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 686, Tue Jul 18 08:58:50 2000 UTC revision 1351, Fri Sep 5 21:35:54 2003 UTC
# Line 8  Line 8 
8  The form of an entry should be:  The form of an entry should be:
9    
10  Name:  Name:
11  Date:  Date: yyyy/mm/dd
12  Tag: <post-commit CVS tag>  Tag: <post-commit CVS tag>
13  Description:  Description:
14    
15    ----------------------------------------------------------------------
16    Name: Matthias Blume (blume@tti-c.org)
17    Date: 2003/09/04 16:30:00 CDT
18    Tag: blume-20030904-slices-etc
19    Description:
20    
21      1. Changed interface to vectors and arrays in Basis to match
22         (draft) Basis spec.
23      2. Added signatures and implementations of slices according to
24         Basis spec.
25      3. Edited source code throughout the system to make it compile again
26         under 1. and 2.  (In some cases code had to be added to have it
27         match the new signatures.)
28      4. MLRISC should be backward-compatible: the copies of the originals
29         of files that needed to change under 3. were retained, the .cm files
30         check the compiler version number and use old versions when
31         appropriate.
32      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
33         dirstream -> string option (in accordance with Basis spec).
34      6. When generating code that counts lines, ml-lex used function
35         CharVector.foldli, taking advantage of its old interface.
36         This has been replaced with the corresponding code from
37         CharVectorSlice.  (html-lex must be re-lexed!)
38      7. BitArray in smlnj-lib/Util has been extended/modified to match the
39         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
40      8. Removed temporary additions (fromInternal, toInternal) from the
41         (now obsolete) IntInf in smlnj-lib/Util.
42      9. Cleaned up structure Byte.
43     10. Added localOffset, scan, and fromString to Date (according to spec).
44         Cleaned/corrected implementation of Date.
45         (Still need to check for correctness; implement better canonicalizeDate.)
46     11. Added "scan" to signature IEEE_REAL.
47     12. Some improvements to IntInf [in particular: efficiency-hack for
48         mod and rem when second operand is 2 (for parity checks).]
49     13. Changed representation of type Time.time, using a single IntInf.int
50         value counting microseconds.  This considerably simplified the
51         implementation of structure Time.  We now support negative time
52         values; scan and fromString handle signs.
53     14. Functor PrimIO now takes two additional arguments (VectorSlice and
54         ArraySlice).
55    
56    ----------------------------------------------------------------------
57    Name: Matthias Blume (blume@tti-c.org)
58    Date: 2003/08/28 17:00:00 CDT
59    Tag: blume-20030828-intinf
60    Description:
61    
62    This is a major update which comes with a version number bump
63    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
64    and an implementation of IntInf in the Basis.
65    
66    There are a fairly large number of related changes and updates throughout
67    the system:
68    
69    Basis:
70            - Implemented IntInf.
71            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
72            - Added some missing Real64 operations, most notably Real.toLargeInt.
73            - Added FixedInt as a synonym for Int32.
74    
75    compiler:
76            * Added support for a built-in intinf type.
77              - literals
78              - pattern matching
79              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
80              - overloading on literals and operations
81    
82            This required adding a primitive type intinf, some additional
83            primops, and implementations for several non-trivial intinf
84            operations in Core.  (The intinf type is completely abstract
85            to the compiler; all operations get delegated back to the Core.)
86    
87            * Intinf equality is handled by polyequal.  However, the compiler
88            does not print its usual warning in this case (since polyequal
89            is the right thing to do there).
90    
91            * Improved the organization of structure InlineT.
92    
93            * A word about conversion primops:
94            If conversions involving intinf do not cancel out during
95            CPS contract, then the compiler must insert calls to Core functions.
96            Since all core access must be resolved already during the FLINT
97            translate phase, it would be too late a the time of CPS contract
98            to add new Core calls.  For this reason, conversion primops
99            for intinf carry two arguments:  1. the numeric argument that
100            they are supposed to convert, and 2. the Core function that
101            can help with this conversion if necessary.  If CPS contract
102            eliminates a primop, then the associated Core function becomes
103            dead and goes away.  Intinf conversion primops that do not get
104            eliminated by CPS contract get rewritten into calls of their
105            core functions by a separate, new phase.
106    
107    interactive system:
108            - Control.Print.intinfDepth controls max length of intinf constants
109            being printed. (Analogous to Control.Print.stringDepth.)
110            - Cleanup in printutil and pputil: got rid of unused stuff and
111            duplicates; replaced some of the code with code that makes better
112            use of library functionality.
113    
114    CM:
115            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
116            are no longer silent.
117    
118    CKIT:
119            Fixed mismatched uses of Int32 and LargeInt.  I always decided
120            in favor of LargeInt -- which is now the same as IntInf.
121            CKIT-knowledgable people should check whether this is what's
122            intended and otherwise change things back to using Int32 or
123            FixedInt.
124    
125    Throughout the code:
126            Started using IntInf.int literals and built-in operations
127            (e.g., comparison with 0) where this seems appropriate.
128    
129    
130    ----------------------------------------------------------------------
131    Name: Dave MacQueen (dbm@cs.uchicago.edu)
132    Date: 2003/08/13 11:36:00 CDT
133    Tag: dbm-20030813-mcz-merge1
134    Description:
135    
136    Merging changes from the mcz-branch development branch into trunk.
137    These changes involve replacement of the emulated old prettyprinter
138    interface with direct use of the SML/NJ Lib PP library, and fixing
139    of a couple of bugs (895, 1186) relating to error messages.  A new
140    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
141    has been added.
142    
143    ----------------------------------------------------------------------
144    Name: Matthias Blume (blume@tti-c.org)
145    Date: 2003/08/11 15:45:00 CDT
146    Tag: blume-20030811-windows
147    Description:
148    
149    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
150    
151        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
152    
153    This patch restores SML/NJ's ability to run under win32.  There are a
154    number of changes, including fixes for several bugs that had gone
155    unnoticed until now:
156    
157      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
158        trying to run the win32 version from within cygwin.)
159    
160      - There are a number of simple .bat scripts that substitute for their
161        corresponding Unix shell-scripts.  (See below.)
162    
163      - The internals of ml-build have been modified slightly.  The main
164        difference is that instead of calling ".link-sml" (or link-sml.bat)
165        using OS.Process.system, the ML process delegates this task back
166        to the script.  Otherwise problems arise in mixed environments such
167        as Cygwin where scripts look and work like Unix scripts, but
168        where OS.Process.system cannot run them.
169    
170      - In CM, the srcpath pickler used native pathname syntax -- which
171        is incorrect in the case of cross-compilation.  The new pickle format
172        is independent of platform-specific naming conventions.
173    
174      - Path configuration files (such as lib/pathconfig) can now choose
175        between native and standard syntax.  Placing a line of the form
176    
177            standard!
178    
179        into the file causes all subsequent paths to be interpreted using
180        CM standard pathname syntax (= Unix conventions); a line
181    
182            native!
183    
184        switches back to native style.  This was needed so that
185        path config files can be written portably, see src/system/pathconfig.
186    
187      - Runtime system:
188    
189          - win32-filesys.c:  get_file_time and set_file_time now
190            access modification time, not creation time.
191    
192          - I/O code made aware of new array representation.
193    
194          - Bug fixes in X86.prim.masm.
195    
196       - src/system/makeml made aware of win32. (For use under cygwin
197            and other Unix-environments for windows.)
198    
199       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
200         which caused BinIO.inputAll to fail consistently.
201    
202    .bat scripts:
203    
204       Windows .bat scripts assume that SMLNJ_HOME is defined.
205    
206       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
207         applications (sml, ml-yacc, ml-lex).
208       - ml-build.bat: analogous to ml-build.
209       - config\install.bat:  Analogous to config/install.sh.  This requires
210         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
211         (nmake etc. must be on the path, and vcvars32 must have been run.)
212         Moreover, sources for ml-lex and ml-yacc need to exist under src,
213         and the bootfile hierarchy must have been unpacked under
214         sml.boot.x86-win32.
215         The script is very primitive and does a poor job at error checking.
216         It only installs the base system, ml-lex, and ml-yacc.  No other
217         libraries are being installed (i.e., you get only those that
218         are part of the compiler.)
219       - link-sml.bat:  analogous to .link-sml, but not currently used
220    
221    Unrelated bug fixes:
222    
223      - ml-nlffigen now exports structures ST_* corresponding to incomplete
224        types.
225      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
226        otherwise.)
227    
228    ----------------------------------------------------------------------
229    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
230    Date: 2003/06/17
231    Tag: macqueen-20030617-bug895
232    Description:
233    
234    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
235    Tag will be used for new development branch (mcz-branch) for
236    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
237    
238    ----------------------------------------------------------------------
239    Name: Matthias Blume (blume@tti-c.org)
240    Date: 2003/05/27 16:55:00 CDT
241    Tag: blume-20030527-polyeq
242    Description:
243    
244    Tried to eliminated most cases of polymorphic equality.
245    
246    ----------------------------------------------------------------------
247    Name: Matthias Blume (blume@tti-c.org)
248    Date: 2003/05/21 17:45:00 CDT
249    Tag: blume-20030517-complete
250    Description:
251    
252    Two changes:
253    
254      1. Added a flag for controlling whether non-exhaustive bindings will
255         be treated as errors (default is false).
256      2. Cleaned up the *entire* source tree so that CMB.make goes through
257         without a single non-exhaustive match- or bind warning.
258    
259    ----------------------------------------------------------------------
260    Name: Matthias Blume (blume@tti-c.org)
261    Date: 2003/05/17 10:20:00 CDT
262    Tag: blume-20030517-absyn
263    Description:
264    
265    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
266    
267       This mainly affects the quality of error messages.  However, some
268       of the code is now more straightforward than before. (Treatment of
269       the above four constructs in translate.sml is much simpler than
270       the "macro-expansion" that was going on before.  Plus, the mach-
271       compiler no longer gets invoked just to be able to compile an
272       if-expression.)
273    
274    2. The ErrorMsg.Error exception is now caught and absorbed by the
275       interactive loop.
276    
277    ----------------------------------------------------------------------
278    Name: Allen Leung
279    Date: 2003/05/16 13:05:00 CDT
280    Tag: leunga-20030516-cygwin-runtime
281    Description:
282    
283        Ported the runtime system to cygwin, which uses the unix
284    x86-unix bin files.  Missing/buggy features:
285    
286       o getnetbyname, getnetbyaddr: these functions seem to be missing in
287         the Cygwin library.
288       o Ctrl-C handling may be flaky.
289       o Windows system calls and Windows I/O are not supported.
290    
291       A new set of binfiles is located at:
292    
293         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
294    
295       This is only needed for bootstrapping the cygwin version of smlnj.
296       Other x86 versions can use the existing binfiles.
297    
298    ----------------------------------------------------------------------
299    Name: Matthias Blume
300    Date: 2003/04/08 15:42:00 CDT
301    Tag: blume-20030408-listpair
302    Description:
303    
304    1. Added a target 'mlrisc' to installer.
305    
306    2. Added missing elements to structure ListPair.
307    
308    ----------------------------------------------------------------------
309    Name: Allen Leung
310    Date: 2003/01/07 10:40:00 EST
311    Tag: leunga-20030107-int-rem
312    Description:
313    
314    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
315    The arguments to the SUBL instruction were swapped.
316    
317    ----------------------------------------------------------------------
318    Name: Matthias Blume
319    Date: 2002/12/12 16:25:00 EST
320    Tag: blume-20021212-risc-ra
321    Description:
322    
323    Fixed a serious bug in the rewrite code for FP spilling/reloading that
324    sent the RA into an infinite loop when floating point registers get
325    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
326    110.37 and 110.38.)
327    There was another set of potential problems related to the handling of
328    MLRISC annotations (but those did not yet cause real problems, apparently).
329    
330    ----------------------------------------------------------------------
331    Name: Matthias Blume
332    Date: 2002/12/06 22:40:00 EST
333    Tag: blume-20021206-cm-fileid
334    Description:
335    
336    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
337    This fixes the problem of CM getting confused by files that suddenly
338    change their identity (e.g., by getting unlinked and recreated by some
339    text editor such as vi).  There might be a better/cheaper/cleaner way
340    of doing this, but for now this will have to do.
341    
342    ----------------------------------------------------------------------
343    Name: Matthias Blume
344    Date: 2002/10/28 09:50:00 EST
345    Tag: blume-20021028-typecheck
346    Description:
347    
348    Exported structure Typecheck from $smlnj/viscomp/core.cm.
349    
350    ----------------------------------------------------------------------
351    Name: Matthias Blume
352    Date: 2002/10/17 09:10:00 EDT
353    Tag: Release_110_42
354    Description:
355    
356    In good old tradition, there has been a slight hiccup so that we have
357    to patch 110.42 after the fact.   The old release tag has been replaced
358    (see below).
359    
360    The change solves a problem with two competing approaches the
361    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
362    each other's way.
363    
364    This change only affects the runtime system code and the installer script.
365    (No new bootfiles.)
366    
367    ----------------------------------------------------------------------
368    Name: Matthias Blume
369    Date: 2002/10/16 12:00:00 EDT
370    Tag: Release_110_42_removed
371    Description:
372    
373    New working release.  New bootfiles.
374    
375    ----------------------------------------------------------------------
376    Name: Matthias Blume
377    Date: 2002/10/10 13:10:00 EDT
378    Tag: blume-20021010-ppc-divs
379    Description:
380    
381    The mltree operator DIVS must be implemented with an overflow check on
382    the PPC because the hardware indicates divide-by-zero using "overflow" as
383    well.
384    
385    ----------------------------------------------------------------------
386    Name: Matthias Blume
387    Date: 2002/07/23 11:45:00
388    Tag: blume-20020723-smlnj-home
389    Description:
390    
391    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
392    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
393    CM_PATHCONFIG is also set), the path configuration file is assumed
394    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
395    the entire tree to some other place and everything will "just work".
396    
397    (Companion commands such as ml-build and ml-makedepend also sense this
398    variable.)
399    
400    ----------------------------------------------------------------------
401    Name: Matthias Blume
402    Date: 2002/07/12 21:19:00 EDT
403    Tag: blume-20020712-liveness
404    Description:
405    
406    Exported two useful "step" functions from liveness module (MLRISC).
407    
408    ----------------------------------------------------------------------
409    Name: Matthias Blume
410    Date: 2002/07/05 16:00 EDT
411    Tag: Release_110_41
412    Description:
413    
414    New working release.  New bootfiles.
415    
416    ----------------------------------------------------------------------
417    Name: Matthias Blume
418    Date: 2002/07/05 10:25:00 EDT
419    Tag: blume-20020705-btimp
420    Description:
421    
422    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
423    other clients can set up backtracing support.
424    
425    ----------------------------------------------------------------------
426    Name: Matthias Blume
427    Date: 2002/06/25 17:23:00 EDT
428    Tag: blume-20020625-fpmax
429    Description:
430    
431    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
432    case.  (The sense of the isNaN test was reversed -- which made min and
433    max always return their first argument.)
434    
435    ----------------------------------------------------------------------
436    Name: Matthias Blume
437    Date: 2002/06/11
438    Tag: blume-20020611-unixpath
439    Description:
440    
441    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
442    
443    ----------------------------------------------------------------------
444    Name: Matthias Blume
445    Date: 2002/06/10 16:35:00 EDT
446    Tag: blume-20020610-ieeereal
447    Description:
448    
449    I back-ported my implementation of IEEEReal.fromString from the
450    idlbasis-devel branch so that we can test it.
451    
452    Another small change is that ppDec tries to give more information
453    than just "<sig>" in the case of functors.  However, this code is
454    broken in some mysterious way if the functor's body's signature
455    has not been declared by ascription but gets inferred from the
456    implementation.  This needs fixing...
457    
458    ----------------------------------------------------------------------
459    Name: Matthias Blume
460    Date: 2002/05/31
461    Tag: blume-20020531-btrace-mode
462    Description:
463    
464    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
465    the wayside when I switched over to using Controls everywhere.)
466    
467    ----------------------------------------------------------------------
468    Name: Lal George
469    Date: 2002/05/23 12:21:40 EDT
470    Tag: george-20020523-visual-labels
471    Description:
472    
473    Labels are now displayed in the graphical output to make
474    the fall-through and target blocks obvious.
