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 1073, Fri Feb 15 22:07:38 2002 UTC revision 1317, Tue Apr 8 20:42:37 2003 UTC
# Line 14  Line 14 
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16  Name: Matthias Blume  Name: Matthias Blume
17    Date: 2003/04/08 15:42:00 CDT
18    Tag: blume-20030408-listpair
19    Description:
20    
21    1. Added a target 'mlrisc' to installer.
22    
23    2. Added missing elements to structure ListPair.
24    
25    ----------------------------------------------------------------------
26    Name: Allen Leung
27    Date: 2003/01/07 10:40:00 EST
28    Tag: leunga-20030107-int-rem
29    Description:
30    
31    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
32    The arguments to the SUBL instruction were swapped.
33    
34    ----------------------------------------------------------------------
35    Name: Matthias Blume
36    Date: 2002/12/12 16:25:00 EST
37    Tag: blume-20021212-risc-ra
38    Description:
39    
40    Fixed a serious bug in the rewrite code for FP spilling/reloading that
41    sent the RA into an infinite loop when floating point registers get
42    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
43    110.37 and 110.38.)
44    There was another set of potential problems related to the handling of
45    MLRISC annotations (but those did not yet cause real problems, apparently).
46    
47    ----------------------------------------------------------------------
48    Name: Matthias Blume
49    Date: 2002/12/06 22:40:00 EST
50    Tag: blume-20021206-cm-fileid
51    Description:
52    
53    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
54    This fixes the problem of CM getting confused by files that suddenly
55    change their identity (e.g., by getting unlinked and recreated by some
56    text editor such as vi).  There might be a better/cheaper/cleaner way
57    of doing this, but for now this will have to do.
58    
59    ----------------------------------------------------------------------
60    Name: Matthias Blume
61    Date: 2002/10/28 09:50:00 EST
62    Tag: blume-20021028-typecheck
63    Description:
64    
65    Exported structure Typecheck from $smlnj/viscomp/core.cm.
66    
67    ----------------------------------------------------------------------
68    Name: Matthias Blume
69    Date: 2002/10/17 09:10:00 EDT
70    Tag: Release_110_42
71    Description:
72    
73    In good old tradition, there has been a slight hiccup so that we have
74    to patch 110.42 after the fact.   The old release tag has been replaced
75    (see below).
76    
77    The change solves a problem with two competing approaches the
78    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
79    each other's way.
80    
81    This change only affects the runtime system code and the installer script.
82    (No new bootfiles.)
83    
84    ----------------------------------------------------------------------
85    Name: Matthias Blume
86    Date: 2002/10/16 12:00:00 EDT
87    Tag: Release_110_42_removed
88    Description:
89    
90    New working release.  New bootfiles.
91    
92    ----------------------------------------------------------------------
93    Name: Matthias Blume
94    Date: 2002/10/10 13:10:00 EDT
95    Tag: blume-20021010-ppc-divs
96    Description:
97    
98    The mltree operator DIVS must be implemented with an overflow check on
99    the PPC because the hardware indicates divide-by-zero using "overflow" as
100    well.
101    
102    ----------------------------------------------------------------------
103    Name: Matthias Blume
104    Date: 2002/07/23 11:45:00
105    Tag: blume-20020723-smlnj-home
106    Description:
107    
108    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
109    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
110    CM_PATHCONFIG is also set), the path configuration file is assumed
111    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
112    the entire tree to some other place and everything will "just work".
113    
114    (Companion commands such as ml-build and ml-makedepend also sense this
115    variable.)
116    
117    ----------------------------------------------------------------------
118    Name: Matthias Blume
119    Date: 2002/07/12 21:19:00 EDT
120    Tag: blume-20020712-liveness
121    Description:
122    
123    Exported two useful "step" functions from liveness module (MLRISC).
124    
125    ----------------------------------------------------------------------
126    Name: Matthias Blume
127    Date: 2002/07/05 16:00 EDT
128    Tag: Release_110_41
129    Description:
130    
131    New working release.  New bootfiles.
132    
133    ----------------------------------------------------------------------
134    Name: Matthias Blume
135    Date: 2002/07/05 10:25:00 EDT
136    Tag: blume-20020705-btimp
137    Description:
138    
139    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
140    other clients can set up backtracing support.
141    
142    ----------------------------------------------------------------------
143    Name: Matthias Blume
144    Date: 2002/06/25 17:23:00 EDT
145    Tag: blume-20020625-fpmax
146    Description:
147    
148    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
149    case.  (The sense of the isNaN test was reversed -- which made min and
150    max always return their first argument.)
151    
152    ----------------------------------------------------------------------
153    Name: Matthias Blume
154    Date: 2002/06/11
155    Tag: blume-20020611-unixpath
156    Description:
157    
158    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
159    
160    ----------------------------------------------------------------------
161    Name: Matthias Blume
162    Date: 2002/06/10 16:35:00 EDT
163    Tag: blume-20020610-ieeereal
164    Description:
165    
166    I back-ported my implementation of IEEEReal.fromString from the
167    idlbasis-devel branch so that we can test it.
168    
169    Another small change is that ppDec tries to give more information
170    than just "<sig>" in the case of functors.  However, this code is
171    broken in some mysterious way if the functor's body's signature
172    has not been declared by ascription but gets inferred from the
173    implementation.  This needs fixing...
174    
175    ----------------------------------------------------------------------
176    Name: Matthias Blume
177    Date: 2002/05/31
178    Tag: blume-20020531-btrace-mode
179    Description:
180    
181    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
182    the wayside when I switched over to using Controls everywhere.)
183    
184    ----------------------------------------------------------------------
185    Name: Lal George
186    Date: 2002/05/23 12:21:40 EDT
187    Tag: george-20020523-visual-labels
188    Description:
189    
190    Labels are now displayed in the graphical output to make
191    the fall-through and target blocks obvious.
192    
193    ----------------------------------------------------------------------
194    Name: Matthias Blume
195    Date: 2002/05/22 11:03:00 EDT
196    Tag: blume-20020522-shrink
197    Description:
198    
199    John tweaked yesterday's fix for 1131 to handle an out-of-memory
200    situation that comes up when allocating huge arrays.
