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 1032, Tue Jan 22 16:59:16 2002 UTC revision 1220, Wed May 22 15:04:10 2002 UTC
# Line 14  Line 14 
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16  Name: Matthias Blume  Name: Matthias Blume
17    Date: 2002/05/22 11:03:00 EDT
18    Tag: blume-20020522-shrink
19    Description:
20    
21    John tweaked yesterday's fix for 1131 to handle an out-of-memory
22    situation that comes up when allocating huge arrays.
23    
24    ----------------------------------------------------------------------
25    Name: Matthias Blume
26    Date: 2002/05/21 16:00:00 EDT
27    Tag: Release_110_40
28    Description:
29    
30    New working release (110.40).  New bootfiles.
31    
32    [Also: John Reppy fixed GC bug 1131.]
33    
34    ----------------------------------------------------------------------
35    Name: Matthias Blume
36    Date: 2002/05/21 12:35:00 EDT
37    Tag: blume-20020521-cmdoc
38    Description:
39    
40    CM documentation update.
41    
42    ----------------------------------------------------------------------
43    Name: Matthias Blume
44    Date: 2002/05/21 10:55:00 EDT
45    Tag: blume-20020521-misc
46    Description:
47    
48    - John tweaked runtime to be silent on heap export (except when
49      GC messages are on).
50    - I added a few more things (cross-compiling versions of CMB) to
51      config/preloads (as suggestions).
52    
53    ----------------------------------------------------------------------
54    Name: Matthias Blume
55    Date: 2002/05/20 22:25:00 EDT
56    Tag: blume-20020520-controls
57    Description:
58    
59    - Added ControlUtil structure to control-lib.cm.
60    - Use it throughout.
61    - Used Controls facility to define MLRISC controls (as opposed to
62      registering MLRISC control ref cells with Controls after the fact)
63    - Fixed messed-up controls priorities.
64    
65    * Removed again all the stuff from config/preloads that one wouldn't
66      be able to preload at the time the initial heap image is built.
67      (Many libraries, e.g., CML, do not exist yet at this time.  The
68       only libraries that can be preloaded via config/preloads are those
69       that come bundled with the bootfiles.)
70    
71    ----------------------------------------------------------------------
72    Name: Matthias Blume
73    Date: 2002/05/20 10:59:00 EDT
74    Tag: blume-20020520-preloads
75    Description:
76    
77    Added a lot of commented-out suggestions for things to be included
78    in config/preloads.
79    
80    ----------------------------------------------------------------------
81    Name: Allen Leung
82    Date: 2002/05/18 14:20:00 EDT
83    Tag: leunga-20020518-mdl
84    Description:
85    
86    o Made the mdl tool stuff compile and run again.
87    o I've disabled all the stuff that depends on RTL specifications; they
88      are all badly broken anyway.
89    
90    ----------------------------------------------------------------------
91    Name: Matthias Blume
92    Date: 2002/05/17 16:49:00 EDT
93    Tag: blume-20020517-controls
94    Description:
95    
96    0. John Reppy made several modifications to the SML/NJ library.
97       In particular, there is a shiny new controls-lib.cm.
98    
99    1. Pushed new controls interface through compiler so that everything
100       compiles again.
101    
102    2. Added FormatComb and FORMAT_COMB to the CML version of the
103       SML/NJ library (so that CML compiles again).
104    
105    3. Modified init scripts because XXX_DEFAULT environment variables
106       are no longer with us.  (Boot-time initialization is now done
107       using the same environment variables that are also used for
108       startup-time initialization of controls.)
109    
110    ----------------------------------------------------------------------
111    Name: Lal George
112    Date: 2002/05/15 09:20:10 EDT
113    Tag: george-20020515-pseudo-op-decls
114    Description:
115    
116            All pseudo-ops emitted before the first segment declaration
117    such as TEXT, DATA, and BSS directives are assumed to be global
118    declarations and are emitted first in the assembly file. This is
119    useful in a number of situations where one has pseudo-ops that are not
120    specific to any segment, and also works around the constraint that one
121    cannot have client pseudo-ops in the TEXT segment.
122    
123            Because no segment is associated with these declarations it is
124    an error to allocate any space or objects before the first segment
125    directive and an exception will be raised. However, we cannot make
126    this check for client pseudo-ops.
127    
128            These top level declarations are a field in the CFG graph_info.
129    In theory you can continue to add to this field after the CFG has been
130    built -- provided you know what you are doing;-)
131    
132    ----------------------------------------------------------------------
133    Name: Matthias Blume
134    Date: 2002/05/13 16:40:00 EDT
135    Tag: blume-20020513-pp-etc
136    Description:
137    
138    A few minor bugfixes:
139    
140      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
141        (Bogus printouts for redefined bindings still occur.  Compiler
142        bug should no longer occur now.  We need to redo the prettyprinter
143        from scratch.)
144    
145      - CM pathname printer now also adds escape sequences for ( and )
146    
147      - commend and docu fixes for ml-nlffi
148    
149    ----------------------------------------------------------------------
150    Name: Matthias Blume
151    Date: 2002/05/10 16:40:00 EDT
152    Tag: blume-20020510-erg-textio
153    Description:
154    
155    Applied the following bugfix provided by Emden Gansner:
156    
157        Output is corrupted when outputSubstr is used rather than output.
158    
159        The problem occurs when a substring
160    
161            ss = (s, dataStart, dataLen)
162    
163        where dataStart > 0, fills a stream buffer with avail bytes left.
164        avail bytes of s, starting at index dataStart, are copied into the
165        buffer, the buffer is flushed, and then the remaining dataLen-avail
166        bytes of ss are copied into the beginning of the buffer. Instead of
167        starting this copy at index dataStart+avail in s, the current code
168        starts the copy at index avail.
169    
170        Fix:
171        In text-io-fn.sml, change line 695 from
172             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
173        to
174             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
175    
176    ----------------------------------------------------------------------
177    Name: Matthias Blume
178    Date: 2002/04/12 13:55:00 EDT
179    Tag: blume-20020412-assyntax
180    Description:
181    
182    1. Grabbed newer assyntax.h from the XFree86 project.
183    2. Fiddled with how to compile X86.prim.asm without warnings.
184    3. (Very) Minor cleanup in CM.
185    
186    ----------------------------------------------------------------------
187    Name: Matthias Blume
188    Date: 2002/04/01 (no joke!) 17:07:00 EST
189    Tag: blume-20020401-x86div
190    Description:
191    
192    Added full support for div/mod/rem/quot on the x86, using the machine
193    instruction's two results (without clumsily recomputing the remainder)
194    directly where appropriate.
195    
196    Some more extensive power-of-two support was added to the x86 instruction
197    selector (avoiding expensive divs, mods, and muls where they can be
198    replaced with cheaper shifts and masks).  However, this sort of thing
199    ought to be done earlier, e.g., within the CPS optimizer so that
200    all architectures benefit from it.
201    
202    The compiler compiles to a fixed point, but changes might be somewhat
203    fragile nevertheless.  Please, report any strange things that you might
204    see wrt. div/mod/quot/rem...