475    
476    ----------------------------------------------------------------------
477    Name: Matthias Blume
478    Date: 2002/05/22 11:03:00 EDT
479    Tag: blume-20020522-shrink
480    Description:
481    
482    John tweaked yesterday's fix for 1131 to handle an out-of-memory
483    situation that comes up when allocating huge arrays.
484    
485    ----------------------------------------------------------------------
486    Name: Matthias Blume
487    Date: 2002/05/21 16:00:00 EDT
488    Tag: Release_110_40
489    Description:
490    
491    New working release (110.40).  New bootfiles.
492    
493    [Also: John Reppy fixed GC bug 1131.]
494    
495    ----------------------------------------------------------------------
496    Name: Matthias Blume
497    Date: 2002/05/21 12:35:00 EDT
498    Tag: blume-20020521-cmdoc
499    Description:
500    
501    CM documentation update.
502    
503    ----------------------------------------------------------------------
504    Name: Matthias Blume
505    Date: 2002/05/21 10:55:00 EDT
506    Tag: blume-20020521-misc
507    Description:
508    
509    - John tweaked runtime to be silent on heap export (except when
510      GC messages are on).
511    - I added a few more things (cross-compiling versions of CMB) to
512      config/preloads (as suggestions).
513    
514    ----------------------------------------------------------------------
515    Name: Matthias Blume
516    Date: 2002/05/20 22:25:00 EDT
517    Tag: blume-20020520-controls
518    Description:
519    
520    - Added ControlUtil structure to control-lib.cm.
521    - Use it throughout.
522    - Used Controls facility to define MLRISC controls (as opposed to
523      registering MLRISC control ref cells with Controls after the fact)
524    - Fixed messed-up controls priorities.
525    
526    * Removed again all the stuff from config/preloads that one wouldn't
527      be able to preload at the time the initial heap image is built.
528      (Many libraries, e.g., CML, do not exist yet at this time.  The
529       only libraries that can be preloaded via config/preloads are those
530       that come bundled with the bootfiles.)
531    
532    ----------------------------------------------------------------------
533    Name: Matthias Blume
534    Date: 2002/05/20 10:59:00 EDT
535    Tag: blume-20020520-preloads
536    Description:
537    
538    Added a lot of commented-out suggestions for things to be included
539    in config/preloads.
540    
541    ----------------------------------------------------------------------
542    Name: Allen Leung
543    Date: 2002/05/18 14:20:00 EDT
544    Tag: leunga-20020518-mdl
545    Description:
546    
547    o Made the mdl tool stuff compile and run again.
548    o I've disabled all the stuff that depends on RTL specifications; they
549      are all badly broken anyway.
550    
551    ----------------------------------------------------------------------
552    Name: Matthias Blume
553    Date: 2002/05/17 16:49:00 EDT
554    Tag: blume-20020517-controls
555    Description:
556    
557    0. John Reppy made several modifications to the SML/NJ library.
558       In particular, there is a shiny new controls-lib.cm.
559    
560    1. Pushed new controls interface through compiler so that everything
561       compiles again.
562    
563    2. Added FormatComb and FORMAT_COMB to the CML version of the
564       SML/NJ library (so that CML compiles again).
565    
566    3. Modified init scripts because XXX_DEFAULT environment variables
567       are no longer with us.  (Boot-time initialization is now done
568       using the same environment variables that are also used for
569       startup-time initialization of controls.)
570    
571    ----------------------------------------------------------------------
572    Name: Lal George
573    Date: 2002/05/15 09:20:10 EDT
574    Tag: george-20020515-pseudo-op-decls
575    Description:
576    
577            All pseudo-ops emitted before the first segment declaration
578    such as TEXT, DATA, and BSS directives are assumed to be global
579    declarations and are emitted first in the assembly file. This is
580    useful in a number of situations where one has pseudo-ops that are not
581    specific to any segment, and also works around the constraint that one
582    cannot have client pseudo-ops in the TEXT segment.
583    
584            Because no segment is associated with these declarations it is
585    an error to allocate any space or objects before the first segment
586    directive and an exception will be raised. However, we cannot make
587    this check for client pseudo-ops.
588    
589            These top level declarations are a field in the CFG graph_info.
590    In theory you can continue to add to this field after the CFG has been
591    built -- provided you know what you are doing;-)
592    
593    ----------------------------------------------------------------------
594    Name: Matthias Blume
595    Date: 2002/05/13 16:40:00 EDT
596    Tag: blume-20020513-pp-etc
597    Description:
598    
599    A few minor bugfixes:
600    
601      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
602        (Bogus printouts for redefined bindings still occur.  Compiler
603        bug should no longer occur now.  We need to redo the prettyprinter
604        from scratch.)
605    
606      - CM pathname printer now also adds escape sequences for ( and )
607    
608      - commend and docu fixes for ml-nlffi
609    
610    ----------------------------------------------------------------------
611    Name: Matthias Blume
612    Date: 2002/05/10 16:40:00 EDT
613    Tag: blume-20020510-erg-textio
614    Description:
615    
616    Applied the following bugfix provided by Emden Gansner:
617    
618        Output is corrupted when outputSubstr is used rather than output.
619    
620        The problem occurs when a substring
621    
622            ss = (s, dataStart, dataLen)
623    
624        where dataStart > 0, fills a stream buffer with avail bytes left.
625        avail bytes of s, starting at index dataStart, are copied into the
626        buffer, the buffer is flushed, and then the remaining dataLen-avail
627        bytes of ss are copied into the beginning of the buffer. Instead of
628        starting this copy at index dataStart+avail in s, the current code
629        starts the copy at index avail.
630    
631        Fix:
632        In text-io-fn.sml, change line 695 from
633             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
634        to
635             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
636    
637    ----------------------------------------------------------------------
638    Name: Matthias Blume
639    Date: 2002/04/12 13:55:00 EDT
640    Tag: blume-20020412-assyntax
641    Description:
642    
643    1. Grabbed newer assyntax.h from the XFree86 project.
644    2. Fiddled with how to compile X86.prim.asm without warnings.
645    3. (Very) Minor cleanup in CM.
646    
647    ----------------------------------------------------------------------
648    Name: Matthias Blume
649    Date: 2002/04/01 (no joke!) 17:07:00 EST
650    Tag: blume-20020401-x86div
651    Description:
652    
653    Added full support for div/mod/rem/quot on the x86, using the machine
654    instruction's two results (without clumsily recomputing the remainder)
655    directly where appropriate.
656    
657    Some more extensive power-of-two support was added to the x86 instruction
658    selector (avoiding expensive divs, mods, and muls where they can be
659    replaced with cheaper shifts and masks).  However, this sort of thing
660    ought to be done earlier, e.g., within the CPS optimizer so that
661    all architectures benefit from it.
662    
663    The compiler compiles to a fixed point, but changes might be somewhat
664    fragile nevertheless.  Please, report any strange things that you might
665    see wrt. div/mod/quot/rem...
666    
667    ----------------------------------------------------------------------
668    Name: Matthias Blume
669    Date: 2002/03/29 17:22:00
670    Tag: blume-20020329-div
671    Description:
672    
673    Fixed my broken div/mod logic.  Unfortunately, this means that the
674    inline code for div/mod now has one more comparison than before.
675    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
676    The problem was with quotient = 0, because that alone does not tell
677    us which way the rounding went.  One then has to look at whether
678    remainder and divisor have the same sign...  :(
679    
680    Anyway, I replaced the bootfiles with fresh ones...
681    
682    ----------------------------------------------------------------------
683    Name: Matthias Blume
684    Date: 2002/03/29 14:10:00 EST
685    Tag: blume-20020329-inlprims
686    Description:
687    
688    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
689    
690    Primops have changed. This means that the bin/boot-file formats have
691    changed as well.
692    
693    To make sure that there is no confusion, I made a new version.
694    
695    
696    CHANGES:
697    
698    * removed REMT from mltree (remainder should never overflow).
699    
700    * added primops to deal with divisions of all flavors to the frontend
701    
702    * handled these primops all the way through so they map to their respective
703      MLRISC support
704    
705    * used these primops in the implementation of Int, Int32, Word, Word32
706    
707    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
708    
709    * parameterized INLMIN, INLMAX, and INLABS by a numkind
710    
711    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
712      floating point
713    
714    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
715      and Real (but Real.abs maps to a separate floating-point-only primop)
716    
717    
718    TODO items:
719    
720    * Hacked Alpha32 instruction selection, disabling the selection of REMx
721      instructions because the machine instruction encoder cannot handle
722      them.  (Hppa, PPC, and Sparc instruction selection did not handle
723      REM in the first place, and REM is supported by the x86 machine coder.)
724    
725    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
726      selection phase.  (The two can be streamlined because the hardware
727      delivers both quotient and remainder at the same time anyway.)
728    
729    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
730      (Currently the behavior is inconsistent both across architectures and
731      wrt. the draft Basis spec.)
732    
733    * Word8 should eventually be handled natively, too.
734    
735    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
736      as if there currently is no execution path that could trigger it in
737      SML/NJ.  (The assumptions underlying functions arith and promotable do not
738      hold for things like multiplication and division.)
739    
740    ----------------------------------------------------------------------
741    Name: Matthias Blume
742    Date: 2002/03/27 16:27:00 EST
743    Tag: blume-20020327-mlrisc-divisions
744    Description:
745    
746    Added support for all four division operations (ML's div, mod, quot,
747    and rem) to MLRISC.  In the course of doing so, I also rationalized
748    the naming (no more annoying switch-around of DIV and QUOT), by
749    parameterizing the operation by div_rounding_mode (which can be either
750    DIV_TO_ZERO or DIV_TO_NEGINF).
751    
752    The generic MLTreeGen functor takes care of compiling all four
753    operations down to only round-to-zero div.
754    
755    Missing pieces:
756    
757      * Doing something smarter than relying on MLTreeGen on architectures
758        like, e.g., the x86 where hardware division delivers both quotient and
759        remainder at the same time.  With this, the implementation of the
760        round-to-neginf operations could be further streamlined.
761    
762      * Remove inlining support for div/mod/rem from the frontend and replace it
763        with primops that get carried through to the backend.  Do this for all
764        int and word types.
765    
766    ----------------------------------------------------------------------
767    Name: Matthias Blume
768    Date: 2002/03/25 17:25:00 EST
769    Tag: blume-20020325-divmod
770    Description:
771    
772    I improved (hopefully without breaking them) the implementation of Int.div,
773    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
774    advantage of the following observations:
775    
776      Let  q = x quot y      r = x rem y
777           d = x div  y      m = x mod y
778    
779    where "quot" is the round-to-zero version of integer division that
780    hardware usually provides.  Then we have:
781    
782         r = x - q * y        where neither the * nor the - will overflow
783         d = if q >= 0 orelse x = q * y then q else q - 1
784                              where neither the * nor the - will overflow
785         m = if q >= 0 orelse r = 0 then r else r + y
786                              where the + will not overflow
787    
788    This results in substantial simplification of the generated code.
789    The following table shows the number of CFG nodes and edges generated
790    for
791            fun f (x, y) = x OPER y
792            (* with OPER \in div, mod, quot, rem *)
793    
794    
795        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
796        --------------------------------------------------------
797         div |         24 |         39 |         12 |         16
798         mod |         41 |         71 |         12 |         16
799        quot |          8 |         10 |          8 |         10
800         rem |         10 |         14 |          8 |         10
801    
802    
803    ----------------------------------------------------------------------
804    Name: Matthias Blume
805    Date: 2002/03/25 22:06:00 EST
806    Tag: blume-20020325-cprotobug
807    Description:
808    
809    Fixed a bug in cproto (c prototype decoder).
810    
811    ----------------------------------------------------------------------
812    Name: Matthias Blume
813    Date: 2002/03/25 16:00:00 EST
814    Tag: blume-20020325-raw-primops
815    Description:
816    
817    I did some cleanup to Allen's new primop code and
818    replaced yesterday's bootfiles with new ones.
819    (But they are stored in the same place.)
820    
821    ----------------------------------------------------------------------
822    Name: Matthias Blume
823    Date: 2002/03/24 22:40:00 EST
824    Tag: blume-20020324-bootfiles
825    Description:
826    
827    Made the bootfiles that Allen asked for.
828    
829    ----------------------------------------------------------------------
830    Name: Allen Leung
831    Date: 2002/03/23 15:50:00 EST
832    Tag: leunga-20020323-flint-cps-rcc-primops
833    Description:
834    
835      1. Changes to FLINT primops:
836    
837        (* make a call to a C-function;
838         * The primop carries C function prototype information and specifies
839         * which of its (ML-) arguments are floating point. C prototype
840         * information is for use by the backend, ML information is for
841         * use by the CPS converter. *)
842      | RAW_CCALL of { c_proto: CTypes.c_proto,
843                       ml_args: ccall_type list,
844                       ml_res_opt: ccall_type option,
845                       reentrant : bool
846                     } option
847       (* Allocate uninitialized storage on the heap.
848        * The record is meant to hold short-lived C objects, i.e., they
849        * are not ML pointers.  With the tag, the representation is
850        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
851        *)
852      | RAW_RECORD of {tag:bool,sz:int}
853      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
854    
855      2.  These CPS primops are now overloaded:
856    
857           rawload of {kind:numkind}
858           rawstore of {kind:numkind}
859    
860          The one argument form is:
861    
862             rawload {kind} address
863    
864          The two argument form is:
865    
866             rawload {kind} [ml object, byte-offset]
867    
868      3. RAW_CCALL/RCC now takes two extra arguments:
869    
870         a. The first is whether the C call is reentrant, i.e., whether
871            ML state should be saved and restored.
872         b. The second argument is a string argument specifying the name of
873            library and the C function.
874    
875         These things are currently not handled in the code generator, yet.
876    
877      4. In CProto,
878    
879         An encoding type of "bool" means "ml object" and is mapped into
880         C prototype of PTR.  Note that "bool" is different than "string",
881         even though "string" is also mapped into PTR, because "bool"
882         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
883    
884      5. Pickler/unpicker
885    
886         Changed to handle RAW_RECORD and newest RAW_CCALL
887    
888      6. MLRiscGen,
889    
890         1. Changed to handle the new rawload/rawstore/rawrecord operators.
891         2. Code for handling C Calls has been moved to a new module CPSCCalls,
892            in the file CodeGen/cpscompile/cps-c-calls.sml
893    
894      7. Added the conditional move operator
895    
896             condmove of branch
897    
898         to cps.  Generation of this is still buggy so it is currently
899         disabled.
900    
901    ----------------------------------------------------------------------
902    Name: Lal George
903    Date: 2002/03/22 14:18:25 EST
904    Tag: george-20020322-cps-branch-prob
905    Description:
906    
907    Implemented the Ball-Larus branch prediction-heuristics, and
908    incorporated graphical viewers for control flow graphs.
909    
910    Ball-Larus Heuristics:
911    ---------------------
912    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
913    
914    By design it uses the Dempster-Shafer theory for combining
915    probabilities.  For example, in the function:
916    
917        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
918    
919    the ball-larus heuristics predicts that the n=0 is unlikely
920    (OH-heuristic), and the 'then' branch is unlikely because of the
921    RH-heuristic -- giving the 'then' branch an even lower combined
922    probability using the Dempster-Shafer theory.
923    
924    Finally, John Reppy's loop analysis in MLRISC, further lowers the
925    probability of the 'then' branch because of the loop in the else
926    branch.
927    
928    
929    Graphical Viewing:
930    ------------------
931    I merely plugged in Allen's graphical viewers into the compiler. The
932    additional code is not much. At the top level, saying:
933    
934            Control.MLRISC.getFlag "cfg-graphical-view" := true;
935    
936    will display the graphical view of the control flow graph just before
937    back-patching.  daVinci must be in your path for this to work. If
938    daVinci is not available, then the default viewer can be changed
939    using:
940    
941            Control.MLRISC.getString "viewer"
942    
943    which can be set to "dot" or "vcg" for the corresponding viewers. Of
944    course, these viewers must be in your path.
945    
946    The above will display the compilation unit at the level of clusters,
947    many of which are small, boring, and un-interesting. Also setting:
948    
949            Control.MLRISC.getInt "cfg-graphical-view_size"
950    
951    will display clusters that are larger than the value set by the above.
