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 1094, Mon Feb 25 09:58:56 2002 UTC revision 1259, Sat Jul 13 01:19:00 2002 UTC
# Line 13  Line 13 
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume
17    Date: 2002/07/12 21:19:00 EDT
18    Tag: blume-20020712-liveness
19    Description:
20    
21    Exported two useful "step" functions from liveness module (MLRISC).
22    
23    ----------------------------------------------------------------------
24    Name: Matthias Blume
25    Date: 2002/07/05 16:00 EDT
26    Tag: Release_110_41
27    Description:
28    
29    New working release.  New bootfiles.
30    
31    ----------------------------------------------------------------------
32    Name: Matthias Blume
33    Date: 2002/07/05 10:25:00 EDT
34    Tag: blume-20020705-btimp
35    Description:
36    
37    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
38    other clients can set up backtracing support.
39    
40    ----------------------------------------------------------------------
41    Name: Matthias Blume
42    Date: 2002/06/25 17:23:00 EDT
43    Tag: blume-20020625-fpmax
44    Description:
45    
46    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
47    case.  (The sense of the isNaN test was reversed -- which made min and
48    max always return their first argument.)
49    
50    ----------------------------------------------------------------------
51    Name: Matthias Blume
52    Date: 2002/06/11
53    Tag: blume-20020611-unixpath
54    Description:
55    
56    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
57    
58    ----------------------------------------------------------------------
59    Name: Matthias Blume
60    Date: 2002/06/10 16:35:00 EDT
61    Tag: blume-20020610-ieeereal
62    Description:
63    
64    I back-ported my implementation of IEEEReal.fromString from the
65    idlbasis-devel branch so that we can test it.
66    
67    Another small change is that ppDec tries to give more information
68    than just "<sig>" in the case of functors.  However, this code is
69    broken in some mysterious way if the functor's body's signature
70    has not been declared by ascription but gets inferred from the
71    implementation.  This needs fixing...
72    
73    ----------------------------------------------------------------------
74    Name: Matthias Blume
75    Date: 2002/05/31
76    Tag: blume-20020531-btrace-mode
77    Description:
78    
79    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
80    the wayside when I switched over to using Controls everywhere.)
81    
82    ----------------------------------------------------------------------
83    Name: Lal George
84    Date: 2002/05/23 12:21:40 EDT
85    Tag: george-20020523-visual-labels
86    Description:
87    
88    Labels are now displayed in the graphical output to make
89    the fall-through and target blocks obvious.
90    
91    ----------------------------------------------------------------------
92    Name: Matthias Blume
93    Date: 2002/05/22 11:03:00 EDT
94    Tag: blume-20020522-shrink
95    Description:
96    
97    John tweaked yesterday's fix for 1131 to handle an out-of-memory
98    situation that comes up when allocating huge arrays.
99    
100    ----------------------------------------------------------------------
101    Name: Matthias Blume
102    Date: 2002/05/21 16:00:00 EDT
103    Tag: Release_110_40
104    Description:
105    
106    New working release (110.40).  New bootfiles.
107    
108    [Also: John Reppy fixed GC bug 1131.]
109    
110    ----------------------------------------------------------------------
111    Name: Matthias Blume
112    Date: 2002/05/21 12:35:00 EDT
113    Tag: blume-20020521-cmdoc
114    Description:
115    
116    CM documentation update.
117    
118    ----------------------------------------------------------------------
119    Name: Matthias Blume
120    Date: 2002/05/21 10:55:00 EDT
121    Tag: blume-20020521-misc
122    Description:
123    
124    - John tweaked runtime to be silent on heap export (except when
125      GC messages are on).
126    - I added a few more things (cross-compiling versions of CMB) to
127      config/preloads (as suggestions).
128    
129    ----------------------------------------------------------------------
130    Name: Matthias Blume
131    Date: 2002/05/20 22:25:00 EDT
132    Tag: blume-20020520-controls
133    Description:
134    
135    - Added ControlUtil structure to control-lib.cm.
136    - Use it throughout.
137    - Used Controls facility to define MLRISC controls (as opposed to
138      registering MLRISC control ref cells with Controls after the fact)
139    - Fixed messed-up controls priorities.
140    
141    * Removed again all the stuff from config/preloads that one wouldn't
142      be able to preload at the time the initial heap image is built.
143      (Many libraries, e.g., CML, do not exist yet at this time.  The
144       only libraries that can be preloaded via config/preloads are those
145       that come bundled with the bootfiles.)
146    
147    ----------------------------------------------------------------------
148    Name: Matthias Blume
149    Date: 2002/05/20 10:59:00 EDT
150    Tag: blume-20020520-preloads
151    Description:
152    
153    Added a lot of commented-out suggestions for things to be included
154    in config/preloads.