201    
202    ----------------------------------------------------------------------
203    Name: Matthias Blume
204    Date: 2002/05/21 16:00:00 EDT
205    Tag: Release_110_40
206    Description:
207    
208    New working release (110.40).  New bootfiles.
209    
210    [Also: John Reppy fixed GC bug 1131.]
211    
212    ----------------------------------------------------------------------
213    Name: Matthias Blume
214    Date: 2002/05/21 12:35:00 EDT
215    Tag: blume-20020521-cmdoc
216    Description:
217    
218    CM documentation update.
219    
220    ----------------------------------------------------------------------
221    Name: Matthias Blume
222    Date: 2002/05/21 10:55:00 EDT
223    Tag: blume-20020521-misc
224    Description:
225    
226    - John tweaked runtime to be silent on heap export (except when
227      GC messages are on).
228    - I added a few more things (cross-compiling versions of CMB) to
229      config/preloads (as suggestions).
230    
231    ----------------------------------------------------------------------
232    Name: Matthias Blume
233    Date: 2002/05/20 22:25:00 EDT
234    Tag: blume-20020520-controls
235    Description:
236    
237    - Added ControlUtil structure to control-lib.cm.
238    - Use it throughout.
239    - Used Controls facility to define MLRISC controls (as opposed to
240      registering MLRISC control ref cells with Controls after the fact)
241    - Fixed messed-up controls priorities.
242    
243    * Removed again all the stuff from config/preloads that one wouldn't
244      be able to preload at the time the initial heap image is built.
245      (Many libraries, e.g., CML, do not exist yet at this time.  The
246       only libraries that can be preloaded via config/preloads are those
247       that come bundled with the bootfiles.)
248    
249    ----------------------------------------------------------------------
250    Name: Matthias Blume
251    Date: 2002/05/20 10:59:00 EDT
252    Tag: blume-20020520-preloads
253    Description:
254    
255    Added a lot of commented-out suggestions for things to be included
256    in config/preloads.
257    
258    ----------------------------------------------------------------------
259    Name: Allen Leung
260    Date: 2002/05/18 14:20:00 EDT
261    Tag: leunga-20020518-mdl
262    Description:
263    
264    o Made the mdl tool stuff compile and run again.
265    o I've disabled all the stuff that depends on RTL specifications; they
266      are all badly broken anyway.
267    
268    ----------------------------------------------------------------------
269    Name: Matthias Blume
270    Date: 2002/05/17 16:49:00 EDT
271    Tag: blume-20020517-controls
272    Description:
273    
274    0. John Reppy made several modifications to the SML/NJ library.
275       In particular, there is a shiny new controls-lib.cm.
276    
277    1. Pushed new controls interface through compiler so that everything
278       compiles again.
279    
280    2. Added FormatComb and FORMAT_COMB to the CML version of the
281       SML/NJ library (so that CML compiles again).
282    
283    3. Modified init scripts because XXX_DEFAULT environment variables
284       are no longer with us.  (Boot-time initialization is now done
285       using the same environment variables that are also used for
286       startup-time initialization of controls.)
287    
288    ----------------------------------------------------------------------
289    Name: Lal George
290    Date: 2002/05/15 09:20:10 EDT
291    Tag: george-20020515-pseudo-op-decls
292    Description:
293    
294            All pseudo-ops emitted before the first segment declaration
295    such as TEXT, DATA, and BSS directives are assumed to be global
296    declarations and are emitted first in the assembly file. This is
297    useful in a number of situations where one has pseudo-ops that are not
298    specific to any segment, and also works around the constraint that one
299    cannot have client pseudo-ops in the TEXT segment.
300    
301            Because no segment is associated with these declarations it is
302    an error to allocate any space or objects before the first segment
303    directive and an exception will be raised. However, we cannot make
304    this check for client pseudo-ops.
305    
306            These top level declarations are a field in the CFG graph_info.
307    In theory you can continue to add to this field after the CFG has been
308    built -- provided you know what you are doing;-)
309    
310    ----------------------------------------------------------------------
311    Name: Matthias Blume
312    Date: 2002/05/13 16:40:00 EDT
313    Tag: blume-20020513-pp-etc
314    Description:
315    
316    A few minor bugfixes:
317    
318      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
319        (Bogus printouts for redefined bindings still occur.  Compiler
320        bug should no longer occur now.  We need to redo the prettyprinter
321        from scratch.)
322    
323      - CM pathname printer now also adds escape sequences for ( and )
324    
325      - commend and docu fixes for ml-nlffi
326    
327    ----------------------------------------------------------------------
328    Name: Matthias Blume
329    Date: 2002/05/10 16:40:00 EDT
330    Tag: blume-20020510-erg-textio
331    Description:
332    
333    Applied the following bugfix provided by Emden Gansner:
334    
335        Output is corrupted when outputSubstr is used rather than output.
336    
337        The problem occurs when a substring
338    
339            ss = (s, dataStart, dataLen)
340    
341        where dataStart > 0, fills a stream buffer with avail bytes left.
342        avail bytes of s, starting at index dataStart, are copied into the
343        buffer, the buffer is flushed, and then the remaining dataLen-avail
344        bytes of ss are copied into the beginning of the buffer. Instead of
345        starting this copy at index dataStart+avail in s, the current code
346        starts the copy at index avail.
347    
348        Fix:
349        In text-io-fn.sml, change line 695 from
350             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
351        to
352             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
353    
354    ----------------------------------------------------------------------
355    Name: Matthias Blume
356    Date: 2002/04/12 13:55:00 EDT
357    Tag: blume-20020412-assyntax
358    Description:
359    
360    1. Grabbed newer assyntax.h from the XFree86 project.
361    2. Fiddled with how to compile X86.prim.asm without warnings.
362    3. (Very) Minor cleanup in CM.
363    
364    ----------------------------------------------------------------------
365    Name: Matthias Blume
366    Date: 2002/04/01 (no joke!) 17:07:00 EST
367    Tag: blume-20020401-x86div
368    Description:
369    
370    Added full support for div/mod/rem/quot on the x86, using the machine
371    instruction's two results (without clumsily recomputing the remainder)
372    directly where appropriate.
373    
374    Some more extensive power-of-two support was added to the x86 instruction
375    selector (avoiding expensive divs, mods, and muls where they can be
376    replaced with cheaper shifts and masks).  However, this sort of thing
377    ought to be done earlier, e.g., within the CPS optimizer so that
378    all architectures benefit from it.
