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 1067, Fri Feb 15 17:08:17 2002 UTC revision 1480, Fri Apr 23 15:37:52 2004 UTC
# Line 13  Line 13 
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume (blume (at) tti - c (dot) org)
17    Date: 2004/04/23 10:40:00 CDT
18    Tag: blume-20040423-ml-build
19    Description:
20    
21    The ml-build script now terminates with a non-0 status when something
22    goes wrong.
23    
24    ----------------------------------------------------------------------
25    Name: Matthias Blume (blume (at) tti - c (dot) org)
26    Date: 2004/04/22 16:35:00 CDT
27    Tag: blume-20040422-Option
28    Description:
29    
30    Made exception Option to be the same as exception Option.Option
31    (as it should be).
32    
33    ----------------------------------------------------------------------
34    Name: Allen Leung (leunga (at) reservoir (dot) com)
35    Date: 2004/03/19 14:40:00 EST
36    Tag: leunga-20040319-cygwin-nlffi
37    Description:
38    
39    Fixed the runtime so that ml-nlffi-lib runs on the cygwin version
40    of SML/NJ.  The problem is that
41    
42       lib = dlopen(NULL, ...)
43       f   = dlsym(lib, "malloc");
44    
45    does not work on Windows unless we explicitly export symbols
46    such as 'malloc' during linking.  We fixed this by explicitly
47    exporting the required symbols with the magic gcc incantation:
48    
49        -Wl,--export-all cygwin.def
50    
51    where cygwin.def is a file containing all the symbols that
52    we wish to export.
53    
54    I suspect this is a Windows problem and we'll have to
55    do the same (somehow with windows compilers) when
56    we build the native win32 version with the system
57    calls LoadLibrary/GetProcAddress.
58    
59    ----------------------------------------------------------------------
60    Name: Matthias Blume (blume (at) tti - c (dot) org)
61    Date: 2004/03/04 16:35:00 CST
62    Tag: blume-20040304-intinf-fmt
63    Description:
64    
65    Fixed problem with IntInf.fmt (sign would show up on the right instead
66    of on the left for BIN, OCT, and HEX).
67    
68    ----------------------------------------------------------------------
69    Name: Matthias Blume (blume (at) tti - c (dot) org)
70    Date: 2004/03/04 11:25:00 CST
71    Tag: blume-20040304-symlinks
72    Description:
73    
74    Fixed problem with installer script (unix only) where bin/ml-yacc and
75    friends pointed (via symlinks) to absolute locations instead of just
76    .run-sml.  This was reported by Vesa A Norrman.
77    
78    ----------------------------------------------------------------------
79    Name: Matthias Blume (blume (at) tti - c (dot) org)
80    Date: 2004/02/13 14:50:00 CST
81    Tag: Release_110_45
82    Description:
83    
84    New working version (110.45). New bootfiles.
85    
86    ----------------------------------------------------------------------
87    Name: Matthias Blume (blume (at) tti - c (dot) org)
88    Date: 2004/01/26 15:15:15 CST
89    Tag: blume-20040126-toplevel
90    Description:
91    
92    Improved handling of exceptions at the interactive toplevel.
93    
94    ----------------------------------------------------------------------
95    Name: Matthias Blume (blume (at) tti - c (dot) org)
96    Date: 2004/01/26 11:25:00
97    Tag: blume-20040126-app
98    Description:
99    
100    Type of top-level "app" corrected.
101    Added code for setting vp_limitPtrMask to Win32-specific runtime.
102    
103    ----------------------------------------------------------------------
104    Name: Matthias Blume (blume (at) tti - c (dot) org)
105    Date: 2003/11/18 17:10 CST
106    Tag: blume-20031118-basis-fiddle
107    Description:
108    
109    - changed Timer interface to what might become the spec
110    - POSIX_FLAGS -> BIT_FLAGS according to spec
111    - some other minor discrepancies wrt. spec eliminated
112    
113    ----------------------------------------------------------------------
114    Name: Matthias Blume (blume (at) tti - c (dot) org)
115    Date: 2003/11/06 12:00:00 CST
116    Tag: Release_110_44
117    Description:
118    
119    New working version (110.44). New bootfiles.
120    
121    ----------------------------------------------------------------------
122    Name: Matthias Blume (blume (at) tti - c (dot) org)
123    Date: 2003/11/04 11:50:00 CST
124    Tag: blume-20031104-move-libraries
125    Description:
126    
127    Eliminated the "dont_move_libraries" directive in config/targets.
128    (The mechanism was broken and could not be fixed easily.  Moreover,
129    there does not seem to be any reason not to move all libraries into
130    lib during installation.  I originally implemented this directive as a
131    backward-compatibility feature when I first introduced the new CM.
132    Now that things have been stable for a long time and going back to the
133    old CM is not an option, there is no reason to keep it around.)
134    
135    ----------------------------------------------------------------------
136    Name: Matthias Blume (blume (at) tti - c (dot) org)
137    Date: 2003/11/03 16:00:00 CST
138    Tag: blume-20031103-installdir
139    Description:
140    
141    Made installer honor INSTALLDIR variable again.  (Thanks to Chris
142    Richards for pointing out the problem and providing the solution.)
143    
144    ----------------------------------------------------------------------
145    Name: Matthias Blume (blume (at) tti - c (dot) org)
146    Date: 2003/10/01 17:05:00 CDT
147    Tag: blume-20031001-lal-mlrisc
148    Description:
149    
150    MLRISC bug fix from Lal.
151    
152    ----------------------------------------------------------------------
153    Name: Matthias Blume (blume (at) tti - c (dot) org)
154    Date: 2003/09/30 16:10:00 CDT
155    Tag: blume-20030930-primio-bat
156    Description:
157    
158    1. Added openVector, nullRd, and nullWr to PRIM_IO.
159    2. Improved .bat files (for Win32 port) to make things work under Win95.
160       (thanks to Aaron S. Hawley for this one)
161    
162    ----------------------------------------------------------------------
163    Name: Matthias Blume (blume (at) tti - c (dot) org)
164    Date: 2003/09/26 16:05:00 CDT
165    Tag: blume-20030926-wrappriv
166    Description:
167    
168    Added missing wrapper for privilege "primitive" in $smlnj/viscomp/core.cm.
169    
170    ----------------------------------------------------------------------
171    Name: Matthias Blume (blume (at) tti - c (dot) org)
172    Date: 2003/09/26 15:00:00 CDT
173    Tag: blume-20030926-110_43_3
174    Description:
175    
176    - additional cleanup
177    - version number bump, NEW BOOTFILES
178    
179    ----------------------------------------------------------------------
180    Name: Matthias Blume (blume (at) tti - c (dot) org)
181    Date: 2003/09/26 12:00:00 CDT
182    Tag: blume-20030926-ppautoload
183    Description:
184    
185    I modified the read-eval-print loop so that the autoloader gets
186    invoked whenever the prettyprinter tries to look up a symbol that
187    is not currently defined in the toplevel environment but which
188    appears in CM's autoload registry.  As a result, we see far fewer of
189    those ?.Foo.Bar.xxx names in the prettyprinter's output.
190    
191    In addition to this I tried to clean up some pieces of the Basis
192    implementation (e.g., Socket, Word8Array) in order to prevent other
193    instances of these ?.Foo.Bar.xxx names from being printed.
194    
195    The mechanism that picks names for types still needs some work, though.
196    (Right now it seems that if there is a type A.t which is defined to
197    be B.u, but B is unavailable at toplevel, then A.t gets printed as
198    "?.B.u" although the perhaps more sensible solution would be to use
199    "A.t" in this case.  In other words, the prettyprinter should follow
200    a chain of DEFtycs not farther than there are corresponding toplevel
201    names in the current environment.)
202    
203    ----------------------------------------------------------------------
204    Name: Matthias Blume (blume (at) tti - c (dot) org)
205    Date: 2003/09/24 16:31:00 CDT
206    Tag: blume-20030924-installer
207    Description:
208    
209    Another installer tweak:  All the ML code for the installer is now
210    compiled during CMB.make and put into a little library called
211    $smlnj/installer.cm.  The installation then simply invokes
212    
213       sml -m $smlnj/installer.cm
214    
215    and everything happens automagically.
216    
217    Win32:  ML code senses value of environment variable SMLNJ_HOME.
218    Unix:   ML code senses values of environment variables ROOT, CONFIGDIR,
219            and BINDIR.
220    
221    The new scheme guarantees that the ML code responsible for the installation
222    is in sync with the APIs of the main system.  Also, the installer is
223    somewhat faster because the installer script is precompiled.
224    
225    ----------------------------------------------------------------------
226    Name: Matthias Blume (blume (at) tti - c (dot) org)
227    Date: 2003/09/24 15:35:00 CDT
228    Tag: blume-20030924-synsock
229    Description:
230    
231    Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
232    but excludes all non-blocking operations.  Defined SOCKET (in Basis)
233    and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
234    implementations of non-blocking operations from CML's Socket
235    structure.
236    
237    ----------------------------------------------------------------------
238    Name: Matthias Blume (blume (at) tti - c (dot) org)
239    Date: 2003/09/24 15:10:05 CDT
240    Tag: blume-20030924-sockets
241    Description:
242    
243    1. Fixed SOCKET API and implementation to match Basis spec.
244       This required changing the internal representation of sockets to one
245       that remembers (for each socket file descriptor) whether it is currently
246       blocking or non-blocking.  This state is maintained lazily (i.e., a system
247       call is made only if the state actually needs to change).
248    
249    2. OS-specific details of sockets were moved into separate files, thus
250       making it possible to unify the bulk of the socket implementations
251       between Unix and Win32.
252    
253    3. CML's socket API changed accordingly.
254       (Note that we need to remove non-blocking functions from this API
255       since they are redundant in the case of CML!)