952    
953    
954    ----------------------------------------------------------------------
955    Name: Matthias Blume
956    Date: 2002/03/21 22:20:00 EST
957    Tag: blume-20020321-kmp-bugfix
958    Description:
959    
960    Changed the interface to the KMP routine in PreString and fixed
961    a minor bug in one place where it was used.
962    
963    ----------------------------------------------------------------------
964    Name: Allen Leung
965    Date: 2002/03/21 20:30:00 EST
966    Tag: leunga-20020321-cfg
967    Description:
968    
969      Fixed a potential problem in cfg edge splitting.
970    
971    ----------------------------------------------------------------------
972    Name: Allen Leung
973    Date: 2002/03/21 17:15:00 EST
974    Tag: leunga-20020321-x86-fp-cfg
975    Description:
976    
977      1. Recoded the buggy parts of x86-fp.
978    
979         a. All the block reordering code has been removed.
980            We now depend on the block placement phases to do this work.
981    
982         b. Critical edge splitting code has been simplified and moved into the
983            CFG modules, as where they belong.
984    
985         Both of these were quite buggy and complex.  The code is now much, much
986         simpler.
987    
988      2. X86 backend.
989    
990         a. Added instructions for 64-bit support.  Instruction selection for
991            64-bit has not been committed, however, since that
992            requires changes to MLTREE which haven't been approved by
993            Lal and John.
994    
995         b. Added support for FUCOMI and FUCOMIP when generating code for
996            PentiumPro and above.  We only generate these instructions in
997            the fast-fp mode.
998    
999         c. Added cases for JP and JNP in X86FreqProps.
1000    
1001      3. CFG
1002    
1003         CFG now has a bunch of methods for edge splitting and merging.
1004    
1005      4. Machine description.
1006    
1007         John's simplification of MLTREE_BASIS.fcond broke a few machine
1008         description things:
1009    
1010         rtl-build.{sig,sml} and hppa.mdl fixed.
1011    
1012         NOTE: the machine description stuff in the repository is still broken.
1013               Again, I can't put my fixes in because that involves
1014               changes to MLTREE.
1015    
1016    ----------------------------------------------------------------------
1017    Name: Matthias Blume
1018    Date: 2002/03/20 15:55:00 EST
1019    Tag: blume-20020320-kmp
1020    Description:
1021    
1022    Implemented Knuth-Morris-Pratt string matching in PreString and used
1023    it for String.isSubstring, Substring.isSubstring, and
1024    Substring.position.
1025    
1026    (Might need some stress-testing.  Simple examples worked fine.)
1027    
1028    ----------------------------------------------------------------------
1029    Name: Matthias Blume
1030    Date: 2002/03/19 16:37:00 EST
1031    Tag: blume-20020319-witnesses
1032    Description:
1033    
1034    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1035    
1036    This implements a generic mechanism for changing constness qualifiers
1037    anywhere within big C types without resorting to outright "casts".
1038    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1039    modify the constness at the outermost level.)
1040    The implementation of "convert" is based on the idea of "witness"
1041    values -- values that are not used by the operation but whose types
1042    "testify" to their applicability.  On the implementation side, "convert"
1043    is simply a projection (returning its second curried argument).  With
1044    cross-module inlining, it should not result in any machine code being
1045    generated.
1046    
1047    ----------------------------------------------------------------------
1048    Name: Matthias Blume
1049    Date: 2002/03/15 16:40:00 EST
1050    Tag: blume-20020315-basis
1051    Description:
1052    
1053    Provided (preliminary?) implementations for
1054    
1055      {String,Substring}.{concatWith,isSuffix,isSubstring}
1056    
1057    and
1058    
1059      Substring.full
1060    
1061    Those are in the Basis spec but they were missing in SML/NJ.
1062    
1063    ----------------------------------------------------------------------
1064    Name: Matthias Blume
1065    Date: 2002/03/14 21:30:00 EST
1066    Tag: blume-20020314-controls
1067    Description:
1068    
1069    Controls:
1070    ---------
1071    
1072    1. Factored out the recently-added Controls : CONTROLS stuff and put
1073       it into its own library $/controls-lib.cm.  The source tree for
1074       this is under src/smlnj-lib/Controls.
1075    
1076    2. Changed the names of types and functions in this interface, so they
1077       make a bit more "sense":
1078    
1079          module -> registry
1080          'a registry -> 'a group
1081    
1082    3. The interface now deals in ref cells only.  The getter/setter interface
1083       is (mostly) gone.
1084    
1085    4. Added a function that lets one register an already-existing ref cell.
1086    
1087    5. Made the corresponding modifications to the rest of the code so that
1088       everything compiles again.
1089    
1090    6. Changed the implementation of Controls.MLRISC back to something closer
1091       to the original.  In particular, this module (and therefore MLRISC)
1092       does not depend on Controls.  There now is some link-time code in
1093       int-sys.sml that registers the MLRISC controls with the Controls
1094       module.
1095    
1096    CM:
1097    ---
1098    
1099      * One can now specify the lambda-split aggressiveness in init.cmi.
1100    
1101    ----------------------------------------------------------------------
1102    Name: Allen Leung
1103    Date: 2002/03/13 17:30:00 EST
1104    Tag: leunga-20020313-x86-fp-unary
1105    Description:
1106    
1107    Bug fix for:
1108    
1109    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1110    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1111    > - fun f(x,(y,z)) = Real.~ y;
1112    > [autoloading]
1113    > [autoloading done]
1114    >       fchsl   (%eax), 184(%esp)
1115    > Error: MLRisc bug: X86MCEmitter.emitInstr
1116    >
1117    > uncaught exception Error
1118    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1119    
1120    The problem was that the code generator did not generate any fp registers
1121    in this case, and the ra didn't know that it needed to run the X86FP phase to
1122    translate the pseudo fp instruction.   This only happened with unary fp
1123    operators in certain situations.
1124    
1125    ----------------------------------------------------------------------
1126    Name: Matthias Blume
1127    Date: 2002/03/13 14:00:00 EST
1128    Tag: blume-20020313-overload-etc
1129    Description:
1130    
1131    1. Added _overload as a synonym for overload for backward compatibility.
1132       (Control.overloadKW must be true for either version to be accepted.)
1133    
1134    2. Fixed bug in install script that caused more things to be installed
1135       than what was requested in config/targets.
1136    
1137    3. Made CM aware of the (_)overload construct so that autoloading
1138       works.
1139    
1140    ----------------------------------------------------------------------
1141    Name: Matthias Blume
1142    Date: 2002/03/12 22:03:00 EST
1143    Tag: blume-20020312-url
1144    Description:
1145    
1146    Forgot to update BOOT and srcarchiveurl.
1147    
1148    ----------------------------------------------------------------------
1149    Name: Matthias Blume
1150    Date: 2002/03/12 17:30:00 EST
1151    Tag: blume-20020312-version110392
1152    Description:
1153    
1154    Yet another version number bump (because of small changes to the
1155    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1156    
1157    Changes:
1158    
1159      The new pid generation scheme described a few weeks ago was overly
1160      complicated.  I implemented a new mechanism that is simpler and
1161      provides a bit more "stability":  Once CM has seen a compilation
1162      unit, it keeps its identity constant (as long as you do not delete
1163      those crucial CM/GUID/* files).  This means that when you change
1164      an interface, compile, then go back to the old interface, and
1165      compile again, you arrive at the original pid.
1166    
1167      There now also is a mechanism that instructs CM to use the plain
1168      environment hash as a module's pid (effectively making its GUID
1169      the empty string).  For this, "noguid" must be specified as an
1170      option to the .sml file in question within its .cm file.
1171      This is most useful for code that is being generated by tools such
1172      as ml-nlffigen (because during development programmers tend to
1173      erase the tool's entire output directory tree including CM's cached
1174      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1175      revert to the old, broken behavior of SML/NJ, but in specific cases
1176      where there is no danger of interface confusion, its use is ok
1177      (I think).
1178    
1179      ml-nlffigen by default generates "noguid" annotations.  They can be
1180      turned off by specifying -guid in its command line.
1181    
1182    ----------------------------------------------------------------------
1183    Name: Lal George
1184    Date: 2002/03/12 12 14:42:36 EST
1185    Tag: george-20020312-frequency-computation
1186    Description:
1187    
1188    Integrated jump chaining and static block frequency into the
1189    compiler. More details and numbers later.
1190    
1191    ----------------------------------------------------------------------
1192    Name: Lal George
1193    Date: 2002/03/11 11 22:38:53 EST
1194    Tag: george-20020311-jump-chain-elim
1195    Description:
1196    
1197    Tested the jump chain elimination on all architectures (except the
1198    hppa).  This is on by default right now and is profitable for the
1199    alpha and x86, however, it may not be profitable for the sparc and ppc
1200    when compiling the compiler.
1201    
1202    The gc test will typically jump to a label at the end of the cluster,
1203    where there is another jump to an external cluster containing the actual
1204    code to invoke gc. This is to allow factoring of common gc invocation
1205    sequences. That is to say, we generate:
1206    
1207            f:
1208               testgc
1209               ja   L1      % jump if above to L1
1210    
1211            L1:
1212               jmp L2
1213    
1214    
1215    After jump chain elimination the 'ja L1' instructions is converted to
1216    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1217    up being implemented in their long form (if L2 is far away) using:
1218    
1219            jbe     L3      % jump if below or equal to L3
1220            jmp     L2
1221         L3:
1222            ...
1223    
1224    
1225    For large compilation units L2  may be far away.
1226    
1227    
1228    ----------------------------------------------------------------------
1229    Name: Matthias Blume
1230    Date: 2002/03/11 13:30:00 EST
1231    Tag: blume-20020311-mltreeeval
1232    Description:
1233    
1234    A functor parameter was missing.
1235    
1236    ----------------------------------------------------------------------
1237    Name: Allen Leung
1238    Date: 2002/03/11 10:30:00 EST
1239    Tag: leunga-20020311-runtime-string0
1240    Description:
1241    
1242       The representation of the empty string now points to a
1243    legal null terminated C string instead of unit.  It is now possible
1244    to convert an ML string into C string with InlineT.CharVector.getData.
1245    This compiles into one single machine instruction.
1246    
1247    ----------------------------------------------------------------------
1248    Name: Allen Leung
1249    Date: 2002/03/10 23:55:00 EST
1250    Tag: leunga-20020310-x86-call
1251    Description:
1252    
1253       Added machine generation for CALL instruction (relative displacement mode)
1254    
1255    ----------------------------------------------------------------------
1256    Name: Matthias Blume
1257    Date: 2002/03/08 16:05:00
1258    Tag: blume-20020308-entrypoints
1259    Description:
1260    
1261    Version number bumped to 110.39.1.  NEW BOOTFILES!
1262    
1263    Entrypoints: non-zero offset into a code object where execution should begin.
1264    
1265    - Added the notion of an entrypoint to CodeObj.
1266    - Added reading/writing of entrypoint info to Binfile.
1267    - Made runtime system bootloader aware of entrypoints.
1268    - Use the address of the label of the first function given to mlriscGen
1269      as the entrypoint.  This address is currently always 0, but it will
1270      not be 0 once we turn on block placement.
1271    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1272      with entry points) from mlriscGen.
1273    
1274    ----------------------------------------------------------------------
1275    Name: Allen Leung
1276    Date: 2002/03/07 20:45:00 EST
1277    Tag: leunga-20020307-x86-cmov
1278    Description:
1279    
1280       Bug fixes for CMOVcc on x86.
1281    
1282       1. Added machine code generation for CMOVcc
1283       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1284       3. CMOVcc cannot have an immediate operand as argument.
1285    
1286    ----------------------------------------------------------------------
1287    Name: Matthias Blume
1288    Date: 2002/03/07 16:15:00 EST
1289    Tag: blume-20020307-controls
1290    Description:
1291    
1292    This is a very large but mostly boring patch which makes (almost)
1293    every tuneable compiler knob (i.e., pretty much everything under
1294    Control.* plus a few other things) configurable via both the command
1295    line and environment variables in the style CM did its configuration
1296    until now.
1297    
1298    Try starting sml with '-h' (or, if you are brave, '-H')
1299    
1300    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1301    implements the underlying generic mechanism.
1302    
1303    The interface to some of the existing such facilities has changed somewhat.
1304    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1305    (The getFoo interface is still there for backward-compatibility, but its
1306    use is deprecated.)
1307    
1308    The ml-build script passes -Cxxx=yyy command-line arguments through so
1309    that one can now twiddle the compiler settings when using this "batch"
1310    compiler.
1311    
1312    TODO items:
1313    
1314    We should go through and throw out all controls that are no longer
1315    connected to anything.  Moreover, we should go through and provide
1316    meaningful (and correct!) documentation strings for those controls
1317    that still are connected.
1318    
1319    Currently, multiple calls to Controls.new are accepted (only the first
1320    has any effect).  Eventually we should make sure that every control
1321    is being made (via Controls.new) exactly once.  Future access can then
1322    be done using Controls.acc.
1323    
1324    Finally, it would probably be a good idea to use the getter-setter
1325    interface to controls rather than ref cells.  For the time being, both
1326    styles are provided by the Controls module, but getter-setter pairs are
1327    better if thread-safety is of any concern because they can be wrapped.
1328    
1329    *****************************************
1330    
1331    One bug fix: The function blockPlacement in three of the MLRISC
1332    backpatch files used to be hard-wired to one of two possibilities at
1333    link time (according to the value of the placementFlag).  But (I
1334    think) it should rather sense the flag every time.
1335    
1336    *****************************************
1337    
1338    Other assorted changes (by other people who did not supply a HISTORY entry):
1339    
1340    1. the cross-module inliner now works much better (Monnier)
1341    2. representation of weights, frequencies, and probabilities in MLRISC
1342       changed in preparation of using those for weighted block placement
1343       (Reppy, George)
1344    
1345    ----------------------------------------------------------------------
1346    Name: Lal George
1347    Date: 2002/03/07 14:44:24 EST 2002
1348    Tag: george-20020307-weighted-block-placement
1349    
1350    Tested the weighted block placement optimization on all architectures
1351    (except the hppa) using AMPL to generate the block and edge frequencies.
1352    Changes were required in the machine properties to correctly
1353    categorize trap instructions. There is an MLRISC flag
1354    "weighted-block-placement" that can be used to enable weighted block
1355    placement, but this will be ineffective without block/edge
1356    frequencies (coming soon).
1357    
1358    
1359    ----------------------------------------------------------------------
1360    Name: Lal George
1361    Date: 2002/03/05 17:24:48 EST
1362    Tag: george-20020305-linkage-cluster
1363    
1364    In order to support the block placement optimization, a new cluster
1365    is generated as the very first cluster (called the linkage cluster).
1366    It contains a single jump to the 'real' entry point for the compilation
1367    unit. Block placement has no effect on the linkage cluster itself, but
1368    all the other clusters  have full freedom in the manner in which they
1369    reorder blocks or functions.
1370    
1371    On the x86 the typical linkage code that is generated is:
1372       ----------------------
1373            .align 2
1374       L0:
1375            addl    $L1-L0, 72(%esp)
1376            jmp     L1
1377    
1378    
1379            .align  2
1380       L1:
1381       ----------------------
1382    
1383    72(%esp) is the memory location for the stdlink register. This
1384    must contain the address of the CPS function being called. In the
1385    above example, it contains the address of  L0; before
1386    calling L1 (the real entry point for the compilation unit), it
1387    must contain the address for L1, and hence
1388    
1389            addl $L1-L0, 72(%esp)
1390    
1391    I have tested this on all architectures except the hppa.The increase
1392    in code size is of course negligible
1393    
1394    ----------------------------------------------------------------------
1395    Name: Allen Leung
1396    Date: 2002/03/03 13:20:00 EST
1397    Tag: leunga-20020303-mlrisc-tools
1398    
1399      Added #[ ... ] expressions to mlrisc tools
1400    
1401    ----------------------------------------------------------------------
1402    Name: Matthias Blume
1403    Date: 2002/02/27 12:29:00 EST
1404    Tag: blume-20020227-cdebug
1405    Description:
1406    
1407    - made types in structure C and C_Debug to be equal
1408    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1409    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1410    
1411    ----------------------------------------------------------------------
1412    Name: Matthias Blume
1413    Date: 2002/02/26 12:00:00 EST
1414    Tag: blume-20020226-ffi
1415    Description:
1416    
1417    1. Fixed a minor bug in CM's "noweb" tool:
1418       If numbering is turned off, then truly don't number (i.e., do not
1419       supply the -L option to noweb).  The previous behavior was to supply
1420       -L'' -- which caused noweb to use the "default" line numbering scheme.