379    
380    The compiler compiles to a fixed point, but changes might be somewhat
381    fragile nevertheless.  Please, report any strange things that you might
382    see wrt. div/mod/quot/rem...
383    
384    ----------------------------------------------------------------------
385    Name: Matthias Blume
386    Date: 2002/03/29 17:22:00
387    Tag: blume-20020329-div
388    Description:
389    
390    Fixed my broken div/mod logic.  Unfortunately, this means that the
391    inline code for div/mod now has one more comparison than before.
392    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
393    The problem was with quotient = 0, because that alone does not tell
394    us which way the rounding went.  One then has to look at whether
395    remainder and divisor have the same sign...  :(
396    
397    Anyway, I replaced the bootfiles with fresh ones...
398    
399    ----------------------------------------------------------------------
400    Name: Matthias Blume
401    Date: 2002/03/29 14:10:00 EST
402    Tag: blume-20020329-inlprims
403    Description:
404    
405    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
406    
407    Primops have changed. This means that the bin/boot-file formats have
408    changed as well.
409    
410    To make sure that there is no confusion, I made a new version.
411    
412    
413    CHANGES:
414    
415    * removed REMT from mltree (remainder should never overflow).
416    
417    * added primops to deal with divisions of all flavors to the frontend
418    
419    * handled these primops all the way through so they map to their respective
420      MLRISC support
421    
422    * used these primops in the implementation of Int, Int32, Word, Word32
423    
424    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
425    
426    * parameterized INLMIN, INLMAX, and INLABS by a numkind
427    
428    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
429      floating point
430    
431    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
432      and Real (but Real.abs maps to a separate floating-point-only primop)
433    
434    
435    TODO items:
436    
437    * Hacked Alpha32 instruction selection, disabling the selection of REMx
438      instructions because the machine instruction encoder cannot handle
439      them.  (Hppa, PPC, and Sparc instruction selection did not handle
440      REM in the first place, and REM is supported by the x86 machine coder.)
441    
442    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
443      selection phase.  (The two can be streamlined because the hardware
444      delivers both quotient and remainder at the same time anyway.)
445    
446    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
447      (Currently the behavior is inconsistent both across architectures and
448      wrt. the draft Basis spec.)
449    
450    * Word8 should eventually be handled natively, too.
451    
452    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
453      as if there currently is no execution path that could trigger it in
454      SML/NJ.  (The assumptions underlying functions arith and promotable do not
455      hold for things like multiplication and division.)
456    
457    ----------------------------------------------------------------------
458    Name: Matthias Blume
459    Date: 2002/03/27 16:27:00 EST
460    Tag: blume-20020327-mlrisc-divisions
461    Description:
462    
463    Added support for all four division operations (ML's div, mod, quot,
464    and rem) to MLRISC.  In the course of doing so, I also rationalized
465    the naming (no more annoying switch-around of DIV and QUOT), by
466    parameterizing the operation by div_rounding_mode (which can be either
467    DIV_TO_ZERO or DIV_TO_NEGINF).
468    
469    The generic MLTreeGen functor takes care of compiling all four
470    operations down to only round-to-zero div.
471    
472    Missing pieces:
473    
474      * Doing something smarter than relying on MLTreeGen on architectures
475        like, e.g., the x86 where hardware division delivers both quotient and
476        remainder at the same time.  With this, the implementation of the
477        round-to-neginf operations could be further streamlined.
478    
479      * Remove inlining support for div/mod/rem from the frontend and replace it
480        with primops that get carried through to the backend.  Do this for all
481        int and word types.
482    
483    ----------------------------------------------------------------------
484    Name: Matthias Blume
485    Date: 2002/03/25 17:25:00 EST
486    Tag: blume-20020325-divmod
487    Description:
488    
489    I improved (hopefully without breaking them) the implementation of Int.div,
490    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
491    advantage of the following observations:
492    
493      Let  q = x quot y      r = x rem y
494           d = x div  y      m = x mod y
495    
496    where "quot" is the round-to-zero version of integer division that
497    hardware usually provides.  Then we have:
498    
499         r = x - q * y        where neither the * nor the - will overflow
500         d = if q >= 0 orelse x = q * y then q else q - 1
501                              where neither the * nor the - will overflow
502         m = if q >= 0 orelse r = 0 then r else r + y
503                              where the + will not overflow
504    
505    This results in substantial simplification of the generated code.
506    The following table shows the number of CFG nodes and edges generated
507    for
508            fun f (x, y) = x OPER y
509            (* with OPER \in div, mod, quot, rem *)
510    
511    
512        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
513        --------------------------------------------------------
514         div |         24 |         39 |         12 |         16
515         mod |         41 |         71 |         12 |         16
516        quot |          8 |         10 |          8 |         10
517         rem |         10 |         14 |          8 |         10
518    
519    
520    ----------------------------------------------------------------------
521    Name: Matthias Blume
522    Date: 2002/03/25 22:06:00 EST
523    Tag: blume-20020325-cprotobug
524    Description:
525    
526    Fixed a bug in cproto (c prototype decoder).
527    
528    ----------------------------------------------------------------------
529    Name: Matthias Blume
530    Date: 2002/03/25 16:00:00 EST
531    Tag: blume-20020325-raw-primops
532    Description:
533    
534    I did some cleanup to Allen's new primop code and
535    replaced yesterday's bootfiles with new ones.
536    (But they are stored in the same place.)
537    
538    ----------------------------------------------------------------------
539    Name: Matthias Blume
540    Date: 2002/03/24 22:40:00 EST
541    Tag: blume-20020324-bootfiles
542    Description:
543    
544    Made the bootfiles that Allen asked for.
545    
546    ----------------------------------------------------------------------
547    Name: Allen Leung
548    Date: 2002/03/23 15:50:00 EST
549    Tag: leunga-20020323-flint-cps-rcc-primops
550    Description:
551    
552      1. Changes to FLINT primops:
553    
554        (* make a call to a C-function;
555         * The primop carries C function prototype information and specifies
556         * which of its (ML-) arguments are floating point. C prototype
557         * information is for use by the backend, ML information is for
558         * use by the CPS converter. *)
559      | RAW_CCALL of { c_proto: CTypes.c_proto,
560                       ml_args: ccall_type list,
561                       ml_res_opt: ccall_type option,
562                       reentrant : bool
563                     } option
564       (* Allocate uninitialized storage on the heap.