256    
257    4. CML's socket implementation now makes use of non-blocking functions
258       provided by Basis, thus removing all OS-dependent code from this part
259       of CML.
260    
261    5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
262    
263    ----------------------------------------------------------------------
264    Name: Matthias Blume (blume (at) tti - c (dot) org)
265    Date: 2003/09/22 12:10:00 CDT
266    Tag: blume-20030922-110_43_2
267    Description:
268    
269    Made a new interim version and bootfiles for developer's bootstrapping
270    convenience.
271    
272       110.43.2 -- NEW BOOTFILES
273    
274    ----------------------------------------------------------------------
275    Name: Matthias Blume (blume (at) tti - c (dot) org)
276    Date: 2003/09/19 15:55:00 CDT
277    Tag: blume-20030919-cmdir
278    Description:
279    
280    1. new-install.sh -> install.sh
281    2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
282    3. tweaked installer so that another name instead of .cm can be chosen
283       at install time (by setting the CM_DIR_ARC environment variable
284       during installation); once installation is complete, the name is
285       fixed
286    
287    ----------------------------------------------------------------------
288    Name: Matthias Blume (blume (at) tti - c (dot) org)
289    Date: 2003/09/18 16:00:00 CDT
290    Tag: blume-20030918-110_43_1
291    Description:
292    
293    Made a new interim version and bootfiles for developer's bootstrapping
294    convenience.
295    
296       110.43.1 -- NEW BOOTFILES
297    
298    ----------------------------------------------------------------------
299    Name: Matthias Blume (blume (at) tti - c (dot) org)
300    Date: 2003/09/18 15:20:00 CDT
301    Tag: blume-20030918-misc
302    Description:
303    
304    1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
305       this seems to be controversial at the moment) and used those in
306       Posix.ProcEnv.times.
307    
308    2. Added Time.{from,to}Nanoseconds to Time.
309    
310    3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
311       For example, fromLargeInt never needs to look at more than 3 "big
312       digits" to get its 53 bits of precision.
313    
314    ----------------------------------------------------------------------
315    Name: Matthias Blume (blume (at) tti - c (dot) org)
316    Date: 2003/09/17 16:30:00 CDT
317    Tag: blume-20030917-real32-slices
318    Description:
319    
320    Added an entry to the primitive environment
321    (compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
322    added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
323    
324    Removed some of the "magic" constants in real64.sml and replaced them
325    with code that generates these values from their corresponding
326    integer counterparts.
327    
328    Made all(?) the slice-related changes to the Basis and made everything
329    compile again...
330    
331    ----------------------------------------------------------------------
332    Name: Matthias Blume (blume (at) tti - c (dot) org)
333    Date: 2003/09/15 17:45:00 CDT
334    Tag: blume-20030915-rbase
335    Description:
336    
337    Fixed bug in Real.fromLargeInt.
338    
339    ----------------------------------------------------------------------
340    Name: Matthias Blume (blume (at) tti - c (dot) org)
341    Date: 2003/09/13 18:11:00 CDT
342    Tag: blume-20030913-libinstall
343    Description:
344    
345    Minor bugfix in config/libinstall (set anchor with path to
346    standalone tool after installing it, otherwise libraries that
347    need ml-lex or ml-yacc won't compile the first time the installer
348    runs).
349    
350    ----------------------------------------------------------------------
351    Name: Matthias Blume (blume (at) tti - c (dot) org)
352    Date: 2003/09/12 11:45:00 CDT
353    Tag: blume-20030912-various
354    Description:
355    
356     - fixed bug in Real.toLargeInt
357     - fixed bug in Posix.ProcEnv.times
358     - changed inputLine functions to return an option
359     - minor installer improvements / bugfixes
360     - changed default @SMLalloc parameter for x86/celeron to 64k
361    
362    ----------------------------------------------------------------------
363    Name: Matthias Blume (blume (at) tti - c (dot) org)
364    Date: 2003/09/09 22:00:00 CDT
365    Tag: Release_110_43
366    Description:
367    
368        New working release 110.43.  New bootfiles.
369    
370    ----------------------------------------------------------------------
371    Name: Matthias Blume (blume (at) tti - c (dot) org)
372    Date: 2003/09/09 19:20:00 CDT
373    Tag: blume-20030909-installer
374    Description:
375    
376    Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
377    Modified config/install.bat to take advantage of it.  Also modified
378    config/install.sh (and called it config/new-install.sh) to take advantage
379    of it on Unix systems.  (The SML code is (supposed to be) platform-
380    independent.)
381    
382    The installer can now install everything under Win32
383    as well as under *nix as long as it compiles.
384    
385    Other changes:
386    
387      - made CML compile again under Win32
388      - made eXene compile under Win32 (by providing a fake structure UnixSock
389        and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
390      - fixed a bug in nowhere: it assumed that type OS.Process.status is the
391        same as type int;  under Win32 it isn't
392      - fixed some slice-related problems in the win32-specific parts of CML
393      - added a functor argument "sameVol" to os-path-fn.sml in the Basis
394        (under Win32, the volume name is case-insensitive, and the
395        OS.Path code compares volume names for equality)
396    
397    ----------------------------------------------------------------------
398    Name: Matthias Blume (blume (at) tti - c (dot) org)
399    Date: 2003/09/08 11:55:00 CDT
400    Tag: blume-20030908-fullpath
401    Description:
402    
403    Made Win32 version of OS.FileSys.fullPath return current directory
404    when given an empty string.  This is what the spec says, and incidentally,
405    CM depends on it.  (CM otherwise goes into an infinite loop in certain
406    cases when presented with the name of a non-existing .cm file.)
407    
408    ----------------------------------------------------------------------
409    Name: Matthias Blume (blume (at) tti - c (dot) org)
410    Date: 2003/09/04 16:30:00 CDT
411    Tag: blume-20030905-slices-etc
412    Description:
413    
414      1. Changed interface to vectors and arrays in Basis to match
415         (draft) Basis spec.
416      2. Added signatures and implementations of slices according to
417         Basis spec.
418      3. Edited source code throughout the system to make it compile again
419         under 1. and 2.  (In some cases code had to be added to have it
420         match the new signatures.)
421      4. MLRISC should be backward-compatible: the copies of the originals
422         of files that needed to change under 3. were retained, the .cm files
423         check the compiler version number and use old versions when
424         appropriate.
425      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
426         dirstream -> string option (in accordance with Basis spec).
427      6. When generating code that counts lines, ml-lex used function
428         CharVector.foldli, taking advantage of its old interface.
429         This has been replaced with the corresponding code from
430         CharVectorSlice.  (html-lex must be re-lexed!)
431      7. BitArray in smlnj-lib/Util has been extended/modified to match the
432         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
433      8. Removed temporary additions (fromInternal, toInternal) from the
434         (now obsolete) IntInf in smlnj-lib/Util.
435      9. Cleaned up structure Byte.
436     10. Added localOffset, scan, and fromString to Date (according to spec).
437         Cleaned/corrected implementation of Date.
438         (Still need to check for correctness; implement better canonicalizeDate.)
439     11. Added "scan" to signature IEEE_REAL.
440     12. Some improvements to IntInf [in particular: efficiency-hack for
441         mod and rem when second operand is 2 (for parity checks).]
442     13. Changed representation of type Time.time, using a single IntInf.int
443         value counting microseconds.  This considerably simplified the
444         implementation of structure Time.  We now support negative time
445         values; scan and fromString handle signs.
446     14. Functor PrimIO now takes two additional arguments (VectorSlice and
447         ArraySlice).
448    
449    ----------------------------------------------------------------------
450    Name: Matthias Blume (blume (at) tti - c (dot) org)
451    Date: 2003/08/28 17:00:00 CDT
452    Tag: blume-20030828-intinf
453    Description:
454    
455    This is a major update which comes with a version number bump
456    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
457    and an implementation of IntInf in the Basis.
458    
459    There are a fairly large number of related changes and updates throughout
460    the system:
461    
462    Basis:
463            - Implemented IntInf.
464            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
465            - Added some missing Real64 operations, most notably Real.toLargeInt.
466            - Added FixedInt as a synonym for Int32.
467    
468    compiler:
469            * Added support for a built-in intinf type.
470              - literals
471              - pattern matching
472              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
473              - overloading on literals and operations
474    
475            This required adding a primitive type intinf, some additional
476            primops, and implementations for several non-trivial intinf
477            operations in Core.  (The intinf type is completely abstract
478            to the compiler; all operations get delegated back to the Core.)
479    
480            * Intinf equality is handled by polyequal.  However, the compiler
481            does not print its usual warning in this case (since polyequal
482            is the right thing to do there).
483    
484            * Improved the organization of structure InlineT.
485    
486            * A word about conversion primops:
487            If conversions involving intinf do not cancel out during
488            CPS contract, then the compiler must insert calls to Core functions.
489            Since all core access must be resolved already during the FLINT
490            translate phase, it would be too late a the time of CPS contract
491            to add new Core calls.  For this reason, conversion primops
492            for intinf carry two arguments:  1. the numeric argument that
493            they are supposed to convert, and 2. the Core function that
494            can help with this conversion if necessary.  If CPS contract
495            eliminates a primop, then the associated Core function becomes
496            dead and goes away.  Intinf conversion primops that do not get
497            eliminated by CPS contract get rewritten into calls of their
498            core functions by a separate, new phase.
499    
500    interactive system:
501            - Control.Print.intinfDepth controls max length of intinf constants
502            being printed. (Analogous to Control.Print.stringDepth.)
503            - Cleanup in printutil and pputil: got rid of unused stuff and
504            duplicates; replaced some of the code with code that makes better
505            use of library functionality.
506    
507    CM:
508            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
509            are no longer silent.