1421       Thanks to Chris Richards for pointing this out (and supplying the fix).
1422    
1423    2. Once again, I reworked some aspects of the FFI:
1424    
1425       A. The incomplete/complete type business:
1426    
1427       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1428         gone!
1429       - ML types representing an incomplete type are now *equal* to
1430         ML types representing their corresponding complete types (just like
1431         in C).  This is still safe because ml-nlffigen will not generate
1432         RTTI for incomplete types, nor will it generate functions that
1433         require access to such RTTI.   But when ML code generated from both
1434         incomplete and complete versions of the C type meet, the ML types
1435         are trivially interoperable.
1436    
1437         NOTE:  These changes restore the full generality of the translation
1438         (which was previously lost when I eliminated functorization)!
1439    
1440       B. Enum types:
1441    
1442       - Structure C now has a type constructor "enum" that is similar to
1443         how the "su" constructor works.  However, "enum" is not a phantom
1444         type because each "T enum" has values (and is isomorphic to
1445         MLRep.Signed.int).
1446       - There are generic access operations for enum objects (using
1447         MLRep.Signed.int).
1448       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1449         * The structure contains the definition of type "mlrep" (the ML-side
1450         representation type of the enum).  Normally, mlrep is the same
1451         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1452         then mlrep will be defined as a datatype -- thus facilitating
1453         pattern matching on mlrep values.
1454         ("-ec" will be suppressed if there are duplicate values in an
1455          enumeration.)
1456         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1457         will be generated for each C enum constant xxx.
1458         * Conversion functions m2i and i2m convert between mlrep and
1459         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1460         * Coversion functions c and ml convert between mlrep and "tag enum".
1461         * Access functions (get/set) fetch and store mlrep values.
1462       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1463         enumerations are merged into one single enumeration represented by
1464         structure E_'.
1465    
1466    ----------------------------------------------------------------------
1467    Name: Allen Leung
1468    Date: 2002/02/25 04:45:00 EST
1469    Tag: leunga-20020225-cps-spill
1470    
1471    This is a new implementation of the CPS spill phase.
1472    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1473    In case of problems, replace it with the old file spill.sml
1474    
1475    The current compiler runs into some serious performance problems when
1476    constructing a large record.  This can happen when we try to compile a
1477    structure with many items.  Even a very simple structure like the following
1478    makes the compiler slow down.
1479    
1480        structure Foo = struct
1481           val x_1 = 0w1 : Word32.int
1482           val x_2 = 0w2 : Word32.int
1483           val x_3 = 0w3 : Word32.int
1484           ...
1485           val x_N = 0wN : Word32.int
1486        end
1487    
1488    The following table shows the compile time, from N=1000 to N=4000,
1489    with the old compiler:
1490    
1491    N
1492    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1493           MLRISC ra                               0.06u  0.00s  0.05g
1494              (spills = 0 reloads = 0)
1495           TOTAL                                   0.63u  0.07s  0.21g
1496    
1497    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1498           MLRISC ra                               5.68u  0.59s  3.93g
1499              (spills = 0 reloads = 0)
1500           TOTAL                                   14.71u  0.99s  4.81g
1501    
1502    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1503           MLRISC ra                               5.54u  0.65s  3.91g
1504              (spills = 543 reloads = 1082)
1505           TOTAL                                   65.40u  3.13s  6.00g
1506    
1507    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1508           MLRISC ra                               0.80u  0.10s  0.55g
1509              (spills = 42 reloads = 84)
1510           TOTAL                                   129.42u  5.10s  4.13g
1511    
1512    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1513           MLRISC ra                               2.69u  0.27s  1.38g
1514              (spills = 62 reloads = 124)
1515           TOTAL                                   682.48u  19.61s  13.99g
1516    
1517    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1518           MLRISC ra                               4.96u  0.27s  2.72g
1519              (spills = 85 reloads = 170)
1520           TOTAL                                   2375.26u  57.21s  48.00g
1521    
1522    As you can see the old cps spill module suffers from some serious
1523    performance problem.  But since I cannot decipher the old code fully,
1524    instead of patching the problems up, I'm reimplementing it
1525    with a different algorithm.  The new code is more modular,
1526    smaller when compiled, and substantially faster
1527    (O(n log n) time and O(n) space).  Timing of the new spill module:
1528    
1529    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1530          MLRISC ra                               0.25u  0.02s  0.15g
1531             (spills=1 reloads=3)
1532          TOTAL                                   7.74u  0.34s  1.62g
1533    
1534    Implementation details:
1535    
1536    As far as I can tell, the purpose of the CPS spill module is to make sure the
1537    number of live variables at any program point (the bandwidth)
1538    does not exceed a certain limit, which is determined by the
1539    size of the spill area.
1540    
1541    When the bandwidth is too large, we decrease the register pressure by
1542    packing live variables into spill records.  How we achieve this is
1543    completely different than what we did in the old code.
1544    
1545    First, there is something about the MLRiscGen code generator
1546    that we should be aware of:
1547    
1548    o MLRiscGen performs code motion!
1549    
1550       In particular, it will move floating point computations and
1551       address computations involving only the heap pointer to
1552       their use sites (if there is only a single use).
1553       What this means is that if we have a CPS record construction
1554       statement
1555    
1556           RECORD(k,vl,w,e)
1557    
1558       we should never count the new record address w as live if w
1559       has only one use (which is often the case).
1560    
1561       We should do something similar to floating point, but the transformation
1562       there is much more complex, so I won't deal with that.
1563    
1564    Secondly, there are now two new cps primops at our disposal:
1565    
1566     1. rawrecord of record_kind option
1567        This pure operator allocates some uninitialized storage from the heap.
1568        There are two forms:
1569    
1570         rawrecord NONE [INT n]  allocates a tagless record of length n
1571         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1572                                     and initializes the tag.
1573    
1574     2. rawupdate of cty
1575          rawupdate cty (v,i,x)
1576          Assigns to x to the ith component of record v.
1577          The storelist is not updated.
1578    
1579    We use these new primops for both spilling and increment record construction.
1580    
1581     1. Spilling.
1582    
1583        This is implemented with a linear scan algorithm (but generalized
1584        to trees).  The algorithm will create a single spill record at the
1585        beginning of the cps function and use rawupdate to spill to it,
1586        and SELECT or SELp to reload from it.  So both spills and reloads
1587        are fine-grain operations.  In contrast, in the old algorithm
1588        "spills" have to be bundled together in records.
1589    
1590        Ideally, we should sink the spill record construction to where
1591        it is needed.  We can even split the spill record into multiple ones
1592        at the places where they are needed.  But CPS is not a good
1593        representation for global code motion, so I'll keep it simple and
1594        am not attempting this.
1595    
1596     2. Incremental record construction (aka record splitting).
1597    
1598        Long records with many component values which are simulatenously live
1599        (recall that single use record addresses are not considered to
1600         be live) are constructed with rawrecord and rawupdate.
1601        We allocate space on the heap with rawrecord first, then gradually
1602        fill it in with rawupdate.  This is the technique suggested to me
1603        by Matthias.
1604    
1605        Some restrictions on when this is applicable:
1606        1. It is not a VECTOR record.  The code generator currently does not handle
1607           this case. VECTOR record uses double indirection like arrays.
1608        2. All the record component values are defined in the same "basic block"
1609           as the record constructor.  This is to prevent speculative
1610           record construction.
1611    
1612    ----------------------------------------------------------------------
1613    Name: Allen Leung
1614    Date: 2002/02/22 01:02:00 EST
1615    Tag: leunga-20020222-mlrisc-tools
1616    
1617    Minor bug fixes in the parser and rewriter
1618    
1619    ----------------------------------------------------------------------
1620    Name: Allen Leung
1621    Date: 2002/02/21 20:20:00 EST
1622    Tag: leunga-20020221-peephole
1623    
1624    Regenerated the peephole files.  Some contained typos in the specification
1625    and some didn't compile because of pretty printing bugs in the old version
1626    of 'nowhere'.
1627    
1628    ----------------------------------------------------------------------
1629    Name: Allen Leung
1630    Date: 2002/02/19 20:20:00 EST
1631    Tag: leunga-20020219-mlrisc-tools
1632    Description:
1633    
1634       Minor bug fixes to the mlrisc-tools library:
1635    
1636       1.  Fixed up parsing colon suffixed keywords
1637       2.  Added the ability to shut the error messages up
1638       3.  Reimplemented the pretty printer and fixed up/improved
1639           the pretty printing of handle and -> types.
1640       4.  Fixed up generation of literal symbols in the nowhere tool.
1641       5.  Added some SML keywords to to sml.sty
1642    
1643    ----------------------------------------------------------------------
1644    Name: Matthias Blume
1645    Date: 2002/02/19 16:20:00 EST
1646    Tag: blume-20020219-cmffi
1647    Description:
1648    
1649    A wild mix of changes, some minor, some major:
1650    
1651    * All C FFI-related libraries are now anchored under $c:
1652        $/c.cm      --> $c/c.cm
1653        $/c-int.cm  --> $c/internals/c-int.cm
1654        $/memory.cm --> $c/memory/memory.cm
1655    
1656    * "make" tool (in CM) now treats its argument pathname slightly
1657      differently:
1658        1. If the native expansion is an absolute name, then before invoking
1659           the "make" command on it, CM will apply OS.Path.mkRelative
1660           (with relativeTo = OS.FileSys.getDir()) to it.
1661        2. The argument will be passed through to subsequent phases of CM
1662           processing without "going native".  In particular, if the argument
1663           was an anchored path, then "make" will not lose track of that anchor.
1664    
1665    * Compiler backends now "know" their respective C calling conventions
1666      instead of having to be told about it by ml-nlffigen.  This relieves
1667      ml-nlffigen from one of its burdens.
1668    
1669    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1670    
1671    * Export C_DEBUG and C_Debug from $c/c.cm.
1672    
1673    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1674      subtyping relationship between incomplete pointers and their complete
1675      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1676      with the convention of instantiating 'o with ('t, 'c) obj whenever
1677      the pointer target type is complete.  In the incomplete case, 'o
1678      will be instantiated with some "'c iobj" -- a type obtained by
1679      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1680    
1681      Operations that work on both incomplete and complete pointer types are
1682      typed as taking an 'o ptr while operations that require the target to
1683      be known are typed as taking some ('t, 'c) obj ptr.
1684    
1685      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1686      where void is an eqtype without any values.  This makes it possible
1687      to work on voidptr values using functions meant to operate on light
1688      incomplete pointers.
1689    
1690    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1691      been vastly simplified.
1692    
1693    ----------------------------------------------------------------------
1694    Name: Matthias Blume
1695    Date: 2002/02/19 10:48:00 EST
1696    Tag: blume-20020219-pqfix
1697    Description:
1698    
1699    Applied Chris Okasaki's bug fix for priority queues.
1700    
1701    ----------------------------------------------------------------------
1702    Name: Matthias Blume
1703    Date: 2002/02/15 17:05:00
1704    Tag: Release_110_39
1705    Description:
1706    
1707    Last-minute retagging is becoming a tradition... :-(
1708    
1709    This is the working release 110.39.
1710    
1711    ----------------------------------------------------------------------
1712    Name: Matthias Blume
1713    Date: 2002/02/15 16:00:00 EST
1714    Tag: Release_110_39-orig
1715    Description:
1716    
1717    Working release 110.39.  New bootfiles.
1718    
1719    (Update: There was a small bug in the installer so it wouldn't work
1720    with all shells.  So I retagged. -Matthias)
1721    
1722    ----------------------------------------------------------------------
1723    Name: Matthias Blume
1724    Date: 2002/02/15 14:17:00 EST
1725    Tag: blume-20020215-showbindings
1726    Description:
1727    
1728    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1729    the latter can be useful for exploring what bindings are available at
1730    the interactive prompt.  (The first function returns only the list
1731    of symbols that are really bound, the second prints those but also the
1732    ones that CM's autoloading mechanism knows about.)
1733    
1734    ----------------------------------------------------------------------
1735    Name: Matthias Blume
1736    Date: 2002/02/15 12:08:00 EST
1737    Tag: blume-20020215-iptrs
1738    Description:
1739    
1740    Two improvements to ml-nlffigen:
1741    
1742      1. Write files only if they do not exist or if their current contents
1743         do not coincide with what's being written.  (That is, avoid messing
1744         with the time stamps unless absolutely necessary.)
1745    
1746      2. Implement a "repository" mechanism for generated files related
1747         to "incomplete pointer types".   See the README file for details.
1748    
1749    ----------------------------------------------------------------------
1750    Name: Matthias Blume
1751    Date: 2002/02/14 11:50:00 EST
1752    Tag: blume-20020214-quote
1753    Description:
1754    
1755    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1756    because of the new and improved tag generation scheme.  (Thanks to Allen
1757    Leung for pointing it out.)
1758    
1759    ----------------------------------------------------------------------
1760    Name: Lal George
1761    Date: 2002/02/14 09:55:27 EST 2002
1762    Tag: george-20020214-isabelle-bug
1763    Description:
1764    
1765    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1766    of Isabelle on the x86.
1767    
1768    From Allen:
1769    -----------
1770     I've found the problem:
1771    
1772         in ra-core.sml, I use the counter "blocked" to keep track of the
1773         true number of elements in the freeze queue.  When the counter goes
1774         to zero, I skip examining the queue.  But I've messed up the
1775         bookkeeping in combine():
1776    
1777             else ();
1778             case !ucol of
1779               PSEUDO => (if !cntv > 0 then
1780                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1781                                        ^^^^^^^^^^^^^^^^^^^^^^^
1782                      moveu := mergeMoveList(!movev, !moveu)
1783                     )
1784                  else ();
1785    
1786         combine() is called to coalesce two nodes u and v.
1787         I think I was thinking that if the move counts of u and v are both
1788         greater than zero then after they are coalesced then one node is
1789         removed from the freeze queue.  Apparently I was thinking that
1790         both u and v are of low degree, but that's clearly not necessarily true.
1791    
1792    
1793    02/12/2002:
1794        Here's the patch.  HOL now compiles.
1795    
1796        I don't know how this impact on performance (compile
1797        time or runtime).  This bug caused the RA (especially on the x86)
1798        to go thru the potential spill phase when there are still nodes on the
1799        freeze queue.
1800    
1801    
1802    
1803    
1804    ----------------------------------------------------------------------
1805    Name: Matthias Blume
1806    Date: 2002/02/13 22:40:00 EST
1807    Tag: blume-20020213-fptr-rtti
1808    Description:
1809    
1810    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1811    updates.
1812    
1813    ----------------------------------------------------------------------
1814    Name: Matthias Blume
1815    Date: 2002/02/13 16:41:00 EST
1816    Tag: blume-20020213-cmlpq
1817    Description:
1818    
1819    Added new priority queue export symbols (which have just been added to
1820    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1821    not compile and the installer would choke.)
1822    
1823    ----------------------------------------------------------------------
1824    Name: Matthias Blume
1825    Date: 2002/02/13 16:15:00 EST
1826    Tag: blume-20020213-various
1827    Description:
1828    
1829    1. More tweaks to ml-nlffigen:
1830    
1831       - better internal datastructures (resulting in slight speedup)
1832       - "-match" option requires exact match
1833       - "localized" gensym counters (untagged structs/unions nested within
1834         other structs/unions or within typedefs get a fresh counter; their
1835         tag will be prefixed by a concatenation of their parents' tags)
1836       - bug fixes (related to calculation of transitive closure of types
1837         to be included in the output)
1838    
1839    2. Minor Basis updates:
1840    
1841       - added implementations for List.collate and Option.app
1842    
1843    ----------------------------------------------------------------------
1844    Name: Matthias Blume
1845    Date: 2002/02/11 15:55:00 EST
1846    Tag: blume-20020211-gensym
1847    Description:
1848    
1849    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1850    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1851    names (ML structure names that correspond to unnamed C structs, unions,
1852    and enums), so that separate runs of ml-nlffigen do not clash.