565        * The record is meant to hold short-lived C objects, i.e., they
566        * are not ML pointers.  With the tag, the representation is
567        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
568        *)
569      | RAW_RECORD of {tag:bool,sz:int}
570      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
571    
572      2.  These CPS primops are now overloaded:
573    
574           rawload of {kind:numkind}
575           rawstore of {kind:numkind}
576    
577          The one argument form is:
578    
579             rawload {kind} address
580    
581          The two argument form is:
582    
583             rawload {kind} [ml object, byte-offset]
584    
585      3. RAW_CCALL/RCC now takes two extra arguments:
586    
587         a. The first is whether the C call is reentrant, i.e., whether
588            ML state should be saved and restored.
589         b. The second argument is a string argument specifying the name of
590            library and the C function.
591    
592         These things are currently not handled in the code generator, yet.
593    
594      4. In CProto,
595    
596         An encoding type of "bool" means "ml object" and is mapped into
597         C prototype of PTR.  Note that "bool" is different than "string",
598         even though "string" is also mapped into PTR, because "bool"
599         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
600    
601      5. Pickler/unpicker
602    
603         Changed to handle RAW_RECORD and newest RAW_CCALL
604    
605      6. MLRiscGen,
606    
607         1. Changed to handle the new rawload/rawstore/rawrecord operators.
608         2. Code for handling C Calls has been moved to a new module CPSCCalls,
609            in the file CodeGen/cpscompile/cps-c-calls.sml
610    
611      7. Added the conditional move operator
612    
613             condmove of branch
614    
615         to cps.  Generation of this is still buggy so it is currently
616         disabled.
617    
618    ----------------------------------------------------------------------
619    Name: Lal George
620    Date: 2002/03/22 14:18:25 EST
621    Tag: george-20020322-cps-branch-prob
622    Description:
623    
624    Implemented the Ball-Larus branch prediction-heuristics, and
625    incorporated graphical viewers for control flow graphs.
626    
627    Ball-Larus Heuristics:
628    ---------------------
629    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
630    
631    By design it uses the Dempster-Shafer theory for combining
632    probabilities.  For example, in the function:
633    
634        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
635    
636    the ball-larus heuristics predicts that the n=0 is unlikely
637    (OH-heuristic), and the 'then' branch is unlikely because of the
638    RH-heuristic -- giving the 'then' branch an even lower combined
639    probability using the Dempster-Shafer theory.
640    
641    Finally, John Reppy's loop analysis in MLRISC, further lowers the
642    probability of the 'then' branch because of the loop in the else
643    branch.
644    
645    
646    Graphical Viewing:
647    ------------------
648    I merely plugged in Allen's graphical viewers into the compiler. The
649    additional code is not much. At the top level, saying:
650    
651            Control.MLRISC.getFlag "cfg-graphical-view" := true;
652    
653    will display the graphical view of the control flow graph just before
654    back-patching.  daVinci must be in your path for this to work. If
655    daVinci is not available, then the default viewer can be changed
656    using:
657    
658            Control.MLRISC.getString "viewer"
659    
660    which can be set to "dot" or "vcg" for the corresponding viewers. Of
661    course, these viewers must be in your path.
662    
663    The above will display the compilation unit at the level of clusters,
664    many of which are small, boring, and un-interesting. Also setting:
665    
666            Control.MLRISC.getInt "cfg-graphical-view_size"
667    
668    will display clusters that are larger than the value set by the above.
669    
670    
671    ----------------------------------------------------------------------
672    Name: Matthias Blume
673    Date: 2002/03/21 22:20:00 EST
674    Tag: blume-20020321-kmp-bugfix
675    Description:
676    
677    Changed the interface to the KMP routine in PreString and fixed
678    a minor bug in one place where it was used.
679    
680    ----------------------------------------------------------------------
681    Name: Allen Leung
682    Date: 2002/03/21 20:30:00 EST
683    Tag: leunga-20020321-cfg
684    Description:
685    
686      Fixed a potential problem in cfg edge splitting.
687    
688    ----------------------------------------------------------------------
689    Name: Allen Leung
690    Date: 2002/03/21 17:15:00 EST
691    Tag: leunga-20020321-x86-fp-cfg
692    Description:
693    
694      1. Recoded the buggy parts of x86-fp.
695    
696         a. All the block reordering code has been removed.
697            We now depend on the block placement phases to do this work.
698    
699         b. Critical edge splitting code has been simplified and moved into the
700            CFG modules, as where they belong.
701    
702         Both of these were quite buggy and complex.  The code is now much, much
703         simpler.
704    
705      2. X86 backend.
706    
707         a. Added instructions for 64-bit support.  Instruction selection for
708            64-bit has not been committed, however, since that
709            requires changes to MLTREE which haven't been approved by
710            Lal and John.
711    
712         b. Added support for FUCOMI and FUCOMIP when generating code for
713            PentiumPro and above.  We only generate these instructions in
714            the fast-fp mode.
715    
716         c. Added cases for JP and JNP in X86FreqProps.
717    
718      3. CFG
719    
720         CFG now has a bunch of methods for edge splitting and merging.
721    
722      4. Machine description.
723    
724         John's simplification of MLTREE_BASIS.fcond broke a few machine
725         description things:
726    
727         rtl-build.{sig,sml} and hppa.mdl fixed.
728    
729         NOTE: the machine description stuff in the repository is still broken.
730               Again, I can't put my fixes in because that involves
731               changes to MLTREE.
732    
733    ----------------------------------------------------------------------
734    Name: Matthias Blume
735    Date: 2002/03/20 15:55:00 EST
736    Tag: blume-20020320-kmp
737    Description:
738    
739    Implemented Knuth-Morris-Pratt string matching in PreString and used
740    it for String.isSubstring, Substring.isSubstring, and
741    Substring.position.
742    
743    (Might need some stress-testing.  Simple examples worked fine.)