510    
511    CKIT:
512            Fixed mismatched uses of Int32 and LargeInt.  I always decided
513            in favor of LargeInt -- which is now the same as IntInf.
514            CKIT-knowledgable people should check whether this is what's
515            intended and otherwise change things back to using Int32 or
516            FixedInt.
517    
518    Throughout the code:
519            Started using IntInf.int literals and built-in operations
520            (e.g., comparison with 0) where this seems appropriate.
521    
522    
523    ----------------------------------------------------------------------
524    Name: Dave MacQueen (dbm@cs.uchicago.edu)
525    Date: 2003/08/13 11:36:00 CDT
526    Tag: dbm-20030813-mcz-merge1
527    Description:
528    
529    Merging changes from the mcz-branch development branch into trunk.
530    These changes involve replacement of the emulated old prettyprinter
531    interface with direct use of the SML/NJ Lib PP library, and fixing
532    of a couple of bugs (895, 1186) relating to error messages.  A new
533    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
534    has been added.
535    
536    ----------------------------------------------------------------------
537    Name: Matthias Blume (blume (at) tti - c (dot) org)
538    Date: 2003/08/11 15:45:00 CDT
539    Tag: blume-20030811-windows
540    Description:
541    
542    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
543    
544        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
545    
546    This patch restores SML/NJ's ability to run under win32.  There are a
547    number of changes, including fixes for several bugs that had gone
548    unnoticed until now:
549    
550      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
551        trying to run the win32 version from within cygwin.)
552    
553      - There are a number of simple .bat scripts that substitute for their
554        corresponding Unix shell-scripts.  (See below.)
555    
556      - The internals of ml-build have been modified slightly.  The main
557        difference is that instead of calling ".link-sml" (or link-sml.bat)
558        using OS.Process.system, the ML process delegates this task back
559        to the script.  Otherwise problems arise in mixed environments such
560        as Cygwin where scripts look and work like Unix scripts, but
561        where OS.Process.system cannot run them.
562    
563      - In CM, the srcpath pickler used native pathname syntax -- which
564        is incorrect in the case of cross-compilation.  The new pickle format
565        is independent of platform-specific naming conventions.
566    
567      - Path configuration files (such as lib/pathconfig) can now choose
568        between native and standard syntax.  Placing a line of the form
569    
570            standard!
571    
572        into the file causes all subsequent paths to be interpreted using
573        CM standard pathname syntax (= Unix conventions); a line
574    
575            native!
576    
577        switches back to native style.  This was needed so that
578        path config files can be written portably, see src/system/pathconfig.
579    
580      - Runtime system:
581    
582          - win32-filesys.c:  get_file_time and set_file_time now
583            access modification time, not creation time.
584    
585          - I/O code made aware of new array representation.
586    
587          - Bug fixes in X86.prim.masm.
588    
589       - src/system/makeml made aware of win32. (For use under cygwin
590            and other Unix-environments for windows.)
591    
592       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
593         which caused BinIO.inputAll to fail consistently.
594    
595    .bat scripts:
596    
597       Windows .bat scripts assume that SMLNJ_HOME is defined.
598    
599       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
600         applications (sml, ml-yacc, ml-lex).
601       - ml-build.bat: analogous to ml-build.
602       - config\install.bat:  Analogous to config/install.sh.  This requires
603         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
604         (nmake etc. must be on the path, and vcvars32 must have been run.)
605         Moreover, sources for ml-lex and ml-yacc need to exist under src,
606         and the bootfile hierarchy must have been unpacked under
607         sml.boot.x86-win32.
608         The script is very primitive and does a poor job at error checking.
609         It only installs the base system, ml-lex, and ml-yacc.  No other
610         libraries are being installed (i.e., you get only those that
611         are part of the compiler.)
612       - link-sml.bat:  analogous to .link-sml, but not currently used
613    
614    Unrelated bug fixes:
615    
616      - ml-nlffigen now exports structures ST_* corresponding to incomplete
617        types.
618      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
619        otherwise.)
620    
621    ----------------------------------------------------------------------
622    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
623    Date: 2003/06/17
624    Tag: macqueen-20030617-bug895
625    Description:
626    
627    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
628    Tag will be used for new development branch (mcz-branch) for
629    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
630    
631    ----------------------------------------------------------------------
632    Name: Matthias Blume (blume (at) tti - c (dot) org)
633    Date: 2003/05/27 16:55:00 CDT
634    Tag: blume-20030527-polyeq
635    Description:
636    
637    Tried to eliminated most cases of polymorphic equality.
638    
639    ----------------------------------------------------------------------
640    Name: Matthias Blume (blume (at) tti - c (dot) org)
641    Date: 2003/05/21 17:45:00 CDT
642    Tag: blume-20030517-complete
643    Description:
644    
645    Two changes:
646    
647      1. Added a flag for controlling whether non-exhaustive bindings will
648         be treated as errors (default is false).
649      2. Cleaned up the *entire* source tree so that CMB.make goes through
650         without a single non-exhaustive match- or bind warning.
651    
652    ----------------------------------------------------------------------
653    Name: Matthias Blume (blume (at) tti - c (dot) org)
654    Date: 2003/05/17 10:20:00 CDT
655    Tag: blume-20030517-absyn
656    Description:
657    
658    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
659    
660       This mainly affects the quality of error messages.  However, some
661       of the code is now more straightforward than before. (Treatment of
662       the above four constructs in translate.sml is much simpler than
663       the "macro-expansion" that was going on before.  Plus, the mach-
664       compiler no longer gets invoked just to be able to compile an
665       if-expression.)
666    
667    2. The ErrorMsg.Error exception is now caught and absorbed by the
668       interactive loop.
669    
670    ----------------------------------------------------------------------
671    Name: Allen Leung
672    Date: 2003/05/16 13:05:00 CDT
673    Tag: leunga-20030516-cygwin-runtime
674    Description:
675    
676        Ported the runtime system to cygwin, which uses the unix
677    x86-unix bin files.  Missing/buggy features:
678    
679       o getnetbyname, getnetbyaddr: these functions seem to be missing in
680         the Cygwin library.
681       o Ctrl-C handling may be flaky.
682       o Windows system calls and Windows I/O are not supported.
683    
684       A new set of binfiles is located at:
685    
686         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
687    
688       This is only needed for bootstrapping the cygwin version of smlnj.
689       Other x86 versions can use the existing binfiles.
690    
691    ----------------------------------------------------------------------
692    Name: Matthias Blume
693    Date: 2003/04/08 15:42:00 CDT
694    Tag: blume-20030408-listpair
695    Description:
696    
697    1. Added a target 'mlrisc' to installer.
698    
699    2. Added missing elements to structure ListPair.
700    
701    ----------------------------------------------------------------------
702    Name: Allen Leung
703    Date: 2003/01/07 10:40:00 EST
704    Tag: leunga-20030107-int-rem
705    Description:
706    
707    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
708    The arguments to the SUBL instruction were swapped.
709    
710    ----------------------------------------------------------------------
711    Name: Matthias Blume
712    Date: 2002/12/12 16:25:00 EST
713    Tag: blume-20021212-risc-ra
714    Description:
715    
716    Fixed a serious bug in the rewrite code for FP spilling/reloading that
717    sent the RA into an infinite loop when floating point registers get
718    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
719    110.37 and 110.38.)
720    There was another set of potential problems related to the handling of
721    MLRISC annotations (but those did not yet cause real problems, apparently).
722    
723    ----------------------------------------------------------------------
724    Name: Matthias Blume
725    Date: 2002/12/06 22:40:00 EST
726    Tag: blume-20021206-cm-fileid
727    Description:
728    
729    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
730    This fixes the problem of CM getting confused by files that suddenly
731    change their identity (e.g., by getting unlinked and recreated by some
732    text editor such as vi).  There might be a better/cheaper/cleaner way
733    of doing this, but for now this will have to do.
734    
735    ----------------------------------------------------------------------
736    Name: Matthias Blume
737    Date: 2002/10/28 09:50:00 EST
738    Tag: blume-20021028-typecheck
739    Description:
740    
741    Exported structure Typecheck from $smlnj/viscomp/core.cm.
742    
743    ----------------------------------------------------------------------
744    Name: Matthias Blume
745    Date: 2002/10/17 09:10:00 EDT
746    Tag: Release_110_42
747    Description:
748    
749    In good old tradition, there has been a slight hiccup so that we have
750    to patch 110.42 after the fact.   The old release tag has been replaced
751    (see below).
752    
753    The change solves a problem with two competing approaches the
754    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
755    each other's way.
756    
757    This change only affects the runtime system code and the installer script.
758    (No new bootfiles.)
759    
760    ----------------------------------------------------------------------
761    Name: Matthias Blume
762    Date: 2002/10/16 12:00:00 EDT
763    Tag: Release_110_42_removed
764    Description:
765    
766    New working release.  New bootfiles.
767    
768    ----------------------------------------------------------------------
769    Name: Matthias Blume
770    Date: 2002/10/10 13:10:00 EDT
771    Tag: blume-20021010-ppc-divs
772    Description:
773    
774    The mltree operator DIVS must be implemented with an overflow check on
775    the PPC because the hardware indicates divide-by-zero using "overflow" as
776    well.
777    
778    ----------------------------------------------------------------------
779    Name: Matthias Blume
780    Date: 2002/07/23 11:45:00
781    Tag: blume-20020723-smlnj-home
782    Description:
783    
784    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
785    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
786    CM_PATHCONFIG is also set), the path configuration file is assumed
787    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
788    the entire tree to some other place and everything will "just work".
789    
790    (Companion commands such as ml-build and ml-makedepend also sense this
791    variable.)