155    
156    ----------------------------------------------------------------------
157    Name: Allen Leung
158    Date: 2002/05/18 14:20:00 EDT
159    Tag: leunga-20020518-mdl
160    Description:
161    
162    o Made the mdl tool stuff compile and run again.
163    o I've disabled all the stuff that depends on RTL specifications; they
164      are all badly broken anyway.
165    
166    ----------------------------------------------------------------------
167    Name: Matthias Blume
168    Date: 2002/05/17 16:49:00 EDT
169    Tag: blume-20020517-controls
170    Description:
171    
172    0. John Reppy made several modifications to the SML/NJ library.
173       In particular, there is a shiny new controls-lib.cm.
174    
175    1. Pushed new controls interface through compiler so that everything
176       compiles again.
177    
178    2. Added FormatComb and FORMAT_COMB to the CML version of the
179       SML/NJ library (so that CML compiles again).
180    
181    3. Modified init scripts because XXX_DEFAULT environment variables
182       are no longer with us.  (Boot-time initialization is now done
183       using the same environment variables that are also used for
184       startup-time initialization of controls.)
185    
186    ----------------------------------------------------------------------
187    Name: Lal George
188    Date: 2002/05/15 09:20:10 EDT
189    Tag: george-20020515-pseudo-op-decls
190    Description:
191    
192            All pseudo-ops emitted before the first segment declaration
193    such as TEXT, DATA, and BSS directives are assumed to be global
194    declarations and are emitted first in the assembly file. This is
195    useful in a number of situations where one has pseudo-ops that are not
196    specific to any segment, and also works around the constraint that one
197    cannot have client pseudo-ops in the TEXT segment.
198    
199            Because no segment is associated with these declarations it is
200    an error to allocate any space or objects before the first segment
201    directive and an exception will be raised. However, we cannot make
202    this check for client pseudo-ops.
203    
204            These top level declarations are a field in the CFG graph_info.
205    In theory you can continue to add to this field after the CFG has been
206    built -- provided you know what you are doing;-)
207    
208    ----------------------------------------------------------------------
209    Name: Matthias Blume
210    Date: 2002/05/13 16:40:00 EDT
211    Tag: blume-20020513-pp-etc
212    Description:
213    
214    A few minor bugfixes:
215    
216      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
217        (Bogus printouts for redefined bindings still occur.  Compiler
218        bug should no longer occur now.  We need to redo the prettyprinter
219        from scratch.)
220    
221      - CM pathname printer now also adds escape sequences for ( and )
222    
223      - commend and docu fixes for ml-nlffi
224    
225    ----------------------------------------------------------------------
226    Name: Matthias Blume
227    Date: 2002/05/10 16:40:00 EDT
228    Tag: blume-20020510-erg-textio
229    Description:
230    
231    Applied the following bugfix provided by Emden Gansner:
232    
233        Output is corrupted when outputSubstr is used rather than output.
234    
235        The problem occurs when a substring
236    
237            ss = (s, dataStart, dataLen)
238    
239        where dataStart > 0, fills a stream buffer with avail bytes left.
240        avail bytes of s, starting at index dataStart, are copied into the
241        buffer, the buffer is flushed, and then the remaining dataLen-avail
242        bytes of ss are copied into the beginning of the buffer. Instead of
243        starting this copy at index dataStart+avail in s, the current code
244        starts the copy at index avail.
245    
246        Fix:
247        In text-io-fn.sml, change line 695 from
248             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
249        to
250             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
251    
252    ----------------------------------------------------------------------
253    Name: Matthias Blume
254    Date: 2002/04/12 13:55:00 EDT
255    Tag: blume-20020412-assyntax
256    Description:
257    
258    1. Grabbed newer assyntax.h from the XFree86 project.
259    2. Fiddled with how to compile X86.prim.asm without warnings.
260    3. (Very) Minor cleanup in CM.
261    
262    ----------------------------------------------------------------------
263    Name: Matthias Blume
264    Date: 2002/04/01 (no joke!) 17:07:00 EST
265    Tag: blume-20020401-x86div
266    Description:
267    
268    Added full support for div/mod/rem/quot on the x86, using the machine
269    instruction's two results (without clumsily recomputing the remainder)
270    directly where appropriate.
271    
272    Some more extensive power-of-two support was added to the x86 instruction
273    selector (avoiding expensive divs, mods, and muls where they can be
274    replaced with cheaper shifts and masks).  However, this sort of thing
275    ought to be done earlier, e.g., within the CPS optimizer so that
276    all architectures benefit from it.
277    
278    The compiler compiles to a fixed point, but changes might be somewhat
279    fragile nevertheless.  Please, report any strange things that you might
280    see wrt. div/mod/quot/rem...