205    
206    ----------------------------------------------------------------------
207    Name: Matthias Blume
208    Date: 2002/03/29 17:22:00
209    Tag: blume-20020329-div
210    Description:
211    
212    Fixed my broken div/mod logic.  Unfortunately, this means that the
213    inline code for div/mod now has one more comparison than before.
214    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
215    The problem was with quotient = 0, because that alone does not tell
216    us which way the rounding went.  One then has to look at whether
217    remainder and divisor have the same sign...  :(
218    
219    Anyway, I replaced the bootfiles with fresh ones...
220    
221    ----------------------------------------------------------------------
222    Name: Matthias Blume
223    Date: 2002/03/29 14:10:00 EST
224    Tag: blume-20020329-inlprims
225    Description:
226    
227    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
228    
229    Primops have changed. This means that the bin/boot-file formats have
230    changed as well.
231    
232    To make sure that there is no confusion, I made a new version.
233    
234    
235    CHANGES:
236    
237    * removed REMT from mltree (remainder should never overflow).
238    
239    * added primops to deal with divisions of all flavors to the frontend
240    
241    * handled these primops all the way through so they map to their respective
242      MLRISC support
243    
244    * used these primops in the implementation of Int, Int32, Word, Word32
245    
246    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
247    
248    * parameterized INLMIN, INLMAX, and INLABS by a numkind
249    
250    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
251      floating point
252    
253    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
254      and Real (but Real.abs maps to a separate floating-point-only primop)
255    
256    
257    TODO items:
258    
259    * Hacked Alpha32 instruction selection, disabling the selection of REMx
260      instructions because the machine instruction encoder cannot handle
261      them.  (Hppa, PPC, and Sparc instruction selection did not handle
262      REM in the first place, and REM is supported by the x86 machine coder.)
263    
264    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
265      selection phase.  (The two can be streamlined because the hardware
266      delivers both quotient and remainder at the same time anyway.)
267    
268    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
269      (Currently the behavior is inconsistent both across architectures and
270      wrt. the draft Basis spec.)
271    
272    * Word8 should eventually be handled natively, too.
273    
274    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
275      as if there currently is no execution path that could trigger it in
276      SML/NJ.  (The assumptions underlying functions arith and promotable do not
277      hold for things like multiplication and division.)
278    
279    ----------------------------------------------------------------------
280    Name: Matthias Blume
281    Date: 2002/03/27 16:27:00 EST
282    Tag: blume-20020327-mlrisc-divisions
283    Description:
284    
285    Added support for all four division operations (ML's div, mod, quot,
286    and rem) to MLRISC.  In the course of doing so, I also rationalized
287    the naming (no more annoying switch-around of DIV and QUOT), by
288    parameterizing the operation by div_rounding_mode (which can be either
289    DIV_TO_ZERO or DIV_TO_NEGINF).
290    
291    The generic MLTreeGen functor takes care of compiling all four
292    operations down to only round-to-zero div.
293    
294    Missing pieces:
295    
296      * Doing something smarter than relying on MLTreeGen on architectures
297        like, e.g., the x86 where hardware division delivers both quotient and
298        remainder at the same time.  With this, the implementation of the
299        round-to-neginf operations could be further streamlined.
300    
301      * Remove inlining support for div/mod/rem from the frontend and replace it
302        with primops that get carried through to the backend.  Do this for all
303        int and word types.
304    
305    ----------------------------------------------------------------------
306    Name: Matthias Blume
307    Date: 2002/03/25 17:25:00 EST
308    Tag: blume-20020325-divmod
309    Description:
310    
311    I improved (hopefully without breaking them) the implementation of Int.div,
312    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
313    advantage of the following observations:
314    
315      Let  q = x quot y      r = x rem y
316           d = x div  y      m = x mod y
317    
318    where "quot" is the round-to-zero version of integer division that
319    hardware usually provides.  Then we have:
320    
321         r = x - q * y        where neither the * nor the - will overflow
322         d = if q >= 0 orelse x = q * y then q else q - 1
323                              where neither the * nor the - will overflow
324         m = if q >= 0 orelse r = 0 then r else r + y
325                              where the + will not overflow
326    
327    This results in substantial simplification of the generated code.
328    The following table shows the number of CFG nodes and edges generated
329    for
330            fun f (x, y) = x OPER y
331            (* with OPER \in div, mod, quot, rem *)
332    
333    
334        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
335        --------------------------------------------------------
336         div |         24 |         39 |         12 |         16
337         mod |         41 |         71 |         12 |         16
338        quot |          8 |         10 |          8 |         10
339         rem |         10 |         14 |          8 |         10
340    
341    
342    ----------------------------------------------------------------------
343    Name: Matthias Blume
344    Date: 2002/03/25 22:06:00 EST
345    Tag: blume-20020325-cprotobug
346    Description:
347    
348    Fixed a bug in cproto (c prototype decoder).
349    
350    ----------------------------------------------------------------------
351    Name: Matthias Blume
352    Date: 2002/03/25 16:00:00 EST
353    Tag: blume-20020325-raw-primops
354    Description:
355    
356    I did some cleanup to Allen's new primop code and
357    replaced yesterday's bootfiles with new ones.
358    (But they are stored in the same place.)
359    
360    ----------------------------------------------------------------------
361    Name: Matthias Blume
362    Date: 2002/03/24 22:40:00 EST
363    Tag: blume-20020324-bootfiles
364    Description:
365    
366    Made the bootfiles that Allen asked for.
367    
368    ----------------------------------------------------------------------
369    Name: Allen Leung
370    Date: 2002/03/23 15:50:00 EST
371    Tag: leunga-20020323-flint-cps-rcc-primops
372    Description:
373    
374      1. Changes to FLINT primops:
375    
376        (* make a call to a C-function;
377         * The primop carries C function prototype information and specifies
378         * which of its (ML-) arguments are floating point. C prototype
379         * information is for use by the backend, ML information is for
380         * use by the CPS converter. *)
381      | RAW_CCALL of { c_proto: CTypes.c_proto,
382                       ml_args: ccall_type list,
383                       ml_res_opt: ccall_type option,
384                       reentrant : bool
385                     } option
386       (* Allocate uninitialized storage on the heap.
387        * The record is meant to hold short-lived C objects, i.e., they
388        * are not ML pointers.  With the tag, the representation is
389        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
390        *)
391      | RAW_RECORD of {tag:bool,sz:int}
392      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
393    
394      2.  These CPS primops are now overloaded:
395    
396           rawload of {kind:numkind}
397           rawstore of {kind:numkind}
398    
399          The one argument form is:
400    
401             rawload {kind} address
402    
403          The two argument form is:
404    
405             rawload {kind} [ml object, byte-offset]
406    
407      3. RAW_CCALL/RCC now takes two extra arguments:
408    
409         a. The first is whether the C call is reentrant, i.e., whether
410            ML state should be saved and restored.
411         b. The second argument is a string argument specifying the name of
412            library and the C function.
