dataLen)
203    
204        where dataStart > 0, fills a stream buffer with avail bytes left.
205        avail bytes of s, starting at index dataStart, are copied into the
206        buffer, the buffer is flushed, and then the remaining dataLen-avail
207        bytes of ss are copied into the beginning of the buffer. Instead of
208        starting this copy at index dataStart+avail in s, the current code
209        starts the copy at index avail.
210    
211        Fix:
212        In text-io-fn.sml, change line 695 from
213             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
214        to
215             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
216    
217    ----------------------------------------------------------------------
218    Name: Matthias Blume
219    Date: 2002/04/12 13:55:00 EDT
220    Tag: blume-20020412-assyntax
221    Description:
222    
223    1. Grabbed newer assyntax.h from the XFree86 project.
224    2. Fiddled with how to compile X86.prim.asm without warnings.
225    3. (Very) Minor cleanup in CM.
226    
227    ----------------------------------------------------------------------
228    Name: Matthias Blume
229    Date: 2002/04/01 (no joke!) 17:07:00 EST
230    Tag: blume-20020401-x86div
231    Description:
232    
233    Added full support for div/mod/rem/quot on the x86, using the machine
234    instruction's two results (without clumsily recomputing the remainder)
235    directly where appropriate.
236    
237    Some more extensive power-of-two support was added to the x86 instruction
238    selector (avoiding expensive divs, mods, and muls where they can be
239    replaced with cheaper shifts and masks).  However, this sort of thing
240    ought to be done earlier, e.g., within the CPS optimizer so that
241    all architectures benefit from it.
242    
243    The compiler compiles to a fixed point, but changes might be somewhat
244    fragile nevertheless.  Please, report any strange things that you might
245    see wrt. div/mod/quot/rem...
246    
247    ----------------------------------------------------------------------
248    Name: Matthias Blume
249    Date: 2002/03/29 17:22:00
250    Tag: blume-20020329-div
251    Description:
252    
253    Fixed my broken div/mod logic.  Unfortunately, this means that the
254    inline code for div/mod now has one more comparison than before.
255    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
256    The problem was with quotient = 0, because that alone does not tell
257    us which way the rounding went.  One then has to look at whether
258    remainder and divisor have the same sign...  :(
259    
260    Anyway, I replaced the bootfiles with fresh ones...
261    
262    ----------------------------------------------------------------------
263    Name: Matthias Blume
264    Date: 2002/03/29 14:10:00 EST
265    Tag: blume-20020329-inlprims
266    Description:
267    
268    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
269    
270    Primops have changed. This means that the bin/boot-file formats have
271    changed as well.
272    
273    To make sure that there is no confusion, I made a new version.
274    
275    
276    CHANGES:
277    
278    * removed REMT from mltree (remainder should never overflow).
279    
280    * added primops to deal with divisions of all flavors to the frontend
281    
282    * handled these primops all the way through so they map to their respective
283      MLRISC support
284    
285    * used these primops in the implementation of Int, Int32, Word, Word32
286    
287    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
288    
289    * parameterized INLMIN, INLMAX, and INLABS by a numkind
290    
291    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
292      floating point
293    
294    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
295      and Real (but Real.abs maps to a separate floating-point-only primop)
296    
297    
298    TODO items:
299    
300    * Hacked Alpha32 instruction selection, disabling the selection of REMx
301      instructions because the machine instruction encoder cannot handle
302      them.  (Hppa, PPC, and Sparc instruction selection did not handle
303      REM in the first place, and REM is supported by the x86 machine coder.)
304    
305    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
306      selection phase.  (The two can be streamlined because the hardware
307      delivers both quotient and remainder at the same time anyway.)
308    
309    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
310      (Currently the behavior is inconsistent both across architectures and
311      wrt. the draft Basis spec.)
312    
313    * Word8 should eventually be handled natively, too.
314    
315    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
316      as if there currently is no execution path that could trigger it in
317      SML/NJ.  (The assumptions underlying functions arith and promotable do not
318      hold for things like multiplication and division.)
319    
320    ----------------------------------------------------------------------
321    Name: Matthias Blume
322    Date: 2002/03/27 16:27:00 EST
323    Tag: blume-20020327-mlrisc-divisions
324    Description:
325    
326    Added support for all four division operations (ML's div, mod, quot,
327    and rem) to MLRISC.  In the course of doing so, I also rationalized
328    the naming (no more annoying switch-around of DIV and QUOT), by
329    parameterizing the operation by div_rounding_mode (which can be either
330    DIV_TO_ZERO or DIV_TO_NEGINF).
331    
332    The generic MLTreeGen functor takes care of compiling all four
333    operations down to only round-to-zero div.
334    
335    Missing pieces:
336    
337      * Doing something smarter than relying on MLTreeGen on architectures
338        like, e.g., the x86 where hardware division delivers both quotient and
339        remainder at the same time.  With this, the implementation of the
340        round-to-neginf operations could be further streamlined.
341    
342      * Remove inlining support for div/mod/rem from the frontend and replace it
343        with primops that get carried through to the backend.  Do this for all
344        int and word types.
345    
346    ----------------------------------------------------------------------
347    Name: Matthias Blume
348    Date: 2002/03/25 17:25:00 EST
349    Tag: blume-20020325-divmod
350    Description:
351    
352    I improved (hopefully without breaking them) the implementation of Int.div,
353    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
354    advantage of the following observations:
355    
356      Let  q = x quot y      r = x rem y
357           d = x div  y      m = x mod y
358    
359    where "quot" is the round-to-zero version of integer division that
360    hardware usually provides.  Then we have:
361    
362         r = x - q * y        where neither the * nor the - will overflow
363         d = if q >= 0 orelse x = q * y then q else q - 1
364                              where neither the * nor the - will overflow
365         m = if q >= 0 orelse r = 0 then r else r + y
366                              where the + will not overflow
367    
368    This results in substantial simplification of the generated code.
369    The following table shows the number of CFG nodes and edges generated
370    for
371            fun f (x, y) = x OPER y
372            (* with OPER \in div, mod, quot, rem *)
373    
374    
375        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
376        --------------------------------------------------------
377         div |         24 |         39 |         12 |         16
378         mod |         41 |         71 |         12 |         16
379        quot |          8 |         10 |          8 |         10
380         rem |         10 |         14 |          8 |         10
381    
382    
383    ----------------------------------------------------------------------
384    Name: Matthias Blume
385    Date: 2002/03/25 22:06:00 EST
386    Tag: blume-20020325-cprotobug
387    Description:
388    
389    Fixed a bug in cproto (c prototype decoder).
390    
391    ----------------------------------------------------------------------
392    Name: Matthias Blume
393    Date: 2002/03/25 16:00:00 EST
394    Tag: blume-20020325-raw-primops
395    Description:
396    
397    I did some cleanup to Allen's new primop code and
398    replaced yesterday's bootfiles with new ones.