281    
282    ----------------------------------------------------------------------
283    Name: Matthias Blume
284    Date: 2002/03/29 17:22:00
285    Tag: blume-20020329-div
286    Description:
287    
288    Fixed my broken div/mod logic.  Unfortunately, this means that the
289    inline code for div/mod now has one more comparison than before.
290    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
291    The problem was with quotient = 0, because that alone does not tell
292    us which way the rounding went.  One then has to look at whether
293    remainder and divisor have the same sign...  :(
294    
295    Anyway, I replaced the bootfiles with fresh ones...
296    
297    ----------------------------------------------------------------------
298    Name: Matthias Blume
299    Date: 2002/03/29 14:10:00 EST
300    Tag: blume-20020329-inlprims
301    Description:
302    
303    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
304    
305    Primops have changed. This means that the bin/boot-file formats have
306    changed as well.
307    
308    To make sure that there is no confusion, I made a new version.
309    
310    
311    CHANGES:
312    
313    * removed REMT from mltree (remainder should never overflow).
314    
315    * added primops to deal with divisions of all flavors to the frontend
316    
317    * handled these primops all the way through so they map to their respective
318      MLRISC support
319    
320    * used these primops in the implementation of Int, Int32, Word, Word32
321    
322    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
323    
324    * parameterized INLMIN, INLMAX, and INLABS by a numkind
325    
326    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
327      floating point
328    
329    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
330      and Real (but Real.abs maps to a separate floating-point-only primop)
331    
332    
333    TODO items:
334    
335    * Hacked Alpha32 instruction selection, disabling the selection of REMx
336      instructions because the machine instruction encoder cannot handle
337      them.  (Hppa, PPC, and Sparc instruction selection did not handle
338      REM in the first place, and REM is supported by the x86 machine coder.)
339    
340    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
341      selection phase.  (The two can be streamlined because the hardware
342      delivers both quotient and remainder at the same time anyway.)
343    
344    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
345      (Currently the behavior is inconsistent both across architectures and
346      wrt. the draft Basis spec.)
347    
348    * Word8 should eventually be handled natively, too.
349    
350    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
351      as if there currently is no execution path that could trigger it in
352      SML/NJ.  (The assumptions underlying functions arith and promotable do not
353      hold for things like multiplication and division.)
354    
355    ----------------------------------------------------------------------
356    Name: Matthias Blume
357    Date: 2002/03/27 16:27:00 EST
358    Tag: blume-20020327-mlrisc-divisions
359    Description:
360    
361    Added support for all four division operations (ML's div, mod, quot,
362    and rem) to MLRISC.  In the course of doing so, I also rationalized
363    the naming (no more annoying switch-around of DIV and QUOT), by
364    parameterizing the operation by div_rounding_mode (which can be either
365    DIV_TO_ZERO or DIV_TO_NEGINF).
366    
367    The generic MLTreeGen functor takes care of compiling all four
368    operations down to only round-to-zero div.
369    
370    Missing pieces:
371    
372      * Doing something smarter than relying on MLTreeGen on architectures
373        like, e.g., the x86 where hardware division delivers both quotient and
374        remainder at the same time.  With this, the implementation of the
375        round-to-neginf operations could be further streamlined.
376    
377      * Remove inlining support for div/mod/rem from the frontend and replace it
378        with primops that get carried through to the backend.  Do this for all
379        int and word types.
380    
381    ----------------------------------------------------------------------
382    Name: Matthias Blume
383    Date: 2002/03/25 17:25:00 EST
384    Tag: blume-20020325-divmod
385    Description:
386    
387    I improved (hopefully without breaking them) the implementation of Int.div,
388    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
389    advantage of the following observations:
390    
391      Let  q = x quot y      r = x rem y
392           d = x div  y      m = x mod y
393    
394    where "quot" is the round-to-zero version of integer division that
395    hardware usually provides.  Then we have:
396    
397         r = x - q * y        where neither the * nor the - will overflow
398         d = if q >= 0 orelse x = q * y then q else q - 1
399                              where neither the * nor the - will overflow
400         m = if q >= 0 orelse r = 0 then r else r + y
401                              where the + will not overflow
402    
403    This results in substantial simplification of the generated code.
404    The following table shows the number of CFG nodes and edges generated
405    for
406            fun f (x, y) = x OPER y
407            (* with OPER \in div, mod, quot, rem *)
408    
409    
410        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
411        --------------------------------------------------------
412         div |         24 |         39 |         12 |         16
413         mod |         41 |         71 |         12 |         16
414        quot |          8 |         10 |          8 |         10
415         rem |         10 |         14 |          8 |         10
416    
417    
418    ----------------------------------------------------------------------
419    Name: Matthias Blume
420    Date: 2002/03/25 22:06:00 EST
421    Tag: blume-20020325-cprotobug
422    Description:
423    
424    Fixed a bug in cproto (c prototype decoder).
425    
426    ----------------------------------------------------------------------
427    Name: Matthias Blume
428    Date: 2002/03/25 16:00:00 EST
429    Tag: blume-20020325-raw-primops
430    Description:
431    
432    I did some cleanup to Allen's new primop code and
433    replaced yesterday's bootfiles with new ones.