744    
745    ----------------------------------------------------------------------
746    Name: Matthias Blume
747    Date: 2002/03/19 16:37:00 EST
748    Tag: blume-20020319-witnesses
749    Description:
750    
751    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
752    
753    This implements a generic mechanism for changing constness qualifiers
754    anywhere within big C types without resorting to outright "casts".
755    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
756    modify the constness at the outermost level.)
757    The implementation of "convert" is based on the idea of "witness"
758    values -- values that are not used by the operation but whose types
759    "testify" to their applicability.  On the implementation side, "convert"
760    is simply a projection (returning its second curried argument).  With
761    cross-module inlining, it should not result in any machine code being
762    generated.
763    
764    ----------------------------------------------------------------------
765    Name: Matthias Blume
766    Date: 2002/03/15 16:40:00 EST
767    Tag: blume-20020315-basis
768    Description:
769    
770    Provided (preliminary?) implementations for
771    
772      {String,Substring}.{concatWith,isSuffix,isSubstring}
773    
774    and
775    
776      Substring.full
777    
778    Those are in the Basis spec but they were missing in SML/NJ.
779    
780    ----------------------------------------------------------------------
781    Name: Matthias Blume
782    Date: 2002/03/14 21:30:00 EST
783    Tag: blume-20020314-controls
784    Description:
785    
786    Controls:
787    ---------
788    
789    1. Factored out the recently-added Controls : CONTROLS stuff and put
790       it into its own library $/controls-lib.cm.  The source tree for
791       this is under src/smlnj-lib/Controls.
792    
793    2. Changed the names of types and functions in this interface, so they
794       make a bit more "sense":
795    
796          module -> registry
797          'a registry -> 'a group
798    
799    3. The interface now deals in ref cells only.  The getter/setter interface
800       is (mostly) gone.
801    
802    4. Added a function that lets one register an already-existing ref cell.
803    
804    5. Made the corresponding modifications to the rest of the code so that
805       everything compiles again.
806    
807    6. Changed the implementation of Controls.MLRISC back to something closer
808       to the original.  In particular, this module (and therefore MLRISC)
809       does not depend on Controls.  There now is some link-time code in
810       int-sys.sml that registers the MLRISC controls with the Controls
811       module.
812    
813    CM:
814    ---
815    
816      * One can now specify the lambda-split aggressiveness in init.cmi.
817    
818    ----------------------------------------------------------------------
819    Name: Allen Leung
820    Date: 2002/03/13 17:30:00 EST
821    Tag: leunga-20020313-x86-fp-unary
822    Description:
823    
824    Bug fix for:
825    
826    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
827    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
828    > - fun f(x,(y,z)) = Real.~ y;
829    > [autoloading]
830    > [autoloading done]
831    >       fchsl   (%eax), 184(%esp)
832    > Error: MLRisc bug: X86MCEmitter.emitInstr
833    >
834    > uncaught exception Error
835    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
836    
837    The problem was that the code generator did not generate any fp registers
838    in this case, and the ra didn't know that it needed to run the X86FP phase to
839    translate the pseudo fp instruction.   This only happened with unary fp
840    operators in certain situations.
841    
842    ----------------------------------------------------------------------
843    Name: Matthias Blume
844    Date: 2002/03/13 14:00:00 EST
845    Tag: blume-20020313-overload-etc
846    Description:
847    
848    1. Added _overload as a synonym for overload for backward compatibility.
849       (Control.overloadKW must be true for either version to be accepted.)
850    
851    2. Fixed bug in install script that caused more things to be installed
852       than what was requested in config/targets.
853    
854    3. Made CM aware of the (_)overload construct so that autoloading
855       works.
856    
857    ----------------------------------------------------------------------
858    Name: Matthias Blume
859    Date: 2002/03/12 22:03:00 EST
860    Tag: blume-20020312-url
861    Description:
862    
863    Forgot to update BOOT and srcarchiveurl.
864    
865    ----------------------------------------------------------------------
866    Name: Matthias Blume
867    Date: 2002/03/12 17:30:00 EST
868    Tag: blume-20020312-version110392
869    Description:
870    
871    Yet another version number bump (because of small changes to the
872    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
873    
874    Changes:
875    
876      The new pid generation scheme described a few weeks ago was overly
877      complicated.  I implemented a new mechanism that is simpler and
878      provides a bit more "stability":  Once CM has seen a compilation
879      unit, it keeps its identity constant (as long as you do not delete
880      those crucial CM/GUID/* files).  This means that when you change
881      an interface, compile, then go back to the old interface, and
882      compile again, you arrive at the original pid.
883    
884      There now also is a mechanism that instructs CM to use the plain
885      environment hash as a module's pid (effectively making its GUID
886      the empty string).  For this, "noguid" must be specified as an
887      option to the .sml file in question within its .cm file.
888      This is most useful for code that is being generated by tools such
889      as ml-nlffigen (because during development programmers tend to
890      erase the tool's entire output directory tree including CM's cached
891      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
892      revert to the old, broken behavior of SML/NJ, but in specific cases
893      where there is no danger of interface confusion, its use is ok
894      (I think).
895    
896      ml-nlffigen by default generates "noguid" annotations.  They can be
897      turned off by specifying -guid in its command line.
898    
899    ----------------------------------------------------------------------
900    Name: Lal George
901    Date: 2002/03/12 12 14:42:36 EST
902    Tag: george-20020312-frequency-computation
903    Description:
904    
905    Integrated jump chaining and static block frequency into the
906    compiler. More details and numbers later.
907    
908    ----------------------------------------------------------------------
909    Name: Lal George
910    Date: 2002/03/11 11 22:38:53 EST
911    Tag: george-20020311-jump-chain-elim
912    Description:
913    
914    Tested the jump chain elimination on all architectures (except the
915    hppa).  This is on by default right now and is profitable for the
916    alpha and x86, however, it may not be profitable for the sparc and ppc
917    when compiling the compiler.
918    
919    The gc test will typically jump to a label at the end of the cluster,
920    where there is another jump to an external cluster containing the actual
921    code to invoke gc. This is to allow factoring of common gc invocation
922    sequences. That is to say, we generate:
923    
924            f:
925               testgc
926               ja   L1      % jump if above to L1
927    
928            L1:
929               jmp L2
930    
931    
932    After jump chain elimination the 'ja L1' instructions is converted to
933    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
934    up being implemented in their long form (if L2 is far away) using:
935    
936            jbe     L3      % jump if below or equal to L3
937            jmp     L2
938         L3:
939            ...