399    (But they are stored in the same place.)
400    
401    ----------------------------------------------------------------------
402    Name: Matthias Blume
403    Date: 2002/03/24 22:40:00 EST
404    Tag: blume-20020324-bootfiles
405    Description:
406    
407    Made the bootfiles that Allen asked for.
408    
409    ----------------------------------------------------------------------
410    Name: Allen Leung
411    Date: 2002/03/23 15:50:00 EST
412    Tag: leunga-20020323-flint-cps-rcc-primops
413    Description:
414    
415      1. Changes to FLINT primops:
416    
417        (* make a call to a C-function;
418         * The primop carries C function prototype information and specifies
419         * which of its (ML-) arguments are floating point. C prototype
420         * information is for use by the backend, ML information is for
421         * use by the CPS converter. *)
422      | RAW_CCALL of { c_proto: CTypes.c_proto,
423                       ml_args: ccall_type list,
424                       ml_res_opt: ccall_type option,
425                       reentrant : bool
426                     } option
427       (* Allocate uninitialized storage on the heap.
428        * The record is meant to hold short-lived C objects, i.e., they
429        * are not ML pointers.  With the tag, the representation is
430        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
431        *)
432      | RAW_RECORD of {tag:bool,sz:int}
433      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
434    
435      2.  These CPS primops are now overloaded:
436    
437           rawload of {kind:numkind}
438           rawstore of {kind:numkind}
439    
440          The one argument form is:
441    
442             rawload {kind} address
443    
444          The two argument form is:
445    
446             rawload {kind} [ml object, byte-offset]
447    
448      3. RAW_CCALL/RCC now takes two extra arguments:
449    
450         a. The first is whether the C call is reentrant, i.e., whether
451            ML state should be saved and restored.
452         b. The second argument is a string argument specifying the name of
453            library and the C function.
454    
455         These things are currently not handled in the code generator, yet.
456    
457      4. In CProto,
458    
459         An encoding type of "bool" means "ml object" and is mapped into
460         C prototype of PTR.  Note that "bool" is different than "string",
461         even though "string" is also mapped into PTR, because "bool"
462         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
463    
464      5. Pickler/unpicker
465    
466         Changed to handle RAW_RECORD and newest RAW_CCALL
467    
468      6. MLRiscGen,
469    
470         1. Changed to handle the new rawload/rawstore/rawrecord operators.
471         2. Code for handling C Calls has been moved to a new module CPSCCalls,
472            in the file CodeGen/cpscompile/cps-c-calls.sml
473    
474      7. Added the conditional move operator
475    
476             condmove of branch
477    
478         to cps.  Generation of this is still buggy so it is currently
479         disabled.
480    
481    ----------------------------------------------------------------------
482    Name: Lal George
483    Date: 2002/03/22 14:18:25 EST
484    Tag: george-20020322-cps-branch-prob
485    Description:
486    
487    Implemented the Ball-Larus branch prediction-heuristics, and
488    incorporated graphical viewers for control flow graphs.
489    
490    Ball-Larus Heuristics:
491    ---------------------
492    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
493    
494    By design it uses the Dempster-Shafer theory for combining
495    probabilities.  For example, in the function:
496    
497        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
498    
499    the ball-larus heuristics predicts that the n=0 is unlikely
500    (OH-heuristic), and the 'then' branch is unlikely because of the
501    RH-heuristic -- giving the 'then' branch an even lower combined
502    probability using the Dempster-Shafer theory.
503    
504    Finally, John Reppy's loop analysis in MLRISC, further lowers the
505    probability of the 'then' branch because of the loop in the else
506    branch.
507    
508    
509    Graphical Viewing:
510    ------------------
511    I merely plugged in Allen's graphical viewers into the compiler. The
512    additional code is not much. At the top level, saying:
513    
514            Control.MLRISC.getFlag "cfg-graphical-view" := true;
515    
516    will display the graphical view of the control flow graph just before
517    back-patching.  daVinci must be in your path for this to work. If
518    daVinci is not available, then the default viewer can be changed
519    using:
520    
521            Control.MLRISC.getString "viewer"
522    
523    which can be set to "dot" or "vcg" for the corresponding viewers. Of
524    course, these viewers must be in your path.
525    
526    The above will display the compilation unit at the level of clusters,
527    many of which are small, boring, and un-interesting. Also setting:
528    
529            Control.MLRISC.getInt "cfg-graphical-view_size"
530    
531    will display clusters that are larger than the value set by the above.
532    
533    
534    ----------------------------------------------------------------------
535    Name: Matthias Blume
536    Date: 2002/03/21 22:20:00 EST
537    Tag: blume-20020321-kmp-bugfix
538    Description:
539    
540    Changed the interface to the KMP routine in PreString and fixed
541    a minor bug in one place where it was used.
542    
543    ----------------------------------------------------------------------
544    Name: Allen Leung
545    Date: 2002/03/21 20:30:00 EST
546    Tag: leunga-20020321-cfg
547    Description:
548    
549      Fixed a potential problem in cfg edge splitting.
550    
551    ----------------------------------------------------------------------
552    Name: Allen Leung
553    Date: 2002/03/21 17:15:00 EST
554    Tag: leunga-20020321-x86-fp-cfg
555    Description:
556    
557      1. Recoded the buggy parts of x86-fp.
558    
559         a. All the block reordering code has been removed.
560            We now depend on the block placement phases to do this work.
561    
562         b. Critical edge splitting code has been simplified and moved into the
563            CFG modules, as where they belong.
564    
565         Both of these were quite buggy and complex.  The code is now much, much
566         simpler.
567    
568      2. X86 backend.
569    
570         a. Added instructions for 64-bit support.  Instruction selection for
571            64-bit has not been committed, however, since that
572            requires changes to MLTREE which haven't been approved by
573            Lal and John.
574    
575         b. Added support for FUCOMI and FUCOMIP when generating code for
576            PentiumPro and above.  We only generate these instructions in
577            the fast-fp mode.
578    
579         c. Added cases for JP and JNP in X86FreqProps.