434    (But they are stored in the same place.)
435    
436    ----------------------------------------------------------------------
437    Name: Matthias Blume
438    Date: 2002/03/24 22:40:00 EST
439    Tag: blume-20020324-bootfiles
440    Description:
441    
442    Made the bootfiles that Allen asked for.
443    
444    ----------------------------------------------------------------------
445    Name: Allen Leung
446    Date: 2002/03/23 15:50:00 EST
447    Tag: leunga-20020323-flint-cps-rcc-primops
448    Description:
449    
450      1. Changes to FLINT primops:
451    
452        (* make a call to a C-function;
453         * The primop carries C function prototype information and specifies
454         * which of its (ML-) arguments are floating point. C prototype
455         * information is for use by the backend, ML information is for
456         * use by the CPS converter. *)
457      | RAW_CCALL of { c_proto: CTypes.c_proto,
458                       ml_args: ccall_type list,
459                       ml_res_opt: ccall_type option,
460                       reentrant : bool
461                     } option
462       (* Allocate uninitialized storage on the heap.
463        * The record is meant to hold short-lived C objects, i.e., they
464        * are not ML pointers.  With the tag, the representation is
465        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
466        *)
467      | RAW_RECORD of {tag:bool,sz:int}
468      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
469    
470      2.  These CPS primops are now overloaded:
471    
472           rawload of {kind:numkind}
473           rawstore of {kind:numkind}
474    
475          The one argument form is:
476    
477             rawload {kind} address
478    
479          The two argument form is:
480    
481             rawload {kind} [ml object, byte-offset]
482    
483      3. RAW_CCALL/RCC now takes two extra arguments:
484    
485         a. The first is whether the C call is reentrant, i.e., whether
486            ML state should be saved and restored.
487         b. The second argument is a string argument specifying the name of
488            library and the C function.
489    
490         These things are currently not handled in the code generator, yet.
491    
492      4. In CProto,
493    
494         An encoding type of "bool" means "ml object" and is mapped into
495         C prototype of PTR.  Note that "bool" is different than "string",
496         even though "string" is also mapped into PTR, because "bool"
497         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
498    
499      5. Pickler/unpicker
500    
501         Changed to handle RAW_RECORD and newest RAW_CCALL
502    
503      6. MLRiscGen,
504    
505         1. Changed to handle the new rawload/rawstore/rawrecord operators.
506         2. Code for handling C Calls has been moved to a new module CPSCCalls,
507            in the file CodeGen/cpscompile/cps-c-calls.sml
508    
509      7. Added the conditional move operator
510    
511             condmove of branch
512    
513         to cps.  Generation of this is still buggy so it is currently
514         disabled.
515    
516    ----------------------------------------------------------------------
517    Name: Lal George
518    Date: 2002/03/22 14:18:25 EST
519    Tag: george-20020322-cps-branch-prob
520    Description:
521    
522    Implemented the Ball-Larus branch prediction-heuristics, and
523    incorporated graphical viewers for control flow graphs.
524    
525    Ball-Larus Heuristics:
526    ---------------------
527    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
528    
529    By design it uses the Dempster-Shafer theory for combining
530    probabilities.  For example, in the function:
531    
532        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
533    
534    the ball-larus heuristics predicts that the n=0 is unlikely
535    (OH-heuristic), and the 'then' branch is unlikely because of the
536    RH-heuristic -- giving the 'then' branch an even lower combined
537    probability using the Dempster-Shafer theory.
538    
539    Finally, John Reppy's loop analysis in MLRISC, further lowers the
540    probability of the 'then' branch because of the loop in the else
541    branch.
542    
543    
544    Graphical Viewing:
545    ------------------
546    I merely plugged in Allen's graphical viewers into the compiler. The
547    additional code is not much. At the top level, saying:
548    
549            Control.MLRISC.getFlag "cfg-graphical-view" := true;
550    
551    will display the graphical view of the control flow graph just before
552    back-patching.  daVinci must be in your path for this to work. If
553    daVinci is not available, then the default viewer can be changed
554    using:
555    
556            Control.MLRISC.getString "viewer"
557    
558    which can be set to "dot" or "vcg" for the corresponding viewers. Of
559    course, these viewers must be in your path.
560    
561    The above will display the compilation unit at the level of clusters,
562    many of which are small, boring, and un-interesting. Also setting:
563    
564            Control.MLRISC.getInt "cfg-graphical-view_size"
565    
566    will display clusters that are larger than the value set by the above.