1853    
1854    ----------------------------------------------------------------------
1855    Name: Matthias Blume
1856    Date: 2002/02/11 12:05:00 EST
1857    Tag: blume-20020211-gensml
1858    Description:
1859    
1860    A quick fix for a problem with GenSML (in the pgraph-util library):
1861    Make generation of toplevel "local" optional.  (Strictly speaking,
1862    signature definitions within "local" are not legal SML.)
1863    
1864    Other than that: updates to INSTALL and cm/TODO.
1865    
1866    ----------------------------------------------------------------------
1867    Name: Matthias Blume
1868    Date: 2002/02/08 15:00:00 EST
1869    Tag: blume-20020208-uniquepid
1870    Description:
1871    
1872    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1873    
1874    1. The installer (config/install.sh) has gotten smarter:
1875    
1876         - Configuration options are a bit easier to specify now
1877           (in config/targets).
1878         - Bug in recognizing .tar.bz2 files fixed.
1879         - Installer automatically resolves dependencies between
1880           configuration options (e.g., if you ask for eXene, you will
1881           also get cml -- regardless whether you asked for it or not).
1882         - Installer can run in "quieter mode" by setting the environment
1883           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1884           "completely silent", though.
1885         - Build HashCons library as part of smlnj-lib.
1886    
1887    2. A new scheme for assigning persistent identifiers to compilation
1888       units (and, by extension, to types etc.) has been put into place.
1889       This fixes a long-standing bug where types and even dynamic values
1890       can get internally confused, thereby compromising type safety
1891       (abstraction) and dynamic correctness.  See
1892    
1893         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1894    
1895       for an example of how things could go wrong until now.
1896    
1897       The downside of the new scheme is that pids are not quite as
1898       persistent as they used to be: CM will generate a fresh pid
1899       for every compilation unit that it thinks it sees for the first
1900       time.  That means that if you compile starting from a clean, fresh
1901       source tree at two different times, you end up with different
1902       binaries.
1903    
1904       Cutoff recompilation, however, has not been compromised because
1905       CM keeps pid information in special caches between runs.
1906    
1907    ----------------------------------------------------------------------
1908    Name: Lal George
1909    Date: 2002/02/07 15:34:13 EST 2002
1910    Tag: <none>
1911    Description:
1912    
1913    Compilers that generate assembly code may produce  global labels
1914    whose value is resolved at link time. The various peephole optimization
1915    modules did not take this in account.
1916    
1917    TODO. The Labels.addrOf function should really return an option
1918    type so that clients are forced to deal with this issue, rather
1919    than an exception being raised.
1920    
1921    ----------------------------------------------------------------------
1922    Name: Lal George
1923    Date: 2002/02/06 13:55:02 EST
1924    Tag: george-20020206-ra-breakup
1925    Description:
1926    
1927    1. A bug fix from Allen.
1928    
1929        A typo causes extra fstp %st(0)'s to be generated at compensation
1930        edges, which might cause stack underflow traps at runtime.  This
1931        occurs in fft where there are extraneous fstps right before the 'into'
1932        trap instruction (in this case they are harmless since none of the
1933        integers overflow.)
1934    
1935    2. Pulled out various utility modules that were embedded in the modules
1936       of the register allocator. I need these modules for other purposes, but
1937       they are not complete enough to put into a library (just yet).
1938    ----------------------------------------------------------------------
1939    Name: Matthias Blume
1940    Date: 2002/01/31 16:05:00 EST
1941    Tag: blume-20020131-sparc-ccalls
1942    Description:
1943    
1944    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1945       of extra stack space by mistake.  Fixed.
1946    
1947    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1948    
1949    ----------------------------------------------------------------------
1950    Name: Allen Leung
1951    Date: 2002/01/30
1952    Tag: leunga-20020130-nowhere-bug-fix
1953    Description:
1954    
1955       MLRISC bug fixes:
1956       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1957       2. MachineInt.fromString was negating its value.
1958    
1959    ----------------------------------------------------------------------
1960    Name: Matthias Blume
1961    Date: 2002/01/29
1962    Tag: blume-20020129-INSTALL
1963    Description:
1964    
1965    - Added somewhat detailed installation instructions (file INSTALL).
1966    - Fixed curl-detection bug in config/install.sh.
1967    - It is now possible to select the URL getter using the URLGETTER
1968      environment variable:
1969    
1970          not set / "unknown"      --> automatic detection (script tries wget,
1971                                       curl, and lynx)
1972          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1973                                       how to properly invoke them)
1974          other                    --> use $URLGETTER directly, it must take
1975                                       precisely two command-line arguments
1976                                       (source URL and destination file name)
1977    
1978    ----------------------------------------------------------------------
1979    Name: Matthias Blume
1980    Date: 2002/01/28
1981    Tag: blume-20020128-sparc-ccalls
1982    Description:
1983    
1984    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1985    - Make use of the allocParam argument in sparc-c-calls.
1986    
1987    ----------------------------------------------------------------------
1988    Name: Matthias Blume
1989    Date: 2002/01/28
1990    Tag: blume-20020128-allocParam
1991    Description:
1992    
1993    John Reppy:  Changes c-calls API to accept client-callback for
1994    allocating extra stack space.
1995    me: Corresponding changes to mlriscGen (using a dummy argument that
1996        does not change the current behavior).
1997    
1998    ----------------------------------------------------------------------
1999    Name: Matthias Blume
2000    Date: 2002/01/28 12:00:00
2001    Tag: Release_110_38
2002    Description:
2003    
2004    This time for real!!!
2005    
2006    ----------------------------------------------------------------------
2007    Name: Matthias Blume
2008    Date: 2002/01/28 10:56:00 EST
2009    Tag: blume-20020128-retraction
2010    Description:
2011    
2012    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
2013       with blume-Release_110_38-retracted.)
2014    
2015    1. Fixed a problem with incorrect rounding modes in real64.sml.
2016       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
2017    
2018    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
2019       and enums fixed.  The naming of corresponding ML identifiers should
2020       now be consistent again.
2021    
2022    ----------------------------------------------------------------------
2023    Name: Allen Leung
2024    Date: 2002/01/27
2025    Tag: leunga-20020127-nowhere
2026    Description:
2027    
2028       Added a target called nowhere in the configuration scripts.
2029       Enabling this will build the MLRISC 'nowhere' tool (for translating
2030       programs with where-clauses into legal SML code) during installation.
2031    
2032    ----------------------------------------------------------------------
2033    Name: Matthias Blume
2034    Date: 2002/01/25 21:27:00 EST
2035    Tag: blume-Release_110_38-retracted
2036    Description:
2037    
2038    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
2039    
2040    README will be added later.
2041    
2042    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
2043    (Release_110_38) removed.  Reason: Last-minute bug fixes.
2044    
2045    ----------------------------------------------------------------------
2046    Name: Matthias Blume
2047    Date: 2002/01/25
2048    Tag: blume-20020125-ffi
2049    Description:
2050    
2051    A large number of tweaks and improvements to ml-nlffi-lib and
2052    ml-nlffigen:
2053    
2054       - ML represenation types have been streamlined
2055       - getter and setter functions work with concrete values, not abstract
2056         ones where possible
2057       - ml-nlffigen command line more flexible (see README file there)
2058       - some bugs have been fixed (hopefully)
2059    
2060    ----------------------------------------------------------------------
2061    Name: Lal George
2062    Date: 2002/01/24
2063    Tag: george-20020124-risc-ra-interface
2064    Description:
2065    
2066       There is a dramatic simplification in the interface to the
2067       register allocator for RISC architectures as a result of making
2068       parallel copy instructions explicit.
2069    
2070    ----------------------------------------------------------------------
2071    Name: Matthias Blume
2072    Date: 2002/01/22
2073    Tag: blume-20020122-x86-ccalls
2074    Description:
2075    
2076    Bug fix for c-calls on x86 (having to do with how char- and
2077    short-arguments are being handled).
2078    
2079    ----------------------------------------------------------------------
2080    Name: Matthias Blume
2081    Date: 2002/01/21
2082    Tag: blume-20020121-ff
2083    Description:
2084    
2085    Another day of fiddling with the FFI...
2086    
2087    1. Bug fix/workaround:  CKIT does not complain about negative array
2088       dimensions, so ml-nlffigen has to guard itself against this possibility.
2089       (Otherwise a negative dimension would send it into an infinite loop.)
2090    
2091    2. Some of the abstract types (light objects, light pointers, most "base"
2092       types) in structure C are now eqtypes.
2093    
2094    3. Added constructors and test functions for NULL function pointers.
2095    
2096    ----------------------------------------------------------------------
2097    Name: Matthias Blume
2098    Date: 2002/01/18
2099    Tag: blume-20020118-ready-for-new-release
2100    Description:
2101    
2102    Made config/srcarchiveurl point to a new place.  (Will provide boot
2103    files shortly.)
2104    
2105    Maybe we christen this to be 110.38?
2106    
2107    ----------------------------------------------------------------------
2108    Name: Matthias Blume
2109    Date: 2002/01/18
2110    Tag: blume-20020118-more-ffifiddle
2111    Description:
2112    
2113    Today's FFI fiddling:
2114    
2115      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
2116        C.Set.* that use concrete (MLRep.*) arguments and results instead
2117        of abstract ones.
2118    
2119      - Provided word-style bit operations etc. for "int" representation
2120        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
2121        and Long.
2122    
2123    ----------------------------------------------------------------------
2124    Name: Matthias Blume
2125    Date: 2002/01/18
2126    Tag: blume-20020118-use-x86-fp
2127    Description:
2128    
2129    Now that x86-fast-fp seems to be working, I turned it back on again
2130    by default.  (Seems to work fine now, even with the FFI.)
2131    
2132    Other than that, I added some documentation about the FFI to
2133    src/ml-nlffigen/README and updated the FFI test examples in
2134    src/ml-nlffi-lib/Tests/*.
2135    
2136    ----------------------------------------------------------------------
2137    Name: Allen Leung
2138    Date: 2002/01/17
2139    Tag: leunga-20020117-x86-fast-fp-call
2140    Description:
2141    
2142       1. Fixed a problem with handling return fp values when x86's fast fp
2143          mode is turned on.
2144    
2145       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
2146          of %f32.
2147    
2148       3. Added a constructor INT32lit to the ast of MLRISC tools.
2149    
2150    ----------------------------------------------------------------------
2151    Name: Matthias Blume
2152    Date: 2002/01/16
2153    Tag: blume-20020116-ffifiddle
2154    Description:
2155    
2156    More fiddling with the FFI interface:
2157    
2158     - Make constness 'c instead of rw wherever possible.  This eliminates
2159       the need for certain explicit coercions.  (However, due to ML's
2160       value polymorphism, there will still be many cases where explicit
2161       coercions are necessary.  Phantom types are not the whole answer
2162       to modeling a subtyping relationship in ML.)
2163    
2164     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
2165    
2166     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
2167       clumsy workaround for varargs problem.)
2168    
2169    ----------------------------------------------------------------------
2170    Name: Lal George
2171    Date: 2002/01/15
2172    Tag: <none>
2173    Description:
2174    
2175    1. Since COPY instructions are no longer native to the architecture,
2176       a generic functor can be used to implement the expandCopies function.
2177    
2178    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
2179       TEXT segment.
2180    
2181    ----------------------------------------------------------------------
2182    Name: Matthias Blume
2183    Date: 2002/01/15
2184    Tag: blume-20020115-ffiupdates
2185    Description:
2186    
2187    1. Fix for bug resulting in single-precision float values being returned
2188       incorrectly from FFI calls.
2189    
2190    2. Small modifications to C FFI API:
2191    
2192        - memory-allocation routines return straight objects (no options)
2193          and raise an exception in out-of-memory situations
2194        - unsafe extensions to cast between function pointers and pointers
2195          from/to ints
2196        - added structure C_Debug as an alternative to structure C where
2197          pointer-dereferencing (|*| and |*!) always check for null-pointers
2198        - added open_lib' to DynLinkage;  open_lib' works like open_lib
2199          but also takes a (possibly empty) list of existing library handles
2200          that the current library depends on
2201    
2202    ----------------------------------------------------------------------
2203    Name: Matthias Blume
2204    Date: 2002/01/10
2205    Tag: blume-20020110-newffigen
2206    Description:
2207    
2208    1. Updates to portable graph code.
2209    
2210    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
2211       more scalable now so that even huge interfaces such as the one
2212       for GTK compile in finite time and space. :-)
2213       See src/ml-nlffigen/README for details on what's new.
2214    
2215    ----------------------------------------------------------------------
2216    Name: Lal George
2217    Date: 2001/01/09 14:31:35 EST 2002
2218    Tag: george-20011206-rm-native-copy
2219    Description:
2220    
2221            Removed the native COPY and FCOPY instructions
2222            from all the architectures and replaced it with the
2223            explicit COPY instruction from the previous commit.
2224    
2225            It is now possible to simplify many of the optimizations
2226            modules that manipulate copies. This has not been
2227            done in this change.
2228    
2229    ----------------------------------------------------------------------
2230    Name: Lal George
2231    Date: 2001/12/06 16:50:13 EST 2001
2232    Tag: george-20011206-mlrisc-instruction
2233    Description:
2234    
2235    Changed the representation of instructions from being fully abstract
2236    to being partially concrete. That is to say:
2237    
2238      from
2239            type instruction
2240    
2241      to
2242            type instr                              (* machine instruction *)
2243    
2244            datatype instruction =
2245                LIVE of {regs: C.cellset, spilled: C.cellset}
2246              | KILL of {regs: C.cellset, spilled: C.cellset}
2247              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
2248              | ANNOTATION of {i: instruction, a: Annotations.annotation}
2249              | INSTR of instr
2250    
2251    This makes the handling of certain special instructions that appear on
2252    all architectures easier and uniform.
2253    
2254    LIVE and KILL say that a list of registers are live or killed at the
2255    program point where they appear. No spill code is generated when an
2256    element of the 'regs' field is spilled, but the register is moved to
2257    the 'spilled' (which is present, more for debugging than anything else).
2258    
2259    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
2260    We used to generate:
2261    
2262            DEFFREG f1
2263            f1 := f2 + f3
2264            trapb
2265    
2266    but now generate:
2267    
2268            f1 := f2 + f3
2269            trapb
2270            LIVE {regs=[f1,f2,f3], spilled=[]}
2271    
2272    Furthermore, the DEFFREG (hack) required that all floating point instruction
2273    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
2274    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
2275    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
2276    this architecture).
2277    
2278    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
2279    all the architectures. This will result in further simplification of the
2280    register allocator that must be aware of them for coalescing purposes, and
2281    will also simplify certain aspects of the machine description that provides
2282    callbacks related to parallel copies.
2283    
2284    ANNOTATION should be obvious, and now INSTR represents the honest to God
2285    machine instruction set!
2286    
2287    The <arch>/instructions/<arch>Instr.sml files define certain utility
2288    functions for making porting easier -- essentially converting upper case
2289    to lower case. All machine instructions (of type instr) are in upper case,
2290    and the lower case form generates an MLRISC instruction. For example on
2291    the alpha we have:
2292    
2293      datatype instr =
2294         LDA of {r:cell, b:cell, d:operand}
2295       | ...
2296    
2297      val lda : {r:cell, b:cell, d:operand} -> instruction
2298        ...
2299    
2300    where lda is just (INSTR o LDA), etc.
2301    
2302    ----------------------------------------------------------------------
2303    Name: Matthias Blume
2304    Date: 2001/11/22 21:40:00 EST
2305    Tag: Release_110_37
2306    Description:
2307    
2308    Release 110.37.  This time for real.
2309    
2310    ----------------------------------------------------------------------
2311    Name: Matthias Blume
2312    Date: 2001/11/21 16:35:00 EST
2313    Tag: blume-20011121-foot-in-mouth
2314    Description:
2315    
2316    Removed the "Release_110_37" tag because of a serious bug.
2317    This will be re-tagged once the bug is fixed.
2318    
2319    ----------------------------------------------------------------------
2320    Name: Matthias Blume
2321    Date: 2001/11/21 16:14:00 EST
2322    Tag: blume-20011121-forgottenfile
2323    Description:
2324    
2325    Forgot to add a file.  (Just a .tex-file -- part of
2326    the CM manual source.)