580    
581      3. CFG
582    
583         CFG now has a bunch of methods for edge splitting and merging.
584    
585      4. Machine description.
586    
587         John's simplification of MLTREE_BASIS.fcond broke a few machine
588         description things:
589    
590         rtl-build.{sig,sml} and hppa.mdl fixed.
591    
592         NOTE: the machine description stuff in the repository is still broken.
593               Again, I can't put my fixes in because that involves
594               changes to MLTREE.
595    
596    ----------------------------------------------------------------------
597    Name: Matthias Blume
598    Date: 2002/03/20 15:55:00 EST
599    Tag: blume-20020320-kmp
600    Description:
601    
602    Implemented Knuth-Morris-Pratt string matching in PreString and used
603    it for String.isSubstring, Substring.isSubstring, and
604    Substring.position.
605    
606    (Might need some stress-testing.  Simple examples worked fine.)
607    
608    ----------------------------------------------------------------------
609    Name: Matthias Blume
610    Date: 2002/03/19 16:37:00 EST
611    Tag: blume-20020319-witnesses
612    Description:
613    
614    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
615    
616    This implements a generic mechanism for changing constness qualifiers
617    anywhere within big C types without resorting to outright "casts".
618    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
619    modify the constness at the outermost level.)
620    The implementation of "convert" is based on the idea of "witness"
621    values -- values that are not used by the operation but whose types
622    "testify" to their applicability.  On the implementation side, "convert"
623    is simply a projection (returning its second curried argument).  With
624    cross-module inlining, it should not result in any machine code being
625    generated.
626    
627    ----------------------------------------------------------------------
628    Name: Matthias Blume
629    Date: 2002/03/15 16:40:00 EST
630    Tag: blume-20020315-basis
631    Description:
632    
633    Provided (preliminary?) implementations for
634    
635      {String,Substring}.{concatWith,isSuffix,isSubstring}
636    
637    and
638    
639      Substring.full
640    
641    Those are in the Basis spec but they were missing in SML/NJ.
642    
643    ----------------------------------------------------------------------
644    Name: Matthias Blume
645    Date: 2002/03/14 21:30:00 EST
646    Tag: blume-20020314-controls
647    Description:
648    
649    Controls:
650    ---------
651    
652    1. Factored out the recently-added Controls : CONTROLS stuff and put
653       it into its own library $/controls-lib.cm.  The source tree for
654       this is under src/smlnj-lib/Controls.
655    
656    2. Changed the names of types and functions in this interface, so they
657       make a bit more "sense":
658    
659          module -> registry
660          'a registry -> 'a group
661    
662    3. The interface now deals in ref cells only.  The getter/setter interface
663       is (mostly) gone.
664    
665    4. Added a function that lets one register an already-existing ref cell.
666    
667    5. Made the corresponding modifications to the rest of the code so that
668       everything compiles again.
669    
670    6. Changed the implementation of Controls.MLRISC back to something closer
671       to the original.  In particular, this module (and therefore MLRISC)
672       does not depend on Controls.  There now is some link-time code in
673       int-sys.sml that registers the MLRISC controls with the Controls
674       module.
675    
676    CM:
677    ---
678    
679      * One can now specify the lambda-split aggressiveness in init.cmi.
680    
681    ----------------------------------------------------------------------
682    Name: Allen Leung
683    Date: 2002/03/13 17:30:00 EST
684    Tag: leunga-20020313-x86-fp-unary
685    Description:
686    
687    Bug fix for:
688    
689    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
690    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
691    > - fun f(x,(y,z)) = Real.~ y;
692    > [autoloading]
693    > [autoloading done]
694    >       fchsl   (%eax), 184(%esp)
695    > Error: MLRisc bug: X86MCEmitter.emitInstr
696    >
697    > uncaught exception Error
698    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
699    
700    The problem was that the code generator did not generate any fp registers
701    in this case, and the ra didn't know that it needed to run the X86FP phase to
702    translate the pseudo fp instruction.   This only happened with unary fp
703    operators in certain situations.
704    
705    ----------------------------------------------------------------------
706    Name: Matthias Blume
707    Date: 2002/03/13 14:00:00 EST
708    Tag: blume-20020313-overload-etc
709    Description:
710    
711    1. Added _overload as a synonym for overload for backward compatibility.
712       (Control.overloadKW must be true for either version to be accepted.)
713    
714    2. Fixed bug in install script that caused more things to be installed
715       than what was requested in config/targets.
716    
717    3. Made CM aware of the (_)overload construct so that autoloading
718       works.
719    
720    ----------------------------------------------------------------------
721    Name: Matthias Blume
722    Date: 2002/03/12 22:03:00 EST
723    Tag: blume-20020312-url
724    Description:
725    
726    Forgot to update BOOT and srcarchiveurl.
727    
728    ----------------------------------------------------------------------
729    Name: Matthias Blume
730    Date: 2002/03/12 17:30:00 EST
731    Tag: blume-20020312-version110392
732    Description:
733    
734    Yet another version number bump (because of small changes to the
735    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
736    
737    Changes:
738    
739      The new pid generation scheme described a few weeks ago was overly
740      complicated.  I implemented a new mechanism that is simpler and
741      provides a bit more "stability":  Once CM has seen a compilation
742      unit, it keeps its identity constant (as long as you do not delete
743      those crucial CM/GUID/* files).  This means that when you change
744      an interface, compile, then go back to the old interface, and
745      compile again, you arrive at the original pid.
746    
747      There now also is a mechanism that instructs CM to use the plain
748      environment hash as a module's pid (effectively making its GUID
749      the empty string).  For this, "noguid" must be specified as an
750      option to the .sml file in question within its .cm file.
751      This is most useful for code that is being generated by tools such
752      as ml-nlffigen (because during development programmers tend to
753      erase the tool's entire output directory tree including CM's cached
754      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
755      revert to the old, broken behavior of SML/NJ, but in specific cases
756      where there is no danger of interface confusion, its use is ok
757      (I think).
758    
759      ml-nlffigen by default generates "noguid" annotations.  They can be
760      turned off by specifying -guid in its command line.
761    
762    ----------------------------------------------------------------------
763    Name: Lal George
764    Date: 2002/03/12 12 14:42:36 EST
765    Tag: george-20020312-frequency-computation
766    Description:
767    
768    Integrated jump chaining and static block frequency into the
769    compiler. More details and numbers later.