567    
568    
569    ----------------------------------------------------------------------
570    Name: Matthias Blume
571    Date: 2002/03/21 22:20:00 EST
572    Tag: blume-20020321-kmp-bugfix
573    Description:
574    
575    Changed the interface to the KMP routine in PreString and fixed
576    a minor bug in one place where it was used.
577    
578    ----------------------------------------------------------------------
579    Name: Allen Leung
580    Date: 2002/03/21 20:30:00 EST
581    Tag: leunga-20020321-cfg
582    Description:
583    
584      Fixed a potential problem in cfg edge splitting.
585    
586    ----------------------------------------------------------------------
587    Name: Allen Leung
588    Date: 2002/03/21 17:15:00 EST
589    Tag: leunga-20020321-x86-fp-cfg
590    Description:
591    
592      1. Recoded the buggy parts of x86-fp.
593    
594         a. All the block reordering code has been removed.
595            We now depend on the block placement phases to do this work.
596    
597         b. Critical edge splitting code has been simplified and moved into the
598            CFG modules, as where they belong.
599    
600         Both of these were quite buggy and complex.  The code is now much, much
601         simpler.
602    
603      2. X86 backend.
604    
605         a. Added instructions for 64-bit support.  Instruction selection for
606            64-bit has not been committed, however, since that
607            requires changes to MLTREE which haven't been approved by
608            Lal and John.
609    
610         b. Added support for FUCOMI and FUCOMIP when generating code for
611            PentiumPro and above.  We only generate these instructions in
612            the fast-fp mode.
613    
614         c. Added cases for JP and JNP in X86FreqProps.
615    
616      3. CFG
617    
618         CFG now has a bunch of methods for edge splitting and merging.
619    
620      4. Machine description.
621    
622         John's simplification of MLTREE_BASIS.fcond broke a few machine
623         description things:
624    
625         rtl-build.{sig,sml} and hppa.mdl fixed.
626    
627         NOTE: the machine description stuff in the repository is still broken.
628               Again, I can't put my fixes in because that involves
629               changes to MLTREE.
630    
631    ----------------------------------------------------------------------
632    Name: Matthias Blume
633    Date: 2002/03/20 15:55:00 EST
634    Tag: blume-20020320-kmp
635    Description:
636    
637    Implemented Knuth-Morris-Pratt string matching in PreString and used
638    it for String.isSubstring, Substring.isSubstring, and
639    Substring.position.
640    
641    (Might need some stress-testing.  Simple examples worked fine.)
642    
643    ----------------------------------------------------------------------
644    Name: Matthias Blume
645    Date: 2002/03/19 16:37:00 EST
646    Tag: blume-20020319-witnesses
647    Description:
648    
649    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
650    
651    This implements a generic mechanism for changing constness qualifiers
652    anywhere within big C types without resorting to outright "casts".
653    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
654    modify the constness at the outermost level.)
655    The implementation of "convert" is based on the idea of "witness"
656    values -- values that are not used by the operation but whose types
657    "testify" to their applicability.  On the implementation side, "convert"
658    is simply a projection (returning its second curried argument).  With
659    cross-module inlining, it should not result in any machine code being
660    generated.
661    
662    ----------------------------------------------------------------------
663    Name: Matthias Blume
664    Date: 2002/03/15 16:40:00 EST
665    Tag: blume-20020315-basis
666    Description:
667    
668    Provided (preliminary?) implementations for
669    
670      {String,Substring}.{concatWith,isSuffix,isSubstring}
671    
672    and
673    
674      Substring.full
675    
676    Those are in the Basis spec but they were missing in SML/NJ.
677    
678    ----------------------------------------------------------------------
679    Name: Matthias Blume
680    Date: 2002/03/14 21:30:00 EST
681    Tag: blume-20020314-controls
682    Description:
683    
684    Controls:
685    ---------
686    
687    1. Factored out the recently-added Controls : CONTROLS stuff and put
688       it into its own library $/controls-lib.cm.  The source tree for
689       this is under src/smlnj-lib/Controls.
690    
691    2. Changed the names of types and functions in this interface, so they
692       make a bit more "sense":
693    
694          module -> registry
695          'a registry -> 'a group
696    
697    3. The interface now deals in ref cells only.  The getter/setter interface
698       is (mostly) gone.
699    
700    4. Added a function that lets one register an already-existing ref cell.
701    
702    5. Made the corresponding modifications to the rest of the code so that
703       everything compiles again.
704    
705    6. Changed the implementation of Controls.MLRISC back to something closer
706       to the original.  In particular, this module (and therefore MLRISC)
707       does not depend on Controls.  There now is some link-time code in
708       int-sys.sml that registers the MLRISC controls with the Controls
709       module.
710    
711    CM:
712    ---
713    
714      * One can now specify the lambda-split aggressiveness in init.cmi.