413    
414         These things are currently not handled in the code generator, yet.
415    
416      4. In CProto,
417    
418         An encoding type of "bool" means "ml object" and is mapped into
419         C prototype of PTR.  Note that "bool" is different than "string",
420         even though "string" is also mapped into PTR, because "bool"
421         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
422    
423      5. Pickler/unpicker
424    
425         Changed to handle RAW_RECORD and newest RAW_CCALL
426    
427      6. MLRiscGen,
428    
429         1. Changed to handle the new rawload/rawstore/rawrecord operators.
430         2. Code for handling C Calls has been moved to a new module CPSCCalls,
431            in the file CodeGen/cpscompile/cps-c-calls.sml
432    
433      7. Added the conditional move operator
434    
435             condmove of branch
436    
437         to cps.  Generation of this is still buggy so it is currently
438         disabled.
439    
440    ----------------------------------------------------------------------
441    Name: Lal George
442    Date: 2002/03/22 14:18:25 EST
443    Tag: george-20020322-cps-branch-prob
444    Description:
445    
446    Implemented the Ball-Larus branch prediction-heuristics, and
447    incorporated graphical viewers for control flow graphs.
448    
449    Ball-Larus Heuristics:
450    ---------------------
451    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
452    
453    By design it uses the Dempster-Shafer theory for combining
454    probabilities.  For example, in the function:
455    
456        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
457    
458    the ball-larus heuristics predicts that the n=0 is unlikely
459    (OH-heuristic), and the 'then' branch is unlikely because of the
460    RH-heuristic -- giving the 'then' branch an even lower combined
461    probability using the Dempster-Shafer theory.
462    
463    Finally, John Reppy's loop analysis in MLRISC, further lowers the
464    probability of the 'then' branch because of the loop in the else
465    branch.
466    
467    
468    Graphical Viewing:
469    ------------------
470    I merely plugged in Allen's graphical viewers into the compiler. The
471    additional code is not much. At the top level, saying:
472    
473            Control.MLRISC.getFlag "cfg-graphical-view" := true;
474    
475    will display the graphical view of the control flow graph just before
476    back-patching.  daVinci must be in your path for this to work. If
477    daVinci is not available, then the default viewer can be changed
478    using:
479    
480            Control.MLRISC.getString "viewer"
481    
482    which can be set to "dot" or "vcg" for the corresponding viewers. Of
483    course, these viewers must be in your path.
484    
485    The above will display the compilation unit at the level of clusters,
486    many of which are small, boring, and un-interesting. Also setting:
487    
488            Control.MLRISC.getInt "cfg-graphical-view_size"
489    
490    will display clusters that are larger than the value set by the above.
491    
492    
493    ----------------------------------------------------------------------
494    Name: Matthias Blume
495    Date: 2002/03/21 22:20:00 EST
496    Tag: blume-20020321-kmp-bugfix
497    Description:
498    
499    Changed the interface to the KMP routine in PreString and fixed
500    a minor bug in one place where it was used.
501    
502    ----------------------------------------------------------------------
503    Name: Allen Leung
504    Date: 2002/03/21 20:30:00 EST
505    Tag: leunga-20020321-cfg
506    Description:
507    
508      Fixed a potential problem in cfg edge splitting.
509    
510    ----------------------------------------------------------------------
511    Name: Allen Leung
512    Date: 2002/03/21 17:15:00 EST
513    Tag: leunga-20020321-x86-fp-cfg
514    Description:
515    
516      1. Recoded the buggy parts of x86-fp.
517    
518         a. All the block reordering code has been removed.
519            We now depend on the block placement phases to do this work.
520    
521         b. Critical edge splitting code has been simplified and moved into the
522            CFG modules, as where they belong.
523    
524         Both of these were quite buggy and complex.  The code is now much, much
525         simpler.
526    
527      2. X86 backend.
528    
529         a. Added instructions for 64-bit support.  Instruction selection for
530            64-bit has not been committed, however, since that
531            requires changes to MLTREE which haven't been approved by
532            Lal and John.
533    
534         b. Added support for FUCOMI and FUCOMIP when generating code for
535            PentiumPro and above.  We only generate these instructions in
536            the fast-fp mode.
537    
538         c. Added cases for JP and JNP in X86FreqProps.
539    
540      3. CFG
541    
542         CFG now has a bunch of methods for edge splitting and merging.
543    
544      4. Machine description.
545    
546         John's simplification of MLTREE_BASIS.fcond broke a few machine
547         description things:
548    
549         rtl-build.{sig,sml} and hppa.mdl fixed.
550    
551         NOTE: the machine description stuff in the repository is still broken.
552               Again, I can't put my fixes in because that involves
553               changes to MLTREE.
554    
555    ----------------------------------------------------------------------
556    Name: Matthias Blume
557    Date: 2002/03/20 15:55:00 EST
558    Tag: blume-20020320-kmp
559    Description:
560    
561    Implemented Knuth-Morris-Pratt string matching in PreString and used
562    it for String.isSubstring, Substring.isSubstring, and
563    Substring.position.
564    
565    (Might need some stress-testing.  Simple examples worked fine.)
566    
567    ----------------------------------------------------------------------
568    Name: Matthias Blume
569    Date: 2002/03/19 16:37:00 EST
570    Tag: blume-20020319-witnesses
571    Description:
572    
573    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
574    
575    This implements a generic mechanism for changing constness qualifiers
576    anywhere within big C types without resorting to outright "casts".
577    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
578    modify the constness at the outermost level.)
579    The implementation of "convert" is based on the idea of "witness"
580    values -- values that are not used by the operation but whose types
581    "testify" to their applicability.  On the implementation side, "convert"
582    is simply a projection (returning its second curried argument).  With
583    cross-module inlining, it should not result in any machine code being
584    generated.
585    
586    ----------------------------------------------------------------------
587    Name: Matthias Blume
588    Date: 2002/03/15 16:40:00 EST
589    Tag: blume-20020315-basis
590    Description:
591    
592    Provided (preliminary?) implementations for
593    
594      {String,Substring}.{concatWith,isSuffix,isSubstring}
595    
596    and
597    
598      Substring.full
599    
600    Those are in the Basis spec but they were missing in SML/NJ.
601    
602    ----------------------------------------------------------------------
603    Name: Matthias Blume
604    Date: 2002/03/14 21:30:00 EST
605    Tag: blume-20020314-controls
606    Description:
607    
608    Controls:
609    ---------
610    
611    1. Factored out the recently-added Controls : CONTROLS stuff and put
612       it into its own library $/controls-lib.cm.  The source tree for
613       this is under src/smlnj-lib/Controls.
614    
615    2. Changed the names of types and functions in this interface, so they
616       make a bit more "sense":
617    
618          module -> registry
619          'a registry -> 'a group
620    
621    3. The interface now deals in ref cells only.  The getter/setter interface
622       is (mostly) gone.
623    
624    4. Added a function that lets one register an already-existing ref cell.
625    
626    5. Made the corresponding modifications to the rest of the code so that
627       everything compiles again.
628    
629    6. Changed the implementation of Controls.MLRISC back to something closer
630       to the original.  In particular, this module (and therefore MLRISC)
631       does not depend on Controls.  There now is some link-time code in
632       int-sys.sml that registers the MLRISC controls with the Controls
633       module.
634    
635    CM:
636    ---
637    
638      * One can now specify the lambda-split aggressiveness in init.cmi.