770    
771    ----------------------------------------------------------------------
772    Name: Lal George
773    Date: 2002/03/11 11 22:38:53 EST
774    Tag: george-20020311-jump-chain-elim
775    Description:
776    
777    Tested the jump chain elimination on all architectures (except the
778    hppa).  This is on by default right now and is profitable for the
779    alpha and x86, however, it may not be profitable for the sparc and ppc
780    when compiling the compiler.
781    
782    The gc test will typically jump to a label at the end of the cluster,
783    where there is another jump to an external cluster containing the actual
784    code to invoke gc. This is to allow factoring of common gc invocation
785    sequences. That is to say, we generate:
786    
787            f:
788               testgc
789               ja   L1      % jump if above to L1
790    
791            L1:
792               jmp L2
793    
794    
795    After jump chain elimination the 'ja L1' instructions is converted to
796    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
797    up being implemented in their long form (if L2 is far away) using:
798    
799            jbe     L3      % jump if below or equal to L3
800            jmp     L2
801         L3:
802            ...
803    
804    
805    For large compilation units L2  may be far away.
806    
807    
808    ----------------------------------------------------------------------
809    Name: Matthias Blume
810    Date: 2002/03/11 13:30:00 EST
811    Tag: blume-20020311-mltreeeval
812    Description:
813    
814    A functor parameter was missing.
815    
816    ----------------------------------------------------------------------
817    Name: Allen Leung
818    Date: 2002/03/11 10:30:00 EST
819    Tag: leunga-20020311-runtime-string0
820    Description:
821    
822       The representation of the empty string now points to a
823    legal null terminated C string instead of unit.  It is now possible
824    to convert an ML string into C string with InlineT.CharVector.getData.
825    This compiles into one single machine instruction.
826    
827    ----------------------------------------------------------------------
828    Name: Allen Leung
829    Date: 2002/03/10 23:55:00 EST
830    Tag: leunga-20020310-x86-call
831    Description:
832    
833       Added machine generation for CALL instruction (relative displacement mode)
834    
835    ----------------------------------------------------------------------
836    Name: Matthias Blume
837    Date: 2002/03/08 16:05:00
838    Tag: blume-20020308-entrypoints
839    Description:
840    
841    Version number bumped to 110.39.1.  NEW BOOTFILES!
842    
843    Entrypoints: non-zero offset into a code object where execution should begin.
844    
845    - Added the notion of an entrypoint to CodeObj.
846    - Added reading/writing of entrypoint info to Binfile.
847    - Made runtime system bootloader aware of entrypoints.
848    - Use the address of the label of the first function given to mlriscGen
849      as the entrypoint.  This address is currently always 0, but it will
850      not be 0 once we turn on block placement.
851    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
852      with entry points) from mlriscGen.
853    
854    ----------------------------------------------------------------------
855    Name: Allen Leung
856    Date: 2002/03/07 20:45:00 EST
857    Tag: leunga-20020307-x86-cmov
858    Description:
859    
860       Bug fixes for CMOVcc on x86.
861    
862       1. Added machine code generation for CMOVcc
863       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
864       3. CMOVcc cannot have an immediate operand as argument.
865    
866    ----------------------------------------------------------------------
867    Name: Matthias Blume
868    Date: 2002/03/07 16:15:00 EST
869    Tag: blume-20020307-controls
870    Description:
871    
872    This is a very large but mostly boring patch which makes (almost)
873    every tuneable compiler knob (i.e., pretty much everything under
874    Control.* plus a few other things) configurable via both the command
875    line and environment variables in the style CM did its configuration
876    until now.
877    
878    Try starting sml with '-h' (or, if you are brave, '-H')
879    
880    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
881    implements the underlying generic mechanism.
882    
883    The interface to some of the existing such facilities has changed somewhat.
884    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
885    (The getFoo interface is still there for backward-compatibility, but its
886    use is deprecated.)
887    
888    The ml-build script passes -Cxxx=yyy command-line arguments through so
889    that one can now twiddle the compiler settings when using this "batch"
890    compiler.
891    
892    TODO items:
893    
894    We should go through and throw out all controls that are no longer
895    connected to anything.  Moreover, we should go through and provide
896    meaningful (and correct!) documentation strings for those controls
897    that still are connected.
898    
899    Currently, multiple calls to Controls.new are accepted (only the first
900    has any effect).  Eventually we should make sure that every control
901    is being made (via Controls.new) exactly once.  Future access can then
902    be done using Controls.acc.
903    
904    Finally, it would probably be a good idea to use the getter-setter
905    interface to controls rather than ref cells.  For the time being, both
906    styles are provided by the Controls module, but getter-setter pairs are
907    better if thread-safety is of any concern because they can be wrapped.
908    
909    *****************************************
910    
911    One bug fix: The function blockPlacement in three of the MLRISC
912    backpatch files used to be hard-wired to one of two possibilities at
913    link time (according to the value of the placementFlag).  But (I
914    think) it should rather sense the flag every time.
915    
916    *****************************************
917    
918    Other assorted changes (by other people who did not supply a HISTORY entry):
919    
920    1. the cross-module inliner now works much better (Monnier)
921    2. representation of weights, frequencies, and probabilities in MLRISC
922       changed in preparation of using those for weighted block placement
923       (Reppy, George)
924    
925    ----------------------------------------------------------------------
926    Name: Lal George
927    Date: 2002/03/07 14:44:24 EST 2002
928    Tag: george-20020307-weighted-block-placement
929    
930    Tested the weighted block placement optimization on all architectures
931    (except the hppa) using AMPL to generate the block and edge frequencies.
932    Changes were required in the machine properties to correctly
933    categorize trap instructions. There is an MLRISC flag
934    "weighted-block-placement" that can be used to enable weighted block
935    placement, but this will be ineffective without block/edge
936    frequencies (coming soon).
937    
938    
939    ----------------------------------------------------------------------
940    Name: Lal George
941    Date: 2002/03/05 17:24:48 EST
942    Tag: george-20020305-linkage-cluster
943    
944    In order to support the block placement optimization, a new cluster
945    is generated as the very first cluster (called the linkage cluster).
946    It contains a single jump to the 'real' entry point for the compilation
947    unit. Block placement has no effect on the linkage cluster itself, but
948    all the other clusters  have full freedom in the manner in which they
949    reorder blocks or functions.
