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 1085, Fri Feb 22 00:15:55 2002 UTC revision 1381, Wed Sep 17 21:28:41 2003 UTC
# Line 13  Line 13 
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name:  Matthias Blume (blume (at) tti - c (dot) org)
17    Date: 2003/09/17 16:30:00 CDT
18    Tag: blume-20030917-real32-slices
19    Description:
20    
21    Added an entry to the primitive environment
22    (compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
23    added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
24    
25    Removed some of the "magic" constants in real64.sml and replaced them
26    with code that generates these values from their corresponding
27    integer counterparts.
28    
29    Made all(?) the slice-related changes to the Basis and made everything
30    compile again...
31    
32    ----------------------------------------------------------------------
33    Name: Matthias Blume (blume (at) tti - c (dot) org)
34    Date: 2003/09/15 17:45:00 CDT
35    Tag: blume-20030915-rbase
36    Description:
37    
38    Fixed bug in Real.fromLargeInt.
39    
40    ----------------------------------------------------------------------
41    Name: Matthias Blume (blume (at) tti - c (dot) org)
42    Date: 2003/09/13 18:11:00 CDT
43    Tag: blume-20030913-libinstall
44    Description:
45    
46    Minor bugfix in config/libinstall (set anchor with path to
47    standalone tool after installing it, otherwise libraries that
48    need ml-lex or ml-yacc won't compile the first time the installer
49    runs).
50    
51    ----------------------------------------------------------------------
52    Name: Matthias Blume (blume (at) tti - c (dot) org)
53    Date: 2003/09/12 11:45:00 CDT
54    Tag: blume-20030912-various
55    Description:
56    
57     - fixed bug in Real.toLargeInt
58     - fixed bug in Posix.ProcEnv.times
59     - changed inputLine functions to return an option
60     - minor installer improvements / bugfixes
61     - changed default @SMLalloc parameter for x86/celeron to 64k
62    
63    ----------------------------------------------------------------------
64    Name: Matthias Blume (blume (at) tti - c (dot) org)
65    Date: 2003/09/09 22:00:00 CDT
66    Tag: Release_110_43
67    Description:
68    
69        New working release 110.43.  New bootfiles.
70    
71    ----------------------------------------------------------------------
72    Name: Matthias Blume (blume (at) tti - c (dot) org)
73    Date: 2003/09/09 19:20:00 CDT
74    Tag: blume-20030909-installer
75    Description:
76    
77    Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
78    Modified config/install.bat to take advantage of it.  Also modified
79    config/install.sh (and called it config/new-install.sh) to take advantage
80    of it on Unix systems.  (The SML code is (supposed to be) platform-
81    independent.)
82    
83    The installer can now install everything under Win32
84    as well as under *nix as long as it compiles.
85    
86    Other changes:
87    
88      - made CML compile again under Win32
89      - made eXene compile under Win32 (by providing a fake structure UnixSock
90        and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
91      - fixed a bug in nowhere: it assumed that type OS.Process.status is the
92        same as type int;  under Win32 it isn't
93      - fixed some slice-related problems in the win32-specific parts of CML
94      - added a functor argument "sameVol" to os-path-fn.sml in the Basis
95        (under Win32, the volume name is case-insensitive, and the
96        OS.Path code compares volume names for equality)
97    
98    ----------------------------------------------------------------------
99    Name: Matthias Blume (blume (at) tti - c (dot) org)
100    Date: 2003/09/08 11:55:00 CDT
101    Tag: blume-20030908-fullpath
102    Description:
103    
104    Made Win32 version of OS.FileSys.fullPath return current directory
105    when given an empty string.  This is what the spec says, and incidentally,
106    CM depends on it.  (CM otherwise goes into an infinite loop in certain
107    cases when presented with the name of a non-existing .cm file.)
108    
109    ----------------------------------------------------------------------
110    Name: Matthias Blume (blume (at) tti - c (dot) org)
111    Date: 2003/09/04 16:30:00 CDT
112    Tag: blume-20030905-slices-etc
113    Description:
114    
115      1. Changed interface to vectors and arrays in Basis to match
116         (draft) Basis spec.
117      2. Added signatures and implementations of slices according to
118         Basis spec.
119      3. Edited source code throughout the system to make it compile again
120         under 1. and 2.  (In some cases code had to be added to have it
121         match the new signatures.)
122      4. MLRISC should be backward-compatible: the copies of the originals
123         of files that needed to change under 3. were retained, the .cm files
124         check the compiler version number and use old versions when
125         appropriate.
126      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
127         dirstream -> string option (in accordance with Basis spec).
128      6. When generating code that counts lines, ml-lex used function
129         CharVector.foldli, taking advantage of its old interface.
130         This has been replaced with the corresponding code from
131         CharVectorSlice.  (html-lex must be re-lexed!)
132      7. BitArray in smlnj-lib/Util has been extended/modified to match the
133         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
134      8. Removed temporary additions (fromInternal, toInternal) from the
135         (now obsolete) IntInf in smlnj-lib/Util.
136      9. Cleaned up structure Byte.
137     10. Added localOffset, scan, and fromString to Date (according to spec).
138         Cleaned/corrected implementation of Date.
139         (Still need to check for correctness; implement better canonicalizeDate.)
140     11. Added "scan" to signature IEEE_REAL.
141     12. Some improvements to IntInf [in particular: efficiency-hack for
142         mod and rem when second operand is 2 (for parity checks).]
143     13. Changed representation of type Time.time, using a single IntInf.int
144         value counting microseconds.  This considerably simplified the
145         implementation of structure Time.  We now support negative time
146         values; scan and fromString handle signs.
147     14. Functor PrimIO now takes two additional arguments (VectorSlice and
148         ArraySlice).
149    
150    ----------------------------------------------------------------------
151    Name: Matthias Blume (blume (at) tti - c (dot) org)
152    Date: 2003/08/28 17:00:00 CDT
153    Tag: blume-20030828-intinf
154    Description:
155    
156    This is a major update which comes with a version number bump
157    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
158    and an implementation of IntInf in the Basis.
159    
160    There are a fairly large number of related changes and updates throughout
161    the system:
162    
163    Basis:
164            - Implemented IntInf.
165            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
166            - Added some missing Real64 operations, most notably Real.toLargeInt.
167            - Added FixedInt as a synonym for Int32.
168    
169    compiler:
170            * Added support for a built-in intinf type.
171              - literals
172              - pattern matching
173              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
174              - overloading on literals and operations
175    
176            This required adding a primitive type intinf, some additional
177            primops, and implementations for several non-trivial intinf
178            operations in Core.  (The intinf type is completely abstract
179            to the compiler; all operations get delegated back to the Core.)
180    
181            * Intinf equality is handled by polyequal.  However, the compiler
182            does not print its usual warning in this case (since polyequal
183            is the right thing to do there).
184    
185            * Improved the organization of structure InlineT.
186    
187            * A word about conversion primops:
188            If conversions involving intinf do not cancel out during
189            CPS contract, then the compiler must insert calls to Core functions.
190            Since all core access must be resolved already during the FLINT
191            translate phase, it would be too late a the time of CPS contract
192            to add new Core calls.  For this reason, conversion primops
193            for intinf carry two arguments:  1. the numeric argument that
194            they are supposed to convert, and 2. the Core function that
195            can help with this conversion if necessary.  If CPS contract
196            eliminates a primop, then the associated Core function becomes
197            dead and goes away.  Intinf conversion primops that do not get
198            eliminated by CPS contract get rewritten into calls of their
199            core functions by a separate, new phase.
200    
201    interactive system:
202            - Control.Print.intinfDepth controls max length of intinf constants
203            being printed. (Analogous to Control.Print.stringDepth.)