639    
640    ----------------------------------------------------------------------
641    Name: Allen Leung
642    Date: 2002/03/13 17:30:00 EST
643    Tag: leunga-20020313-x86-fp-unary
644    Description:
645    
646    Bug fix for:
647    
648    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
649    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
650    > - fun f(x,(y,z)) = Real.~ y;
651    > [autoloading]
652    > [autoloading done]
653    >       fchsl   (%eax), 184(%esp)
654    > Error: MLRisc bug: X86MCEmitter.emitInstr
655    >
656    > uncaught exception Error
657    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
658    
659    The problem was that the code generator did not generate any fp registers
660    in this case, and the ra didn't know that it needed to run the X86FP phase to
661    translate the pseudo fp instruction.   This only happened with unary fp
662    operators in certain situations.
663    
664    ----------------------------------------------------------------------
665    Name: Matthias Blume
666    Date: 2002/03/13 14:00:00 EST
667    Tag: blume-20020313-overload-etc
668    Description:
669    
670    1. Added _overload as a synonym for overload for backward compatibility.
671       (Control.overloadKW must be true for either version to be accepted.)
672    
673    2. Fixed bug in install script that caused more things to be installed
674       than what was requested in config/targets.
675    
676    3. Made CM aware of the (_)overload construct so that autoloading
677       works.
678    
679    ----------------------------------------------------------------------
680    Name: Matthias Blume
681    Date: 2002/03/12 22:03:00 EST
682    Tag: blume-20020312-url
683    Description:
684    
685    Forgot to update BOOT and srcarchiveurl.
686    
687    ----------------------------------------------------------------------
688    Name: Matthias Blume
689    Date: 2002/03/12 17:30:00 EST
690    Tag: blume-20020312-version110392
691    Description:
692    
693    Yet another version number bump (because of small changes to the
694    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
695    
696    Changes:
697    
698      The new pid generation scheme described a few weeks ago was overly
699      complicated.  I implemented a new mechanism that is simpler and
700      provides a bit more "stability":  Once CM has seen a compilation
701      unit, it keeps its identity constant (as long as you do not delete
702      those crucial CM/GUID/* files).  This means that when you change
703      an interface, compile, then go back to the old interface, and
704      compile again, you arrive at the original pid.
705    
706      There now also is a mechanism that instructs CM to use the plain
707      environment hash as a module's pid (effectively making its GUID
708      the empty string).  For this, "noguid" must be specified as an
709      option to the .sml file in question within its .cm file.
710      This is most useful for code that is being generated by tools such
711      as ml-nlffigen (because during development programmers tend to
712      erase the tool's entire output directory tree including CM's cached
713      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
714      revert to the old, broken behavior of SML/NJ, but in specific cases
715      where there is no danger of interface confusion, its use is ok
716      (I think).
717    
718      ml-nlffigen by default generates "noguid" annotations.  They can be
719      turned off by specifying -guid in its command line.
720    
721    ----------------------------------------------------------------------
722    Name: Lal George
723    Date: 2002/03/12 12 14:42:36 EST
724    Tag: george-20020312-frequency-computation
725    Description:
726    
727    Integrated jump chaining and static block frequency into the
728    compiler. More details and numbers later.
729    
730    ----------------------------------------------------------------------
731    Name: Lal George
732    Date: 2002/03/11 11 22:38:53 EST
733    Tag: george-20020311-jump-chain-elim
734    Description:
735    
736    Tested the jump chain elimination on all architectures (except the
737    hppa).  This is on by default right now and is profitable for the
738    alpha and x86, however, it may not be profitable for the sparc and ppc
739    when compiling the compiler.
740    
741    The gc test will typically jump to a label at the end of the cluster,
742    where there is another jump to an external cluster containing the actual
743    code to invoke gc. This is to allow factoring of common gc invocation
744    sequences. That is to say, we generate:
745    
746            f:
747               testgc
748               ja   L1      % jump if above to L1
749    
750            L1:
751               jmp L2
752    
753    
754    After jump chain elimination the 'ja L1' instructions is converted to
755    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
756    up being implemented in their long form (if L2 is far away) using:
757    
758            jbe     L3      % jump if below or equal to L3
759            jmp     L2
760         L3:
761            ...
762    
763    
764    For large compilation units L2  may be far away.
765    
766    
767    ----------------------------------------------------------------------
768    Name: Matthias Blume
769    Date: 2002/03/11 13:30:00 EST
770    Tag: blume-20020311-mltreeeval
771    Description:
772    
773    A functor parameter was missing.
774    
775    ----------------------------------------------------------------------
776    Name: Allen Leung
777    Date: 2002/03/11 10:30:00 EST
778    Tag: leunga-20020311-runtime-string0
779    Description:
780    
781       The representation of the empty string now points to a
782    legal null terminated C string instead of unit.  It is now possible
783    to convert an ML string into C string with InlineT.CharVector.getData.
784    This compiles into one single machine instruction.
785    
786    ----------------------------------------------------------------------
787    Name: Allen Leung
788    Date: 2002/03/10 23:55:00 EST
789    Tag: leunga-20020310-x86-call
790    Description:
791    
792       Added machine generation for CALL instruction (relative displacement mode)
793    
794    ----------------------------------------------------------------------
795    Name: Matthias Blume
796    Date: 2002/03/08 16:05:00
797    Tag: blume-20020308-entrypoints
798    Description:
799    
800    Version number bumped to 110.39.1.  NEW BOOTFILES!
801    
802    Entrypoints: non-zero offset into a code object where execution should begin.
803    
804    - Added the notion of an entrypoint to CodeObj.
805    - Added reading/writing of entrypoint info to Binfile.
806    - Made runtime system bootloader aware of entrypoints.
807    - Use the address of the label of the first function given to mlriscGen
808      as the entrypoint.  This address is currently always 0, but it will
809      not be 0 once we turn on block placement.
810    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
811      with entry points) from mlriscGen.
812    
813    ----------------------------------------------------------------------
814    Name: Allen Leung
815    Date: 2002/03/07 20:45:00 EST
816    Tag: leunga-20020307-x86-cmov
817    Description:
818    
819       Bug fixes for CMOVcc on x86.
820    
821       1. Added machine code generation for CMOVcc
822       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
823       3. CMOVcc cannot have an immediate operand as argument.
824    
825    ----------------------------------------------------------------------
826    Name: Matthias Blume
827    Date: 2002/03/07 16:15:00 EST
828    Tag: blume-20020307-controls
829    Description:
830    
831    This is a very large but mostly boring patch which makes (almost)
832    every tuneable compiler knob (i.e., pretty much everything under
833    Control.* plus a few other things) configurable via both the command
834    line and environment variables in the style CM did its configuration
835    until now.
836    
837    Try starting sml with '-h' (or, if you are brave, '-H')
838    
839    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
840    implements the underlying generic mechanism.
841    
842    The interface to some of the existing such facilities has changed somewhat.
843    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
844    (The getFoo interface is still there for backward-compatibility, but its
845    use is deprecated.)
846    
847    The ml-build script passes -Cxxx=yyy command-line arguments through so
848    that one can now twiddle the compiler settings when using this "batch"
849    compiler.