940    
941    
942    For large compilation units L2  may be far away.
943    
944    
945    ----------------------------------------------------------------------
946    Name: Matthias Blume
947    Date: 2002/03/11 13:30:00 EST
948    Tag: blume-20020311-mltreeeval
949    Description:
950    
951    A functor parameter was missing.
952    
953    ----------------------------------------------------------------------
954    Name: Allen Leung
955    Date: 2002/03/11 10:30:00 EST
956    Tag: leunga-20020311-runtime-string0
957    Description:
958    
959       The representation of the empty string now points to a
960    legal null terminated C string instead of unit.  It is now possible
961    to convert an ML string into C string with InlineT.CharVector.getData.
962    This compiles into one single machine instruction.
963    
964    ----------------------------------------------------------------------
965    Name: Allen Leung
966    Date: 2002/03/10 23:55:00 EST
967    Tag: leunga-20020310-x86-call
968    Description:
969    
970       Added machine generation for CALL instruction (relative displacement mode)
971    
972    ----------------------------------------------------------------------
973    Name: Matthias Blume
974    Date: 2002/03/08 16:05:00
975    Tag: blume-20020308-entrypoints
976    Description:
977    
978    Version number bumped to 110.39.1.  NEW BOOTFILES!
979    
980    Entrypoints: non-zero offset into a code object where execution should begin.
981    
982    - Added the notion of an entrypoint to CodeObj.
983    - Added reading/writing of entrypoint info to Binfile.
984    - Made runtime system bootloader aware of entrypoints.
985    - Use the address of the label of the first function given to mlriscGen
986      as the entrypoint.  This address is currently always 0, but it will
987      not be 0 once we turn on block placement.
988    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
989      with entry points) from mlriscGen.
990    
991    ----------------------------------------------------------------------
992    Name: Allen Leung
993    Date: 2002/03/07 20:45:00 EST
994    Tag: leunga-20020307-x86-cmov
995    Description:
996    
997       Bug fixes for CMOVcc on x86.
998    
999       1. Added machine code generation for CMOVcc
1000       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1001       3. CMOVcc cannot have an immediate operand as argument.
1002    
1003    ----------------------------------------------------------------------
1004    Name: Matthias Blume
1005    Date: 2002/03/07 16:15:00 EST
1006    Tag: blume-20020307-controls
1007    Description:
1008    
1009    This is a very large but mostly boring patch which makes (almost)
1010    every tuneable compiler knob (i.e., pretty much everything under
1011    Control.* plus a few other things) configurable via both the command
1012    line and environment variables in the style CM did its configuration
1013    until now.
1014    
1015    Try starting sml with '-h' (or, if you are brave, '-H')
1016    
1017    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1018    implements the underlying generic mechanism.
1019    
1020    The interface to some of the existing such facilities has changed somewhat.
1021    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1022    (The getFoo interface is still there for backward-compatibility, but its
1023    use is deprecated.)
1024    
1025    The ml-build script passes -Cxxx=yyy command-line arguments through so
1026    that one can now twiddle the compiler settings when using this "batch"
1027    compiler.
1028    
1029    TODO items:
1030    
1031    We should go through and throw out all controls that are no longer
1032    connected to anything.  Moreover, we should go through and provide
1033    meaningful (and correct!) documentation strings for those controls
1034    that still are connected.
1035    
1036    Currently, multiple calls to Controls.new are accepted (only the first
1037    has any effect).  Eventually we should make sure that every control
1038    is being made (via Controls.new) exactly once.  Future access can then
1039    be done using Controls.acc.
1040    
1041    Finally, it would probably be a good idea to use the getter-setter
1042    interface to controls rather than ref cells.  For the time being, both
1043    styles are provided by the Controls module, but getter-setter pairs are
1044    better if thread-safety is of any concern because they can be wrapped.
1045    
1046    *****************************************
1047    
1048    One bug fix: The function blockPlacement in three of the MLRISC
1049    backpatch files used to be hard-wired to one of two possibilities at
1050    link time (according to the value of the placementFlag).  But (I
1051    think) it should rather sense the flag every time.
1052    
1053    *****************************************
1054    
1055    Other assorted changes (by other people who did not supply a HISTORY entry):
1056    
1057    1. the cross-module inliner now works much better (Monnier)
1058    2. representation of weights, frequencies, and probabilities in MLRISC
1059       changed in preparation of using those for weighted block placement
1060       (Reppy, George)
1061    
1062    ----------------------------------------------------------------------
1063    Name: Lal George
1064    Date: 2002/03/07 14:44:24 EST 2002
1065    Tag: george-20020307-weighted-block-placement
1066    
1067    Tested the weighted block placement optimization on all architectures
1068    (except the hppa) using AMPL to generate the block and edge frequencies.
1069    Changes were required in the machine properties to correctly
1070    categorize trap instructions. There is an MLRISC flag
1071    "weighted-block-placement" that can be used to enable weighted block
1072    placement, but this will be ineffective without block/edge
1073    frequencies (coming soon).
1074    
1075    
1076    ----------------------------------------------------------------------
1077    Name: Lal George
1078    Date: 2002/03/05 17:24:48 EST
1079    Tag: george-20020305-linkage-cluster
1080    
1081    In order to support the block placement optimization, a new cluster
1082    is generated as the very first cluster (called the linkage cluster).
1083    It contains a single jump to the 'real' entry point for the compilation
1084    unit. Block placement has no effect on the linkage cluster itself, but
1085    all the other clusters  have full freedom in the manner in which they
1086    reorder blocks or functions.