204            - Cleanup in printutil and pputil: got rid of unused stuff and
205            duplicates; replaced some of the code with code that makes better
206            use of library functionality.
207    
208    CM:
209            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
210            are no longer silent.
211    
212    CKIT:
213            Fixed mismatched uses of Int32 and LargeInt.  I always decided
214            in favor of LargeInt -- which is now the same as IntInf.
215            CKIT-knowledgable people should check whether this is what's
216            intended and otherwise change things back to using Int32 or
217            FixedInt.
218    
219    Throughout the code:
220            Started using IntInf.int literals and built-in operations
221            (e.g., comparison with 0) where this seems appropriate.
222    
223    
224    ----------------------------------------------------------------------
225    Name: Dave MacQueen (dbm@cs.uchicago.edu)
226    Date: 2003/08/13 11:36:00 CDT
227    Tag: dbm-20030813-mcz-merge1
228    Description:
229    
230    Merging changes from the mcz-branch development branch into trunk.
231    These changes involve replacement of the emulated old prettyprinter
232    interface with direct use of the SML/NJ Lib PP library, and fixing
233    of a couple of bugs (895, 1186) relating to error messages.  A new
234    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
235    has been added.
236    
237    ----------------------------------------------------------------------
238    Name: Matthias Blume (blume (at) tti - c (dot) org)
239    Date: 2003/08/11 15:45:00 CDT
240    Tag: blume-20030811-windows
241    Description:
242    
243    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
244    
245        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
246    
247    This patch restores SML/NJ's ability to run under win32.  There are a
248    number of changes, including fixes for several bugs that had gone
249    unnoticed until now:
250    
251      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
252        trying to run the win32 version from within cygwin.)
253    
254      - There are a number of simple .bat scripts that substitute for their
255        corresponding Unix shell-scripts.  (See below.)
256    
257      - The internals of ml-build have been modified slightly.  The main
258        difference is that instead of calling ".link-sml" (or link-sml.bat)
259        using OS.Process.system, the ML process delegates this task back
260        to the script.  Otherwise problems arise in mixed environments such
261        as Cygwin where scripts look and work like Unix scripts, but
262        where OS.Process.system cannot run them.
263    
264      - In CM, the srcpath pickler used native pathname syntax -- which
265        is incorrect in the case of cross-compilation.  The new pickle format
266        is independent of platform-specific naming conventions.
267    
268      - Path configuration files (such as lib/pathconfig) can now choose
269        between native and standard syntax.  Placing a line of the form
270    
271            standard!
272    
273        into the file causes all subsequent paths to be interpreted using
274        CM standard pathname syntax (= Unix conventions); a line
275    
276            native!
277    
278        switches back to native style.  This was needed so that
279        path config files can be written portably, see src/system/pathconfig.
280    
281      - Runtime system:
282    
283          - win32-filesys.c:  get_file_time and set_file_time now
284            access modification time, not creation time.
285    
286          - I/O code made aware of new array representation.
287    
288          - Bug fixes in X86.prim.masm.
289    
290       - src/system/makeml made aware of win32. (For use under cygwin
291            and other Unix-environments for windows.)
292    
293       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
294         which caused BinIO.inputAll to fail consistently.
295    
296    .bat scripts:
297    
298       Windows .bat scripts assume that SMLNJ_HOME is defined.
299    
300       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
301         applications (sml, ml-yacc, ml-lex).
302       - ml-build.bat: analogous to ml-build.
303       - config\install.bat:  Analogous to config/install.sh.  This requires
304         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
305         (nmake etc. must be on the path, and vcvars32 must have been run.)
306         Moreover, sources for ml-lex and ml-yacc need to exist under src,
307         and the bootfile hierarchy must have been unpacked under
308         sml.boot.x86-win32.
309         The script is very primitive and does a poor job at error checking.
310         It only installs the base system, ml-lex, and ml-yacc.  No other
311         libraries are being installed (i.e., you get only those that
312         are part of the compiler.)
313       - link-sml.bat:  analogous to .link-sml, but not currently used
314    
315    Unrelated bug fixes:
316    
317      - ml-nlffigen now exports structures ST_* corresponding to incomplete
318        types.
319      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
320        otherwise.)
321    
322    ----------------------------------------------------------------------
323    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
324    Date: 2003/06/17
325    Tag: macqueen-20030617-bug895
326    Description:
327    
328    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
329    Tag will be used for new development branch (mcz-branch) for
330    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
331    
332    ----------------------------------------------------------------------
333    Name: Matthias Blume (blume (at) tti - c (dot) org)
334    Date: 2003/05/27 16:55:00 CDT
335    Tag: blume-20030527-polyeq
336    Description:
337    
338    Tried to eliminated most cases of polymorphic equality.
339    
340    ----------------------------------------------------------------------
341    Name: Matthias Blume (blume (at) tti - c (dot) org)
342    Date: 2003/05/21 17:45:00 CDT
343    Tag: blume-20030517-complete
344    Description:
345    
346    Two changes:
347    
348      1. Added a flag for controlling whether non-exhaustive bindings will
349         be treated as errors (default is false).
350      2. Cleaned up the *entire* source tree so that CMB.make goes through
351         without a single non-exhaustive match- or bind warning.
352    
353    ----------------------------------------------------------------------
354    Name: Matthias Blume (blume (at) tti - c (dot) org)
355    Date: 2003/05/17 10:20:00 CDT
356    Tag: blume-20030517-absyn
357    Description:
358    
359    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
360    
361       This mainly affects the quality of error messages.  However, some
362       of the code is now more straightforward than before. (Treatment of
363       the above four constructs in translate.sml is much simpler than
364       the "macro-expansion" that was going on before.  Plus, the mach-
365       compiler no longer gets invoked just to be able to compile an
366       if-expression.)
367    
368    2. The ErrorMsg.Error exception is now caught and absorbed by the
369       interactive loop.
370    
371    ----------------------------------------------------------------------
372    Name: Allen Leung
373    Date: 2003/05/16 13:05:00 CDT
374    Tag: leunga-20030516-cygwin-runtime
375    Description:
376    
377        Ported the runtime system to cygwin, which uses the unix
378    x86-unix bin files.  Missing/buggy features:
379    
380       o getnetbyname, getnetbyaddr: these functions seem to be missing in
381         the Cygwin library.
382       o Ctrl-C handling may be flaky.
383       o Windows system calls and Windows I/O are not supported.
384    
385       A new set of binfiles is located at:
386    
387         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
388    
389       This is only needed for bootstrapping the cygwin version of smlnj.
390       Other x86 versions can use the existing binfiles.
391    
392    ----------------------------------------------------------------------
393    Name: Matthias Blume
394    Date: 2003/04/08 15:42:00 CDT
395    Tag: blume-20030408-listpair
396    Description:
397    
398    1. Added a target 'mlrisc' to installer.
399    
400    2. Added missing elements to structure ListPair.
401    
402    ----------------------------------------------------------------------
403    Name: Allen Leung
404    Date: 2003/01/07 10:40:00 EST
405    Tag: leunga-20030107-int-rem
406    Description:
407    
408    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
409    The arguments to the SUBL instruction were swapped.
410    
411    ----------------------------------------------------------------------
412    Name: Matthias Blume
413    Date: 2002/12/12 16:25:00 EST
414    Tag: blume-20021212-risc-ra
415    Description:
416    
417    Fixed a serious bug in the rewrite code for FP spilling/reloading that
418    sent the RA into an infinite loop when floating point registers get
419    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
420    110.37 and 110.38.)