2327    
2328    ----------------------------------------------------------------------
2329    Name: Matthias Blume
2330    Date: 2001/11/21 16:10:00 EST
2331    Tag: blume-20011121-invalid_110_37
2332    Description:
2333    
2334    Note: I removed the original tag "Release_110_37" from this commit
2335          because we found a serious bug in all non-x86 backends.
2336          - Matthias
2337    
2338    1. Modifications to the SML/NJ code generator and to the runtime system
2339       so that code object name strings are directly inserted into code
2340       objects at code generation time.  The only business the runtime system
2341       has with this is now to read the name strings on occasions.
2342       (The encoding of the name string has also changed somewhat.)
2343    
2344    2. CM now implements a simple "set calculus" for specifying export lists.
2345       In particular, it is now possible to refer to the export lists of
2346       other libraries/groups/sources and form unions as well as differences.
2347       See the latest CM manual for details.
2348    
2349    3. An separate notion of "proxy" libraries has again be eliminated from
2350       CM's model.  (Proxy libraries are now simply a special case of using
2351       the export list calculus.)
2352    
2353    4. Some of the existing libraries now take advantage of the new set
2354       calculus.
2355       (Notice that not all libraries have been converted because some
2356       of the existing .cm-files are supposed to be backward compatible
2357       with 110.0.x.)
2358    
2359    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2360       "General.exnMessage"!  The former relies on a certain hook to be
2361       initialized, and that often does not happen in the stand-alone case.)
2362    
2363    ----------------------------------------------------------------------
2364    Name: Lal George
2365    Date: 2001/11/21  13:56:18 EST
2366    Tag: george-2001121-pseudo-ops
2367    Description:
2368    
2369      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2370      ought to never be any question of incompatabilities with
2371      pseudo-op syntax expected by host assemblers.
2372    
2373      For now, only modules supporting GAS syntax are implemented
2374      but more should follow, such as MASM, and vendor assembler
2375      syntax, e.g. IBM as, Sun as, etc.
2376    
2377    ----------------------------------------------------------------------
2378    Name: Matthias Blume
2379    Date: 2001/11/14 11:52:00 EST
2380    Tag: blume-20011114-srcname
2381    Description:
2382    
2383    1. Routed the name of the current source file to mlriscgen where it
2384       should be directly emitted into the code object.  (This last part
2385       is yet to be done.)
2386    
2387    2. Some cleanup of the pgraph code to make it match the proposal that
2388       I put out the other day.  (The proposal notwithstanding, things are
2389       still in flux here.)
2390    
2391    ----------------------------------------------------------------------
2392    Name: Lal George
2393    Date: 2001/11/14 09:44:04 EST
2394    Tag:
2395    Description:
2396    
2397      Fix for a backpatching bug reported by Allen.
2398    
2399      Because the boundary between short and long span-dependent
2400      instructions is +/- 128, there are an astounding number of
2401      span-dependent instructions whose size is over estimated.
2402    
2403      Allen came up with the idea of letting the size of span
2404      dependent instructions be non-monotonic, for a maxIter
2405      number of times, after which the size must be monotonically
2406      increasing.
2407    
2408      This table shows the number of span-dependent instructions
2409      whose size was over-estimated as a function of maxIter, for the
2410      file Parse/parse/ml.grm.sml:
2411    
2412         maxIter            # of instructions:
2413            10                      687
2414            20                      438
2415            30                      198
2416            40                        0
2417    
2418      In compiling the compiler, there is no significant difference in
2419      compilation speed between maxIter=10 and maxIter=40. Actually,
2420      my measurements showed that maxIter=40 was a tad faster than
2421      maxIter=10! Also 96% of the  files in the compiler reach a fix
2422      point within 13 iterations, so fixing maxIter at 40, while high,
2423      is okay.
2424    
2425    ----------------------------------------------------------------------
2426    Name: Matthias Blume
2427    Date: 2001/10/31 15:25:00 EST
2428    Tag: blume-20011031-pgraph
2429    Description:
2430    
2431    CKIT:
2432    * Changed the "Function" constructor of type Ast.ctype to carry optional
2433      argument identifiers.
2434    * Changed the return type of TypeUtil.getFunction accordingly.
2435    * Type equality ignores the argument names.
2436    * TypeUtil.composite tries to preserve argument names but gives up quickly
2437      if there is a mismatch.
2438    
2439    installation script:
2440    * attempts to use "curl" if available (unless "wget" is available as well)
2441    
2442    CM:
2443    * has an experimental implementation of "portable graphs" which I will
2444      soon propose as an implementation-independent library format
2445    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2446    
2447    NLFFI-LIB:
2448    * some cleanup (all cosmetic)
2449    
2450    NLFFIGEN:
2451    * temporarily disabled the mechanism that suppresses ML output for
2452      C definitions whose identifiers start with an underscore character
2453    * generate val bindings for enum constants
2454    * user can request that only one style (light or heavy) is being used;
2455      default is to use both (command-line arguments: -heavy and -light)
2456    * fixed bug in handling of function types involving incomplete pointers
2457    * generate ML entry points that take record arguments (i.e., using
2458      named arguments) for C functions that have a prototype with named
2459      arguments
2460      (see changes to CKIT)
2461    
2462    ----------------------------------------------------------------------
2463    Name: Allen Leung
2464    Date: 2001/10/27 20:34:00 EDT
2465    Tag: leunga-20011027-x86-fast-fp-call
2466    Description:
2467    
2468       Fixed the bug described in blume-20010920-slowfp.
2469    
2470       The fix involves
2471          1. generating FCOPYs in FSTP in ia32-svid
2472          2. marking a CALL with the appropriate annotation
2473    
2474    ----------------------------------------------------------------------
2475    Name: Matthias Blume
2476    Date: 2001/10/16 11:32:00 EDT
2477    Tag: blume-20011016-netbsd
2478    Description:
2479    
2480    Underscore patch from Chris Richards (fixing problem with compiling
2481    runtime system under recent NetBSD).
2482    
2483    ----------------------------------------------------------------------
2484    Name: Allen Leung
2485    Date: 2001/10/12 17:18:32 EDT 2001
2486    Tag: leung-20011012-x86-printflowgraph
2487    Description:
2488    
2489    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2490    
2491    ----------------------------------------------------------------------
2492    Name: Lal George
2493    Date: 2001/10/11 23:51:34 EDT
2494    Tag: george-20011011-too-many-instrs
2495    Description:
2496    
2497    The representation of a program point never expected to see more
2498    than 65536 instructions in a basic block!
2499    
2500    ----------------------------------------------------------------------
2501    Name: Lal George
2502    Date: 2001/10/09 09:41:37 EDT
2503    Tag: george-20011008-mlrisc-labels
2504    Description:
2505    
2506    Changed the machine description files to support printing of
2507    local and global labels in assembly code, based on host assembler
2508    conventions.
2509    
2510    ----------------------------------------------------------------------
2511    Name: Matthias Blume
2512    Date: 2001/09/25 15:25:00 EDT
2513    Tag: blume-20010925-exninfo
2514    Description:
2515    
2516    I provided a non-hook implementation of exnName (at the toplevel) and
2517    made the "dummy" implementation of exnMessage (at the toplevel) more
2518    useful: if nothing gets "hooked in", then at least you are going to
2519    see the exception name and a message indicating why you don't see more.
2520    
2521    [For the time being, programs that need exnMessage and want to use
2522    ml-build should either use General.exnMessage (strongly recommended) or
2523    refer to structure General at some other point so that CM sees a
2524    static dependency.]
2525    
2526    [Similar remarks go for "print" and "use":  If you want to use their
2527    functionality in stand-alone programs generated by ml-build, then use
2528    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2529    
2530    ----------------------------------------------------------------------
2531    Name: Matthias Blume
2532    Date: 2001/09/20 17:28:00 EDT
2533    Tag: blume-20010920-slowfp
2534    Description:
2535    
2536    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2537    
2538    ----------------------------------------------------------------------
2539    Name: Matthias Blume
2540    Date: 2001/09/20 17:20:00 EDT
2541    Tag: blume-20010920-canonicalpaths
2542    Description:
2543    
2544    0. Updated the BOOT file (something that I forgot to do earlier).
2545    
2546    1. Small internal change to CM so that it avoids "/../" in filenames
2547       as much as possible (but only where it is safe).
2548    
2549    2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2550       that arguments that contain delimiters are passed through correctly.
2551       This change also means that all "special" arguments of the form
2552       @SMLxxx... must come first.
2553    
2554    3. Changed install script to put relative anchor names for tool commands
2555       into pathconfig.
2556    
2557    ----------------------------------------------------------------------
2558    Name: Matthias Blume
2559    Date: 2001/09/18 15:35:00 EDT
2560    Tag: blume-20010918-readme11036
2561    Description:
2562    
2563    Added README files.
2564    
2565    ----------------------------------------------------------------------
2566    Name: Matthias Blume
2567    Date: 2001/09/18 11:45:00 EDT
2568    Tag: Release_110_36 (retag)
2569    Description:
2570    
2571    Fixed mistake in config/preloads. Retagged as 110.36.
2572    
2573    ----------------------------------------------------------------------
2574    Name: Matthias Blume
2575    Date: 2001/09/18 09:40:00 EDT
2576    Tag: Release_110_36_orig (tag changed)
2577    Description:
2578    
2579    New version (110.36).  New bootfiles.
2580    
2581    ----------------------------------------------------------------------
2582    Name: Matthias Blume
2583    Date: 2001/09/14 16:15:00 EDT
2584    Tag: blume-20010914-x86fastfp
2585    Description:
2586    
2587    John committed some changes that Allen made, in particular a (hopefully)
2588    correctly working version of the x86-fp module.
2589    
2590    I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
2591    flag to "true".  Everything seems to compile to a fixpoint ok, and
2592    "mandelbrot" speeds up by about 15%.
2593    
2594    ----------------------------------------------------------------------
2595    Name: Matthias Blume
2596    Date: 2001/09/13 11:20:00 EDT
2597    Tag: blume-20010913-minimal
2598    Description:
2599    
2600    1. Stefan Monnier's patch to fix a miscompilation problem that
2601       was brought to light by John Reppy's work on Moby.
2602    
2603    2. Implemented a minimal "structure Compiler" that contains just
2604       "version" and "architecture".  The minimal version will be
2605       available when the full version is not.  This is for backward-
2606       compatibility with code that wants to test Compiler.version.
2607    
2608    ----------------------------------------------------------------------
2609    Name: Matthias Blume
2610    Date: 2001/08/28 14:03:00 EDT
2611    Tag: blume-20010828-ml-lex
2612    Description:
2613    
2614    Fix for bug 1581, received from Neophytos Michael.
2615    
2616    ----------------------------------------------------------------------
2617    Name: Matthias Blume
2618    Date: 2001/08/27 11:20:00 EDT
2619    Tag: blume-20010827-readme11035
2620    Description:
2621    
2622    Fleshed out the README file for 110.35.
2623    
2624    ----------------------------------------------------------------------
2625    Name: Matthias Blume
2626    Date: 2001/08/24 17:10:00 EDT
2627    Tag: Release_110_35
2628    Description:
2629    
2630    New version number (110.35).  New bootfiles.
2631    
2632    ----------------------------------------------------------------------
2633    Name: Lal George
2634    Date: 2001/08/24 13:47:18 EDT 2001
2635    Tag: george-20010824-MLRISC-graphs
2636    Description:
2637    
2638     removed clusters from MLRISC completely and replaced with graphs.
2639    
2640    ----------------------------------------------------------------------
2641    Name: Matthias Blume
2642    Date: 2001/08/23 17:50:00 EDT
2643    Tag: blume-20010823-toplevel
2644    Description:
2645    
2646    - some reorganization of the code that implements various kinds of
2647      environments in the compiler (static, dynamic, symbolic, combined)
2648    - re-implemented the EnvRef module so that evalStream works properly
2649      (if the stream contains references to "use", "CM.make", etc.)
2650    - cleaned up evalloop.sml and interact.sml (but they need more cleaning)
2651    
2652    ----------------------------------------------------------------------
2653    Name: Matthias Blume
2654    Date: 2001/08/20 15:50 EDT
2655    Tag: blume20010820-slipup
2656    Description:
2657    
2658    I forgot to commit a few files.  Here they are...
2659    
2660    ----------------------------------------------------------------------
2661    Name: Matthias Blume
2662    Date: 2001/08/20 15:35:00 EDT
2663    Tag: blume-20010820-debugprof
2664    Description:
2665    
2666    !!!! NEW BOOTFILES !!!!
2667    
2668    This is another round of reorganizing the compiler sources.  This
2669    time the main goal was to factor out all the "instrumentation"
2670    passes (for profiling and backtracing) into their own library.
2671    The difficulty was to do it in such a way that it does not depend
2672    on elaborate.cm but only on elabdata.cm.
2673    
2674    Therefore there have been further changes to both elaborate.cm and
2675    elabdata.cm -- more "generic" things have been moved from the former
2676    to the latter.  As a result, I was forced to split the assignment
2677    of numbers indicating "primtyc"s into two portions: SML-generic and
2678    SML/NJ-specific.  Since it would have been awkward to maintain,
2679    I bit the bullet and actually _changed_ the mapping between these
2680    numbers and primtycs.  The bottom line of this is that you need
2681    a new set of bin- and bootfiles.
2682    
2683    I have built new bootfiles for all architectures, so doing a fresh
2684    checkout and config/install.sh should be all you need.
2685    
2686    The newly created library's name is
2687    
2688        $smlnj/viscomp/debugprof.cm
2689    
2690    and its sources live under
2691    
2692        src/compiler/DebugProf
2693    
2694    ----------------------------------------------------------------------
2695    Name: Matthias Blume
2696    Date: 2001/08/15 17:15:00 EDT
2697    Tag: blume-20010815-compreorg
2698    Description:
2699    
2700    This is a first cut at reorganizing the CM libraries that make up the
2701    core of the compiler.  The idea is to separate out pieces that could
2702    be used independently by tools, e.g., the parser, the typechecker, etc.
2703    
2704    The current status is a step in this direction, but it is not quite
2705    satisfactory yet.  Expect more changes in the future.
2706    
2707    Here is the current (new) organization...
2708    
2709        What used to be $smlnj/viscomp/core.cm is now divided into
2710        six CM libraries:
2711    
2712             $smlnj/viscomp/basics.cm
2713                           /parser.cm
2714                           /elabdata.cm
2715                           /elaborate.cm
2716                           /execute.cm
2717                           /core.cm
2718    
2719        The CM files for these libraries live under src/system/smlnj/viscomp.
2720        All these libraries are proxy libraries that contain precisely
2721        one CM library component.  Here are the locations of the components
2722        (all within the src/compiler tree):
2723    
2724             Basics/basics.cm
2725             Parse/parser.cm
2726             ElabData/elabdata.cm
2727             Elaborator/elaborate.cm
2728             Execution/execute.cm
2729             core.cm
2730    
2731         [This organization is the same that has been used already
2732         for a while for the architecture-specific parts of the visible
2733         compiler and for the old version of core.cm.]
2734    
2735         As you will notice, many source files have been moved from their
2736         respective original locations to a new home in one of the above
2737         subtrees.
2738    
2739         The division of labor between the new libraries is the following:
2740    
2741             basics.cm:
2742                - Simple, basic definitions that pertain to many (or all) of
2743                  the other libraries.
2744             parser.cm:
2745                - The SML parser, producing output of type Ast.dec.
2746                - The type family for Ast is also defined and exported here.
2747             elabdata.cm:
2748                - The datatypes that describe input and output of the elaborator.
2749                  This includes types, absyn, and static environments.
2750             elaborator.cm:
2751                - The SML/NJ type checker and elaborator.
2752                  This maps an Ast.dec (with a given static environment) to
2753                  an Absyn.dec (with a new static environment).
2754                - This libraries implements certain modules that used to be
2755                  structures as functors (to remove dependencies on FLINT).
2756             execute.cm:
2757                - Everything having to do with executing binary code objects.
2758                - Dynamic environments.
2759             core.cm:
2760                - SML/NJ-specific instantiations of the elaborator and MLRISC.
2761                - Top-level modules.
2762                - FLINT (this should eventually become its own library)
2763    
2764    Notes:
2765    
2766    I am not 100% happy with the way I separated the elaborator (and its
2767    data structures) from FLINT.  Two instances of the same problem:
2768    
2769        1. Data structures contain certain fields that carry FLINT-specific
2770           information.  I hacked around this using exn and the property list
2771           module from smlnj-lib.  But the fact that there are middle-end
2772           specific fields around at all is a bit annoying.