950    
951    On the x86 the typical linkage code that is generated is:
952       ----------------------
953            .align 2
954       L0:
955            addl    $L1-L0, 72(%esp)
956            jmp     L1
957    
958    
959            .align  2
960       L1:
961       ----------------------
962    
963    72(%esp) is the memory location for the stdlink register. This
964    must contain the address of the CPS function being called. In the
965    above example, it contains the address of  L0; before
966    calling L1 (the real entry point for the compilation unit), it
967    must contain the address for L1, and hence
968    
969            addl $L1-L0, 72(%esp)
970    
971    I have tested this on all architectures except the hppa.The increase
972    in code size is of course negligible
973    
974    ----------------------------------------------------------------------
975    Name: Allen Leung
976    Date: 2002/03/03 13:20:00 EST
977    Tag: leunga-20020303-mlrisc-tools
978    
979      Added #[ ... ] expressions to mlrisc tools
980    
981    ----------------------------------------------------------------------
982    Name: Matthias Blume
983    Date: 2002/02/27 12:29:00 EST
984    Tag: blume-20020227-cdebug
985    Description:
986    
987    - made types in structure C and C_Debug to be equal
988    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
989    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
990    
991    ----------------------------------------------------------------------
992    Name: Matthias Blume
993    Date: 2002/02/26 12:00:00 EST
994    Tag: blume-20020226-ffi
995    Description:
996    
997    1. Fixed a minor bug in CM's "noweb" tool:
998       If numbering is turned off, then truly don't number (i.e., do not
999       supply the -L option to noweb).  The previous behavior was to supply
1000       -L'' -- which caused noweb to use the "default" line numbering scheme.
1001       Thanks to Chris Richards for pointing this out (and supplying the fix).
1002    
1003    2. Once again, I reworked some aspects of the FFI:
1004    
1005       A. The incomplete/complete type business:
1006    
1007       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1008         gone!
1009       - ML types representing an incomplete type are now *equal* to
1010         ML types representing their corresponding complete types (just like
1011         in C).  This is still safe because ml-nlffigen will not generate
1012         RTTI for incomplete types, nor will it generate functions that
1013         require access to such RTTI.   But when ML code generated from both
1014         incomplete and complete versions of the C type meet, the ML types
1015         are trivially interoperable.
1016    
1017         NOTE:  These changes restore the full generality of the translation
1018         (which was previously lost when I eliminated functorization)!
1019    
1020       B. Enum types:
1021    
1022       - Structure C now has a type constructor "enum" that is similar to
1023         how the "su" constructor works.  However, "enum" is not a phantom
1024         type because each "T enum" has values (and is isomorphic to
1025         MLRep.Signed.int).
1026       - There are generic access operations for enum objects (using
1027         MLRep.Signed.int).
1028       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1029         * The structure contains the definition of type "mlrep" (the ML-side
1030         representation type of the enum).  Normally, mlrep is the same
1031         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1032         then mlrep will be defined as a datatype -- thus facilitating
1033         pattern matching on mlrep values.
1034         ("-ec" will be suppressed if there are duplicate values in an
1035          enumeration.)
1036         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1037         will be generated for each C enum constant xxx.
1038         * Conversion functions m2i and i2m convert between mlrep and
1039         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1040         * Coversion functions c and ml convert between mlrep and "tag enum".
1041         * Access functions (get/set) fetch and store mlrep values.
1042       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1043         enumerations are merged into one single enumeration represented by
1044         structure E_'.
1045    
1046    ----------------------------------------------------------------------
1047    Name: Allen Leung
1048    Date: 2002/02/25 04:45:00 EST
1049    Tag: leunga-20020225-cps-spill
1050    
1051    This is a new implementation of the CPS spill phase.
1052    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1053    In case of problems, replace it with the old file spill.sml
1054    
1055    The current compiler runs into some serious performance problems when
1056    constructing a large record.  This can happen when we try to compile a
1057    structure with many items.  Even a very simple structure like the following
1058    makes the compiler slow down.
1059    
1060        structure Foo = struct
1061           val x_1 = 0w1 : Word32.int
1062           val x_2 = 0w2 : Word32.int
1063           val x_3 = 0w3 : Word32.int
1064           ...
1065           val x_N = 0wN : Word32.int
1066        end
1067    
1068    The following table shows the compile time, from N=1000 to N=4000,
1069    with the old compiler:
1070    
1071    N
1072    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1073           MLRISC ra                               0.06u  0.00s  0.05g
1074              (spills = 0 reloads = 0)
1075           TOTAL                                   0.63u  0.07s  0.21g
1076    
1077    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1078           MLRISC ra                               5.68u  0.59s  3.93g
1079              (spills = 0 reloads = 0)
1080           TOTAL                                   14.71u  0.99s  4.81g
1081    
1082    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1083           MLRISC ra                               5.54u  0.65s  3.91g
1084              (spills = 543 reloads = 1082)
1085           TOTAL                                   65.40u  3.13s  6.00g
1086    
1087    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1088           MLRISC ra                               0.80u  0.10s  0.55g
1089              (spills = 42 reloads = 84)
1090           TOTAL                                   129.42u  5.10s  4.13g
1091    
1092    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1093           MLRISC ra                               2.69u  0.27s  1.38g
1094              (spills = 62 reloads = 124)
1095           TOTAL                                   682.48u  19.61s  13.99g
1096    
1097    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1098           MLRISC ra                               4.96u  0.27s  2.72g
1099              (spills = 85 reloads = 170)
1100           TOTAL                                   2375.26u  57.21s  48.00g
1101    
1102    As you can see the old cps spill module suffers from some serious
1103    performance problem.  But since I cannot decipher the old code fully,
1104    instead of patching the problems up, I'm reimplementing it
1105    with a different algorithm.  The new code is more modular,
1106    smaller when compiled, and substantially faster
1107    (O(n log n) time and O(n) space).  Timing of the new spill module:
1108    
1109    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1110          MLRISC ra                               0.25u  0.02s  0.15g
1111             (spills=1 reloads=3)
1112          TOTAL                                   7.74u  0.34s  1.62g
1113    
1114    Implementation details:
1115    
1116    As far as I can tell, the purpose of the CPS spill module is to make sure the
1117    number of live variables at any program point (the bandwidth)
1118    does not exceed a certain limit, which is determined by the
1119    size of the spill area.
1120    
1121    When the bandwidth is too large, we decrease the register pressure by
1122    packing live variables into spill records.  How we achieve this is
1123    completely different than what we did in the old code.
1124    
1125    First, there is something about the MLRiscGen code generator
1126    that we should be aware of:
1127    
1128    o MLRiscGen performs code motion!