421    There was another set of potential problems related to the handling of
422    MLRISC annotations (but those did not yet cause real problems, apparently).
423    
424    ----------------------------------------------------------------------
425    Name: Matthias Blume
426    Date: 2002/12/06 22:40:00 EST
427    Tag: blume-20021206-cm-fileid
428    Description:
429    
430    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
431    This fixes the problem of CM getting confused by files that suddenly
432    change their identity (e.g., by getting unlinked and recreated by some
433    text editor such as vi).  There might be a better/cheaper/cleaner way
434    of doing this, but for now this will have to do.
435    
436    ----------------------------------------------------------------------
437    Name: Matthias Blume
438    Date: 2002/10/28 09:50:00 EST
439    Tag: blume-20021028-typecheck
440    Description:
441    
442    Exported structure Typecheck from $smlnj/viscomp/core.cm.
443    
444    ----------------------------------------------------------------------
445    Name: Matthias Blume
446    Date: 2002/10/17 09:10:00 EDT
447    Tag: Release_110_42
448    Description:
449    
450    In good old tradition, there has been a slight hiccup so that we have
451    to patch 110.42 after the fact.   The old release tag has been replaced
452    (see below).
453    
454    The change solves a problem with two competing approaches the
455    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
456    each other's way.
457    
458    This change only affects the runtime system code and the installer script.
459    (No new bootfiles.)
460    
461    ----------------------------------------------------------------------
462    Name: Matthias Blume
463    Date: 2002/10/16 12:00:00 EDT
464    Tag: Release_110_42_removed
465    Description:
466    
467    New working release.  New bootfiles.
468    
469    ----------------------------------------------------------------------
470    Name: Matthias Blume
471    Date: 2002/10/10 13:10:00 EDT
472    Tag: blume-20021010-ppc-divs
473    Description:
474    
475    The mltree operator DIVS must be implemented with an overflow check on
476    the PPC because the hardware indicates divide-by-zero using "overflow" as
477    well.
478    
479    ----------------------------------------------------------------------
480    Name: Matthias Blume
481    Date: 2002/07/23 11:45:00
482    Tag: blume-20020723-smlnj-home
483    Description:
484    
485    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
486    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
487    CM_PATHCONFIG is also set), the path configuration file is assumed
488    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
489    the entire tree to some other place and everything will "just work".
490    
491    (Companion commands such as ml-build and ml-makedepend also sense this
492    variable.)
493    
494    ----------------------------------------------------------------------
495    Name: Matthias Blume
496    Date: 2002/07/12 21:19:00 EDT
497    Tag: blume-20020712-liveness
498    Description:
499    
500    Exported two useful "step" functions from liveness module (MLRISC).
501    
502    ----------------------------------------------------------------------
503    Name: Matthias Blume
504    Date: 2002/07/05 16:00 EDT
505    Tag: Release_110_41
506    Description:
507    
508    New working release.  New bootfiles.
509    
510    ----------------------------------------------------------------------
511    Name: Matthias Blume
512    Date: 2002/07/05 10:25:00 EDT
513    Tag: blume-20020705-btimp
514    Description:
515    
516    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
517    other clients can set up backtracing support.
518    
519    ----------------------------------------------------------------------
520    Name: Matthias Blume
521    Date: 2002/06/25 17:23:00 EDT
522    Tag: blume-20020625-fpmax
523    Description:
524    
525    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
526    case.  (The sense of the isNaN test was reversed -- which made min and
527    max always return their first argument.)
528    
529    ----------------------------------------------------------------------
530    Name: Matthias Blume
531    Date: 2002/06/11
532    Tag: blume-20020611-unixpath
533    Description:
534    
535    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
536    
537    ----------------------------------------------------------------------
538    Name: Matthias Blume
539    Date: 2002/06/10 16:35:00 EDT
540    Tag: blume-20020610-ieeereal
541    Description:
542    
543    I back-ported my implementation of IEEEReal.fromString from the
544    idlbasis-devel branch so that we can test it.
545    
546    Another small change is that ppDec tries to give more information
547    than just "<sig>" in the case of functors.  However, this code is
548    broken in some mysterious way if the functor's body's signature
549    has not been declared by ascription but gets inferred from the
550    implementation.  This needs fixing...
551    
552    ----------------------------------------------------------------------
553    Name: Matthias Blume
554    Date: 2002/05/31
555    Tag: blume-20020531-btrace-mode
556    Description:
557    
558    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
559    the wayside when I switched over to using Controls everywhere.)
560    
561    ----------------------------------------------------------------------
562    Name: Lal George
563    Date: 2002/05/23 12:21:40 EDT
564    Tag: george-20020523-visual-labels
565    Description:
566    
567    Labels are now displayed in the graphical output to make
568    the fall-through and target blocks obvious.
569    
570    ----------------------------------------------------------------------
571    Name: Matthias Blume
572    Date: 2002/05/22 11:03:00 EDT
573    Tag: blume-20020522-shrink
574    Description:
575    
576    John tweaked yesterday's fix for 1131 to handle an out-of-memory
577    situation that comes up when allocating huge arrays.
578    
579    ----------------------------------------------------------------------
580    Name: Matthias Blume
581    Date: 2002/05/21 16:00:00 EDT
582    Tag: Release_110_40
583    Description:
584    
585    New working release (110.40).  New bootfiles.
586    
587    [Also: John Reppy fixed GC bug 1131.]
588    
589    ----------------------------------------------------------------------
590    Name: Matthias Blume
591    Date: 2002/05/21 12:35:00 EDT
592    Tag: blume-20020521-cmdoc
593    Description:
594    
595    CM documentation update.
596    
597    ----------------------------------------------------------------------
598    Name: Matthias Blume
599    Date: 2002/05/21 10:55:00 EDT
600    Tag: blume-20020521-misc
601    Description:
602    
603    - John tweaked runtime to be silent on heap export (except when
604      GC messages are on).
605    - I added a few more things (cross-compiling versions of CMB) to
606      config/preloads (as suggestions).
607    
608    ----------------------------------------------------------------------
609    Name: Matthias Blume
610    Date: 2002/05/20 22:25:00 EDT
611    Tag: blume-20020520-controls
612    Description:
613    
614    - Added ControlUtil structure to control-lib.cm.
615    - Use it throughout.
616    - Used Controls facility to define MLRISC controls (as opposed to
617      registering MLRISC control ref cells with Controls after the fact)
618    - Fixed messed-up controls priorities.
619    
620    * Removed again all the stuff from config/preloads that one wouldn't
621      be able to preload at the time the initial heap image is built.
622      (Many libraries, e.g., CML, do not exist yet at this time.  The
623       only libraries that can be preloaded via config/preloads are those
624       that come bundled with the bootfiles.)
625    
626    ----------------------------------------------------------------------
627    Name: Matthias Blume
628    Date: 2002/05/20 10:59:00 EDT
629    Tag: blume-20020520-preloads
630    Description:
631    
632    Added a lot of commented-out suggestions for things to be included
633    in config/preloads.
634    
635    ----------------------------------------------------------------------
636    Name: Allen Leung
637    Date: 2002/05/18 14:20:00 EDT
638    Tag: leunga-20020518-mdl
639    Description:
640    
641    o Made the mdl tool stuff compile and run again.
642    o I've disabled all the stuff that depends on RTL specifications; they
643      are all badly broken anyway.
644    
645    ----------------------------------------------------------------------
646    Name: Matthias Blume
647    Date: 2002/05/17 16:49:00 EDT
648    Tag: blume-20020517-controls
649    Description:
650    
651    0. John Reppy made several modifications to the SML/NJ library.
652       In particular, there is a shiny new controls-lib.cm.
653    
654    1. Pushed new controls interface through compiler so that everything
655       compiles again.