792    
793    ----------------------------------------------------------------------
794    Name: Matthias Blume
795    Date: 2002/07/12 21:19:00 EDT
796    Tag: blume-20020712-liveness
797    Description:
798    
799    Exported two useful "step" functions from liveness module (MLRISC).
800    
801    ----------------------------------------------------------------------
802    Name: Matthias Blume
803    Date: 2002/07/05 16:00 EDT
804    Tag: Release_110_41
805    Description:
806    
807    New working release.  New bootfiles.
808    
809    ----------------------------------------------------------------------
810    Name: Matthias Blume
811    Date: 2002/07/05 10:25:00 EDT
812    Tag: blume-20020705-btimp
813    Description:
814    
815    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
816    other clients can set up backtracing support.
817    
818    ----------------------------------------------------------------------
819    Name: Matthias Blume
820    Date: 2002/06/25 17:23:00 EDT
821    Tag: blume-20020625-fpmax
822    Description:
823    
824    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
825    case.  (The sense of the isNaN test was reversed -- which made min and
826    max always return their first argument.)
827    
828    ----------------------------------------------------------------------
829    Name: Matthias Blume
830    Date: 2002/06/11
831    Tag: blume-20020611-unixpath
832    Description:
833    
834    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
835    
836    ----------------------------------------------------------------------
837    Name: Matthias Blume
838    Date: 2002/06/10 16:35:00 EDT
839    Tag: blume-20020610-ieeereal
840    Description:
841    
842    I back-ported my implementation of IEEEReal.fromString from the
843    idlbasis-devel branch so that we can test it.
844    
845    Another small change is that ppDec tries to give more information
846    than just "<sig>" in the case of functors.  However, this code is
847    broken in some mysterious way if the functor's body's signature
848    has not been declared by ascription but gets inferred from the
849    implementation.  This needs fixing...
850    
851    ----------------------------------------------------------------------
852    Name: Matthias Blume
853    Date: 2002/05/31
854    Tag: blume-20020531-btrace-mode
855    Description:
856    
857    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
858    the wayside when I switched over to using Controls everywhere.)
859    
860    ----------------------------------------------------------------------
861    Name: Lal George
862    Date: 2002/05/23 12:21:40 EDT
863    Tag: george-20020523-visual-labels
864    Description:
865    
866    Labels are now displayed in the graphical output to make
867    the fall-through and target blocks obvious.
868    
869    ----------------------------------------------------------------------
870    Name: Matthias Blume
871    Date: 2002/05/22 11:03:00 EDT
872    Tag: blume-20020522-shrink
873    Description:
874    
875    John tweaked yesterday's fix for 1131 to handle an out-of-memory
876    situation that comes up when allocating huge arrays.
877    
878    ----------------------------------------------------------------------
879    Name: Matthias Blume
880    Date: 2002/05/21 16:00:00 EDT
881    Tag: Release_110_40
882    Description:
883    
884    New working release (110.40).  New bootfiles.
885    
886    [Also: John Reppy fixed GC bug 1131.]
887    
888    ----------------------------------------------------------------------
889    Name: Matthias Blume
890    Date: 2002/05/21 12:35:00 EDT
891    Tag: blume-20020521-cmdoc
892    Description:
893    
894    CM documentation update.
895    
896    ----------------------------------------------------------------------
897    Name: Matthias Blume
898    Date: 2002/05/21 10:55:00 EDT
899    Tag: blume-20020521-misc
900    Description:
901    
902    - John tweaked runtime to be silent on heap export (except when
903      GC messages are on).
904    - I added a few more things (cross-compiling versions of CMB) to
905      config/preloads (as suggestions).
906    
907    ----------------------------------------------------------------------
908    Name: Matthias Blume
909    Date: 2002/05/20 22:25:00 EDT
910    Tag: blume-20020520-controls
911    Description:
912    
913    - Added ControlUtil structure to control-lib.cm.
914    - Use it throughout.
915    - Used Controls facility to define MLRISC controls (as opposed to
916      registering MLRISC control ref cells with Controls after the fact)
917    - Fixed messed-up controls priorities.
918    
919    * Removed again all the stuff from config/preloads that one wouldn't
920      be able to preload at the time the initial heap image is built.
921      (Many libraries, e.g., CML, do not exist yet at this time.  The
922       only libraries that can be preloaded via config/preloads are those
923       that come bundled with the bootfiles.)
924    
925    ----------------------------------------------------------------------
926    Name: Matthias Blume
927    Date: 2002/05/20 10:59:00 EDT
928    Tag: blume-20020520-preloads
929    Description:
930    
931    Added a lot of commented-out suggestions for things to be included
932    in config/preloads.
933    
934    ----------------------------------------------------------------------
935    Name: Allen Leung
936    Date: 2002/05/18 14:20:00 EDT
937    Tag: leunga-20020518-mdl
938    Description:
939    
940    o Made the mdl tool stuff compile and run again.
941    o I've disabled all the stuff that depends on RTL specifications; they
942      are all badly broken anyway.
943    
944    ----------------------------------------------------------------------
945    Name: Matthias Blume
946    Date: 2002/05/17 16:49:00 EDT
947    Tag: blume-20020517-controls
948    Description:
949    
950    0. John Reppy made several modifications to the SML/NJ library.
951       In particular, there is a shiny new controls-lib.cm.
952    
953    1. Pushed new controls interface through compiler so that everything
954       compiles again.
955    
956    2. Added FormatComb and FORMAT_COMB to the CML version of the
957       SML/NJ library (so that CML compiles again).
958    
959    3. Modified init scripts because XXX_DEFAULT environment variables
960       are no longer with us.  (Boot-time initialization is now done
961       using the same environment variables that are also used for
962       startup-time initialization of controls.)
963    
964    ----------------------------------------------------------------------
965    Name: Lal George
966    Date: 2002/05/15 09:20:10 EDT
967    Tag: george-20020515-pseudo-op-decls
968    Description:
969    
970            All pseudo-ops emitted before the first segment declaration
971    such as TEXT, DATA, and BSS directives are assumed to be global
972    declarations and are emitted first in the assembly file. This is
973    useful in a number of situations where one has pseudo-ops that are not
974    specific to any segment, and also works around the constraint that one
975    cannot have client pseudo-ops in the TEXT segment.
976    
977            Because no segment is associated with these declarations it is
978    an error to allocate any space or objects before the first segment
979    directive and an exception will be raised. However, we cannot make
980    this check for client pseudo-ops.
981    
982            These top level declarations are a field in the CFG graph_info.
983    In theory you can continue to add to this field after the CFG has been
984    built -- provided you know what you are doing;-)
985    
986    ----------------------------------------------------------------------
987    Name: Matthias Blume
988    Date: 2002/05/13 16:40:00 EDT
989    Tag: blume-20020513-pp-etc
990    Description:
991    
992    A few minor bugfixes:
993    
994      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
995        (Bogus printouts for redefined bindings still occur.  Compiler
996        bug should no longer occur now.  We need to redo the prettyprinter
997        from scratch.)
998    
999      - CM pathname printer now also adds escape sequences for ( and )
1000    
1001      - commend and docu fixes for ml-nlffi
1002    
1003    ----------------------------------------------------------------------
1004    Name: Matthias Blume
1005    Date: 2002/05/10 16:40:00 EDT
1006    Tag: blume-20020510-erg-textio
1007    Description:
1008    
1009    Applied the following bugfix provided by Emden Gansner:
1010    
1011        Output is corrupted when outputSubstr is used rather than output.
1012    
1013        The problem occurs when a substring
1014    
1015            ss = (s, dataStart, dataLen)
1016    
1017        where dataStart > 0, fills a stream buffer with avail bytes left.
1018        avail bytes of s, starting at index dataStart, are copied into the
1019        buffer, the buffer is flushed, and then the remaining dataLen-avail
1020        bytes of ss are copied into the beginning of the buffer. Instead of
1021        starting this copy at index dataStart+avail in s, the current code
1022        starts the copy at index avail.
1023    
1024        Fix:
1025        In text-io-fn.sml, change line 695 from
1026             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
1027        to
1028             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
1029    
1030    ----------------------------------------------------------------------
1031    Name: Matthias Blume
1032    Date: 2002/04/12 13:55:00 EDT
1033    Tag: blume-20020412-assyntax
1034    Description:
1035    
1036    1. Grabbed newer assyntax.h from the XFree86 project.
1037    2. Fiddled with how to compile X86.prim.asm without warnings.
1038    3. (Very) Minor cleanup in CM.
1039    
1040    ----------------------------------------------------------------------
1041    Name: Matthias Blume
1042    Date: 2002/04/01 (no joke!) 17:07:00 EST
1043    Tag: blume-20020401-x86div
1044    Description:
1045    
1046    Added full support for div/mod/rem/quot on the x86, using the machine
1047    instruction's two results (without clumsily recomputing the remainder)
1048    directly where appropriate.
1049    
1050    Some more extensive power-of-two support was added to the x86 instruction
1051    selector (avoiding expensive divs, mods, and muls where they can be
1052    replaced with cheaper shifts and masks).  However, this sort of thing
1053    ought to be done earlier, e.g., within the CPS optimizer so that
1054    all architectures benefit from it.
1055    
1056    The compiler compiles to a fixed point, but changes might be somewhat
1057    fragile nevertheless.  Please, report any strange things that you might
1058    see wrt. div/mod/quot/rem...
1059    
1060    ----------------------------------------------------------------------
1061    Name: Matthias Blume
1062    Date: 2002/03/29 17:22:00
1063    Tag: blume-20020329-div
1064    Description:
1065    
1066    Fixed my broken div/mod logic.  Unfortunately, this means that the
1067    inline code for div/mod now has one more comparison than before.