715    
716    ----------------------------------------------------------------------
717    Name: Allen Leung
718    Date: 2002/03/13 17:30:00 EST
719    Tag: leunga-20020313-x86-fp-unary
720    Description:
721    
722    Bug fix for:
723    
724    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
725    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
726    > - fun f(x,(y,z)) = Real.~ y;
727    > [autoloading]
728    > [autoloading done]
729    >       fchsl   (%eax), 184(%esp)
730    > Error: MLRisc bug: X86MCEmitter.emitInstr
731    >
732    > uncaught exception Error
733    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
734    
735    The problem was that the code generator did not generate any fp registers
736    in this case, and the ra didn't know that it needed to run the X86FP phase to
737    translate the pseudo fp instruction.   This only happened with unary fp
738    operators in certain situations.
739    
740    ----------------------------------------------------------------------
741    Name: Matthias Blume
742    Date: 2002/03/13 14:00:00 EST
743    Tag: blume-20020313-overload-etc
744    Description:
745    
746    1. Added _overload as a synonym for overload for backward compatibility.
747       (Control.overloadKW must be true for either version to be accepted.)
748    
749    2. Fixed bug in install script that caused more things to be installed
750       than what was requested in config/targets.
751    
752    3. Made CM aware of the (_)overload construct so that autoloading
753       works.
754    
755    ----------------------------------------------------------------------
756    Name: Matthias Blume
757    Date: 2002/03/12 22:03:00 EST
758    Tag: blume-20020312-url
759    Description:
760    
761    Forgot to update BOOT and srcarchiveurl.
762    
763    ----------------------------------------------------------------------
764    Name: Matthias Blume
765    Date: 2002/03/12 17:30:00 EST
766    Tag: blume-20020312-version110392
767    Description:
768    
769    Yet another version number bump (because of small changes to the
770    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
771    
772    Changes:
773    
774      The new pid generation scheme described a few weeks ago was overly
775      complicated.  I implemented a new mechanism that is simpler and
776      provides a bit more "stability":  Once CM has seen a compilation
777      unit, it keeps its identity constant (as long as you do not delete
778      those crucial CM/GUID/* files).  This means that when you change
779      an interface, compile, then go back to the old interface, and
780      compile again, you arrive at the original pid.
781    
782      There now also is a mechanism that instructs CM to use the plain
783      environment hash as a module's pid (effectively making its GUID
784      the empty string).  For this, "noguid" must be specified as an
785      option to the .sml file in question within its .cm file.
786      This is most useful for code that is being generated by tools such
787      as ml-nlffigen (because during development programmers tend to
788      erase the tool's entire output directory tree including CM's cached
789      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
790      revert to the old, broken behavior of SML/NJ, but in specific cases
791      where there is no danger of interface confusion, its use is ok
792      (I think).
793    
794      ml-nlffigen by default generates "noguid" annotations.  They can be
795      turned off by specifying -guid in its command line.
796    
797    ----------------------------------------------------------------------
798    Name: Lal George
799    Date: 2002/03/12 12 14:42:36 EST
800    Tag: george-20020312-frequency-computation
801    Description:
802    
803    Integrated jump chaining and static block frequency into the
804    compiler. More details and numbers later.
805    
806    ----------------------------------------------------------------------
807    Name: Lal George
808    Date: 2002/03/11 11 22:38:53 EST
809    Tag: george-20020311-jump-chain-elim
810    Description:
811    
812    Tested the jump chain elimination on all architectures (except the
813    hppa).  This is on by default right now and is profitable for the
814    alpha and x86, however, it may not be profitable for the sparc and ppc
815    when compiling the compiler.
816    
817    The gc test will typically jump to a label at the end of the cluster,
818    where there is another jump to an external cluster containing the actual
819    code to invoke gc. This is to allow factoring of common gc invocation
820    sequences. That is to say, we generate:
821    
822            f:
823               testgc
824               ja   L1      % jump if above to L1
825    
826            L1:
827               jmp L2
828    
829    
830    After jump chain elimination the 'ja L1' instructions is converted to
831    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
832    up being implemented in their long form (if L2 is far away) using:
833    
834            jbe     L3      % jump if below or equal to L3
835            jmp     L2
836         L3:
837            ...
838    
839    
840    For large compilation units L2  may be far away.
841    
842    
843    ----------------------------------------------------------------------
844    Name: Matthias Blume
845    Date: 2002/03/11 13:30:00 EST
846    Tag: blume-20020311-mltreeeval
847    Description:
848    
849    A functor parameter was missing.
850    
851    ----------------------------------------------------------------------
852    Name: Allen Leung
853    Date: 2002/03/11 10:30:00 EST
854    Tag: leunga-20020311-runtime-string0
855    Description:
856    
857       The representation of the empty string now points to a
858    legal null terminated C string instead of unit.  It is now possible
859    to convert an ML string into C string with InlineT.CharVector.getData.