1087    
1088    On the x86 the typical linkage code that is generated is:
1089       ----------------------
1090            .align 2
1091       L0:
1092            addl    $L1-L0, 72(%esp)
1093            jmp     L1
1094    
1095    
1096            .align  2
1097       L1:
1098       ----------------------
1099    
1100    72(%esp) is the memory location for the stdlink register. This
1101    must contain the address of the CPS function being called. In the
1102    above example, it contains the address of  L0; before
1103    calling L1 (the real entry point for the compilation unit), it
1104    must contain the address for L1, and hence
1105    
1106            addl $L1-L0, 72(%esp)
1107    
1108    I have tested this on all architectures except the hppa.The increase
1109    in code size is of course negligible
1110    
1111    ----------------------------------------------------------------------
1112    Name: Allen Leung
1113    Date: 2002/03/03 13:20:00 EST
1114    Tag: leunga-20020303-mlrisc-tools
1115    
1116      Added #[ ... ] expressions to mlrisc tools
1117    
1118    ----------------------------------------------------------------------
1119    Name: Matthias Blume
1120    Date: 2002/02/27 12:29:00 EST
1121    Tag: blume-20020227-cdebug
1122    Description:
1123    
1124    - made types in structure C and C_Debug to be equal
1125    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1126    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1127    
1128    ----------------------------------------------------------------------
1129    Name: Matthias Blume
1130    Date: 2002/02/26 12:00:00 EST
1131    Tag: blume-20020226-ffi
1132    Description:
1133    
1134    1. Fixed a minor bug in CM's "noweb" tool:
1135       If numbering is turned off, then truly don't number (i.e., do not
1136       supply the -L option to noweb).  The previous behavior was to supply
1137       -L'' -- which caused noweb to use the "default" line numbering scheme.
1138       Thanks to Chris Richards for pointing this out (and supplying the fix).
1139    
1140    2. Once again, I reworked some aspects of the FFI:
1141    
1142       A. The incomplete/complete type business:
1143    
1144       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1145         gone!
1146       - ML types representing an incomplete type are now *equal* to
1147         ML types representing their corresponding complete types (just like
1148         in C).  This is still safe because ml-nlffigen will not generate
1149         RTTI for incomplete types, nor will it generate functions that
1150         require access to such RTTI.   But when ML code generated from both
1151         incomplete and complete versions of the C type meet, the ML types
1152         are trivially interoperable.
1153    
1154         NOTE:  These changes restore the full generality of the translation
1155         (which was previously lost when I eliminated functorization)!
1156    
1157       B. Enum types:
1158    
1159       - Structure C now has a type constructor "enum" that is similar to
1160         how the "su" constructor works.  However, "enum" is not a phantom
1161         type because each "T enum" has values (and is isomorphic to
1162         MLRep.Signed.int).
1163       - There are generic access operations for enum objects (using
1164         MLRep.Signed.int).
1165       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1166         * The structure contains the definition of type "mlrep" (the ML-side
1167         representation type of the enum).  Normally, mlrep is the same
1168         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1169         then mlrep will be defined as a datatype -- thus facilitating
1170         pattern matching on mlrep values.
1171         ("-ec" will be suppressed if there are duplicate values in an
1172          enumeration.)
1173         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1174         will be generated for each C enum constant xxx.
1175         * Conversion functions m2i and i2m convert between mlrep and
1176         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1177         * Coversion functions c and ml convert between mlrep and "tag enum".
1178         * Access functions (get/set) fetch and store mlrep values.
1179       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1180         enumerations are merged into one single enumeration represented by
1181         structure E_'.
1182    
1183    ----------------------------------------------------------------------
1184    Name: Allen Leung
1185    Date: 2002/02/25 04:45:00 EST
1186    Tag: leunga-20020225-cps-spill
1187    
1188    This is a new implementation of the CPS spill phase.
1189    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1190    In case of problems, replace it with the old file spill.sml
1191    
1192    The current compiler runs into some serious performance problems when
1193    constructing a large record.  This can happen when we try to compile a
1194    structure with many items.  Even a very simple structure like the following
1195    makes the compiler slow down.
1196    
1197        structure Foo = struct
1198           val x_1 = 0w1 : Word32.int
1199           val x_2 = 0w2 : Word32.int
1200           val x_3 = 0w3 : Word32.int
1201           ...
1202           val x_N = 0wN : Word32.int
1203        end
1204    
1205    The following table shows the compile time, from N=1000 to N=4000,
1206    with the old compiler:
1207    
1208    N
1209    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1210           MLRISC ra                               0.06u  0.00s  0.05g
1211              (spills = 0 reloads = 0)
1212           TOTAL                                   0.63u  0.07s  0.21g
1213    
1214    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1215           MLRISC ra                               5.68u  0.59s  3.93g
1216              (spills = 0 reloads = 0)
1217           TOTAL                                   14.71u  0.99s  4.81g
1218    
1219    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1220           MLRISC ra                               5.54u  0.65s  3.91g
1221              (spills = 543 reloads = 1082)
1222           TOTAL                                   65.40u  3.13s  6.00g
1223    
1224    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1225           MLRISC ra                               0.80u  0.10s  0.55g
1226              (spills = 42 reloads = 84)
1227           TOTAL                                   129.42u  5.10s  4.13g
1228    
1229    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1230           MLRISC ra                               2.69u  0.27s  1.38g
1231              (spills = 62 reloads = 124)
1232           TOTAL                                   682.48u  19.61s  13.99g
1233    
1234    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1235           MLRISC ra                               4.96u  0.27s  2.72g
1236              (spills = 85 reloads = 170)
1237           TOTAL                                   2375.26u  57.21s  48.00g
1238    
1239    As you can see the old cps spill module suffers from some serious
1240    performance problem.  But since I cannot decipher the old code fully,
1241    instead of patching the problems up, I'm reimplementing it
1242    with a different algorithm.  The new code is more modular,
1243    smaller when compiled, and substantially faster
1244    (O(n log n) time and O(n) space).  Timing of the new spill module:
1245    
1246    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1247          MLRISC ra                               0.25u  0.02s  0.15g
1248             (spills=1 reloads=3)
1249          TOTAL                                   7.74u  0.34s  1.62g
1250    
1251    Implementation details:
1252    
1253    As far as I can tell, the purpose of the CPS spill module is to make sure the
1254    number of live variables at any program point (the bandwidth)
1255    does not exceed a certain limit, which is determined by the
1256    size of the spill area.
1257    
1258    When the bandwidth is too large, we decrease the register pressure by
1259    packing live variables into spill records.  How we achieve this is
1260    completely different than what we did in the old code.
1261    
1262    First, there is something about the MLRiscGen code generator
1263    that we should be aware of:
1264    
1265    o MLRiscGen performs code motion!