1068    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
1069    The problem was with quotient = 0, because that alone does not tell
1070    us which way the rounding went.  One then has to look at whether
1071    remainder and divisor have the same sign...  :(
1072    
1073    Anyway, I replaced the bootfiles with fresh ones...
1074    
1075    ----------------------------------------------------------------------
1076    Name: Matthias Blume
1077    Date: 2002/03/29 14:10:00 EST
1078    Tag: blume-20020329-inlprims
1079    Description:
1080    
1081    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
1082    
1083    Primops have changed. This means that the bin/boot-file formats have
1084    changed as well.
1085    
1086    To make sure that there is no confusion, I made a new version.
1087    
1088    
1089    CHANGES:
1090    
1091    * removed REMT from mltree (remainder should never overflow).
1092    
1093    * added primops to deal with divisions of all flavors to the frontend
1094    
1095    * handled these primops all the way through so they map to their respective
1096      MLRISC support
1097    
1098    * used these primops in the implementation of Int, Int32, Word, Word32
1099    
1100    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
1101    
1102    * parameterized INLMIN, INLMAX, and INLABS by a numkind
1103    
1104    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
1105      floating point
1106    
1107    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
1108      and Real (but Real.abs maps to a separate floating-point-only primop)
1109    
1110    
1111    TODO items:
1112    
1113    * Hacked Alpha32 instruction selection, disabling the selection of REMx
1114      instructions because the machine instruction encoder cannot handle
1115      them.  (Hppa, PPC, and Sparc instruction selection did not handle
1116      REM in the first place, and REM is supported by the x86 machine coder.)
1117    
1118    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
1119      selection phase.  (The two can be streamlined because the hardware
1120      delivers both quotient and remainder at the same time anyway.)
1121    
1122    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
1123      (Currently the behavior is inconsistent both across architectures and
1124      wrt. the draft Basis spec.)
1125    
1126    * Word8 should eventually be handled natively, too.
1127    
1128    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
1129      as if there currently is no execution path that could trigger it in
1130      SML/NJ.  (The assumptions underlying functions arith and promotable do not
1131      hold for things like multiplication and division.)
1132    
1133    ----------------------------------------------------------------------
1134    Name: Matthias Blume
1135    Date: 2002/03/27 16:27:00 EST
1136    Tag: blume-20020327-mlrisc-divisions
1137    Description:
1138    
1139    Added support for all four division operations (ML's div, mod, quot,
1140    and rem) to MLRISC.  In the course of doing so, I also rationalized
1141    the naming (no more annoying switch-around of DIV and QUOT), by
1142    parameterizing the operation by div_rounding_mode (which can be either
1143    DIV_TO_ZERO or DIV_TO_NEGINF).
1144    
1145    The generic MLTreeGen functor takes care of compiling all four
1146    operations down to only round-to-zero div.
1147    
1148    Missing pieces:
1149    
1150      * Doing something smarter than relying on MLTreeGen on architectures
1151        like, e.g., the x86 where hardware division delivers both quotient and
1152        remainder at the same time.  With this, the implementation of the
1153        round-to-neginf operations could be further streamlined.
1154    
1155      * Remove inlining support for div/mod/rem from the frontend and replace it
1156        with primops that get carried through to the backend.  Do this for all
1157        int and word types.
1158    
1159    ----------------------------------------------------------------------
1160    Name: Matthias Blume
1161    Date: 2002/03/25 17:25:00 EST
1162    Tag: blume-20020325-divmod
1163    Description:
1164    
1165    I improved (hopefully without breaking them) the implementation of Int.div,
1166    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
1167    advantage of the following observations:
1168    
1169      Let  q = x quot y      r = x rem y
1170           d = x div  y      m = x mod y
1171    
1172    where "quot" is the round-to-zero version of integer division that
1173    hardware usually provides.  Then we have:
1174    
1175         r = x - q * y        where neither the * nor the - will overflow
1176         d = if q >= 0 orelse x = q * y then q else q - 1
1177                              where neither the * nor the - will overflow
1178         m = if q >= 0 orelse r = 0 then r else r + y
1179                              where the + will not overflow
1180    
1181    This results in substantial simplification of the generated code.
1182    The following table shows the number of CFG nodes and edges generated
1183    for
1184            fun f (x, y) = x OPER y
1185            (* with OPER \in div, mod, quot, rem *)
1186    
1187    
1188        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
1189        --------------------------------------------------------
1190         div |         24 |         39 |         12 |         16
1191         mod |         41 |         71 |         12 |         16
1192        quot |          8 |         10 |          8 |         10
1193         rem |         10 |         14 |          8 |         10
1194    
1195    
1196    ----------------------------------------------------------------------
1197    Name: Matthias Blume
1198    Date: 2002/03/25 22:06:00 EST
1199    Tag: blume-20020325-cprotobug
1200    Description:
1201    
1202    Fixed a bug in cproto (c prototype decoder).
1203    
1204    ----------------------------------------------------------------------
1205    Name: Matthias Blume
1206    Date: 2002/03/25 16:00:00 EST
1207    Tag: blume-20020325-raw-primops
1208    Description:
1209    
1210    I did some cleanup to Allen's new primop code and
1211    replaced yesterday's bootfiles with new ones.
1212    (But they are stored in the same place.)
1213    
1214    ----------------------------------------------------------------------
1215    Name: Matthias Blume
1216    Date: 2002/03/24 22:40:00 EST
1217    Tag: blume-20020324-bootfiles
1218    Description:
1219    
1220    Made the bootfiles that Allen asked for.
1221    
1222    ----------------------------------------------------------------------
1223    Name: Allen Leung
1224    Date: 2002/03/23 15:50:00 EST
1225    Tag: leunga-20020323-flint-cps-rcc-primops
1226    Description:
1227    
1228      1. Changes to FLINT primops:
1229    
1230        (* make a call to a C-function;
1231         * The primop carries C function prototype information and specifies
1232         * which of its (ML-) arguments are floating point. C prototype
1233         * information is for use by the backend, ML information is for
1234         * use by the CPS converter. *)
1235      | RAW_CCALL of { c_proto: CTypes.c_proto,
1236                       ml_args: ccall_type list,
1237                       ml_res_opt: ccall_type option,
1238                       reentrant : bool
1239                     } option
1240       (* Allocate uninitialized storage on the heap.
1241        * The record is meant to hold short-lived C objects, i.e., they
1242        * are not ML pointers.  With the tag, the representation is
1243        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
1244        *)
1245      | RAW_RECORD of {tag:bool,sz:int}
1246      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
1247    
1248      2.  These CPS primops are now overloaded:
1249    
1250           rawload of {kind:numkind}
1251           rawstore of {kind:numkind}
1252    
1253          The one argument form is:
1254    
1255             rawload {kind} address
1256    
1257          The two argument form is:
1258    
1259             rawload {kind} [ml object, byte-offset]
1260    
1261      3. RAW_CCALL/RCC now takes two extra arguments:
1262    
1263         a. The first is whether the C call is reentrant, i.e., whether
1264            ML state should be saved and restored.
1265         b. The second argument is a string argument specifying the name of
1266            library and the C function.
1267    
1268         These things are currently not handled in the code generator, yet.
1269    
1270      4. In CProto,
1271    
1272         An encoding type of "bool" means "ml object" and is mapped into
1273         C prototype of PTR.  Note that "bool" is different than "string",
1274         even though "string" is also mapped into PTR, because "bool"
1275         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
1276    
1277      5. Pickler/unpicker
1278    
1279         Changed to handle RAW_RECORD and newest RAW_CCALL
1280    
1281      6. MLRiscGen,
1282    
1283         1. Changed to handle the new rawload/rawstore/rawrecord operators.
1284         2. Code for handling C Calls has been moved to a new module CPSCCalls,
1285            in the file CodeGen/cpscompile/cps-c-calls.sml
1286    
1287      7. Added the conditional move operator
1288    
1289             condmove of branch
1290    
1291         to cps.  Generation of this is still buggy so it is currently
1292         disabled.
1293    
1294    ----------------------------------------------------------------------
1295    Name: Lal George
1296    Date: 2002/03/22 14:18:25 EST
1297    Tag: george-20020322-cps-branch-prob
1298    Description:
1299    
1300    Implemented the Ball-Larus branch prediction-heuristics, and
1301    incorporated graphical viewers for control flow graphs.
1302    
1303    Ball-Larus Heuristics:
1304    ---------------------
1305    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1306    
1307    By design it uses the Dempster-Shafer theory for combining
1308    probabilities.  For example, in the function:
1309    
1310        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1311    
1312    the ball-larus heuristics predicts that the n=0 is unlikely
1313    (OH-heuristic), and the 'then' branch is unlikely because of the
1314    RH-heuristic -- giving the 'then' branch an even lower combined
1315    probability using the Dempster-Shafer theory.
1316    
1317    Finally, John Reppy's loop analysis in MLRISC, further lowers the
1318    probability of the 'then' branch because of the loop in the else
1319    branch.
1320    
1321    
1322    Graphical Viewing:
1323    ------------------
1324    I merely plugged in Allen's graphical viewers into the compiler. The
1325    additional code is not much. At the top level, saying:
1326    
1327            Control.MLRISC.getFlag "cfg-graphical-view" := true;
1328    
1329    will display the graphical view of the control flow graph just before
1330    back-patching.  daVinci must be in your path for this to work. If
1331    daVinci is not available, then the default viewer can be changed
1332    using:
1333    
1334            Control.MLRISC.getString "viewer"
1335    
1336    which can be set to "dot" or "vcg" for the corresponding viewers. Of
1337    course, these viewers must be in your path.
1338    
1339    The above will display the compilation unit at the level of clusters,
1340    many of which are small, boring, and un-interesting. Also setting:
1341    
1342            Control.MLRISC.getInt "cfg-graphical-view_size"
1343    
1344    will display clusters that are larger than the value set by the above.