860    This compiles into one single machine instruction.
861    
862    ----------------------------------------------------------------------
863    Name: Allen Leung
864    Date: 2002/03/10 23:55:00 EST
865    Tag: leunga-20020310-x86-call
866    Description:
867    
868       Added machine generation for CALL instruction (relative displacement mode)
869    
870    ----------------------------------------------------------------------
871    Name: Matthias Blume
872    Date: 2002/03/08 16:05:00
873    Tag: blume-20020308-entrypoints
874    Description:
875    
876    Version number bumped to 110.39.1.  NEW BOOTFILES!
877    
878    Entrypoints: non-zero offset into a code object where execution should begin.
879    
880    - Added the notion of an entrypoint to CodeObj.
881    - Added reading/writing of entrypoint info to Binfile.
882    - Made runtime system bootloader aware of entrypoints.
883    - Use the address of the label of the first function given to mlriscGen
884      as the entrypoint.  This address is currently always 0, but it will
885      not be 0 once we turn on block placement.
886    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
887      with entry points) from mlriscGen.
888    
889    ----------------------------------------------------------------------
890    Name: Allen Leung
891    Date: 2002/03/07 20:45:00 EST
892    Tag: leunga-20020307-x86-cmov
893    Description:
894    
895       Bug fixes for CMOVcc on x86.
896    
897       1. Added machine code generation for CMOVcc
898       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
899       3. CMOVcc cannot have an immediate operand as argument.
900    
901    ----------------------------------------------------------------------
902    Name: Matthias Blume
903    Date: 2002/03/07 16:15:00 EST
904    Tag: blume-20020307-controls
905    Description:
906    
907    This is a very large but mostly boring patch which makes (almost)
908    every tuneable compiler knob (i.e., pretty much everything under
909    Control.* plus a few other things) configurable via both the command
910    line and environment variables in the style CM did its configuration
911    until now.
912    
913    Try starting sml with '-h' (or, if you are brave, '-H')
914    
915    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
916    implements the underlying generic mechanism.
917    
918    The interface to some of the existing such facilities has changed somewhat.
919    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
920    (The getFoo interface is still there for backward-compatibility, but its
921    use is deprecated.)
922    
923    The ml-build script passes -Cxxx=yyy command-line arguments through so
924    that one can now twiddle the compiler settings when using this "batch"
925    compiler.
926    
927    TODO items:
928    
929    We should go through and throw out all controls that are no longer
930    connected to anything.  Moreover, we should go through and provide
931    meaningful (and correct!) documentation strings for those controls
932    that still are connected.
933    
934    Currently, multiple calls to Controls.new are accepted (only the first
935    has any effect).  Eventually we should make sure that every control
936    is being made (via Controls.new) exactly once.  Future access can then
937    be done using Controls.acc.
938    
939    Finally, it would probably be a good idea to use the getter-setter
940    interface to controls rather than ref cells.  For the time being, both
941    styles are provided by the Controls module, but getter-setter pairs are
942    better if thread-safety is of any concern because they can be wrapped.
943    
944    *****************************************
945    
946    One bug fix: The function blockPlacement in three of the MLRISC
947    backpatch files used to be hard-wired to one of two possibilities at
948    link time (according to the value of the placementFlag).  But (I
949    think) it should rather sense the flag every time.
950    
951    *****************************************
952    
953    Other assorted changes (by other people who did not supply a HISTORY entry):
954    
955    1. the cross-module inliner now works much better (Monnier)
956    2. representation of weights, frequencies, and probabilities in MLRISC
957       changed in preparation of using those for weighted block placement
958       (Reppy, George)
959    
960    ----------------------------------------------------------------------
961    Name: Lal George
962    Date: 2002/03/07 14:44:24 EST 2002
963    Tag: george-20020307-weighted-block-placement
964    
965    Tested the weighted block placement optimization on all architectures
966    (except the hppa) using AMPL to generate the block and edge frequencies.
967    Changes were required in the machine properties to correctly
968    categorize trap instructions. There is an MLRISC flag
969    "weighted-block-placement" that can be used to enable weighted block
970    placement, but this will be ineffective without block/edge
971    frequencies (coming soon).
972    
973    
974    ----------------------------------------------------------------------
975    Name: Lal George
976    Date: 2002/03/05 17:24:48 EST
977    Tag: george-20020305-linkage-cluster
978    
979    In order to support the block placement optimization, a new cluster
980    is generated as the very first cluster (called the linkage cluster).
981    It contains a single jump to the 'real' entry point for the compilation
982    unit. Block placement has no effect on the linkage cluster itself, but
983    all the other clusters  have full freedom in the manner in which they
984    reorder blocks or functions.