1266    
1267       In particular, it will move floating point computations and
1268       address computations involving only the heap pointer to
1269       their use sites (if there is only a single use).
1270       What this means is that if we have a CPS record construction
1271       statement
1272    
1273           RECORD(k,vl,w,e)
1274    
1275       we should never count the new record address w as live if w
1276       has only one use (which is often the case).
1277    
1278       We should do something similar to floating point, but the transformation
1279       there is much more complex, so I won't deal with that.
1280    
1281    Secondly, there are now two new cps primops at our disposal:
1282    
1283     1. rawrecord of record_kind option
1284        This pure operator allocates some uninitialized storage from the heap.
1285        There are two forms:
1286    
1287         rawrecord NONE [INT n]  allocates a tagless record of length n
1288         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1289                                     and initializes the tag.
1290    
1291     2. rawupdate of cty
1292          rawupdate cty (v,i,x)
1293          Assigns to x to the ith component of record v.
1294          The storelist is not updated.
1295    
1296    We use these new primops for both spilling and increment record construction.
1297    
1298     1. Spilling.
1299    
1300        This is implemented with a linear scan algorithm (but generalized
1301        to trees).  The algorithm will create a single spill record at the
1302        beginning of the cps function and use rawupdate to spill to it,
1303        and SELECT or SELp to reload from it.  So both spills and reloads
1304        are fine-grain operations.  In contrast, in the old algorithm
1305        "spills" have to be bundled together in records.
1306    
1307        Ideally, we should sink the spill record construction to where
1308        it is needed.  We can even split the spill record into multiple ones
1309        at the places where they are needed.  But CPS is not a good
1310        representation for global code motion, so I'll keep it simple and
1311        am not attempting this.
1312    
1313     2. Incremental record construction (aka record splitting).
1314    
1315        Long records with many component values which are simulatenously live
1316        (recall that single use record addresses are not considered to
1317         be live) are constructed with rawrecord and rawupdate.
1318        We allocate space on the heap with rawrecord first, then gradually
1319        fill it in with rawupdate.  This is the technique suggested to me
1320        by Matthias.
1321    
1322        Some restrictions on when this is applicable:
1323        1. It is not a VECTOR record.  The code generator currently does not handle
1324           this case. VECTOR record uses double indirection like arrays.
1325        2. All the record component values are defined in the same "basic block"
1326           as the record constructor.  This is to prevent speculative
1327           record construction.
1328    
1329    ----------------------------------------------------------------------
1330    Name: Allen Leung
1331    Date: 2002/02/22 01:02:00 EST
1332    Tag: leunga-20020222-mlrisc-tools
1333    
1334    Minor bug fixes in the parser and rewriter
1335    
1336    ----------------------------------------------------------------------
1337    Name: Allen Leung
1338    Date: 2002/02/21 20:20:00 EST
1339    Tag: leunga-20020221-peephole
1340    
1341    Regenerated the peephole files.  Some contained typos in the specification
1342    and some didn't compile because of pretty printing bugs in the old version
1343    of 'nowhere'.
1344    
1345    ----------------------------------------------------------------------
1346    Name: Allen Leung
1347    Date: 2002/02/19 20:20:00 EST
1348    Tag: leunga-20020219-mlrisc-tools
1349    Description:
1350    
1351       Minor bug fixes to the mlrisc-tools library:
1352    
1353       1.  Fixed up parsing colon suffixed keywords
1354       2.  Added the ability to shut the error messages up
1355       3.  Reimplemented the pretty printer and fixed up/improved
1356           the pretty printing of handle and -> types.
1357       4.  Fixed up generation of literal symbols in the nowhere tool.
1358       5.  Added some SML keywords to to sml.sty
1359    
1360    ----------------------------------------------------------------------
1361    Name: Matthias Blume
1362    Date: 2002/02/19 16:20:00 EST
1363    Tag: blume-20020219-cmffi
1364    Description:
1365    
1366    A wild mix of changes, some minor, some major:
1367    
1368    * All C FFI-related libraries are now anchored under $c:
1369        $/c.cm      --> $c/c.cm
1370        $/c-int.cm  --> $c/internals/c-int.cm
1371        $/memory.cm --> $c/memory/memory.cm
1372    
1373    * "make" tool (in CM) now treats its argument pathname slightly
1374      differently:
1375        1. If the native expansion is an absolute name, then before invoking
1376           the "make" command on it, CM will apply OS.Path.mkRelative
1377           (with relativeTo = OS.FileSys.getDir()) to it.
1378        2. The argument will be passed through to subsequent phases of CM
1379           processing without "going native".  In particular, if the argument
1380           was an anchored path, then "make" will not lose track of that anchor.
1381    
1382    * Compiler backends now "know" their respective C calling conventions
1383      instead of having to be told about it by ml-nlffigen.  This relieves
1384      ml-nlffigen from one of its burdens.
1385    
1386    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1387    
1388    * Export C_DEBUG and C_Debug from $c/c.cm.
1389    
1390    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1391      subtyping relationship between incomplete pointers and their complete
1392      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1393      with the convention of instantiating 'o with ('t, 'c) obj whenever
1394      the pointer target type is complete.  In the incomplete case, 'o
1395      will be instantiated with some "'c iobj" -- a type obtained by
1396      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1397    
1398      Operations that work on both incomplete and complete pointer types are
1399      typed as taking an 'o ptr while operations that require the target to
1400      be known are typed as taking some ('t, 'c) obj ptr.
1401    
1402      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1403      where void is an eqtype without any values.  This makes it possible
1404      to work on voidptr values using functions meant to operate on light
1405      incomplete pointers.
1406    
1407    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1408      been vastly simplified.
1409    
1410    ----------------------------------------------------------------------
1411    Name: Matthias Blume
1412    Date: 2002/02/19 10:48:00 EST
1413    Tag: blume-20020219-pqfix
1414    Description:
1415    
1416    Applied Chris Okasaki's bug fix for priority queues.
1417    
1418    ----------------------------------------------------------------------
1419    Name: Matthias Blume
1420  Date: 2002/02/15 17:05:00  Date: 2002/02/15 17:05:00
1421  Tag: Release_110_39  Tag: Release_110_39
1422  Description:  Description:

Legend:
Removed from v.1073  
changed lines
  Added in v.1317

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