850    
851    TODO items:
852    
853    We should go through and throw out all controls that are no longer
854    connected to anything.  Moreover, we should go through and provide
855    meaningful (and correct!) documentation strings for those controls
856    that still are connected.
857    
858    Currently, multiple calls to Controls.new are accepted (only the first
859    has any effect).  Eventually we should make sure that every control
860    is being made (via Controls.new) exactly once.  Future access can then
861    be done using Controls.acc.
862    
863    Finally, it would probably be a good idea to use the getter-setter
864    interface to controls rather than ref cells.  For the time being, both
865    styles are provided by the Controls module, but getter-setter pairs are
866    better if thread-safety is of any concern because they can be wrapped.
867    
868    *****************************************
869    
870    One bug fix: The function blockPlacement in three of the MLRISC
871    backpatch files used to be hard-wired to one of two possibilities at
872    link time (according to the value of the placementFlag).  But (I
873    think) it should rather sense the flag every time.
874    
875    *****************************************
876    
877    Other assorted changes (by other people who did not supply a HISTORY entry):
878    
879    1. the cross-module inliner now works much better (Monnier)
880    2. representation of weights, frequencies, and probabilities in MLRISC
881       changed in preparation of using those for weighted block placement
882       (Reppy, George)
883    
884    ----------------------------------------------------------------------
885    Name: Lal George
886    Date: 2002/03/07 14:44:24 EST 2002
887    Tag: george-20020307-weighted-block-placement
888    
889    Tested the weighted block placement optimization on all architectures
890    (except the hppa) using AMPL to generate the block and edge frequencies.
891    Changes were required in the machine properties to correctly
892    categorize trap instructions. There is an MLRISC flag
893    "weighted-block-placement" that can be used to enable weighted block
894    placement, but this will be ineffective without block/edge
895    frequencies (coming soon).
896    
897    
898    ----------------------------------------------------------------------
899    Name: Lal George
900    Date: 2002/03/05 17:24:48 EST
901    Tag: george-20020305-linkage-cluster
902    
903    In order to support the block placement optimization, a new cluster
904    is generated as the very first cluster (called the linkage cluster).
905    It contains a single jump to the 'real' entry point for the compilation
906    unit. Block placement has no effect on the linkage cluster itself, but
907    all the other clusters  have full freedom in the manner in which they
908    reorder blocks or functions.
909    
910    On the x86 the typical linkage code that is generated is:
911       ----------------------
912            .align 2
913       L0:
914            addl    $L1-L0, 72(%esp)
915            jmp     L1
916    
917    
918            .align  2
919       L1:
920       ----------------------
921    
922    72(%esp) is the memory location for the stdlink register. This
923    must contain the address of the CPS function being called. In the
924    above example, it contains the address of  L0; before
925    calling L1 (the real entry point for the compilation unit), it
926    must contain the address for L1, and hence
927    
928            addl $L1-L0, 72(%esp)
929    
930    I have tested this on all architectures except the hppa.The increase
931    in code size is of course negligible
932    
933    ----------------------------------------------------------------------
934    Name: Allen Leung
935    Date: 2002/03/03 13:20:00 EST
936    Tag: leunga-20020303-mlrisc-tools
937    
938      Added #[ ... ] expressions to mlrisc tools
939    
940    ----------------------------------------------------------------------
941    Name: Matthias Blume
942    Date: 2002/02/27 12:29:00 EST
943    Tag: blume-20020227-cdebug
944    Description:
945    
946    - made types in structure C and C_Debug to be equal
947    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
948    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
949    
950    ----------------------------------------------------------------------
951    Name: Matthias Blume
952    Date: 2002/02/26 12:00:00 EST
953    Tag: blume-20020226-ffi
954    Description:
955    
956    1. Fixed a minor bug in CM's "noweb" tool:
957       If numbering is turned off, then truly don't number (i.e., do not
958       supply the -L option to noweb).  The previous behavior was to supply
959       -L'' -- which caused noweb to use the "default" line numbering scheme.
960       Thanks to Chris Richards for pointing this out (and supplying the fix).
961    
962    2. Once again, I reworked some aspects of the FFI:
963    
964       A. The incomplete/complete type business:
965    
966       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
967         gone!
968       - ML types representing an incomplete type are now *equal* to
969         ML types representing their corresponding complete types (just like
970         in C).  This is still safe because ml-nlffigen will not generate
971         RTTI for incomplete types, nor will it generate functions that
972         require access to such RTTI.   But when ML code generated from both
973         incomplete and complete versions of the C type meet, the ML types
974         are trivially interoperable.
975    
976         NOTE:  These changes restore the full generality of the translation
977         (which was previously lost when I eliminated functorization)!
978    
979       B. Enum types:
980    
981       - Structure C now has a type constructor "enum" that is similar to
982         how the "su" constructor works.  However, "enum" is not a phantom
983         type because each "T enum" has values (and is isomorphic to
984         MLRep.Signed.int).
985       - There are generic access operations for enum objects (using
986         MLRep.Signed.int).
987       - ml-nlffigen will generate a structure E_foo for each "enum foo".
988         * The structure contains the definition of type "mlrep" (the ML-side
989         representation type of the enum).  Normally, mlrep is the same
990         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
991         then mlrep will be defined as a datatype -- thus facilitating
992         pattern matching on mlrep values.
993         ("-ec" will be suppressed if there are duplicate values in an
994          enumeration.)
995         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
996         will be generated for each C enum constant xxx.
997         * Conversion functions m2i and i2m convert between mlrep and
998         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
999         * Coversion functions c and ml convert between mlrep and "tag enum".
1000         * Access functions (get/set) fetch and store mlrep values.
1001       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1002         enumerations are merged into one single enumeration represented by
1003         structure E_'.
1004    
1005    ----------------------------------------------------------------------
1006    Name: Allen Leung
1007    Date: 2002/02/25 04:45:00 EST
1008    Tag: leunga-20020225-cps-spill
1009    
1010    This is a new implementation of the CPS spill phase.
1011    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1012    In case of problems, replace it with the old file spill.sml
1013    
1014    The current compiler runs into some serious performance problems when
1015    constructing a large record.  This can happen when we try to compile a
1016    structure with many items.  Even a very simple structure like the following
1017    makes the compiler slow down.
1018    
1019        structure Foo = struct
1020           val x_1 = 0w1 : Word32.int
1021           val x_2 = 0w2 : Word32.int
1022           val x_3 = 0w3 : Word32.int
1023           ...