1129    
1130       In particular, it will move floating point computations and
1131       address computations involving only the heap pointer to
1132       their use sites (if there is only a single use).
1133       What this means is that if we have a CPS record construction
1134       statement
1135    
1136           RECORD(k,vl,w,e)
1137    
1138       we should never count the new record address w as live if w
1139       has only one use (which is often the case).
1140    
1141       We should do something similar to floating point, but the transformation
1142       there is much more complex, so I won't deal with that.
1143    
1144    Secondly, there are now two new cps primops at our disposal:
1145    
1146     1. rawrecord of record_kind option
1147        This pure operator allocates some uninitialized storage from the heap.
1148        There are two forms:
1149    
1150         rawrecord NONE [INT n]  allocates a tagless record of length n
1151         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1152                                     and initializes the tag.
1153    
1154     2. rawupdate of cty
1155          rawupdate cty (v,i,x)
1156          Assigns to x to the ith component of record v.
1157          The storelist is not updated.
1158    
1159    We use these new primops for both spilling and increment record construction.
1160    
1161     1. Spilling.
1162    
1163        This is implemented with a linear scan algorithm (but generalized
1164        to trees).  The algorithm will create a single spill record at the
1165        beginning of the cps function and use rawupdate to spill to it,
1166        and SELECT or SELp to reload from it.  So both spills and reloads
1167        are fine-grain operations.  In contrast, in the old algorithm
1168        "spills" have to be bundled together in records.
1169    
1170        Ideally, we should sink the spill record construction to where
1171        it is needed.  We can even split the spill record into multiple ones
1172        at the places where they are needed.  But CPS is not a good
1173        representation for global code motion, so I'll keep it simple and
1174        am not attempting this.
1175    
1176     2. Incremental record construction (aka record splitting).
1177    
1178        Long records with many component values which are simulatenously live
1179        (recall that single use record addresses are not considered to
1180         be live) are constructed with rawrecord and rawupdate.
1181        We allocate space on the heap with rawrecord first, then gradually
1182        fill it in with rawupdate.  This is the technique suggested to me
1183        by Matthias.
1184    
1185        Some restrictions on when this is applicable:
1186        1. It is not a VECTOR record.  The code generator currently does not handle
1187           this case. VECTOR record uses double indirection like arrays.
1188        2. All the record component values are defined in the same "basic block"
1189           as the record constructor.  This is to prevent speculative
1190           record construction.
1191    
1192    ----------------------------------------------------------------------
1193    Name: Allen Leung
1194    Date: 2002/02/22 01:02:00 EST
1195    Tag: leunga-20020222-mlrisc-tools
1196    
1197    Minor bug fixes in the parser and rewriter
1198    
1199    ----------------------------------------------------------------------
1200    Name: Allen Leung
1201    Date: 2002/02/21 20:20:00 EST
1202    Tag: leunga-20020221-peephole
1203    
1204    Regenerated the peephole files.  Some contained typos in the specification
1205    and some didn't compile because of pretty printing bugs in the old version
1206    of 'nowhere'.
1207    
1208    ----------------------------------------------------------------------
1209    Name: Allen Leung
1210    Date: 2002/02/19 20:20:00 EST
1211    Tag: leunga-20020219-mlrisc-tools
1212    Description:
1213    
1214       Minor bug fixes to the mlrisc-tools library:
1215    
1216       1.  Fixed up parsing colon suffixed keywords
1217       2.  Added the ability to shut the error messages up
1218       3.  Reimplemented the pretty printer and fixed up/improved
1219           the pretty printing of handle and -> types.
1220       4.  Fixed up generation of literal symbols in the nowhere tool.
1221       5.  Added some SML keywords to to sml.sty
1222    
1223    ----------------------------------------------------------------------
1224    Name: Matthias Blume
1225    Date: 2002/02/19 16:20:00 EST
1226    Tag: blume-20020219-cmffi
1227    Description:
1228    
1229    A wild mix of changes, some minor, some major:
1230    
1231    * All C FFI-related libraries are now anchored under $c:
1232        $/c.cm      --> $c/c.cm
1233        $/c-int.cm  --> $c/internals/c-int.cm
1234        $/memory.cm --> $c/memory/memory.cm
1235    
1236    * "make" tool (in CM) now treats its argument pathname slightly
1237      differently:
1238        1. If the native expansion is an absolute name, then before invoking
1239           the "make" command on it, CM will apply OS.Path.mkRelative
1240           (with relativeTo = OS.FileSys.getDir()) to it.
1241        2. The argument will be passed through to subsequent phases of CM
1242           processing without "going native".  In particular, if the argument
1243           was an anchored path, then "make" will not lose track of that anchor.
1244    
1245    * Compiler backends now "know" their respective C calling conventions
1246      instead of having to be told about it by ml-nlffigen.  This relieves
1247      ml-nlffigen from one of its burdens.
1248    
1249    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1250    
1251    * Export C_DEBUG and C_Debug from $c/c.cm.
1252    
1253    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1254      subtyping relationship between incomplete pointers and their complete
1255      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1256      with the convention of instantiating 'o with ('t, 'c) obj whenever
1257      the pointer target type is complete.  In the incomplete case, 'o
1258      will be instantiated with some "'c iobj" -- a type obtained by
1259      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1260    
1261      Operations that work on both incomplete and complete pointer types are
1262      typed as taking an 'o ptr while operations that require the target to
1263      be known are typed as taking some ('t, 'c) obj ptr.
1264    
1265      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1266      where void is an eqtype without any values.  This makes it possible
1267      to work on voidptr values using functions meant to operate on light
1268      incomplete pointers.
1269    
1270    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1271      been vastly simplified.
1272    
1273    ----------------------------------------------------------------------
1274    Name: Matthias Blume
1275    Date: 2002/02/19 10:48:00 EST
1276    Tag: blume-20020219-pqfix
1277    Description:
1278    
1279    Applied Chris Okasaki's bug fix for priority queues.
1280    
1281    ----------------------------------------------------------------------
1282    Name: Matthias Blume
1283    Date: 2002/02/15 17:05:00
1284    Tag: Release_110_39
1285    Description:
1286    
1287    Last-minute retagging is becoming a tradition... :-(
1288    
1289    This is the working release 110.39.
1290    
1291    ----------------------------------------------------------------------
1292    Name: Matthias Blume
1293    Date: 2002/02/15 16:00:00 EST
1294    Tag: Release_110_39-orig
1295    Description:
1296    
1297    Working release 110.39.  New bootfiles.
1298    
1299    (Update: There was a small bug in the installer so it wouldn't work
1300    with all shells.  So I retagged. -Matthias)
1301    
1302    ----------------------------------------------------------------------
1303    Name: Matthias Blume
1304    Date: 2002/02/15 14:17:00 EST
1305    Tag: blume-20020215-showbindings
1306    Description:
1307    
1308    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1309    the latter can be useful for exploring what bindings are available at
1310    the interactive prompt.  (The first function returns only the list
1311    of symbols that are really bound, the second prints those but also the
1312    ones that CM's autoloading mechanism knows about.)