656    
657    2. Added FormatComb and FORMAT_COMB to the CML version of the
658       SML/NJ library (so that CML compiles again).
659    
660    3. Modified init scripts because XXX_DEFAULT environment variables
661       are no longer with us.  (Boot-time initialization is now done
662       using the same environment variables that are also used for
663       startup-time initialization of controls.)
664    
665    ----------------------------------------------------------------------
666    Name: Lal George
667    Date: 2002/05/15 09:20:10 EDT
668    Tag: george-20020515-pseudo-op-decls
669    Description:
670    
671            All pseudo-ops emitted before the first segment declaration
672    such as TEXT, DATA, and BSS directives are assumed to be global
673    declarations and are emitted first in the assembly file. This is
674    useful in a number of situations where one has pseudo-ops that are not
675    specific to any segment, and also works around the constraint that one
676    cannot have client pseudo-ops in the TEXT segment.
677    
678            Because no segment is associated with these declarations it is
679    an error to allocate any space or objects before the first segment
680    directive and an exception will be raised. However, we cannot make
681    this check for client pseudo-ops.
682    
683            These top level declarations are a field in the CFG graph_info.
684    In theory you can continue to add to this field after the CFG has been
685    built -- provided you know what you are doing;-)
686    
687    ----------------------------------------------------------------------
688    Name: Matthias Blume
689    Date: 2002/05/13 16:40:00 EDT
690    Tag: blume-20020513-pp-etc
691    Description:
692    
693    A few minor bugfixes:
694    
695      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
696        (Bogus printouts for redefined bindings still occur.  Compiler
697        bug should no longer occur now.  We need to redo the prettyprinter
698        from scratch.)
699    
700      - CM pathname printer now also adds escape sequences for ( and )
701    
702      - commend and docu fixes for ml-nlffi
703    
704    ----------------------------------------------------------------------
705    Name: Matthias Blume
706    Date: 2002/05/10 16:40:00 EDT
707    Tag: blume-20020510-erg-textio
708    Description:
709    
710    Applied the following bugfix provided by Emden Gansner:
711    
712        Output is corrupted when outputSubstr is used rather than output.
713    
714        The problem occurs when a substring
715    
716            ss = (s, dataStart, dataLen)
717    
718        where dataStart > 0, fills a stream buffer with avail bytes left.
719        avail bytes of s, starting at index dataStart, are copied into the
720        buffer, the buffer is flushed, and then the remaining dataLen-avail
721        bytes of ss are copied into the beginning of the buffer. Instead of
722        starting this copy at index dataStart+avail in s, the current code
723        starts the copy at index avail.
724    
725        Fix:
726        In text-io-fn.sml, change line 695 from
727             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
728        to
729             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
730    
731    ----------------------------------------------------------------------
732    Name: Matthias Blume
733    Date: 2002/04/12 13:55:00 EDT
734    Tag: blume-20020412-assyntax
735    Description:
736    
737    1. Grabbed newer assyntax.h from the XFree86 project.
738    2. Fiddled with how to compile X86.prim.asm without warnings.
739    3. (Very) Minor cleanup in CM.
740    
741    ----------------------------------------------------------------------
742    Name: Matthias Blume
743    Date: 2002/04/01 (no joke!) 17:07:00 EST
744    Tag: blume-20020401-x86div
745    Description:
746    
747    Added full support for div/mod/rem/quot on the x86, using the machine
748    instruction's two results (without clumsily recomputing the remainder)
749    directly where appropriate.
750    
751    Some more extensive power-of-two support was added to the x86 instruction
752    selector (avoiding expensive divs, mods, and muls where they can be
753    replaced with cheaper shifts and masks).  However, this sort of thing
754    ought to be done earlier, e.g., within the CPS optimizer so that
755    all architectures benefit from it.
756    
757    The compiler compiles to a fixed point, but changes might be somewhat
758    fragile nevertheless.  Please, report any strange things that you might
759    see wrt. div/mod/quot/rem...
760    
761    ----------------------------------------------------------------------
762    Name: Matthias Blume
763    Date: 2002/03/29 17:22:00
764    Tag: blume-20020329-div
765    Description:
766    
767    Fixed my broken div/mod logic.  Unfortunately, this means that the
768    inline code for div/mod now has one more comparison than before.
769    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
770    The problem was with quotient = 0, because that alone does not tell
771    us which way the rounding went.  One then has to look at whether
772    remainder and divisor have the same sign...  :(
773    
774    Anyway, I replaced the bootfiles with fresh ones...
775    
776    ----------------------------------------------------------------------
777    Name: Matthias Blume
778    Date: 2002/03/29 14:10:00 EST
779    Tag: blume-20020329-inlprims
780    Description:
781    
782    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
783    
784    Primops have changed. This means that the bin/boot-file formats have
785    changed as well.
786    
787    To make sure that there is no confusion, I made a new version.
788    
789    
790    CHANGES:
791    
792    * removed REMT from mltree (remainder should never overflow).
793    
794    * added primops to deal with divisions of all flavors to the frontend
795    
796    * handled these primops all the way through so they map to their respective
797      MLRISC support
798    
799    * used these primops in the implementation of Int, Int32, Word, Word32
800    
801    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
802    
803    * parameterized INLMIN, INLMAX, and INLABS by a numkind
804    
805    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
806      floating point
807    
808    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
809      and Real (but Real.abs maps to a separate floating-point-only primop)
810    
811    
812    TODO items:
813    
814    * Hacked Alpha32 instruction selection, disabling the selection of REMx
815      instructions because the machine instruction encoder cannot handle
816      them.  (Hppa, PPC, and Sparc instruction selection did not handle
817      REM in the first place, and REM is supported by the x86 machine coder.)
818    
819    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
820      selection phase.  (The two can be streamlined because the hardware
821      delivers both quotient and remainder at the same time anyway.)
822    
823    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
824      (Currently the behavior is inconsistent both across architectures and
825      wrt. the draft Basis spec.)
826    
827    * Word8 should eventually be handled natively, too.
828    
829    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
830      as if there currently is no execution path that could trigger it in
831      SML/NJ.  (The assumptions underlying functions arith and promotable do not
832      hold for things like multiplication and division.)
833    
834    ----------------------------------------------------------------------
835    Name: Matthias Blume
836    Date: 2002/03/27 16:27:00 EST
837    Tag: blume-20020327-mlrisc-divisions
838    Description:
839    
840    Added support for all four division operations (ML's div, mod, quot,
841    and rem) to MLRISC.  In the course of doing so, I also rationalized
842    the naming (no more annoying switch-around of DIV and QUOT), by
843    parameterizing the operation by div_rounding_mode (which can be either
844    DIV_TO_ZERO or DIV_TO_NEGINF).
845    
846    The generic MLTreeGen functor takes care of compiling all four
847    operations down to only round-to-zero div.
848    
849    Missing pieces:
850    
851      * Doing something smarter than relying on MLTreeGen on architectures
852        like, e.g., the x86 where hardware division delivers both quotient and
853        remainder at the same time.  With this, the implementation of the
854        round-to-neginf operations could be further streamlined.
855    
856      * Remove inlining support for div/mod/rem from the frontend and replace it
857        with primops that get carried through to the backend.  Do this for all
858        int and word types.