2773    
2774        2. The elaborator calculates certain FLINT-related information.  I tried
2775           to make this as abstract as I could using functorization, but, again,
2776           the fact that the elaborator has to perform calculations on behalf
2777           of the middle-end at all is not nice.
2778    
2779        3. Having to used exn and property lists is unfortunate because it
2780           weakens type checking.  The other alternative (parameterizing
2781           nearly *everything*) is not appealing, though.
2782    
2783    I removed the "rebinding =" warning hack because due to the new organization
2784    it was awkward to maintain it.  As a result, the compiler now issues some of
2785    these warnings when compiling init.cmi during bootstrap compilation. On
2786    the plus side, you also get a warning when you do, for example:
2787       val op = = Int32.+
2788    which was not the case up to now.
2789    
2790    I placed "assign" and "deref" into the _Core structure so that the
2791    code that deals with the "lazy" keyword can find them there.  This
2792    removes the need for having access to the primitive environment
2793    during elaboration.
2794    
2795    ----------------------------------------------------------------------
2796    Name: Matthias Blume
2797    Date: 2001/08/13
2798    Tag: blume-20010813-closures
2799    Description:
2800    
2801    This fix was sent to us by Zhong Shao.  It is supposed to improve the
2802    performance of certain loops by avoiding needless closure allocation.
2803    
2804    ----------------------------------------------------------------------
2805    Name: Lal George
2806    Date: 2001/07/31 10:03:23 EDT 2001
2807    Tag: george-20010731-x86-fmalloc
2808    Description: Fixed bug in x86 calls
2809    
2810        There was a bug where call instructions would mysteriously
2811        vanish. The call instruction had to be one that returned
2812        a floating point value.
2813    
2814    ----------------------------------------------------------------------
2815    Name: Lal George
2816    Date: 2001/07/19 16:36:29 EDT 2001
2817    Tag: george-20010719-simple-cells
2818    Description:
2819    
2820    I have dramatically simplified the interface for CELLS in MLRISC.
2821    
2822    In summary, the cells interface is broken up into three parts:
2823    
2824      1. CellsBasis : CELLS_BASIS
2825    
2826            CellsBasis is a top level structure and common for all
2827            architectures.  it contains the definitions of basic datatypes
2828            and utility  functions over these types.
2829    
2830      2. functor Cells() : CELLS
2831    
2832            Cells generates an interface for CELLS that incorporates the
2833            specific resources on the target architecture, such as the
2834            presence of special register classes, their number and size,
2835            and various useful substructures.
2836    
2837      3. <ARCH>CELLS
2838    
2839            e.g. SparcCells: SPARCCELLS
2840    
2841            <ARCH>CELLS usually contains additional bindings for special
2842            registers  on the architecture, such as:
2843    
2844                    val r0 : cell           (* register zero *)
2845                    val y : cell            (* Y register *)
2846                    val psr : cell          (* processor status register *)
2847                    ...
2848    
2849            The structure returned by applying the Cells functor is opened
2850            in this interface.
2851    
2852    The main implication of all this is that the datatypes for cells is
2853    split between CellsBasis and CELLS -- a fairly simple change for user
2854    code.
2855    
2856    In the old scheme the CELLS interface had a definitional binding of
2857    the form:
2858    
2859            signature CELLS = sig
2860    
2861               structure CellsBasis = CellsBasis
2862    
2863               ...
2864    
2865            end
2866    
2867    With all the sharing constraints that goes on in MLRISC, this old
2868    design  quickly leads to errors such as:
2869    
2870            "structure definition spec inside of sharing ... "
2871    
2872    
2873    and appears to require an unacceptable amount of sharing and where
2874    constraint hackery.
2875    
2876    I think this error message (the interaction of definitional specs and
2877    sharing) requires more explanation on our web page.
2878    
2879    ----------------------------------------------------------------------
2880    Name: Matthias Blume
2881    Date: 2001/07/19 15:00:00 EDT
2882    Tag: blume-20010719-libreorg
2883    Description:
2884    
2885    This update puts together a fairly extensive but straightforward change
2886    to the way the libraries that implement the interactive system are
2887    organized:
2888    
2889       The biggest change is the elimination of structure Compiler.  As a
2890       replacement for this structure, there is now a CM library
2891       (known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
2892       that exports all the substructures of the original structure Compiler
2893       directly.  So instead of saying Compiler.Foo.bar one now simply
2894       says Foo.bar.  (The CM libraries actually export a collection of
2895       structures that is richer than the collection of substructures of
2896       structure Compiler.)
2897    
2898       To make the transition smooth, there is a separate library called
2899       $smlnj/compiler/compiler.cm which puts together and exports the
2900       original structure Compiler (or at least something very close to it).
2901    
2902       There are five members of the original structure Compiler
2903       that are not exported directly but which instead became members
2904       of a new structure Backend (described by signature BACKEND).  These are:
2905       structure Profile (: PROFILE), structure Compile (: COMPILE), structure
2906       Interact (: INTERACT), structure Machine (: MACHINE), and val
2907       architecture (: string).
2908    
2909       Structure Compiler.Version has become structure CompilerVersion.
2910    
2911       Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
2912       by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
2913       or x86, respectively.
2914       Each of these exports the same frontend structures that
2915       $smlnj/compiler.cm exports.  But they do not have a structure Backend
2916       and instead export some structure <Arch>Backend where <Arch> is Alpha32,
2917       Hppa, PPC, Sparc, or X86, respectively.
2918    
2919       Library $smlnj/compiler/all.cm exports the union of the exports of
2920       $smlnj/compiler/<arch>.cm
2921    
2922       There are no structures <Arch>Compiler anymore, use
2923       $smlnj/compiler/<arch>.cm instead.
2924    
2925       Library host-compiler-0.cm is gone.  Instead, the internal library
2926       that instantiates CM is now called cm0.cm.  Selection of the host
2927       compiler (backend) is no longer done here but. (Responsibility for it
2928       now lies with $smlnj/compiler/current.cm.  This seems to be more
2929       logical.)
2930    
2931       Many individual files have been moved or renamed.  Some files have
2932       been split into multiple files, and some "dead" files have been deleted.
2933    
2934    Aside from these changes to library organization, there are also changes
2935    to the way the code itself is organized:
2936    
2937       Structure Binfile has been re-implemented in such a way that it no
2938       longer needs any knowledge of the compiler.  It exclusively deals
2939       with the details of binfile layout.  It no longer invokes the
2940       compiler (for the purpose of creating new prospective binfile
2941       content), and it no longer has any knowledge of how to interpret
2942       pickles.
2943    
2944       Structure Compile (: COMPILE) has been stripped down to the bare
2945       essentials of compilation.  It no longer deals with linking/execution.
2946       The interface has been cleaned up considerably.
2947    
2948       Utility routines for dealing with linking and execution have been
2949       moved into their own substructures.
2950    
2951       (The ultimate goal of these changes is to provide a light-weight
2952       binfile loader/linker (at least for, e.g., stable libraries) that
2953       does not require CM or the compiler to be present.)
2954    
2955    CM documentation has been updated to reflect the changes to library
2956    organization.
2957    
2958    ----------------------------------------------------------------------
2959    Name: Matthias Blume
2960    Date: 2001/07/10 17:30:00 EDT
2961    Tag: Release_110_34
2962    Description:
2963    
2964    Minor tweak to 110.34 (re-tagged):
2965    
2966      - README.html file added to CVS repository
2967      - runtime compiles properly under FreeBSD 3.X and 4.X
2968    
2969    ----------------------------------------------------------------------
2970    Name: Matthias Blume
2971    Date: 2001/07/10 17:30:00 EDT
2972    Tag: Release_110_34
2973    Description:
2974    
2975    New version number (110.34). New bootfiles.
2976    
2977    ----------------------------------------------------------------------
2978    Name: Matthias Blume
2979    Date: 2001/07/09 16:00:00 EDT
2980    Tag: blume-20010709-more-varargs
2981    Description:
2982    
2983    I changed the handling of varargs in ml-nlffigen again:
2984    The ellipsis ... will now simply be ignored (with an accompanying warning).
2985    
2986    The immediate effect is that you can actually call a varargs function
2987    from ML -- but you can't actually supply any arguments beyond the ones
2988    specified explicitly.  (For example, you can call printf with its format
2989    string, but you cannot pass additional arguments.)
2990    
2991    This behavior is only marginally more useful than the one before, but
2992    it has the advantage that a function or, more importantly, a function
2993    type never gets dropped on the floor, thus avoiding follow-up problems with
2994    other types that refer to the offending one.
2995    
2996    ----------------------------------------------------------------------
2997    Name: Matthias Blume
2998    Date: 2001/07/09 11:25:00 EDT
2999    Tag: blume-20010709-varargs
3000    Description:
3001    
3002    1. ckit-lib.cm now exports structure Error
3003    2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
3004       with a warning accompanied by a source location.  Moreover, it
3005       merely skips the offending function or type and proceeds with the
3006       rest of its work.u  As a result, one can safely feed C code containing
3007       "..." to ml-nlffigen.
3008    3. There are some internal improvements to CM, providing slightly
3009       more general string substitutions in the tools subsystem.
3010    
3011    ----------------------------------------------------------------------
3012    Name: Matthias Blume
3013    Date: 2001/06/27 15:10:00 EDT
3014    Tag: blume-20010627-concur
3015    Description:
3016    
3017    Fixed a small bug in CM's handling of parallel compilation.
3018    (You could observe the bug by Control-C-interrupting an ordinary
3019    CMB.make or CM.stabilize and then attaching some compile servers.
3020    The result was that all of a sudden the previously interrupted
3021    compilation would continue on its own.  This was because of
3022    an over-optimization: CM did not bother to clean out certain queues
3023    when no servers were attached "anyway", resulting in the contents
3024    of these queues to grab control when new servers did get attached.)
3025    
3026    There is also another minor update to the CM manual.
3027    
3028    ----------------------------------------------------------------------
3029    Name: Matthias Blume
3030    Date: 2001/06/26 16:15:00 EDT
3031    Tag: blume-20010626-cmdoc
3032    Description:
3033    
3034    Minor typo fixed in CM manual (syntax diagram for libraries).
3035    
3036    ----------------------------------------------------------------------
3037    Name: Matthias Blume
3038    Date: 2001/06/25 22:55:00 EDT
3039    Tag: blume-20010625-x86pc
3040    Description:
3041    
3042    Fixed a nasty bug in the X86 assembly code that caused signal
3043    handlers to fail (crash) randomly.
3044    
3045    ----------------------------------------------------------------------
3046    Name: Matthias Blume
3047    Date: 2001/06/25 12:05:00 EDT
3048    Tag: blume-20010625-nlffigen
3049    Description:
3050    
3051    This update fixes a number of minor bugs in ml-nlffigen as reported by
3052    Nick Carter <nbc@andrew.cmu.edu>.
3053    
3054      1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
3055      2. Default names for generated files are now derived from the name of
3056         the C file *without its directory*.  In particular, this causes generated
3057         files to be placed locally even if the C file is in some system directory.
3058      3. Default names for generated signatures and structures are also derived
3059         from the C file name without its directory.  This avoids silly things
3060         like "structure GL/GL".
3061         (Other silly names are still possible because ml-nlffigen does not do
3062          a thorough check of whether generated names are legal ML identifiers.
3063          When in doubt, use command line arguments to force particular names.)
3064    
3065    ----------------------------------------------------------------------
3066    Name: Matthias Blume
3067    Date: 2001/06/21 12:25:00 EDT
3068    Tag: blume-20010621-eXene
3069    Description:
3070    
3071    eXene now compiles and (sort of) works again.
3072    
3073    The library name (for version > 110.33) is $/eXene.cm.
3074    
3075    I also added an new example in src/eXene/examples/nbody.  See the
3076    README file there for details.
3077    
3078    ----------------------------------------------------------------------
3079    Name: Matthias Blume
3080    Date: 2001/06/20 16:40:00 EDT
3081    Tag: blume-20010620-cml
3082    Description:
3083    
3084    CML now compiles and works again.
3085    
3086    Libraries (for version > 110.33):
3087    
3088      $cml/cml.cm            Main CML library.
3089      $cml/basis.cm          CML's version of $/basis.cm.
3090      $cml/cml-internal.cm   Internal helper library.
3091      $cml/core-cml.cm       Internal helper library.
3092      $cml-lib/trace-cml.cm  Tracing facility.
3093      $cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
3094    
3095    The installer (config/install.sh) has been taught how to properly
3096    install this stuff.
3097    
3098    ----------------------------------------------------------------------
3099    Name: Matthias Blume
3100    Date: 2001/06/19 17:55:00 EDT
3101    Tag: blume-20010619-instantiate
3102    Description:
3103    
3104    This un-breaks the fix for bug 1432.
3105    (The bug was originally fixed in 110.9 but I broke it again some
3106    time after that.)
3107    
3108    ----------------------------------------------------------------------
3109    Name: Matthias Blume
3110    Date: 2001/06/19 17:25:00 EDT
3111    Tag: blume-20010619-signals
3112    Description:
3113    
3114    This should (hopefully) fix the long-standing signal handling bug.
3115    (The runtime system was constructing a continuation record with an
3116    incorrect descriptor which would cause the GC to drop data on the floor...)
3117    
3118    ----------------------------------------------------------------------
3119    Name: Matthias Blume
3120    Date: 2001/06/15 15:05:00 EDT
3121    Tag: blume-20010615-moresparc
3122    Description:
3123    
3124    Here is a short late-hour update related to Sparc c-calls:
3125    
3126     -- made handling of double-word arguments a bit smarter
3127    
3128     -- instruction selection phase tries to collapse certain clumsily
3129        constructed ML-Trees; typical example:
3130    
3131            ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
3132    
3133        This currently has no further impact on SML/NJ since mlriscGen does
3134        not seem to generate such patterns in the first place, and c-calls
3135        (which did generate them in the beginning) has meanwhile been fixed
3136        so as to avoid them as well.
3137    
3138    ----------------------------------------------------------------------
3139    Name: Matthias Blume
3140    Date: 2001/06/15 15:05:00 EDT
3141    Tag: blume-20010615-sparc
3142    Description:
3143    
3144    The purpose of this update is to provide an implementation of NLFFI
3145    on Sparc machines.
3146    
3147    Here are the changes in detail:
3148    
3149       * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
3150       the Sparc implementation of the c-calls API.
3151       * The Sparc backend of SML/NJ has been modified to uniformely use %fp
3152       for accessing the ML frame.  Thus, we have a real frame pointer and
3153       can freely modify %sp without need for an omit-frame-ptr phase.
3154       The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
3155       this case.
3156       * ml-nlffigen has been taught to produce code for different architectures
3157       and calling conventions.
3158       * In a way similar to what was done in the x86 case, the Sparc
3159       backend uses its own specific extension to mltree.  (For example,
3160       it needs to be able to generate UNIMP instructions which are part
3161       of the calling convention.)
3162       * ml-nlffi-lib was reorganized to make it more modular (in particular,
3163       to make it easier to plug in new machine- and os-dependent parts).
3164    
3165    There are some other fairly unrelated bug fixes and cleanups as well:
3166    
3167       * I further hacked the .cm files for MLRISC tools (like MDLGen) so
3168       that they properly share their libraries with existing SML/NJ libraries.
3169       * I fixed a minor cosmetic bug in CM, supressing certain spurious
3170       follow-up error messages.
3171       * Updates to CM/CMB documentation.
3172    
3173    TODO items:
3174    
3175       * MLRISC should use a different register as its asmTemp on the Sparc.
3176         (The current %o2 is a really bad choice because it is part of the
3177         calling conventions, so things might interfere in unexpected ways.)
3178    
3179    ----------------------------------------------------------------------
3180    Name: Matthias Blume
3181    Date: 2001/06/07
3182    Tag: blume-20010607-calls
3183    Description:
3184    
3185    A number of internal changes related to C calls and calling conventions:
3186    
3187    1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
3188       number of bytes popped implicitly (by the callee).  In most cases
3189       this field is 0 but on x86/win32 it is some non-zero value.  This
3190       is information provided for the benefit of the "omit-frameptr" pass.