1313    
1314    ----------------------------------------------------------------------
1315    Name: Matthias Blume
1316    Date: 2002/02/15 12:08:00 EST
1317    Tag: blume-20020215-iptrs
1318    Description:
1319    
1320    Two improvements to ml-nlffigen:
1321    
1322      1. Write files only if they do not exist or if their current contents
1323         do not coincide with what's being written.  (That is, avoid messing
1324         with the time stamps unless absolutely necessary.)
1325    
1326      2. Implement a "repository" mechanism for generated files related
1327         to "incomplete pointer types".   See the README file for details.
1328    
1329    ----------------------------------------------------------------------
1330    Name: Matthias Blume
1331    Date: 2002/02/14 11:50:00 EST
1332    Tag: blume-20020214-quote
1333    Description:
1334    
1335    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1336    because of the new and improved tag generation scheme.  (Thanks to Allen
1337    Leung for pointing it out.)
1338    
1339    ----------------------------------------------------------------------
1340    Name: Lal George
1341    Date: 2002/02/14 09:55:27 EST 2002
1342    Tag: george-20020214-isabelle-bug
1343    Description:
1344    
1345    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1346    of Isabelle on the x86.
1347    
1348    From Allen:
1349    -----------
1350     I've found the problem:
1351    
1352         in ra-core.sml, I use the counter "blocked" to keep track of the
1353         true number of elements in the freeze queue.  When the counter goes
1354         to zero, I skip examining the queue.  But I've messed up the
1355         bookkeeping in combine():
1356    
1357             else ();
1358             case !ucol of
1359               PSEUDO => (if !cntv > 0 then
1360                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1361                                        ^^^^^^^^^^^^^^^^^^^^^^^
1362                      moveu := mergeMoveList(!movev, !moveu)
1363                     )
1364                  else ();
1365    
1366         combine() is called to coalesce two nodes u and v.
1367         I think I was thinking that if the move counts of u and v are both
1368         greater than zero then after they are coalesced then one node is
1369         removed from the freeze queue.  Apparently I was thinking that
1370         both u and v are of low degree, but that's clearly not necessarily true.
1371    
1372    
1373    02/12/2002:
1374        Here's the patch.  HOL now compiles.
1375    
1376        I don't know how this impact on performance (compile
1377        time or runtime).  This bug caused the RA (especially on the x86)
1378        to go thru the potential spill phase when there are still nodes on the
1379        freeze queue.
1380    
1381    
1382    
1383    
1384    ----------------------------------------------------------------------
1385    Name: Matthias Blume
1386    Date: 2002/02/13 22:40:00 EST
1387    Tag: blume-20020213-fptr-rtti
1388    Description:
1389    
1390    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1391    updates.
1392    
1393    ----------------------------------------------------------------------
1394    Name: Matthias Blume
1395    Date: 2002/02/13 16:41:00 EST
1396    Tag: blume-20020213-cmlpq
1397    Description:
1398    
1399    Added new priority queue export symbols (which have just been added to
1400    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1401    not compile and the installer would choke.)
1402    
1403    ----------------------------------------------------------------------
1404    Name: Matthias Blume
1405    Date: 2002/02/13 16:15:00 EST
1406    Tag: blume-20020213-various
1407    Description:
1408    
1409    1. More tweaks to ml-nlffigen:
1410    
1411       - better internal datastructures (resulting in slight speedup)
1412       - "-match" option requires exact match
1413       - "localized" gensym counters (untagged structs/unions nested within
1414         other structs/unions or within typedefs get a fresh counter; their
1415         tag will be prefixed by a concatenation of their parents' tags)
1416       - bug fixes (related to calculation of transitive closure of types
1417         to be included in the output)
1418    
1419    2. Minor Basis updates:
1420    
1421       - added implementations for List.collate and Option.app
1422    
1423    ----------------------------------------------------------------------
1424    Name: Matthias Blume
1425    Date: 2002/02/11 15:55:00 EST
1426    Tag: blume-20020211-gensym
1427    Description:
1428    
1429    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1430    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1431    names (ML structure names that correspond to unnamed C structs, unions,
1432    and enums), so that separate runs of ml-nlffigen do not clash.
1433    
1434    ----------------------------------------------------------------------
1435    Name: Matthias Blume
1436    Date: 2002/02/11 12:05:00 EST
1437    Tag: blume-20020211-gensml
1438    Description:
1439    
1440    A quick fix for a problem with GenSML (in the pgraph-util library):
1441    Make generation of toplevel "local" optional.  (Strictly speaking,
1442    signature definitions within "local" are not legal SML.)
1443    
1444    Other than that: updates to INSTALL and cm/TODO.
1445    
1446    ----------------------------------------------------------------------
1447    Name: Matthias Blume
1448    Date: 2002/02/08 15:00:00 EST
1449    Tag: blume-20020208-uniquepid
1450    Description:
1451    
1452    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1453    
1454    1. The installer (config/install.sh) has gotten smarter:
1455    
1456         - Configuration options are a bit easier to specify now
1457           (in config/targets).
1458         - Bug in recognizing .tar.bz2 files fixed.
1459         - Installer automatically resolves dependencies between
1460           configuration options (e.g., if you ask for eXene, you will
1461           also get cml -- regardless whether you asked for it or not).
1462         - Installer can run in "quieter mode" by setting the environment
1463           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1464           "completely silent", though.
1465         - Build HashCons library as part of smlnj-lib.
1466    
1467    2. A new scheme for assigning persistent identifiers to compilation
1468       units (and, by extension, to types etc.) has been put into place.
1469       This fixes a long-standing bug where types and even dynamic values
1470       can get internally confused, thereby compromising type safety
1471       (abstraction) and dynamic correctness.  See
1472    
1473         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1474    
1475       for an example of how things could go wrong until now.