985    
986    On the x86 the typical linkage code that is generated is:
987       ----------------------
988            .align 2
989       L0:
990            addl    $L1-L0, 72(%esp)
991            jmp     L1
992    
993    
994            .align  2
995       L1:
996       ----------------------
997    
998    72(%esp) is the memory location for the stdlink register. This
999    must contain the address of the CPS function being called. In the
1000    above example, it contains the address of  L0; before
1001    calling L1 (the real entry point for the compilation unit), it
1002    must contain the address for L1, and hence
1003    
1004            addl $L1-L0, 72(%esp)
1005    
1006    I have tested this on all architectures except the hppa.The increase
1007    in code size is of course negligible
1008    
1009    ----------------------------------------------------------------------
1010    Name: Allen Leung
1011    Date: 2002/03/03 13:20:00 EST
1012    Tag: leunga-20020303-mlrisc-tools
1013    
1014      Added #[ ... ] expressions to mlrisc tools
1015    
1016    ----------------------------------------------------------------------
1017    Name: Matthias Blume
1018    Date: 2002/02/27 12:29:00 EST
1019    Tag: blume-20020227-cdebug
1020    Description:
1021    
1022    - made types in structure C and C_Debug to be equal
1023    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1024    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1025    
1026    ----------------------------------------------------------------------
1027    Name: Matthias Blume
1028    Date: 2002/02/26 12:00:00 EST
1029    Tag: blume-20020226-ffi
1030    Description:
1031    
1032    1. Fixed a minor bug in CM's "noweb" tool:
1033       If numbering is turned off, then truly don't number (i.e., do not
1034       supply the -L option to noweb).  The previous behavior was to supply
1035       -L'' -- which caused noweb to use the "default" line numbering scheme.
1036       Thanks to Chris Richards for pointing this out (and supplying the fix).
1037    
1038    2. Once again, I reworked some aspects of the FFI:
1039    
1040       A. The incomplete/complete type business:
1041    
1042       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1043         gone!
1044       - ML types representing an incomplete type are now *equal* to
1045         ML types representing their corresponding complete types (just like
1046         in C).  This is still safe because ml-nlffigen will not generate
1047         RTTI for incomplete types, nor will it generate functions that
1048         require access to such RTTI.   But when ML code generated from both
1049         incomplete and complete versions of the C type meet, the ML types
1050         are trivially interoperable.
1051    
1052         NOTE:  These changes restore the full generality of the translation
1053         (which was previously lost when I eliminated functorization)!
1054    
1055       B. Enum types:
1056    
1057       - Structure C now has a type constructor "enum" that is similar to
1058         how the "su" constructor works.  However, "enum" is not a phantom
1059         type because each "T enum" has values (and is isomorphic to
1060         MLRep.Signed.int).
1061       - There are generic access operations for enum objects (using
1062         MLRep.Signed.int).
1063       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1064         * The structure contains the definition of type "mlrep" (the ML-side
1065         representation type of the enum).  Normally, mlrep is the same
1066         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1067         then mlrep will be defined as a datatype -- thus facilitating
1068         pattern matching on mlrep values.
1069         ("-ec" will be suppressed if there are duplicate values in an
1070          enumeration.)
1071         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1072         will be generated for each C enum constant xxx.
1073         * Conversion functions m2i and i2m convert between mlrep and
1074         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1075         * Coversion functions c and ml convert between mlrep and "tag enum".
1076         * Access functions (get/set) fetch and store mlrep values.
1077       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1078         enumerations are merged into one single enumeration represented by
1079         structure E_'.
1080    
1081    ----------------------------------------------------------------------
1082  Name: Allen Leung  Name: Allen Leung
1083  Date: 2002/02/25 04:45:00 EST  Date: 2002/02/25 04:45:00 EST
1084  Tag: leunga-20020225-cps-spill  Tag: leunga-20020225-cps-spill
# Line 69  Line 1135 
1135         TOTAL                                   2375.26u  57.21s  48.00g         TOTAL                                   2375.26u  57.21s  48.00g
1136    
1137  As you can see the old cps spill module suffers from some serious  As you can see the old cps spill module suffers from some serious
1138  performance problem but since I cannot decipher the old code fully,  performance problem.  But since I cannot decipher the old code fully,
1139  innstead of patching the problems up, I'm reimplementing it  instead of patching the problems up, I'm reimplementing it
1140  with a different algorithm.  The new code is more modular,  with a different algorithm.  The new code is more modular,
1141  smaller when compiled, and substantially faster  smaller when compiled, and substantially faster
1142  (O(n log n) time and O(n) space).  Timing of the new spill module:  (O(n log n) time and O(n) space).  Timing of the new spill module:

Legend:
Removed from v.1094  
changed lines
  Added in v.1259

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