3191    2. The CALL instruction on the x86 carries a similar "pops" field.
3192       The instruction selection phase copies its value from the ML-Tree
3193       CALL statement.
3194    3. On all other architectures, the instruction selection phase checks
3195       whether "pops=0" and complains if not.
3196    4. The c-calls implementation for x86 now accepts two calling conventions:
3197       "ccall" and "stdcall".  When "ccall" is selected, the caller cleans
3198       up after the call and pops is set to 0.  For "stdcall", the caller
3199       does nothing, leaving the cleanup to the callee; pops is set to
3200       the number of bytes that were pushed onto the stack.
3201    5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
3202       distinguish between "ccall" and "stdcall".
3203    6. The UNIMP instruction has been added to the supported Sparc instruction
3204       set. (This is needed for implementing the official C calling convention
3205       on this architecture.)
3206    7. I fixed some of the .cm files under src/MLRISC/Tools to make them
3207       work with the latest CM.
3208    
3209    ----------------------------------------------------------------------
3210    Name: Matthias Blume
3211    Date: 2001/06/05 15:10:00 EDT
3212    Tag: blume-20010605-cm-index
3213    Description:
3214    
3215    0. The "lambdasplit" parameter for class "sml" in CM has been documented.
3216    
3217    1. CM can now generate "index files".  These are human-readable files
3218       that list on a per-.cm-file basis each toplevel symbol defined or
3219       imported.  The location of the index file for
3220       <p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
3221       To enable index-file generation, set CM.Control.generate_index to true
3222       or export an environment-symbol: export CM_GENERATE_INDEX=true.
3223    
3224       The CM manual has been updated accordingly.
3225    
3226    2. I made some slight modifications to the c-calls API in MLRISC.
3227    
3228         a) There is now a callback to support saving/restoring of
3229            dedicated but caller-save registers around the actual call
3230            instruction.
3231         b) One can optionally specify a comment-annotation for the
3232            call instruction.
3233    
3234    3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
3235       (For example, the comment annotation shows the C prototype of
3236        the function being called.)
3237    
3238    ----------------------------------------------------------------------
3239    Name: Matthias Blume
3240    Date: 2001/06/01 13:30:00 EDT
3241    Tag: blume-20010601-nlffi-cleanup
3242    Description:
3243    
3244    This is mostly a cleanup of MLFFI stuff:
3245    
3246       - some signature files have been put into a more exposed place
3247       - the ugly 'f type parameter is gone (simplifies types tremendously!)
3248       - ml-nlffigen changed accordingly
3249       - tutorial updated
3250    
3251    Other changes:
3252    
3253       - author's affiliation in CM manual(s) updated
3254       - some more recognized keywords added to Allen's sml.sty
3255    
3256    ----------------------------------------------------------------------
3257    Name: Matthias Blume
3258    Date: 2001/05/25 15:30:00 EDT
3259    Tag: blume-20010525-iptr
3260    Description:
3261    
3262      - put the official 110.33-README (as it appears on the ftp server) under
3263        CVS
3264      - fixed a small bug related to incomplete pointer types in
3265        ml-nlffigen
3266      - small cosmetic change to the ml-nlffi-lib's "arr" type constructor
3267        (it does not need the 'f type parameter)
3268    
3269    ----------------------------------------------------------------------
3270    Name: Matthias Blume
3271    Date: 2001/05/23 14:30:00 EDT
3272    Tag: Release_110_33
3273    Description:
3274    
3275    New version number (110.33).  New bootfiles.
3276    
3277    ----------------------------------------------------------------------
3278    Name: Matthias Blume
3279    Date: 2001/05/22 18:06:00 EDT
3280    Tag: blume-20010522-targets
3281    Description:
3282    
3283    Made install.sh use file config/targets.customized if it exists, falling
3284    back to config/targets if it doesn't.  This way one can have a customized
3285    version of the targets file without touching the "real thing", thus
3286    eliminating the constant fear of accidentally checking something bogus
3287    back into the CVS repository...  (File config/targets.customized must
3288    not be added to the repository!)
3289    
3290    ----------------------------------------------------------------------
3291    Name: Matthias Blume
3292    Date: 2001/05/22 16:30:00 EDT
3293    Tag: blume-20010522-minitut
3294    Description:
3295    
3296    1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
3297       struct returns.
3298    2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
3299       incomplete, preliminary documentation for NLFFI.
3300    
3301    ----------------------------------------------------------------------
3302    Name: Matthias Blume
3303    Date: 2001/05/14 11:30:00 EDT
3304    Tag: blume-20010514-script
3305    Description:
3306    
3307    Some bugs in install script fixed.
3308    
3309    In addition to that I also made a slight change to the NLFFI API:
3310    Functors generated by ml-nlffigen now take the dynamic library as a
3311    straight functor argument, not as a suspended one.  (The original
3312    functor code used to force the suspension right away anyway, so there
3313    was nothing gained by this complication of the interface.)
3314    
3315    ----------------------------------------------------------------------
3316    Name: Matthias Blume
3317    Date: 2001/05/11 14:35:00 EDT
3318    Tag: blume-20010511-ml-nlffi
3319    Description:
3320    
3321    I finally took the plunge and added my new FFI code to the main
3322    repository.  For x86-linux it is now ready for prime-time.
3323    
3324    There are two new subdirectories of "src":
3325    
3326      - ml-nlffi-lib:
3327           The utility library for programs using the FFI interface.
3328           Here is the implementation of $/c.cm and its associated low-level
3329           partners $/c-int.cm and $/memory.cm.
3330      - ml-nlffigen:
3331           A stand-alone program for generating ML glue code from C source
3332           code.
3333    
3334    Building ml-nlffigen requires $/ckit-lib.cm.
3335    
3336    The config/install.sh script has been updates to do the Right Thing
3337    (hopefully).
3338    
3339    Notice that the source tree for the C-Kit will not be put under "src"
3340    but directly under the installation root directory.  (This is the
3341    structure that currently exists on the CVS server when you check out
3342    module "sml".)  Fortunately, config/install.sh knows about this oddity.
3343    
3344    Bugs: No documentation yet.
3345    
3346    ----------------------------------------------------------------------
3347    Name: Matthias Blume
3348    Date: 2001/05/09 16:35:00 EDT
3349    Tag: blume-20010509-cpscontract
3350    Description:
3351    
3352    Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
3353    wrapper/unwrapper elimination did not decrement usage counts and some
3354    dead variables got overlooked by the dead-up logic.)
3355    
3356    ----------------------------------------------------------------------
3357    Name: Lal George
3358    Date: 2001/05/08  17:26:09 EDT
3359    Tag: george-20010508-omit-frameptr
3360    Description:
3361    
3362    Changes to implement the omit-frame-pointer optimization to support
3363    raw C calls. For now, there is only support on the Intel x86, but
3364    other architectures will follow as more experience is gained with this.
3365    
3366    
3367    ----------------------------------------------------------------------
3368    Name: Matthias Blume
3369    Date: 2001/05/07 14:40:00 EDT
3370    Tag: blume-20010507-proxies
3371    Description:
3372    
3373    I made into "proxy libraries" all libraries that qualify for such a
3374    change.  (A qualifying library is a library that has another library or
3375    groups as its sole member and repeats that member's export list
3376    verbatim.  A proxy library avoids this repetition by omitting its export
3377    list, effectively inheriting the list that its (only) member exports.
3378    See the CM manual for more explanation.)
3379    The main effect is that explicit export lists for these libraries
3380    do not have to be kepts in sync, making maintenance a bit easier.
3381    
3382    I also added copyright notices to many .cm-files.
3383    
3384    Last but not least, I made a new set of bootfiles.
3385    
3386    ----------------------------------------------------------------------
3387    Name: Matthias Blume
3388    Date: 2001/05/04 17:00:00 EDT
3389    Tag: blume-20010504-cm-lsplit
3390    Description:
3391    
3392    0. John merged pending changes to $/smlnj-lib.cm
3393    
3394    1. Allen's previous change accidentally backed out of one of Lal's
3395       earlier changes.  I undid this mistake (re-introducing Lal's change).
3396    
3397    2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
3398       within the compiler where applicable.  There is some code simplification
3399       because of that.
3400    
3401    3. The "split" phase (in FLINT) is now part of the default list of phases.
3402       Compiler.Control.LambdaSplitting.* can be used to globally control the
3403       lambda-splitting (cross-module-inlining) engine.  In addition to that,
3404       it can now also be controlled on a per-source basis: CM has been taught
3405       a new tool parameter applicable to ML source files.
3406    
3407       - To turn lambda-splitting off completely:
3408            local open Compiler.Control.LambdaSplitting in
3409                val _ = set Off
3410            end
3411       - To make "no lambda-splitting" the global default (but allow per-source
3412         overriding); this is the initial setting:
3413            local open Compiler.Control.LambdaSplitting in
3414                val _ = set (Default NONE)
3415            end
3416       - To make "lambda-splitting with aggressiveness a" the global default
3417         (and allow per-source overriding):
3418            local open Compiler.Control.LambdaSplitting in
3419                val _ = set (Default (SOME a))
3420            end
3421    
3422       - To turn lambda-splitting off for a given ML souce file (say: a.sml)
3423         write (in the respective .cm-file):
3424            a.sml (lambdasplitting:off)
3425       - To turn lambda-splitting for a.sml on with minimal aggressiveness:
3426            a.sml (lambdasplitting:on)
3427       - To turn lambda-splitting for a.sml on with aggressiveness <a> (where
3428         <a> is a decimal non-negative integer):
3429            a.sml (lambdasplitting:<a>)
3430       - To turn lambda-splitting for a.sml on with maximal aggressiveness:
3431            a.sml (lambdasplitting:infinity)
3432       - To use the global default for a.sml:
3433            a.sml (lambdasplitting:default)
3434         or simply
3435            a.sml
3436    
3437    ----------------------------------------------------------------------
3438    Name: Allen Leung
3439    Date: 2001/05/04 01:57:00 EDT
3440    Tag: leunga-20010504-sync
3441    Description:
3442    
3443      MLRISC features.
3444    
3445      1. Fix to CMPXCHG instructions.
3446      2. Changed RA interface to allow annotations in callbacks.
3447      3. Added a new method to the stream interface to allow annotations updates.
3448    
3449    ----------------------------------------------------------------------
3450    Name: Matthias Blume
3451    Date: 2001/05/01 11:45:00 EDT
3452    Tag: blume-20010501-pcedittmp
3453    Description:
3454    
3455    Changed install.sh to use the current working directory instead of
3456    /usr/tmp for a temporary file (pcedittmp).  The previous choice
3457    of /usr/tmp caused trouble with MacOS X because of file premission
3458    problems.
3459    
3460    ----------------------------------------------------------------------
3461    Name: Matthias Blume
3462    Date: 2001/04/20 11:10:00 EDT
3463    Tag: blume-20010420-inMLflag
3464    Description:
3465    
3466     - added vp_limitPtrMask to vproc-state.h
3467       (for use by the raw-C-calls mechanism to implement proper interrupt
3468        handling)
3469     - made the ML compiler aware of various data-structure offsets so it
3470       can generate code for accessing the vp_inML flag and vp_limitPtrMask
3471     - tweaked mlriscGen.sml to have it emit interrupt-handling code for
3472       raw C-calls
3473    
3474    ----------------------------------------------------------------------
3475    Name: Lal George
3476    Date: 2001/04/20 09:15:28 EDT
3477    Tag: george-20010420-macosX
3478    Description:
3479    
3480     - Changes to port to Mac OS X; Darwin.
3481    
3482     - In the process I found that sqrt was broken on the PPC, because the
3483       fsqrt instruction is not implemented.
3484    
3485    ----------------------------------------------------------------------
3486    Name: Matthias Blume
3487    Date: 2001/04/18 12:45:00 EDT
3488    Tag: blume-20010418-ccalls
3489    Description:
3490    
3491     - fixed two off-by-4 errors in the x86-specific c-calls implementation
3492       (this bug prevented structure arguments containing pointers from being
3493        passed correctly)
3494     - changed the raw-C-call code in mlriscGen.sml in such a way that
3495       structure arguments are represented as a pointer to the beginning
3496       of the structure (instead of having a series of synthesized arguments,
3497       one for each structure member)
3498    
3499     - made makeml script's verbosity level configurable via environment
3500       variable (MAKEML_VERBOSITY)
3501    
3502     - eliminated placeholder implementations for f32l, w16s, i16s, and f32s
3503       in rawmem-x86.sml; we are now using the real thing
3504    
3505    ----------------------------------------------------------------------
3506    Name: Matthias Blume
3507    Date: 2001/03/22 16:25:00 EST
3508    Tag: blume-20010322-bootfiles
3509    Description:
3510    
3511    Created a new set of bootfiles (for your automatic installation convenience).
3512    
3513    ----------------------------------------------------------------------
3514    Name: Matthias Blume
3515    Date: 2001/03/22 15:10:00 EST
3516    Tag: blume-20010322-rawmem-parcm
3517    Description:
3518    
3519    1. All "raw memory access" primitives for the new FFI are implemented now
3520       (at least on the x86).
3521    2. Some further cleanup of CM's parallel make mechanism.
3522    
3523    ----------------------------------------------------------------------
3524    Name: Matthias Blume
3525    Date: 2001/03/19 17:53:00 EST
3526    Tag: blume-20010319-parallel
3527    Description:
3528    
3529    Parallel make (using compile servers) now works again.
3530    
3531    To this end, CM.stabilize and CMB.make have been modified to work in
3532    two passes when compile servers are attached:
3533       1. Compile everything, do not perform stabilization; this pass
3534          uses compile servers
3535       2. Stabilize everything; this pass does not use compile servers
3536    If there are no compile servers, the two passes are combined into one
3537    (as before).  Splitting the passes increases the inherent parallelism
3538    in the dependency graph because the entire graph including all
3539    libraries is available at the same time.  This, in turn, improves
3540    server utilization.  The downside is that the master process will
3541    have to do some extra work after compilation is done (because for
3542    technical reasons it must re-read all the binfiles during stabilization).
3543    
3544    ----------------------------------------------------------------------
3545    Name: Matthias Blume
3546    Date: 2001/03/16 12:22:00 EST
3547    Tag: blume-20010316-bootfiles
3548    Description:
3549    
3550    Created a new set of bootfiles (for your automatic installation convenience).
3551    
3552    ----------------------------------------------------------------------
3553    Name: Matthias Blume
3554    Date: 2001/03/16 11:00:00 EST
3555    Tag: blume-20010316-MLTREE-fixup
3556    Description:
3557    
3558    This is a minor fixup for an (untagged) earlier commit by Allen.
3559    (A file was missing).
3560    
3561    ----------------------------------------------------------------------
3562    Name: Allen Leung
3563    Date: Mon Mar  5 18:54:57 EST 2001
3564    Tag: leunga-20010305-cut-support
3565    
3566    1. New support for alternative control-flow in MLTREE.
3567       Currently we support
3568    
3569          FLOW_TO(CALL ...., [k1,...,kn])
3570    
3571       This is needed for 'cuts to' in C-- and try/handle-like constructs
3572       in Moby
3573    
3574       New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
3575    
3576    2. Register Allocator
3577    
3578       Changes in interface [from Fermin, John]
3579    
3580    3. Alpha 8-bit SLL support [Fermin]
3581    
3582    4. All architectures
3583    
3584       A new module (ClusterExpandCopies) for expanding parallel copies.
3585    
3586    ----------------------------------------------------------------------
3587    Name: Allen Leung
3588    Date: 2001/02/27 23:07:00 EST
3589    Tag: leunga-20010227-minor-stuff
3590    
3591    1. Alpha bug fix for CMOVNE
3592    2. Handle mltree COND(..,FCMP ...,...)
3593    3. Bug fix in simplifier
3594    
3595    ----------------------------------------------------------------------
3596    Name: Matthias Blume
3597    Date: 2001/01/30 17:50:00 EST
3598    Tag: blume-20010130-sync
3599    Description:
3600    
3601    This is just a minor update to sync my devel branch with the main brach.
3602    The only visible change is the addition of some README files.
3603    
3604    ----------------------------------------------------------------------
3605    Name: Matthias Blume
3606    Date: 2001/01/12 23:30:00 JST
3607    Tag: blume-20010112-bootfiles
3608    Description:
3609    
3610    Made a new set