1024           val x_N = 0wN : Word32.int
1025        end
1026    
1027    The following table shows the compile time, from N=1000 to N=4000,
1028    with the old compiler:
1029    
1030    N
1031    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1032           MLRISC ra                               0.06u  0.00s  0.05g
1033              (spills = 0 reloads = 0)
1034           TOTAL                                   0.63u  0.07s  0.21g
1035    
1036    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1037           MLRISC ra                               5.68u  0.59s  3.93g
1038              (spills = 0 reloads = 0)
1039           TOTAL                                   14.71u  0.99s  4.81g
1040    
1041    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1042           MLRISC ra                               5.54u  0.65s  3.91g
1043              (spills = 543 reloads = 1082)
1044           TOTAL                                   65.40u  3.13s  6.00g
1045    
1046    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1047           MLRISC ra                               0.80u  0.10s  0.55g
1048              (spills = 42 reloads = 84)
1049           TOTAL                                   129.42u  5.10s  4.13g
1050    
1051    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1052           MLRISC ra                               2.69u  0.27s  1.38g
1053              (spills = 62 reloads = 124)
1054           TOTAL                                   682.48u  19.61s  13.99g
1055    
1056    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1057           MLRISC ra                               4.96u  0.27s  2.72g
1058              (spills = 85 reloads = 170)
1059           TOTAL                                   2375.26u  57.21s  48.00g
1060    
1061    As you can see the old cps spill module suffers from some serious
1062    performance problem.  But since I cannot decipher the old code fully,
1063    instead of patching the problems up, I'm reimplementing it
1064    with a different algorithm.  The new code is more modular,
1065    smaller when compiled, and substantially faster
1066    (O(n log n) time and O(n) space).  Timing of the new spill module:
1067    
1068    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1069          MLRISC ra                               0.25u  0.02s  0.15g
1070             (spills=1 reloads=3)
1071          TOTAL                                   7.74u  0.34s  1.62g
1072    
1073    Implementation details:
1074    
1075    As far as I can tell, the purpose of the CPS spill module is to make sure the
1076    number of live variables at any program point (the bandwidth)
1077    does not exceed a certain limit, which is determined by the
1078    size of the spill area.
1079    
1080    When the bandwidth is too large, we decrease the register pressure by
1081    packing live variables into spill records.  How we achieve this is
1082    completely different than what we did in the old code.
1083    
1084    First, there is something about the MLRiscGen code generator
1085    that we should be aware of:
1086    
1087    o MLRiscGen performs code motion!
1088    
1089       In particular, it will move floating point computations and
1090       address computations involving only the heap pointer to
1091       their use sites (if there is only a single use).
1092       What this means is that if we have a CPS record construction
1093       statement
1094    
1095           RECORD(k,vl,w,e)
1096    
1097       we should never count the new record address w as live if w
1098       has only one use (which is often the case).
1099    
1100       We should do something similar to floating point, but the transformation
1101       there is much more complex, so I won't deal with that.
1102    
1103    Secondly, there are now two new cps primops at our disposal:
1104    
1105     1. rawrecord of record_kind option
1106        This pure operator allocates some uninitialized storage from the heap.
1107        There are two forms:
1108    
1109         rawrecord NONE [INT n]  allocates a tagless record of length n
1110         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1111                                     and initializes the tag.
1112    
1113     2. rawupdate of cty
1114          rawupdate cty (v,i,x)
1115          Assigns to x to the ith component of record v.
1116          The storelist is not updated.
1117    
1118    We use these new primops for both spilling and increment record construction.
1119    
1120     1. Spilling.
1121    
1122        This is implemented with a linear scan algorithm (but generalized
1123        to trees).  The algorithm will create a single spill record at the
1124        beginning of the cps function and use rawupdate to spill to it,
1125        and SELECT or SELp to reload from it.  So both spills and reloads
1126        are fine-grain operations.  In contrast, in the old algorithm
1127        "spills" have to be bundled together in records.
1128    
1129        Ideally, we should sink the spill record construction to where
1130        it is needed.  We can even split the spill record into multiple ones
1131        at the places where they are needed.  But CPS is not a good
1132        representation for global code motion, so I'll keep it simple and
1133        am not attempting this.
1134    
1135     2. Incremental record construction (aka record splitting).
1136    
1137        Long records with many component values which are simulatenously live
1138        (recall that single use record addresses are not considered to
1139         be live) are constructed with rawrecord and rawupdate.
1140        We allocate space on the heap with rawrecord first, then gradually
1141        fill it in with rawupdate.  This is the technique suggested to me
1142        by Matthias.
1143    
1144        Some restrictions on when this is applicable:
1145        1. It is not a VECTOR record.  The code generator currently does not handle
1146           this case. VECTOR record uses double indirection like arrays.
1147        2. All the record component values are defined in the same "basic block"
1148           as the record constructor.  This is to prevent speculative
1149           record construction.
1150    
1151    ----------------------------------------------------------------------
1152    Name: Allen Leung
1153    Date: 2002/02/22 01:02:00 EST
1154    Tag: leunga-20020222-mlrisc-tools
1155    
1156    Minor bug fixes in the parser and rewriter
1157    
1158    ----------------------------------------------------------------------
1159    Name: Allen Leung
1160    Date: 2002/02/21 20:20:00 EST
1161    Tag: leunga-20020221-peephole
1162    
1163    Regenerated the peephole files.  Some contained typos in the specification
1164    and some didn't compile because of pretty printing bugs in the old version
1165    of 'nowhere'.
1166    
1167    ----------------------------------------------------------------------
1168    Name: Allen Leung
1169    Date: 2002/02/19 20:20:00 EST
1170    Tag: leunga-20020219-mlrisc-tools
1171    Description:
1172    
1173       Minor bug fixes to the mlrisc-tools library:
1174    
1175       1.  Fixed up parsing colon suffixed keywords
1176       2.  Added the ability to shut the error messages up
1177       3.  Reimplemented the pretty printer and fixed up/improved
1178           the pretty printing of handle and -> types.
1179       4.  Fixed up generation of literal symbols in the nowhere tool.
1180       5.  Added some SML keywords to to sml.sty
1181    
1182    ----------------------------------------------------------------------
1183    Name: Matthias Blume
1184    Date: 2002/02/19 16:20:00 EST
1185    Tag: blume-20020219-cmffi
1186    Description:
1187    
1188    A wild mix of changes, some minor, some major:
1189    
1190    * All C FFI-related libraries are now anchored under $c:
1191        $/c.cm      --> $c/c.cm
1192        $/c-int.cm  --> $c/internals/c-int.cm
1193        $/memory.cm --> $c/memory/memory.cm
1194    
1195    * "make" tool (in CM) now treats its argument pathname slightly
1196      differently:
1197        1. If the native expansion is an absolute name, then before invoking
1198           the "make" command on it, CM will apply OS.Path.mkRelative
1199           (with relativeTo = OS.FileSys.getDir()) to it.
1200        2. The argument will be passed through to subsequent phases of CM
1201           processing without "going native".  In particular, if the argument
1202           was an anchored path, then "make" will not lose track of that anchor.
1203    
1204    * Compiler backends now "know" their respective C calling conventions
1205      instead of having to be told about it by ml-nlffigen.  This relieves
1206      ml-nlffigen from one of its burdens.
1207    
1208    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1209    
1210    * Export C_DEBUG and C_Debug from $c/c.cm.
1211    
1212    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1213      subtyping relationship between incomplete pointers and their complete
1214      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1215      with the convention of instantiating 'o with ('t, 'c) obj whenever
1216      the pointer target type is complete.  In the incomplete case, 'o
1217      will be instantiated with some "'c iobj" -- a type obtained by
1218      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1219    
1220      Operations that work on both incomplete and complete pointer types are
1221      typed as taking an 'o ptr while operations that require the target to
1222      be known are typed as taking some ('t, 'c) obj ptr.