1345    
1346    
1347    ----------------------------------------------------------------------
1348    Name: Matthias Blume
1349    Date: 2002/03/21 22:20:00 EST
1350    Tag: blume-20020321-kmp-bugfix
1351    Description:
1352    
1353    Changed the interface to the KMP routine in PreString and fixed
1354    a minor bug in one place where it was used.
1355    
1356    ----------------------------------------------------------------------
1357    Name: Allen Leung
1358    Date: 2002/03/21 20:30:00 EST
1359    Tag: leunga-20020321-cfg
1360    Description:
1361    
1362      Fixed a potential problem in cfg edge splitting.
1363    
1364    ----------------------------------------------------------------------
1365    Name: Allen Leung
1366    Date: 2002/03/21 17:15:00 EST
1367    Tag: leunga-20020321-x86-fp-cfg
1368    Description:
1369    
1370      1. Recoded the buggy parts of x86-fp.
1371    
1372         a. All the block reordering code has been removed.
1373            We now depend on the block placement phases to do this work.
1374    
1375         b. Critical edge splitting code has been simplified and moved into the
1376            CFG modules, as where they belong.
1377    
1378         Both of these were quite buggy and complex.  The code is now much, much
1379         simpler.
1380    
1381      2. X86 backend.
1382    
1383         a. Added instructions for 64-bit support.  Instruction selection for
1384            64-bit has not been committed, however, since that
1385            requires changes to MLTREE which haven't been approved by
1386            Lal and John.
1387    
1388         b. Added support for FUCOMI and FUCOMIP when generating code for
1389            PentiumPro and above.  We only generate these instructions in
1390            the fast-fp mode.
1391    
1392         c. Added cases for JP and JNP in X86FreqProps.
1393    
1394      3. CFG
1395    
1396         CFG now has a bunch of methods for edge splitting and merging.
1397    
1398      4. Machine description.
1399    
1400         John's simplification of MLTREE_BASIS.fcond broke a few machine
1401         description things:
1402    
1403         rtl-build.{sig,sml} and hppa.mdl fixed.
1404    
1405         NOTE: the machine description stuff in the repository is still broken.
1406               Again, I can't put my fixes in because that involves
1407               changes to MLTREE.
1408    
1409    ----------------------------------------------------------------------
1410    Name: Matthias Blume
1411    Date: 2002/03/20 15:55:00 EST
1412    Tag: blume-20020320-kmp
1413    Description:
1414    
1415    Implemented Knuth-Morris-Pratt string matching in PreString and used
1416    it for String.isSubstring, Substring.isSubstring, and
1417    Substring.position.
1418    
1419    (Might need some stress-testing.  Simple examples worked fine.)
1420    
1421    ----------------------------------------------------------------------
1422    Name: Matthias Blume
1423    Date: 2002/03/19 16:37:00 EST
1424    Tag: blume-20020319-witnesses
1425    Description:
1426    
1427    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1428    
1429    This implements a generic mechanism for changing constness qualifiers
1430    anywhere within big C types without resorting to outright "casts".
1431    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1432    modify the constness at the outermost level.)
1433    The implementation of "convert" is based on the idea of "witness"
1434    values -- values that are not used by the operation but whose types
1435    "testify" to their applicability.  On the implementation side, "convert"
1436    is simply a projection (returning its second curried argument).  With
1437    cross-module inlining, it should not result in any machine code being
1438    generated.
1439    
1440    ----------------------------------------------------------------------
1441    Name: Matthias Blume
1442    Date: 2002/03/15 16:40:00 EST
1443    Tag: blume-20020315-basis
1444    Description:
1445    
1446    Provided (preliminary?) implementations for
1447    
1448      {String,Substring}.{concatWith,isSuffix,isSubstring}
1449    
1450    and
1451    
1452      Substring.full
1453    
1454    Those are in the Basis spec but they were missing in SML/NJ.
1455    
1456    ----------------------------------------------------------------------
1457    Name: Matthias Blume
1458    Date: 2002/03/14 21:30:00 EST
1459    Tag: blume-20020314-controls
1460    Description:
1461    
1462    Controls:
1463    ---------
1464    
1465    1. Factored out the recently-added Controls : CONTROLS stuff and put
1466       it into its own library $/controls-lib.cm.  The source tree for
1467       this is under src/smlnj-lib/Controls.
1468    
1469    2. Changed the names of types and functions in this interface, so they
1470       make a bit more "sense":
1471    
1472          module -> registry
1473          'a registry -> 'a group
1474    
1475    3. The interface now deals in ref cells only.  The getter/setter interface
1476       is (mostly) gone.
1477    
1478    4. Added a function that lets one register an already-existing ref cell.
1479    
1480    5. Made the corresponding modifications to the rest of the code so that
1481       everything compiles again.
1482    
1483    6. Changed the implementation of Controls.MLRISC back to something closer
1484       to the original.  In particular, this module (and therefore MLRISC)
1485       does not depend on Controls.  There now is some link-time code in
1486       int-sys.sml that registers the MLRISC controls with the Controls
1487       module.
1488    
1489    CM:
1490    ---
1491    
1492      * One can now specify the lambda-split aggressiveness in init.cmi.
1493    
1494    ----------------------------------------------------------------------
1495    Name: Allen Leung
1496    Date: 2002/03/13 17:30:00 EST
1497    Tag: leunga-20020313-x86-fp-unary
1498    Description:
1499    
1500    Bug fix for:
1501    
1502    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1503    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1504    > - fun f(x,(y,z)) = Real.~ y;
1505    > [autoloading]
1506    > [autoloading done]
1507    >       fchsl   (%eax), 184(%esp)
1508    > Error: MLRisc bug: X86MCEmitter.emitInstr
1509    >
1510    > uncaught exception Error
1511    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1512    
1513    The problem was that the code generator did not generate any fp registers
1514    in this case, and the ra didn't know that it needed to run the X86FP phase to
1515    translate the pseudo fp instruction.   This only happened with unary fp
1516    operators in certain situations.
1517    
1518    ----------------------------------------------------------------------
1519    Name: Matthias Blume
1520    Date: 2002/03/13 14:00:00 EST
1521    Tag: blume-20020313-overload-etc
1522    Description:
1523    
1524    1. Added _overload as a synonym for overload for backward compatibility.
1525       (Control.overloadKW must be true for either version to be accepted.)
1526    
1527    2. Fixed bug in install script that caused more things to be installed
1528       than what was requested in config/targets.
1529    
1530    3. Made CM aware of the (_)overload construct so that autoloading
1531       works.
1532    
1533    ----------------------------------------------------------------------
1534    Name: Matthias Blume
1535    Date: 2002/03/12 22:03:00 EST
1536    Tag: blume-20020312-url
1537    Description:
1538    
1539    Forgot to update BOOT and srcarchiveurl.
1540    
1541    ----------------------------------------------------------------------
1542    Name: Matthias Blume
1543    Date: 2002/03/12 17:30:00 EST
1544    Tag: blume-20020312-version110392
1545    Description:
1546    
1547    Yet another version number bump (because of small changes to the
1548    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1549    
1550    Changes:
1551    
1552      The new pid generation scheme described a few weeks ago was overly
1553      complicated.  I implemented a new mechanism that is simpler and
1554      provides a bit more "stability":  Once CM has seen a compilation
1555      unit, it keeps its identity constant (as long as you do not delete
1556      those crucial CM/GUID/* files).  This means that when you change
1557      an interface, compile, then go back to the old interface, and
1558      compile again, you arrive at the original pid.
1559    
1560      There now also is a mechanism that instructs CM to use the plain
1561      environment hash as a module's pid (effectively making its GUID
1562      the empty string).  For this, "noguid" must be specified as an
1563      option to the .sml file in question within its .cm file.
1564      This is most useful for code that is being generated by tools such
1565      as ml-nlffigen (because during development programmers tend to
1566      erase the tool's entire output directory tree including CM's cached
1567      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1568      revert to the old, broken behavior of SML/NJ, but in specific cases
1569      where there is no danger of interface confusion, its use is ok
1570      (I think).
1571    
1572      ml-nlffigen by default generates "noguid" annotations.  They can be
1573      turned off by specifying -guid in its command line.
1574    
1575    ----------------------------------------------------------------------
1576    Name: Lal George
1577    Date: 2002/03/12 12 14:42:36 EST
1578    Tag: george-20020312-frequency-computation
1579    Description:
1580    
1581    Integrated jump chaining and static block frequency into the
1582    compiler. More details and numbers later.
1583    
1584    ----------------------------------------------------------------------
1585    Name: Lal George
1586    Date: 2002/03/11 11 22:38:53 EST
1587    Tag: george-20020311-jump-chain-elim
1588    Description:
1589    
1590    Tested the jump chain elimination on all architectures (except the
1591    hppa).  This is on by default right now and is profitable for the
1592    alpha and x86, however, it may not be profitable for the sparc and ppc
1593    when compiling the compiler.
1594    
1595    The gc test will typically jump to a label at the end of the cluster,
1596    where there is another jump to an external cluster containing the actual
1597    code to invoke gc. This is to allow factoring of common gc invocation
1598    sequences. That is to say, we generate:
1599    
1600            f:
1601               testgc
1602               ja   L1      % jump if above to L1
1603    
1604            L1:
1605               jmp L2
1606    
1607    
1608    After jump chain elimination the 'ja L1' instructions is converted to
1609    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1610    up being implemented in their long form (if L2 is far away) using:
1611    
1612            jbe     L3      % jump if below or equal to L3
1613            jmp     L2
1614         L3:
1615            ...
