Home My Page Projects Code Snippets Project Openings SML/NJ
Summary Activity Forums Tracker Lists Tasks Docs Surveys News SCM Files

SCM Repository

[smlnj] Diff of /sml/trunk/HISTORY
ViewVC logotype

Diff of /sml/trunk/HISTORY

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

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

Legend:
Removed from v.1053  
changed lines
  Added in v.1222

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