1223    
1224      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1225      where void is an eqtype without any values.  This makes it possible
1226      to work on voidptr values using functions meant to operate on light
1227      incomplete pointers.
1228    
1229    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1230      been vastly simplified.
1231    
1232    ----------------------------------------------------------------------
1233    Name: Matthias Blume
1234    Date: 2002/02/19 10:48:00 EST
1235    Tag: blume-20020219-pqfix
1236    Description:
1237    
1238    Applied Chris Okasaki's bug fix for priority queues.
1239    
1240    ----------------------------------------------------------------------
1241    Name: Matthias Blume
1242    Date: 2002/02/15 17:05:00
1243    Tag: Release_110_39
1244    Description:
1245    
1246    Last-minute retagging is becoming a tradition... :-(
1247    
1248    This is the working release 110.39.
1249    
1250    ----------------------------------------------------------------------
1251    Name: Matthias Blume
1252    Date: 2002/02/15 16:00:00 EST
1253    Tag: Release_110_39-orig
1254    Description:
1255    
1256    Working release 110.39.  New bootfiles.
1257    
1258    (Update: There was a small bug in the installer so it wouldn't work
1259    with all shells.  So I retagged. -Matthias)
1260    
1261    ----------------------------------------------------------------------
1262    Name: Matthias Blume
1263    Date: 2002/02/15 14:17:00 EST
1264    Tag: blume-20020215-showbindings
1265    Description:
1266    
1267    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1268    the latter can be useful for exploring what bindings are available at
1269    the interactive prompt.  (The first function returns only the list
1270    of symbols that are really bound, the second prints those but also the
1271    ones that CM's autoloading mechanism knows about.)
1272    
1273    ----------------------------------------------------------------------
1274    Name: Matthias Blume
1275    Date: 2002/02/15 12:08:00 EST
1276    Tag: blume-20020215-iptrs
1277    Description:
1278    
1279    Two improvements to ml-nlffigen:
1280    
1281      1. Write files only if they do not exist or if their current contents
1282         do not coincide with what's being written.  (That is, avoid messing
1283         with the time stamps unless absolutely necessary.)
1284    
1285      2. Implement a "repository" mechanism for generated files related
1286         to "incomplete pointer types".   See the README file for details.
1287    
1288    ----------------------------------------------------------------------
1289    Name: Matthias Blume
1290    Date: 2002/02/14 11:50:00 EST
1291    Tag: blume-20020214-quote
1292    Description:
1293    
1294    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1295    because of the new and improved tag generation scheme.  (Thanks to Allen
1296    Leung for pointing it out.)
1297    
1298    ----------------------------------------------------------------------
1299    Name: Lal George
1300    Date: 2002/02/14 09:55:27 EST 2002
1301    Tag: george-20020214-isabelle-bug
1302    Description:
1303    
1304    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1305    of Isabelle on the x86.
1306    
1307    From Allen:
1308    -----------
1309     I've found the problem:
1310    
1311         in ra-core.sml, I use the counter "blocked" to keep track of the
1312         true number of elements in the freeze queue.  When the counter goes
1313         to zero, I skip examining the queue.  But I've messed up the
1314         bookkeeping in combine():
1315    
1316             else ();
1317             case !ucol of
1318               PSEUDO => (if !cntv > 0 then
1319                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1320                                        ^^^^^^^^^^^^^^^^^^^^^^^
1321                      moveu := mergeMoveList(!movev, !moveu)
1322                     )
1323                  else ();
1324    
1325         combine() is called to coalesce two nodes u and v.
1326         I think I was thinking that if the move counts of u and v are both
1327         greater than zero then after they are coalesced then one node is
1328         removed from the freeze queue.  Apparently I was thinking that
1329         both u and v are of low degree, but that's clearly not necessarily true.
1330    
1331    
1332    02/12/2002:
1333        Here's the patch.  HOL now compiles.
1334    
1335        I don't know how this impact on performance (compile
1336        time or runtime).  This bug caused the RA (especially on the x86)
1337        to go thru the potential spill phase when there are still nodes on the
1338        freeze queue.
1339    
1340    
1341    
1342    
1343    ----------------------------------------------------------------------
1344    Name: Matthias Blume
1345    Date: 2002/02/13 22:40:00 EST
1346    Tag: blume-20020213-fptr-rtti
1347    Description:
1348    
1349    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1350    updates.
1351    
1352    ----------------------------------------------------------------------
1353    Name: Matthias Blume
1354    Date: 2002/02/13 16:41:00 EST
1355    Tag: blume-20020213-cmlpq
1356    Description:
1357    
1358    Added new priority queue export symbols (which have just been added to
1359    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1360    not compile and the installer would choke.)
1361    
1362    ----------------------------------------------------------------------
1363    Name: Matthias Blume
1364    Date: 2002/02/13 16:15:00 EST
1365    Tag: blume-20020213-various
1366    Description:
1367    
1368    1. More tweaks to ml-nlffigen:
1369    
1370       - better internal datastructures (resulting in slight speedup)
1371       - "-match" option requires exact match
1372       - "localized" gensym counters (untagged structs/unions nested within
1373         other structs/unions or within typedefs get a fresh counter; their
1374         tag will be prefixed by a concatenation of their parents' tags)
1375       - bug fixes (related to calculation of transitive closure of types
1376         to be included in the output)
1377    
1378    2. Minor Basis updates:
1379    
1380       - added implementations for List.collate and Option.app
1381    
1382    ----------------------------------------------------------------------
1383    Name: Matthias Blume
1384    Date: 2002/02/11 15:55:00 EST
1385    Tag: blume-20020211-gensym
1386    Description:
1387    
1388    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1389    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1390    names (ML structure names that correspond to unnamed C structs, unions,
1391    and enums), so that separate runs of ml-nlffigen do not clash.
1392    
1393    ----------------------------------------------------------------------
1394    Name: Matthias Blume
1395    Date: 2002/02/11 12:05:00 EST
1396    Tag: blume-20020211-gensml
1397    Description:
1398    
1399    A quick fix for a problem with GenSML (in the pgraph-util library):
1400    Make generation of toplevel "local" optional.  (Strictly speaking,
1401    signature definitions within "local" are not legal SML.)
1402    
1403    Other than that: updates to INSTALL and cm/TODO.
1404    
1405    ----------------------------------------------------------------------
1406    Name: Matthias Blume
1407    Date: 2002/02/08 15:00:00 EST
1408    Tag: blume-20020208-uniquepid
1409    Description:
1410    
1411    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1412    
1413    1. The installer (config/install.sh) has gotten smarter:
1414    
1415         - Configuration options are a bit easier to specify now
1416           (in config/targets).
1417         - Bug in recognizing .tar.bz2 files fixed.
1418         - Installer automatically resolves dependencies between
1419           configuration options (e.g., if you ask for eXene, you will
1420           also get cml -- regardless whether you asked for it or not).
1421         - Installer can run in "quieter mode" by setting the environment
1422           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1423           "completely silent", though.
1424         - Build HashCons library as part of smlnj-lib.