1616    
1617    
1618    For large compilation units L2  may be far away.
1619    
1620    
1621    ----------------------------------------------------------------------
1622    Name: Matthias Blume
1623    Date: 2002/03/11 13:30:00 EST
1624    Tag: blume-20020311-mltreeeval
1625    Description:
1626    
1627    A functor parameter was missing.
1628    
1629    ----------------------------------------------------------------------
1630    Name: Allen Leung
1631    Date: 2002/03/11 10:30:00 EST
1632    Tag: leunga-20020311-runtime-string0
1633    Description:
1634    
1635       The representation of the empty string now points to a
1636    legal null terminated C string instead of unit.  It is now possible
1637    to convert an ML string into C string with InlineT.CharVector.getData.
1638    This compiles into one single machine instruction.
1639    
1640    ----------------------------------------------------------------------
1641    Name: Allen Leung
1642    Date: 2002/03/10 23:55:00 EST
1643    Tag: leunga-20020310-x86-call
1644    Description:
1645    
1646       Added machine generation for CALL instruction (relative displacement mode)
1647    
1648    ----------------------------------------------------------------------
1649    Name: Matthias Blume
1650    Date: 2002/03/08 16:05:00
1651    Tag: blume-20020308-entrypoints
1652    Description:
1653    
1654    Version number bumped to 110.39.1.  NEW BOOTFILES!
1655    
1656    Entrypoints: non-zero offset into a code object where execution should begin.
1657    
1658    - Added the notion of an entrypoint to CodeObj.
1659    - Added reading/writing of entrypoint info to Binfile.
1660    - Made runtime system bootloader aware of entrypoints.
1661    - Use the address of the label of the first function given to mlriscGen
1662      as the entrypoint.  This address is currently always 0, but it will
1663      not be 0 once we turn on block placement.
1664    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1665      with entry points) from mlriscGen.
1666    
1667    ----------------------------------------------------------------------
1668    Name: Allen Leung
1669    Date: 2002/03/07 20:45:00 EST
1670    Tag: leunga-20020307-x86-cmov
1671    Description:
1672    
1673       Bug fixes for CMOVcc on x86.
1674    
1675       1. Added machine code generation for CMOVcc
1676       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1677       3. CMOVcc cannot have an immediate operand as argument.
1678    
1679    ----------------------------------------------------------------------
1680    Name: Matthias Blume
1681    Date: 2002/03/07 16:15:00 EST
1682    Tag: blume-20020307-controls
1683    Description:
1684    
1685    This is a very large but mostly boring patch which makes (almost)
1686    every tuneable compiler knob (i.e., pretty much everything under
1687    Control.* plus a few other things) configurable via both the command
1688    line and environment variables in the style CM did its configuration
1689    until now.
1690    
1691    Try starting sml with '-h' (or, if you are brave, '-H')
1692    
1693    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1694    implements the underlying generic mechanism.
1695    
1696    The interface to some of the existing such facilities has changed somewhat.
1697    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1698    (The getFoo interface is still there for backward-compatibility, but its
1699    use is deprecated.)
1700    
1701    The ml-build script passes -Cxxx=yyy command-line arguments through so
1702    that one can now twiddle the compiler settings when using this "batch"
1703    compiler.
1704    
1705    TODO items:
1706    
1707    We should go through and throw out all controls that are no longer
1708    connected to anything.  Moreover, we should go through and provide
1709    meaningful (and correct!) documentation strings for those controls
1710    that still are connected.
1711    
1712    Currently, multiple calls to Controls.new are accepted (only the first
1713    has any effect).  Eventually we should make sure that every control
1714    is being made (via Controls.new) exactly once.  Future access can then
1715    be done using Controls.acc.
1716    
1717    Finally, it would probably be a good idea to use the getter-setter
1718    interface to controls rather than ref cells.  For the time being, both
1719    styles are provided by the Controls module, but getter-setter pairs are
1720    better if thread-safety is of any concern because they can be wrapped.
1721    
1722    *****************************************
1723    
1724    One bug fix: The function blockPlacement in three of the MLRISC
1725    backpatch files used to be hard-wired to one of two possibilities at
1726    link time (according to the value of the placementFlag).  But (I
1727    think) it should rather sense the flag every time.
1728    
1729    *****************************************
1730    
1731    Other assorted changes (by other people who did not supply a HISTORY entry):
1732    
1733    1. the cross-module inliner now works much better (Monnier)
1734    2. representation of weights, frequencies, and probabilities in MLRISC
1735       changed in preparation of using those for weighted block placement
1736       (Reppy, George)
1737    
1738    ----------------------------------------------------------------------
1739    Name: Lal George
1740    Date: 2002/03/07 14:44:24 EST 2002
1741    Tag: george-20020307-weighted-block-placement
1742    
1743    Tested the weighted block placement optimization on all architectures
1744    (except the hppa) using AMPL to generate the block and edge frequencies.
1745    Changes were required in the machine properties to correctly
1746    categorize trap instructions. There is an MLRISC flag
1747    "weighted-block-placement" that can be used to enable weighted block
1748    placement, but this will be ineffective without block/edge
1749    frequencies (coming soon).
1750    
1751    
1752    ----------------------------------------------------------------------
1753    Name: Lal George
1754    Date: 2002/03/05 17:24:48 EST
1755    Tag: george-20020305-linkage-cluster
1756    
1757    In order to support the block placement optimization, a new cluster
1758    is generated as the very first cluster (called the linkage cluster).
1759    It contains a single jump to the 'real' entry point for the compilation
1760    unit. Block placement has no effect on the linkage cluster itself, but
1761    all the other clusters  have full freedom in the manner in which they
1762    reorder blocks or functions.
1763    
1764    On the x86 the typical linkage code that is generated is:
1765       ----------------------
1766            .align 2
1767       L0:
1768            addl    $L1-L0, 72(%esp)
1769            jmp     L1
1770    
1771    
1772            .align  2
1773       L1:
1774       ----------------------
1775    
1776    72(%esp) is the memory location for the stdlink register. This
1777    must contain the address of the CPS function being called. In the
1778    above example, it contains the address of  L0; before
1779    calling L1 (the real entry point for the compilation unit), it
1780    must contain the address for L1, and hence
1781    
1782            addl $L1-L0, 72(%esp)
1783    
1784    I have tested this on all architectures except the hppa.The increase
1785    in code size is of course negligible
1786    
1787    ----------------------------------------------------------------------
1788    Name: Allen Leung
1789    Date: 2002/03/03 13:20:00 EST
1790    Tag: leunga-20020303-mlrisc-tools
1791    
1792      Added #[ ... ] expressions to mlrisc tools
1793    
1794    ----------------------------------------------------------------------
1795    Name: Matthias Blume
1796    Date: 2002/02/27 12:29:00 EST
1797    Tag: blume-20020227-cdebug
1798    Description:
1799    
1800    - made types in structure C and C_Debug to be equal
1801    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1802    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1803    
1804    ----------------------------------------------------------------------
1805    Name: Matthias Blume
1806    Date: 2002/02/26 12:00:00 EST
1807    Tag: blume-20020226-ffi
1808    Description:
1809    
1810    1. Fixed a minor bug in CM's "noweb" tool:
1811       If numbering is turned off, then truly don't number (i.e., do not
1812       supply the -L option to noweb).  The previous behavior was to supply
1813       -L'' -- which caused noweb to use the "default" line numbering scheme.
1814       Thanks to Chris Richards for pointing this out (and supplying the fix).
1815    
1816    2. Once again, I reworked some aspects of the FFI:
1817    
1818       A. The incomplete/complete type business:
1819    
1820       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1821         gone!
1822       - ML types representing an incomplete type are now *equal* to
1823         ML types representing their corresponding complete types (just like
1824         in C).  This is still safe because ml-nlffigen will not generate
1825         RTTI for incomplete types, nor will it generate functions that
1826         require access to such RTTI.   But when ML code generated from both
1827         incomplete and complete versions of the C type meet, the ML types
1828         are trivially interoperable.
1829    
1830         NOTE:  These changes restore the full generality of the translation
1831         (which was previously lost when I eliminated functorization)!
1832    
1833       B. Enum types:
1834    
1835       - Structure C now has a type constructor "enum" that is similar to
1836         how the "su" constructor works.  However, "enum" is not a phantom
1837         type because each "T enum" has values (and is isomorphic to
1838         MLRep.Signed.int).
1839       - There are generic access operations for enum objects (using
1840         MLRep.Signed.int).
1841       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1842         * The structure contains the definition of type "mlrep" (the ML-side
1843         representation type of the enum).  Normally, mlrep is the same
1844         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1845         then mlrep will be defined as a datatype -- thus facilitating
1846         pattern matching on mlrep values.
1847         ("-ec" will be suppressed if there are duplicate values in an
1848          enumeration.)
1849         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1850         will be generated for each C enum constant xxx.
1851         * Conversion functions m2i and i2m convert between mlrep and
1852         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1853         * Coversion functions c and ml convert between mlrep and "tag enum".
1854         * Access functions (get/set) fetch and store mlrep values.
1855       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1856         enumerations are merged into one single enumeration represented by
1857         structure E_'.
1858    
1859    ----------------------------------------------------------------------
1860    Name: Allen Leung
1861    Date: 2002/02/25 04:45:00 EST
1862    Tag: leunga-20020225-cps-spill
1863    
1864    This is a new implementation of the CPS spill phase.
1865    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1866    In case of problems, replace it with the old file spill.sml
1867    
1868    The current compiler runs into some serious performance problems when
1869    constructing a large record.  This can happen when we try to compile a
1870    structure with many items.  Even a very simple structure like the following
1871    makes the compiler slow down.
1872    
1873        structure Foo = struct
1874           val x_1 = 0w1 : Word32.int
1875           val x_2 = 0w2 : Word32.int
1876           val x_3 = 0w3 : Word32.int
1877           ...