859    
860    ----------------------------------------------------------------------
861    Name: Matthias Blume
862    Date: 2002/03/25 17:25:00 EST
863    Tag: blume-20020325-divmod
864    Description:
865    
866    I improved (hopefully without breaking them) the implementation of Int.div,
867    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
868    advantage of the following observations:
869    
870      Let  q = x quot y      r = x rem y
871           d = x div  y      m = x mod y
872    
873    where "quot" is the round-to-zero version of integer division that
874    hardware usually provides.  Then we have:
875    
876         r = x - q * y        where neither the * nor the - will overflow
877         d = if q >= 0 orelse x = q * y then q else q - 1
878                              where neither the * nor the - will overflow
879         m = if q >= 0 orelse r = 0 then r else r + y
880                              where the + will not overflow
881    
882    This results in substantial simplification of the generated code.
883    The following table shows the number of CFG nodes and edges generated
884    for
885            fun f (x, y) = x OPER y
886            (* with OPER \in div, mod, quot, rem *)
887    
888    
889        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
890        --------------------------------------------------------
891         div |         24 |         39 |         12 |         16
892         mod |         41 |         71 |         12 |         16
893        quot |          8 |         10 |          8 |         10
894         rem |         10 |         14 |          8 |         10
895    
896    
897    ----------------------------------------------------------------------
898    Name: Matthias Blume
899    Date: 2002/03/25 22:06:00 EST
900    Tag: blume-20020325-cprotobug
901    Description:
902    
903    Fixed a bug in cproto (c prototype decoder).
904    
905    ----------------------------------------------------------------------
906    Name: Matthias Blume
907    Date: 2002/03/25 16:00:00 EST
908    Tag: blume-20020325-raw-primops
909    Description:
910    
911    I did some cleanup to Allen's new primop code and
912    replaced yesterday's bootfiles with new ones.
913    (But they are stored in the same place.)
914    
915    ----------------------------------------------------------------------
916    Name: Matthias Blume
917    Date: 2002/03/24 22:40:00 EST
918    Tag: blume-20020324-bootfiles
919    Description:
920    
921    Made the bootfiles that Allen asked for.
922    
923    ----------------------------------------------------------------------
924    Name: Allen Leung
925    Date: 2002/03/23 15:50:00 EST
926    Tag: leunga-20020323-flint-cps-rcc-primops
927    Description:
928    
929      1. Changes to FLINT primops:
930    
931        (* make a call to a C-function;
932         * The primop carries C function prototype information and specifies
933         * which of its (ML-) arguments are floating point. C prototype
934         * information is for use by the backend, ML information is for
935         * use by the CPS converter. *)
936      | RAW_CCALL of { c_proto: CTypes.c_proto,
937                       ml_args: ccall_type list,
938                       ml_res_opt: ccall_type option,
939                       reentrant : bool
940                     } option
941       (* Allocate uninitialized storage on the heap.
942        * The record is meant to hold short-lived C objects, i.e., they
943        * are not ML pointers.  With the tag, the representation is
944        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
945        *)
946      | RAW_RECORD of {tag:bool,sz:int}
947      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
948    
949      2.  These CPS primops are now overloaded:
950    
951           rawload of {kind:numkind}
952           rawstore of {kind:numkind}
953    
954          The one argument form is:
955    
956             rawload {kind} address
957    
958          The two argument form is:
959    
960             rawload {kind} [ml object, byte-offset]
961    
962      3. RAW_CCALL/RCC now takes two extra arguments:
963    
964         a. The first is whether the C call is reentrant, i.e., whether
965            ML state should be saved and restored.
966         b. The second argument is a string argument specifying the name of
967            library and the C function.
968    
969         These things are currently not handled in the code generator, yet.
970    
971      4. In CProto,
972    
973         An encoding type of "bool" means "ml object" and is mapped into
974         C prototype of PTR.  Note that "bool" is different than "string",
975         even though "string" is also mapped into PTR, because "bool"
976         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
977    
978      5. Pickler/unpicker
979    
980         Changed to handle RAW_RECORD and newest RAW_CCALL
981    
982      6. MLRiscGen,
983    
984         1. Changed to handle the new rawload/rawstore/rawrecord operators.
985         2. Code for handling C Calls has been moved to a new module CPSCCalls,
986            in the file CodeGen/cpscompile/cps-c-calls.sml
987    
988      7. Added the conditional move operator
989    
990             condmove of branch
991    
992         to cps.  Generation of this is still buggy so it is currently
993         disabled.
994    
995    ----------------------------------------------------------------------
996    Name: Lal George
997    Date: 2002/03/22 14:18:25 EST
998    Tag: george-20020322-cps-branch-prob
999    Description:
1000    
1001    Implemented the Ball-Larus branch prediction-heuristics, and
1002    incorporated graphical viewers for control flow graphs.
1003    
1004    Ball-Larus Heuristics:
1005    ---------------------
1006    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1007    
1008    By design it uses the Dempster-Shafer theory for combining
1009    probabilities.  For example, in the function:
1010    
1011        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1012    
1013    the ball-larus heuristics predicts that the n=0 is unlikely
1014    (OH-heuristic), and the 'then' branch is unlikely because of the
1015    RH-heuristic -- giving the 'then' branch an even lower combined
1016    probability using the Dempster-Shafer theory.
1017    
1018    Finally, John Reppy's loop analysis in MLRISC, further lowers the
1019    probability of the 'then' branch because of the loop in the else
1020    branch.
1021    
1022    
1023    Graphical Viewing:
1024    ------------------
1025    I merely plugged in Allen's graphical viewers into the compiler. The
1026    additional code is not much. At the top level, saying:
1027    
1028            Control.MLRISC.getFlag "cfg-graphical-view" := true;
1029    
1030    will display the graphical view of the control flow graph just before
1031    back-patching.  daVinci must be in your path for this to work. If
1032    daVinci is not available, then the default viewer can be changed
1033    using:
1034    
1035            Control.MLRISC.getString "viewer"
1036    
1037    which can be set to "dot" or "vcg" for the corresponding viewers. Of
1038    course, these viewers must be in your path.
1039    
1040    The above will display the compilation unit at the level of clusters,
1041    many of which are small, boring, and un-interesting. Also setting:
1042    
1043            Control.MLRISC.getInt "cfg-graphical-view_size"
1044    
1045    will display clusters that are larger than the value set by the above.
1046    
1047    
1048    ----------------------------------------------------------------------
1049    Name: Matthias Blume
1050    Date: 2002/03/21 22:20:00 EST
1051    Tag: blume-20020321-kmp-bugfix
1052    Description:
1053    
1054    Changed the interface to the KMP routine in PreString and fixed
1055    a minor bug in one place where it was used.
1056    
1057    ----------------------------------------------------------------------
1058    Name: Allen Leung
1059    Date: 2002/03/21 20:30:00 EST
1060    Tag: leunga-20020321-cfg
1061    Description:
1062    
1063      Fixed a potential problem in cfg edge splitting.
1064    
1065    ----------------------------------------------------------------------
1066    Name: Allen Leung
1067    Date: 2002/03/21 17:15:00 EST
1068    Tag: leunga-20020321-x86-fp-cfg
1069    Description:
1070    
1071      1. Recoded the buggy parts of x86-fp.
1072    
1073         a. All the block reordering code has been removed.
1074            We now depend on the block placement phases to do this work.
1075    
1076         b. Critical edge splitting code has been simplified and moved into the
1077            CFG modules, as where they belong.
1078    
1079         Both of these were quite buggy and complex.  The code is now much, much
1080         simpler.
1081    
1082      2. X86 backend.
1083    
1084         a. Added instructions for 64-bit support.  Instruction selection for
1085            64-bit has not been committed, however, since that
1086            requires changes to MLTREE which haven't been approved by
1087            Lal and John.
1088    
1089         b. Added support for FUCOMI and FUCOMIP when generating code for
1090            PentiumPro and above.  We only generate these instructions in
1091            the fast-fp mode.
1092    
1093         c. Added cases for JP and JNP in X86FreqProps.
1094    
1095      3. CFG
1096    
1097         CFG now has a bunch of methods for edge splitting and merging.
1098    
1099      4. Machine description.