1425    
1426    2. A new scheme for assigning persistent identifiers to compilation
1427       units (and, by extension, to types etc.) has been put into place.
1428       This fixes a long-standing bug where types and even dynamic values
1429       can get internally confused, thereby compromising type safety
1430       (abstraction) and dynamic correctness.  See
1431    
1432         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1433    
1434       for an example of how things could go wrong until now.
1435    
1436       The downside of the new scheme is that pids are not quite as
1437       persistent as they used to be: CM will generate a fresh pid
1438       for every compilation unit that it thinks it sees for the first
1439       time.  That means that if you compile starting from a clean, fresh
1440       source tree at two different times, you end up with different
1441       binaries.
1442    
1443       Cutoff recompilation, however, has not been compromised because
1444       CM keeps pid information in special caches between runs.
1445    
1446    ----------------------------------------------------------------------
1447    Name: Lal George
1448    Date: 2002/02/07 15:34:13 EST 2002
1449    Tag: <none>
1450    Description:
1451    
1452    Compilers that generate assembly code may produce  global labels
1453    whose value is resolved at link time. The various peephole optimization
1454    modules did not take this in account.
1455    
1456    TODO. The Labels.addrOf function should really return an option
1457    type so that clients are forced to deal with this issue, rather
1458    than an exception being raised.
1459    
1460    ----------------------------------------------------------------------
1461    Name: Lal George
1462    Date: 2002/02/06 13:55:02 EST
1463    Tag: george-20020206-ra-breakup
1464    Description:
1465    
1466    1. A bug fix from Allen.
1467    
1468        A typo causes extra fstp %st(0)'s to be generated at compensation
1469        edges, which might cause stack underflow traps at runtime.  This
1470        occurs in fft where there are extraneous fstps right before the 'into'
1471        trap instruction (in this case they are harmless since none of the
1472        integers overflow.)
1473    
1474    2. Pulled out various utility modules that were embedded in the modules
1475       of the register allocator. I need these modules for other purposes, but
1476       they are not complete enough to put into a library (just yet).
1477    ----------------------------------------------------------------------
1478    Name: Matthias Blume
1479    Date: 2002/01/31 16:05:00 EST
1480    Tag: blume-20020131-sparc-ccalls
1481    Description:
1482    
1483    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1484       of extra stack space by mistake.  Fixed.
1485    
1486    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1487    
1488    ----------------------------------------------------------------------
1489    Name: Allen Leung
1490    Date: 2002/01/30
1491    Tag: leunga-20020130-nowhere-bug-fix
1492    Description:
1493    
1494       MLRISC bug fixes:
1495       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1496       2. MachineInt.fromString was negating its value.
1497    
1498    ----------------------------------------------------------------------
1499    Name: Matthias Blume
1500    Date: 2002/01/29
1501    Tag: blume-20020129-INSTALL
1502    Description:
1503    
1504    - Added somewhat detailed installation instructions (file INSTALL).
1505    - Fixed curl-detection bug in config/install.sh.
1506    - It is now possible to select the URL getter using the URLGETTER
1507      environment variable:
1508    
1509          not set / "unknown"      --> automatic detection (script tries wget,
1510                                       curl, and lynx)
1511          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1512                                       how to properly invoke them)
1513          other                    --> use $URLGETTER directly, it must take
1514                                       precisely two command-line arguments
1515                                       (source URL and destination file name)
1516    
1517    ----------------------------------------------------------------------
1518    Name: Matthias Blume
1519    Date: 2002/01/28
1520    Tag: blume-20020128-sparc-ccalls
1521    Description:
1522    
1523    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1524    - Make use of the allocParam argument in sparc-c-calls.
1525    
1526    ----------------------------------------------------------------------
1527    Name: Matthias Blume
1528    Date: 2002/01/28
1529    Tag: blume-20020128-allocParam
1530    Description:
1531    
1532    John Reppy:  Changes c-calls API to accept client-callback for
1533    allocating extra stack space.
1534    me: Corresponding changes to mlriscGen (using a dummy argument that
1535        does not change the current behavior).
1536    
1537    ----------------------------------------------------------------------
1538    Name: Matthias Blume
1539    Date: 2002/01/28 12:00:00
1540    Tag: Release_110_38
1541    Description:
1542    
1543    This time for real!!!
1544    
1545    ----------------------------------------------------------------------
1546    Name: Matthias Blume
1547    Date: 2002/01/28 10:56:00 EST
1548    Tag: blume-20020128-retraction
1549    Description:
1550    
1551    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1552       with blume-Release_110_38-retracted.)
1553    
1554    1. Fixed a problem with incorrect rounding modes in real64.sml.
1555       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1556    
1557    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1558       and enums fixed.  The naming of corresponding ML identifiers should
1559       now be consistent again.
1560    
1561    ----------------------------------------------------------------------
1562    Name: Allen Leung
1563    Date: 2002/01/27
1564    Tag: leunga-20020127-nowhere
1565    Description:
1566    
1567       Added a target called nowhere in the configuration scripts.
1568       Enabling this will build the MLRISC 'nowhere' tool (for translating
1569       programs with where-clauses into legal SML code) during installation.
1570    
1571    ----------------------------------------------------------------------
1572    Name: Matthias Blume
1573    Date: 2002/01/25 21:27:00 EST
1574    Tag: blume-Release_110_38-retracted
1575    Description:
1576    
1577    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1578    
1579    README will be added later.
1580    
1581    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1582    (Release_110_38) removed.  Reason: Last-minute bug fixes.
1583    
1584    ----------------------------------------------------------------------
1585    Name: Matthias Blume
1586    Date: 2002/01/25
1587    Tag: blume-20020125-ffi
1588    Description:
1589    
1590    A large number of tweaks and improvements to ml-nlffi-lib and
1591    ml-nlffigen:
1592    
1593       - ML represenation types have been streamlined
1594       - getter and setter functions work with concrete values, not abstract
1595         ones where possible
1596       - ml-nlffigen command line more flexible (see README file there)
1597       - some bugs have been fixed (hopefully)
1598    
1599    ----------------------------------------------------------------------
1600    Name: Lal George
1601    Date: 2002/01/24
1602    Tag: george-20020124-risc-ra-interface
1603    Description:
1604    
1605       There is a dramatic simplification in the interface to the
1606       register allocator for RISC architectures as a result of making
1607       parallel copy instructions explicit.
1608    
1609    ----------------------------------------------------------------------
1610    Name: Matthias Blume
1611  Date: 2002/01/22  Date: 2002/01/22
1612  Tag: blume-20020122-x86-ccalls  Tag: blume-20020122-x86-ccalls
1613  Description:  Description:
# Line 501  Line 2095 
2095    
2096  ----------------------------------------------------------------------  ----------------------------------------------------------------------
2097  Name: Matthias Blume  Name: Matthias Blume
 >>>>>>> 1.169  
2098  Date: 2001/09/18 15:35:00 EDT  Date: 2001/09/18 15:35:00 EDT
2099  Tag: blume-20010918-readme11036  Tag: blume-20010918-readme11036
2100  Description:  Description:

Legend:
Removed from v.1032  
changed lines
  Added in v.1220

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