1878           val x_N = 0wN : Word32.int
1879        end
1880    
1881    The following table shows the compile time, from N=1000 to N=4000,
1882    with the old compiler:
1883    
1884    N
1885    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1886           MLRISC ra                               0.06u  0.00s  0.05g
1887              (spills = 0 reloads = 0)
1888           TOTAL                                   0.63u  0.07s  0.21g
1889    
1890    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1891           MLRISC ra                               5.68u  0.59s  3.93g
1892              (spills = 0 reloads = 0)
1893           TOTAL                                   14.71u  0.99s  4.81g
1894    
1895    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1896           MLRISC ra                               5.54u  0.65s  3.91g
1897              (spills = 543 reloads = 1082)
1898           TOTAL                                   65.40u  3.13s  6.00g
1899    
1900    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1901           MLRISC ra                               0.80u  0.10s  0.55g
1902              (spills = 42 reloads = 84)
1903           TOTAL                                   129.42u  5.10s  4.13g
1904    
1905    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1906           MLRISC ra                               2.69u  0.27s  1.38g
1907              (spills = 62 reloads = 124)
1908           TOTAL                                   682.48u  19.61s  13.99g
1909    
1910    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1911           MLRISC ra                               4.96u  0.27s  2.72g
1912              (spills = 85 reloads = 170)
1913           TOTAL                                   2375.26u  57.21s  48.00g
1914    
1915    As you can see the old cps spill module suffers from some serious
1916    performance problem.  But since I cannot decipher the old code fully,
1917    instead of patching the problems up, I'm reimplementing it
1918    with a different algorithm.  The new code is more modular,
1919    smaller when compiled, and substantially faster
1920    (O(n log n) time and O(n) space).  Timing of the new spill module:
1921    
1922    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1923          MLRISC ra                               0.25u  0.02s  0.15g
1924             (spills=1 reloads=3)
1925          TOTAL                                   7.74u  0.34s  1.62g
1926    
1927    Implementation details:
1928    
1929    As far as I can tell, the purpose of the CPS spill module is to make sure the
1930    number of live variables at any program point (the bandwidth)
1931    does not exceed a certain limit, which is determined by the
1932    size of the spill area.
1933    
1934    When the bandwidth is too large, we decrease the register pressure by
1935    packing live variables into spill records.  How we achieve this is
1936    completely different than what we did in the old code.
1937    
1938    First, there is something about the MLRiscGen code generator
1939    that we should be aware of:
1940    
1941    o MLRiscGen performs code motion!
1942    
1943       In particular, it will move floating point computations and
1944       address computations involving only the heap pointer to
1945       their use sites (if there is only a single use).
1946       What this means is that if we have a CPS record construction
1947       statement
1948    
1949           RECORD(k,vl,w,e)
1950    
1951       we should never count the new record address w as live if w
1952       has only one use (which is often the case).
1953    
1954       We should do something similar to floating point, but the transformation
1955       there is much more complex, so I won't deal with that.
1956    
1957    Secondly, there are now two new cps primops at our disposal:
1958    
1959     1. rawrecord of record_kind option
1960        This pure operator allocates some uninitialized storage from the heap.
1961        There are two forms:
1962    
1963         rawrecord NONE [INT n]  allocates a tagless record of length n
1964         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1965                                     and initializes the tag.
1966    
1967     2. rawupdate of cty
1968          rawupdate cty (v,i,x)
1969          Assigns to x to the ith component of record v.
1970          The storelist is not updated.
1971    
1972    We use these new primops for both spilling and increment record construction.
1973    
1974     1. Spilling.
1975    
1976        This is implemented with a linear scan algorithm (but generalized
1977        to trees).  The algorithm will create a single spill record at the
1978        beginning of the cps function and use rawupdate to spill to it,
1979        and SELECT or SELp to reload from it.  So both spills and reloads
1980        are fine-grain operations.  In contrast, in the old algorithm
1981        "spills" have to be bundled together in records.
1982    
1983        Ideally, we should sink the spill record construction to where
1984        it is needed.  We can even split the spill record into multiple ones
1985        at the places where they are needed.  But CPS is not a good
1986        representation for global code motion, so I'll keep it simple and
1987        am not attempting this.
1988    
1989     2. Incremental record construction (aka record splitting).
1990    
1991        Long records with many component values which are simulatenously live
1992        (recall that single use record addresses are not considered to
1993         be live) are constructed with rawrecord and rawupdate.
1994        We allocate space on the heap with rawrecord first, then gradually
1995        fill it in with rawupdate.  This is the technique suggested to me
1996        by Matthias.
1997    
1998        Some restrictions on when this is applicable:
1999        1. It is not a VECTOR record.  The code generator currently does not handle
2000           this case. VECTOR record uses double indirection like arrays.
2001        2. All the record component values are defined in the same "basic block"
2002           as the record constructor.  This is to prevent speculative
2003           record construction.
2004    
2005    ----------------------------------------------------------------------
2006    Name: Allen Leung
2007    Date: 2002/02/22 01:02:00 EST
2008    Tag: leunga-20020222-mlrisc-tools
2009    
2010    Minor bug fixes in the parser and rewriter
2011    
2012    ----------------------------------------------------------------------
2013    Name: Allen Leung
2014    Date: 2002/02/21 20:20:00 EST
2015    Tag: leunga-20020221-peephole
2016    
2017    Regenerated the peephole files.  Some contained typos in the specification
2018    and some didn't compile because of pretty printing bugs in the old version
2019    of 'nowhere'.
2020    
2021    ----------------------------------------------------------------------
2022    Name: Allen Leung
2023    Date: 2002/02/19 20:20:00 EST
2024    Tag: leunga-20020219-mlrisc-tools
2025    Description:
2026    
2027       Minor bug fixes to the mlrisc-tools library:
2028    
2029       1.  Fixed up parsing colon suffixed keywords
2030       2.  Added the ability to shut the error messages up
2031       3.  Reimplemented the pretty printer and fixed up/improved
2032           the pretty printing of handle and -> types.
2033       4.  Fixed up generation of literal symbols in the nowhere tool.
2034       5.  Added some SML keywords to to sml.sty
2035    
2036    ----------------------------------------------------------------------
2037    Name: Matthias Blume
2038    Date: 2002/02/19 16:20:00 EST
2039    Tag: blume-20020219-cmffi
2040    Description:
2041    
2042    A wild mix of changes, some minor, some major:
2043    
2044    * All C FFI-related libraries are now anchored under $c:
2045        $/c.cm      --> $c/c.cm
2046        $/c-int.cm  --> $c/internals/c-int.cm
2047        $/memory.cm --> $c/memory/memory.cm
2048    
2049    * "make" tool (in CM) now treats its argument pathname slightly
2050      differently:
2051        1. If the native expansion is an absolute name, then before invoking
2052           the "make" command on it, CM will apply OS.Path.mkRelative
2053           (with relativeTo = OS.FileSys.getDir()) to it.
2054        2. The argument will be passed through to subsequent phases of CM
2055           processing without "going native".  In particular, if the argument
2056           was an anchored path, then "make" will not lose track of that anchor.
2057    
2058    * Compiler backends now "know" their respective C calling conventions
2059      instead of having to be told about it by ml-nlffigen.  This relieves
2060      ml-nlffigen from one of its burdens.
2061    
2062    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
2063    
2064    * Export C_DEBUG and C_Debug from $c/c.cm.
2065    
2066    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
2067      subtyping relationship between incomplete pointers and their complete
2068      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
2069      with the convention of instantiating 'o with ('t, 'c) obj whenever
2070      the pointer target type is complete.  In the incomplete case, 'o
2071      will be instantiated with some "'c iobj" -- a type obtained by
2072      using one of the functors PointerToIncompleteType or PointerToCompleteType.
2073    
2074      Operations that work on both incomplete and complete pointer types are
2075      typed as taking an 'o ptr while operations that require the target to
2076      be known are typed as taking some ('t, 'c) obj ptr.
2077    
2078      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
2079      where void is an eqtype without any values.  This makes it possible
2080      to work on voidptr values using functions meant to operate on light
2081      incomplete pointers.
2082    
2083    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
2084      been vastly simplified.
2085    
2086    ----------------------------------------------------------------------
2087    Name: Matthias Blume
2088    Date: 2002/02/19 10:48:00 EST
2089    Tag: blume-20020219-pqfix
2090    Description:
2091    
2092    Applied Chris Okasaki's bug fix for priority queues.
2093    
2094    ----------------------------------------------------------------------
2095    Name: Matthias Blume
2096    Date: 2002/02/15 17:05:00
2097    Tag: Release_110_39
2098    Description:
2099    
2100    Last-minute retagging is becoming a tradition... :-(
2101    
2102    This is the working release 110.39.
2103    
2104    ----------------------------------------------------------------------
2105    Name: Matthias Blume
2106    Date: 2002/02/15 16:00:00 EST
2107    Tag: Release_110_39-orig
2108    Description:
2109    
2110    Working release 110.39.  New bootfiles.
2111    
2112    (Update: There was a small bug in the installer so it wouldn't work
2113    with all shells.  So I retagged. -Matthias)
2114    
2115    ----------------------------------------------------------------------
2116    Name: Matthias Blume
2117    Date: 2002/02/15 14:17:00 EST
2118    Tag: blume-20020215-showbindings
2119    Description:
2120    
2121    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
2122    the latter can be useful for exploring what bindings are available at
2123    the interactive prompt.  (The first function returns only the list
2124    of symbols that are really bound, the second prints those but also the
2125    ones that CM's autoloading mechanism knows about.)
2126    
2127    ----------------------------------------------------------------------
2128  Name: Matthias Blume  Name: Matthias Blume
2129  Date: 2002/02/15 12:08:00 EST  Date: 2002/02/15 12:08:00 EST
2130  Tag: blume-20020215-iptrs  Tag: blume-20020215-iptrs

Legend:
Removed from v.1067  
changed lines
  Added in v.1480

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