1476    
1477       The downside of the new scheme is that pids are not quite as
1478       persistent as they used to be: CM will generate a fresh pid
1479       for every compilation unit that it thinks it sees for the first
1480       time.  That means that if you compile starting from a clean, fresh
1481       source tree at two different times, you end up with different
1482       binaries.
1483    
1484       Cutoff recompilation, however, has not been compromised because
1485       CM keeps pid information in special caches between runs.
1486    
1487    ----------------------------------------------------------------------
1488    Name: Lal George
1489    Date: 2002/02/07 15:34:13 EST 2002
1490    Tag: <none>
1491    Description:
1492    
1493    Compilers that generate assembly code may produce  global labels
1494    whose value is resolved at link time. The various peephole optimization
1495    modules did not take this in account.
1496    
1497    TODO. The Labels.addrOf function should really return an option
1498    type so that clients are forced to deal with this issue, rather
1499    than an exception being raised.
1500    
1501    ----------------------------------------------------------------------
1502    Name: Lal George
1503    Date: 2002/02/06 13:55:02 EST
1504    Tag: george-20020206-ra-breakup
1505    Description:
1506    
1507    1. A bug fix from Allen.
1508    
1509        A typo causes extra fstp %st(0)'s to be generated at compensation
1510        edges, which might cause stack underflow traps at runtime.  This
1511        occurs in fft where there are extraneous fstps right before the 'into'
1512        trap instruction (in this case they are harmless since none of the
1513        integers overflow.)
1514    
1515    2. Pulled out various utility modules that were embedded in the modules
1516       of the register allocator. I need these modules for other purposes, but
1517       they are not complete enough to put into a library (just yet).
1518    ----------------------------------------------------------------------
1519    Name: Matthias Blume
1520    Date: 2002/01/31 16:05:00 EST
1521    Tag: blume-20020131-sparc-ccalls
1522    Description:
1523    
1524    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1525       of extra stack space by mistake.  Fixed.
1526    
1527    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1528    
1529    ----------------------------------------------------------------------
1530    Name: Allen Leung
1531    Date: 2002/01/30
1532    Tag: leunga-20020130-nowhere-bug-fix
1533    Description:
1534    
1535       MLRISC bug fixes:
1536       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1537       2. MachineInt.fromString was negating its value.
1538    
1539    ----------------------------------------------------------------------
1540    Name: Matthias Blume
1541    Date: 2002/01/29
1542    Tag: blume-20020129-INSTALL
1543    Description:
1544    
1545    - Added somewhat detailed installation instructions (file INSTALL).
1546    - Fixed curl-detection bug in config/install.sh.
1547    - It is now possible to select the URL getter using the URLGETTER
1548      environment variable:
1549    
1550          not set / "unknown"      --> automatic detection (script tries wget,
1551                                       curl, and lynx)
1552          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1553                                       how to properly invoke them)
1554          other                    --> use $URLGETTER directly, it must take
1555                                       precisely two command-line arguments
1556                                       (source URL and destination file name)
1557    
1558    ----------------------------------------------------------------------
1559    Name: Matthias Blume
1560    Date: 2002/01/28
1561    Tag: blume-20020128-sparc-ccalls
1562    Description:
1563    
1564    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1565    - Make use of the allocParam argument in sparc-c-calls.
1566    
1567    ----------------------------------------------------------------------
1568    Name: Matthias Blume
1569    Date: 2002/01/28
1570    Tag: blume-20020128-allocParam
1571    Description:
1572    
1573    John Reppy:  Changes c-calls API to accept client-callback for
1574    allocating extra stack space.
1575    me: Corresponding changes to mlriscGen (using a dummy argument that
1576        does not change the current behavior).
1577    
1578    ----------------------------------------------------------------------
1579    Name: Matthias Blume
1580    Date: 2002/01/28 12:00:00
1581    Tag: Release_110_38
1582    Description:
1583    
1584    This time for real!!!
1585    
1586    ----------------------------------------------------------------------
1587    Name: Matthias Blume
1588    Date: 2002/01/28 10:56:00 EST
1589    Tag: blume-20020128-retraction
1590    Description:
1591    
1592    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1593       with blume-Release_110_38-retracted.)
1594    
1595    1. Fixed a problem with incorrect rounding modes in real64.sml.
1596       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1597    
1598    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1599       and enums fixed.  The naming of corresponding ML identifiers should
1600       now be consistent again.
1601    
1602    ----------------------------------------------------------------------
1603    Name: Allen Leung
1604    Date: 2002/01/27
1605    Tag: leunga-20020127-nowhere
1606    Description:
1607    
1608       Added a target called nowhere in the configuration scripts.
1609       Enabling this will build the MLRISC 'nowhere' tool (for translating
1610       programs with where-clauses into legal SML code) during installation.
1611    
1612    ----------------------------------------------------------------------
1613    Name: Matthias Blume
1614    Date: 2002/01/25 21:27:00 EST
1615    Tag: blume-Release_110_38-retracted
1616    Description:
1617    
1618    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1619    
1620    README will be added later.
1621    
1622    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1623    (Release_110_38) removed.  Reason: Last-minute bug fixes.
1624    
1625    ----------------------------------------------------------------------
1626    Name: Matthias Blume
1627    Date: 2002/01/25
1628    Tag: blume-20020125-ffi
1629    Description:
1630    
1631    A large number of tweaks and improvements to ml-nlffi-lib and
1632    ml-nlffigen:
1633    
1634       - ML represenation types have been streamlined
1635       - getter and setter functions work with concrete values, not abstract
1636         ones where possible
1637       - ml-nlffigen command line more flexible (see README file there)
1638       - some bugs have been fixed (hopefully)
1639    
1640    ----------------------------------------------------------------------
1641  Name: Lal George  Name: Lal George
1642  Date: 2002/01/24  Date: 2002/01/24
1643  Tag: george-20020124-risc-ra-interface  Tag: george-20020124-risc-ra-interface
# Line 511  Line 2136 
2136    
2137  ----------------------------------------------------------------------  ----------------------------------------------------------------------
2138  Name: Matthias Blume  Name: Matthias Blume
 >>>>>>> 1.169  
2139  Date: 2001/09/18 15:35:00 EDT  Date: 2001/09/18 15:35:00 EDT
2140  Tag: blume-20010918-readme11036  Tag: blume-20010918-readme11036
2141  Description:  Description:

Legend:
Removed from v.1034  
changed lines
  Added in v.1240

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