1100    
1101         John's simplification of MLTREE_BASIS.fcond broke a few machine
1102         description things:
1103    
1104         rtl-build.{sig,sml} and hppa.mdl fixed.
1105    
1106         NOTE: the machine description stuff in the repository is still broken.
1107               Again, I can't put my fixes in because that involves
1108               changes to MLTREE.
1109    
1110    ----------------------------------------------------------------------
1111    Name: Matthias Blume
1112    Date: 2002/03/20 15:55:00 EST
1113    Tag: blume-20020320-kmp
1114    Description:
1115    
1116    Implemented Knuth-Morris-Pratt string matching in PreString and used
1117    it for String.isSubstring, Substring.isSubstring, and
1118    Substring.position.
1119    
1120    (Might need some stress-testing.  Simple examples worked fine.)
1121    
1122    ----------------------------------------------------------------------
1123    Name: Matthias Blume
1124    Date: 2002/03/19 16:37:00 EST
1125    Tag: blume-20020319-witnesses
1126    Description:
1127    
1128    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1129    
1130    This implements a generic mechanism for changing constness qualifiers
1131    anywhere within big C types without resorting to outright "casts".
1132    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1133    modify the constness at the outermost level.)
1134    The implementation of "convert" is based on the idea of "witness"
1135    values -- values that are not used by the operation but whose types
1136    "testify" to their applicability.  On the implementation side, "convert"
1137    is simply a projection (returning its second curried argument).  With
1138    cross-module inlining, it should not result in any machine code being
1139    generated.
1140    
1141    ----------------------------------------------------------------------
1142    Name: Matthias Blume
1143    Date: 2002/03/15 16:40:00 EST
1144    Tag: blume-20020315-basis
1145    Description:
1146    
1147    Provided (preliminary?) implementations for
1148    
1149      {String,Substring}.{concatWith,isSuffix,isSubstring}
1150    
1151    and
1152    
1153      Substring.full
1154    
1155    Those are in the Basis spec but they were missing in SML/NJ.
1156    
1157    ----------------------------------------------------------------------
1158    Name: Matthias Blume
1159    Date: 2002/03/14 21:30:00 EST
1160    Tag: blume-20020314-controls
1161    Description:
1162    
1163    Controls:
1164    ---------
1165    
1166    1. Factored out the recently-added Controls : CONTROLS stuff and put
1167       it into its own library $/controls-lib.cm.  The source tree for
1168       this is under src/smlnj-lib/Controls.
1169    
1170    2. Changed the names of types and functions in this interface, so they
1171       make a bit more "sense":
1172    
1173          module -> registry
1174          'a registry -> 'a group
1175    
1176    3. The interface now deals in ref cells only.  The getter/setter interface
1177       is (mostly) gone.
1178    
1179    4. Added a function that lets one register an already-existing ref cell.
1180    
1181    5. Made the corresponding modifications to the rest of the code so that
1182       everything compiles again.
1183    
1184    6. Changed the implementation of Controls.MLRISC back to something closer
1185       to the original.  In particular, this module (and therefore MLRISC)
1186       does not depend on Controls.  There now is some link-time code in
1187       int-sys.sml that registers the MLRISC controls with the Controls
1188       module.
1189    
1190    CM:
1191    ---
1192    
1193      * One can now specify the lambda-split aggressiveness in init.cmi.
1194    
1195    ----------------------------------------------------------------------
1196    Name: Allen Leung
1197    Date: 2002/03/13 17:30:00 EST
1198    Tag: leunga-20020313-x86-fp-unary
1199    Description:
1200    
1201    Bug fix for:
1202    
1203    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1204    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1205    > - fun f(x,(y,z)) = Real.~ y;
1206    > [autoloading]
1207    > [autoloading done]
1208    >       fchsl   (%eax), 184(%esp)
1209    > Error: MLRisc bug: X86MCEmitter.emitInstr
1210    >
1211    > uncaught exception Error
1212    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1213    
1214    The problem was that the code generator did not generate any fp registers
1215    in this case, and the ra didn't know that it needed to run the X86FP phase to
1216    translate the pseudo fp instruction.   This only happened with unary fp
1217    operators in certain situations.
1218    
1219    ----------------------------------------------------------------------
1220    Name: Matthias Blume
1221    Date: 2002/03/13 14:00:00 EST
1222    Tag: blume-20020313-overload-etc
1223    Description:
1224    
1225    1. Added _overload as a synonym for overload for backward compatibility.
1226       (Control.overloadKW must be true for either version to be accepted.)
1227    
1228    2. Fixed bug in install script that caused more things to be installed
1229       than what was requested in config/targets.
1230    
1231    3. Made CM aware of the (_)overload construct so that autoloading
1232       works.
1233    
1234    ----------------------------------------------------------------------
1235    Name: Matthias Blume
1236    Date: 2002/03/12 22:03:00 EST
1237    Tag: blume-20020312-url
1238    Description:
1239    
1240    Forgot to update BOOT and srcarchiveurl.
1241    
1242    ----------------------------------------------------------------------
1243    Name: Matthias Blume
1244    Date: 2002/03/12 17:30:00 EST
1245    Tag: blume-20020312-version110392
1246    Description:
1247    
1248    Yet another version number bump (because of small changes to the
1249    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1250    
1251    Changes:
1252    
1253      The new pid generation scheme described a few weeks ago was overly
1254      complicated.  I implemented a new mechanism that is simpler and
1255      provides a bit more "stability":  Once CM has seen a compilation
1256      unit, it keeps its identity constant (as long as you do not delete
1257      those crucial CM/GUID/* files).  This means that when you change
1258      an interface, compile, then go back to the old interface, and
1259      compile again, you arrive at the original pid.
1260    
1261      There now also is a mechanism that instructs CM to use the plain
1262      environment hash as a module's pid (effectively making its GUID
1263      the empty string).  For this, "noguid" must be specified as an
1264      option to the .sml file in question within its .cm file.
1265      This is most useful for code that is being generated by tools such
1266      as ml-nlffigen (because during development programmers tend to
1267      erase the tool's entire output directory tree including CM's cached
1268      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1269      revert to the old, broken behavior of SML/NJ, but in specific cases
1270      where there is no danger of interface confusion, its use is ok
1271      (I think).
1272    
1273      ml-nlffigen by default generates "noguid" annotations.  They can be
1274      turned off by specifying -guid in its command line.
1275    
1276    ----------------------------------------------------------------------
1277    Name: Lal George
1278    Date: 2002/03/12 12 14:42:36 EST
1279    Tag: george-20020312-frequency-computation
1280    Description:
1281    
1282    Integrated jump chaining and static block frequency into the
1283    compiler. More details and numbers later.
1284    
1285    ----------------------------------------------------------------------
1286    Name: Lal George
1287    Date: 2002/03/11 11 22:38:53 EST
1288    Tag: george-20020311-jump-chain-elim
1289    Description:
1290    
1291    Tested the jump chain elimination on all architectures (except the
1292    hppa).  This is on by default right now and is profitable for the
1293    alpha and x86, however, it may not be profitable for the sparc and ppc
1294    when compiling the compiler.
1295    
1296    The gc test will typically jump to a label at the end of the cluster,
1297    where there is another jump to an external cluster containing the actual
1298    code to invoke gc. This is to allow factoring of common gc invocation
1299    sequences. That is to say, we generate:
1300    
1301            f:
1302               testgc
1303               ja   L1      % jump if above to L1
1304    
1305            L1:
1306               jmp L2
1307    
1308    
1309    After jump chain elimination the 'ja L1' instructions is converted to
1310    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1311    up being implemented in their long form (if L2 is far away) using:
1312    
1313            jbe     L3      % jump if below or equal to L3
1314            jmp     L2
1315         L3:
1316            ...
1317    
1318    
1319    For large compilation units L2  may be far away.
1320    
1321    
1322    ----------------------------------------------------------------------
1323    Name: Matthias Blume
1324    Date: 2002/03/11 13:30:00 EST
1325    Tag: blume-20020311-mltreeeval
1326    Description:
1327    
1328    A functor parameter was missing.
1329    
1330    ----------------------------------------------------------------------
1331    Name: Allen Leung
1332    Date: 2002/03/11 10:30:00 EST
1333    Tag: leunga-20020311-runtime-string0
1334    Description:
1335    
1336       The representation of the empty string now points to a
1337    legal null terminated C string instead of unit.  It is now possible
1338    to convert an ML string into C string with InlineT.CharVector.getData.
1339    This compiles into one single machine instruction.
1340    
1341    ----------------------------------------------------------------------
1342    Name: Allen Leung
1343    Date: 2002/03/10 23:55:00 EST
1344    Tag: leunga-20020310-x86-call
1345    Description:
1346    
1347       Added machine generation for CALL instruction (relative displacement mode)
1348    
1349    ----------------------------------------------------------------------
1350    Name: Matthias Blume
1351    Date: 2002/03/08 16:05:00
1352    Tag: blume-20020308-entrypoints
1353    Description:
1354    
1355    Version number bumped to 110.39.1.  NEW BOOTFILES!
1356    
1357    Entrypoints: non-zero offset into a code object where execution should begin.
1358    
1359    - Added the notion of an entrypoint to CodeObj.
1360    - Added reading/writing of entrypoint info to Binfile.
1361    - Made runtime system bootloader aware of entrypoints.
1362    - Use the address of the label of the first function given to mlriscGen
1363      as the entrypoint.  This address is currently always 0, but it will
1364      not be 0 once we turn on block placement.
1365    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1366      with entry points) from mlriscGen.
1367    
1368    ----------------------------------------------------------------------
1369    Name: Allen Leung
1370    Date: 2002/03/07 20:45:00 EST
1371    Tag: leunga-20020307-x86-cmov
1372    Description:
1373    
1374       Bug fixes for CMOVcc on x86.
1375    
1376       1. Added machine code generation for CMOVcc
1377       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1378       3. CMOVcc cannot have an immediate operand as argument.
1379    
1380    ----------------------------------------------------------------------
1381    Name: Matthias Blume
1382    Date: 2002/03/07 16:15:00 EST
1383    Tag: blume-20020307-controls
1384    Description:
1385    
1386    This is a very large but mostly boring patch which makes (almost)
1387    every tuneable compiler knob (i.e., pretty much everything under
1388    Control.* plus a few other things) configurable via both the command
1389    line and environment variables in the style CM did its configuration
1390    until now.
1391    
1392    Try starting sml with '-h' (or, if you are brave, '-H')
1393    
1394    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1395    implements the underlying generic mechanism.
1396    
1397    The interface to some of the existing such facilities has changed somewhat.
1398    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1399    (The getFoo interface is still there for backward-compatibility, but its
1400    use is deprecated.)
1401    
1402    The ml-build script passes -Cxxx=yyy command-line arguments through so
1403    that one can now twiddle the compiler settings when using this "batch"
1404    compiler.
1405    
1406    TODO items:
1407    
1408    We should go through and throw out all controls that are no longer
1409    connected to anything.  Moreover, we should go through and provide
1410    meaningful (and correct!) documentation strings for those controls
1411    that still are connected.
1412    
1413    Currently, multiple calls to Controls.new are accepted (only the first
1414    has any effect).  Eventually we should make sure that every control
1415    is being made (via Controls.new) exactly once.  Future access can then
1416    be done using Controls.acc.
1417    
1418    Finally, it would probably be a good idea to use the getter-setter
1419    interface to controls rather than ref cells.  For the time being, both
1420    styles are provided by the Controls module, but getter-setter pairs are
1421    better if thread-safety is of any concern because they can be wrapped.
1422    
1423    *****************************************
1424    
1425    One bug fix: The function blockPlacement in three of the MLRISC
1426    backpatch files used to be hard-wired to one of two possibilities at
1427    link time (according to the value of the placementFlag).  But (I
1428    think) it should rather sense the flag every time.
1429    
1430    *****************************************
1431    
1432    Other assorted changes (by other people who did not supply a HISTORY entry):
1433    
1434    1. the cross-module inliner now works much better (Monnier)
1435    2. representation of weights, frequencies, and probabilities in MLRISC
1436       changed in preparation of using those for weighted block placement
1437       (Reppy, George)
1438    
1439    ----------------------------------------------------------------------
1440    Name: Lal George
1441    Date: 2002/03/07 14:44:24 EST 2002
1442    Tag: george-20020307-weighted-block-placement
1443    
1444    Tested the weighted block placement optimization on all architectures
1445    (except the hppa) using AMPL to generate the block and edge frequencies.
1446    Changes were required in the machine properties to correctly
1447    categorize trap instructions. There is an MLRISC flag
1448    "weighted-block-placement" that can be used to enable weighted block
1449    placement, but this will be ineffective without block/edge
1450    frequencies (coming soon).
1451    
1452    
1453    ----------------------------------------------------------------------
1454    Name: Lal George
1455    Date: 2002/03/05 17:24:48 EST
1456    Tag: george-20020305-linkage-cluster
1457    
1458    In order to support the block placement optimization, a new cluster
1459    is generated as the very first cluster (called the linkage cluster).
1460    It contains a single jump to the 'real' entry point for the compilation
1461    unit. Block placement has no effect on the linkage cluster itself, but
1462    all the other clusters  have full freedom in the manner in which they
1463    reorder blocks or functions.
1464    
1465    On the x86 the typical linkage code that is generated is:
1466       ----------------------
1467            .align 2
1468       L0:
1469            addl    $L1-L0, 72(%esp)
1470            jmp     L1
1471    
1472    
1473            .align  2
1474       L1:
1475       ----------------------
1476    
1477    72(%esp) is the memory location for the stdlink register. This
1478    must contain the address of the CPS function being called. In the
1479    above example, it contains the address of  L0; before
1480    calling L1 (the real entry point for the compilation unit), it
1481    must contain the address for L1, and hence
1482    
1483            addl $L1-L0, 72(%esp)
1484    
1485    I have tested this on all architectures except the hppa.The increase
1486    in code size is of course negligible
1487    
1488    ----------------------------------------------------------------------
1489    Name: Allen Leung
1490    Date: 2002/03/03 13:20:00 EST
1491    Tag: leunga-20020303-mlrisc-tools
1492    
1493      Added #[ ... ] expressions to mlrisc tools
1494    
1495    ----------------------------------------------------------------------
1496    Name: Matthias Blume
1497    Date: 2002/02/27 12:29:00 EST
1498    Tag: blume-20020227-cdebug
1499    Description:
1500    
1501    - made types in structure C and C_Debug to be equal
1502    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1503    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1504    
1505    ----------------------------------------------------------------------
1506    Name: Matthias Blume
1507    Date: 2002/02/26 12:00:00 EST
1508    Tag: blume-20020226-ffi
1509    Description:
1510    
1511    1. Fixed a minor bug in CM's "noweb" tool:
1512       If numbering is turned off, then truly don't number (i.e., do not
1513       supply the -L option to noweb).  The previous behavior was to supply
1514       -L'' -- which caused noweb to use the "default" line numbering scheme.
1515       Thanks to Chris Richards for pointing this out (and supplying the fix).
1516    
1517    2. Once again, I reworked some aspects of the FFI:
1518    
1519       A. The incomplete/complete type business:
1520    
1521       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1522         gone!
1523       - ML types representing an incomplete type are now *equal* to
1524         ML types representing their corresponding complete types (just like
1525         in C).  This is still safe because ml-nlffigen will not generate
1526         RTTI for incomplete types, nor will it generate functions that
1527         require access to such RTTI.   But when ML code generated from both
1528         incomplete and complete versions of the C type meet, the ML types
1529         are trivially interoperable.
1530    
1531         NOTE:  These changes restore the full generality of the translation
1532         (which was previously lost when I eliminated functorization)!
1533    
1534       B. Enum types:
1535    
1536       - Structure C now has a type constructor "enum" that is similar to
1537         how the "su" constructor works.  However, "enum" is not a phantom
1538         type because each "T enum" has values (and is isomorphic to
1539         MLRep.Signed.int).
1540       - There are generic access operations for enum objects (using
1541         MLRep.Signed.int).
1542       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1543         * The structure contains the definition of type "mlrep" (the ML-side
1544         representation type of the enum).  Normally, mlrep is the same
1545         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1546         then mlrep will be defined as a datatype -- thus facilitating
1547         pattern matching on mlrep values.
1548         ("-ec" will be suppressed if there are duplicate values in an
1549          enumeration.)
1550         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1551         will be generated for each C enum constant xxx.
1552         * Conversion functions m2i and i2m convert between mlrep and
1553         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1554         * Coversion functions c and ml convert between mlrep and "tag enum".
1555         * Access functions (get/set) fetch and store mlrep values.
1556       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1557         enumerations are merged into one single enumeration represented by
1558         structure E_'.
1559    
1560    ----------------------------------------------------------------------
1561    Name: Allen Leung
1562    Date: 2002/02/25 04:45:00 EST
1563    Tag: leunga-20020225-cps-spill
1564    
1565    This is a new implementation of the CPS spill phase.
1566    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1567    In case of problems, replace it with the old file spill.sml
1568    
1569    The current compiler runs into some serious performance problems when
1570    constructing a large record.  This can happen when we try to compile a
1571    structure with many items.  Even a very simple structure like the following
1572    makes the compiler slow down.
1573    
1574        structure Foo = struct
1575           val x_1 = 0w1 : Word32.int
1576           val x_2 = 0w2 : Word32.int
1577           val x_3 = 0w3 : Word32.int
1578           ...
1579           val x_N = 0wN : Word32.int
1580        end
1581    
1582    The following table shows the compile time, from N=1000 to N=4000,
1583    with the old compiler:
1584    
1585    N
1586    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1587           MLRISC ra                               0.06u  0.00s  0.05g
1588              (spills = 0 reloads = 0)
1589           TOTAL                                   0.63u  0.07s  0.21g
1590    
1591    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1592           MLRISC ra                               5.68u  0.59s  3.93g
1593              (spills = 0 reloads = 0)
1594           TOTAL                                   14.71u  0.99s  4.81g
1595    
1596    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1597           MLRISC ra                               5.54u  0.65s  3.91g
1598              (spills = 543 reloads = 1082)
1599           TOTAL                                   65.40u  3.13s  6.00g
1600    
1601    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1602           MLRISC ra                               0.80u  0.10s  0.55g
1603              (spills = 42 reloads = 84)
1604           TOTAL                                   129.42u  5.10s  4.13g
1605    
1606    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1607           MLRISC ra                               2.69u  0.27s  1.38g
1608              (spills = 62 reloads = 124)
1609           TOTAL                                   682.48u  19.61s  13.99g
1610    
1611    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1612           MLRISC ra                               4.96u  0.27s  2.72g
1613              (spills = 85 reloads = 170)
1614           TOTAL                                   2375.26u  57.21s  48.00g
1615    
1616    As you can see the old cps spill module suffers from some serious
1617    performance problem.  But since I cannot decipher the old code fully,
1618    instead of patching the problems up, I'm reimplementing it
1619    with a different algorithm.  The new code is more modular,
1620    smaller when compiled, and substantially faster
1621    (O(n log n) time and O(n) space).  Timing of the new spill module:
1622    
1623    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1624          MLRISC ra                               0.25u  0.02s  0.15g
1625             (spills=1 reloads=3)
1626          TOTAL                                   7.74u  0.34s  1.62g
1627    
1628    Implementation details:
1629    
1630    As far as I can tell, the purpose of the CPS spill module is to make sure the
1631    number of live variables at any program point (the bandwidth)
1632    does not exceed a certain limit, which is determined by the
1633    size of the spill area.
1634    
1635    When the bandwidth is too large, we decrease the register pressure by
1636    packing live variables into spill records.  How we achieve this is
1637    completely different than what we did in the old code.
1638    
1639    First, there is something about the MLRiscGen code generator
1640    that we should be aware of:
1641    
1642    o MLRiscGen performs code motion!
1643    
1644       In particular, it will move floating point computations and
1645       address computations involving only the heap pointer to
1646       their use sites (if there is only a single use).
1647       What this means is that if we have a CPS record construction
1648       statement
1649    
1650           RECORD(k,vl,w,e)
1651    
1652       we should never count the new record address w as live if w
1653       has only one use (which is often the case).
1654    
1655       We should do something similar to floating point, but the transformation
1656       there is much more complex, so I won't deal with that.
1657    
1658    Secondly, there are now two new cps primops at our disposal:
1659    
1660     1. rawrecord of record_kind option
1661        This pure operator allocates some uninitialized storage from the heap.
1662        There are two forms:
1663    
1664         rawrecord NONE [INT n]  allocates a tagless record of length n
1665         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1666                                     and initializes the tag.
1667    
1668     2. rawupdate of cty
1669          rawupdate cty (v,i,x)
1670          Assigns to x to the ith component of record v.
1671          The storelist is not updated.
1672    
1673    We use these new primops for both spilling and increment record construction.
1674    
1675     1. Spilling.
1676    
1677        This is implemented with a linear scan algorithm (but generalized
1678        to trees).  The algorithm will create a single spill record at the
1679        beginning of the cps function and use rawupdate to spill to it,
1680        and SELECT or SELp to reload from it.  So both spills and reloads
1681        are fine-grain operations.  In contrast, in the old algorithm
1682        "spills" have to be bundled together in records.
1683    
1684        Ideally, we should sink the spill record construction to where
1685        it is needed.  We can even split the spill record into multiple ones
1686        at the places where they are needed.  But CPS is not a good
1687        representation for global code motion, so I'll keep it simple and
1688        am not attempting this.
1689    
1690     2. Incremental record construction (aka record splitting).
1691    
1692        Long records with many component values which are simulatenously live
1693        (recall that single use record addresses are not considered to
1694         be live) are constructed with rawrecord and rawupdate.
1695        We allocate space on the heap with rawrecord first, then gradually
1696        fill it in with rawupdate.  This is the technique suggested to me
1697        by Matthias.
1698    
1699        Some restrictions on when this is applicable:
1700        1. It is not a VECTOR record.  The code generator currently does not handle
1701           this case. VECTOR record uses double indirection like arrays.
1702        2. All the record component values are defined in the same "basic block"
1703           as the record constructor.  This is to prevent speculative
1704           record construction.
1705    
1706    ----------------------------------------------------------------------
1707    Name: Allen Leung
1708    Date: 2002/02/22 01:02:00 EST
1709    Tag: leunga-20020222-mlrisc-tools
1710    
1711    Minor bug fixes in the parser and rewriter
1712    
1713    ----------------------------------------------------------------------
1714  Name: Allen Leung  Name: Allen Leung
1715  Date: 2002/02/21 20:20:00 EST  Date: 2002/02/21 20:20:00 EST
1716  Tag: leunga-20020221-peephole  Tag: leunga-20020221-peephole

Legend:
Removed from v.1085  
changed lines
  Added in v.1381

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