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 628, Tue Apr 25 21:03:34 2000 UTC revision 1485, Tue May 11 16:36:07 2004 UTC
# Line 8  Line 8 
8  The form of an entry should be:  The form of an entry should be:
9    
10  Name:  Name:
11  Date:  Date: yyyy/mm/dd
12  Tag: <post-commit CVS tag>  Tag: <post-commit CVS tag>
13  Description:  Description:
14    
15    ----------------------------------------------------------------------
16    Name: Matthias Blume (blume (at) tti - c (dot) org)
17    Date: 2004/05/11 11:35:00 CDT
18    Tag: blume-20040511-allsource
19    Description:
20    
21    The installer now honors the "src-smlnj" target again, although its meaning
22    has changed from "all sources required for the compiler" to "all sources
23    the installer knows about".  In other words, if you enable "src-smlnj"
24    in the "targets" file, then the installer will pull in sources for
25    everything.  (Notice that this refers to source code only.  Compiled
26    code is still only installed for modules that were requested explicitly
27    or which are required for other modules that were requested explicitly.)
28    
29    ----------------------------------------------------------------------
30    Name: Matthias Blume (blume (at) tti - c (dot) org)
31    Date: 2004/04/23 17:40:00 CDT
32    Tag: blume-20040423-ieee-scan
33    Description:
34    
35    Fixed IEEEReal.scan (and .fromString) so that if there is an overflow
36    in the exponent calculation we get INF or ZERO (depending on the mantissa
37    and the sign of the exponent).
38    
39    ----------------------------------------------------------------------
40    Name: Matthias Blume (blume (at) tti - c (dot) org)
41    Date: 2004/04/23 10:40:00 CDT
42    Tag: blume-20040423-ml-build
43    Description:
44    
45    The ml-build script now terminates with a non-0 status when something
46    goes wrong.
47    
48    ----------------------------------------------------------------------
49    Name: Matthias Blume (blume (at) tti - c (dot) org)
50    Date: 2004/04/22 16:35:00 CDT
51    Tag: blume-20040422-Option
52    Description:
53    
54    Made exception Option to be the same as exception Option.Option
55    (as it should be).
56    
57    ----------------------------------------------------------------------
58    Name: Allen Leung (leunga (at) reservoir (dot) com)
59    Date: 2004/03/19 14:40:00 EST
60    Tag: leunga-20040319-cygwin-nlffi
61    Description:
62    
63    Fixed the runtime so that ml-nlffi-lib runs on the cygwin version
64    of SML/NJ.  The problem is that
65    
66       lib = dlopen(NULL, ...)
67       f   = dlsym(lib, "malloc");
68    
69    does not work on Windows unless we explicitly export symbols
70    such as 'malloc' during linking.  We fixed this by explicitly
71    exporting the required symbols with the magic gcc incantation:
72    
73        -Wl,--export-all cygwin.def
74    
75    where cygwin.def is a file containing all the symbols that
76    we wish to export.
77    
78    I suspect this is a Windows problem and we'll have to
79    do the same (somehow with windows compilers) when
80    we build the native win32 version with the system
81    calls LoadLibrary/GetProcAddress.
82    
83    ----------------------------------------------------------------------
84    Name: Matthias Blume (blume (at) tti - c (dot) org)
85    Date: 2004/03/04 16:35:00 CST
86    Tag: blume-20040304-intinf-fmt
87    Description:
88    
89    Fixed problem with IntInf.fmt (sign would show up on the right instead
90    of on the left for BIN, OCT, and HEX).
91    
92    ----------------------------------------------------------------------
93    Name: Matthias Blume (blume (at) tti - c (dot) org)
94    Date: 2004/03/04 11:25:00 CST
95    Tag: blume-20040304-symlinks
96    Description:
97    
98    Fixed problem with installer script (unix only) where bin/ml-yacc and
99    friends pointed (via symlinks) to absolute locations instead of just
100    .run-sml.  This was reported by Vesa A Norrman.
101    
102    ----------------------------------------------------------------------
103    Name: Matthias Blume (blume (at) tti - c (dot) org)
104    Date: 2004/02/13 14:50:00 CST
105    Tag: Release_110_45
106    Description:
107    
108    New working version (110.45). New bootfiles.
109    
110    ----------------------------------------------------------------------
111    Name: Matthias Blume (blume (at) tti - c (dot) org)
112    Date: 2004/01/26 15:15:15 CST
113    Tag: blume-20040126-toplevel
114    Description:
115    
116    Improved handling of exceptions at the interactive toplevel.
117    
118    ----------------------------------------------------------------------
119    Name: Matthias Blume (blume (at) tti - c (dot) org)
120    Date: 2004/01/26 11:25:00
121    Tag: blume-20040126-app
122    Description:
123    
124    Type of top-level "app" corrected.
125    Added code for setting vp_limitPtrMask to Win32-specific runtime.
126    
127    ----------------------------------------------------------------------
128    Name: Matthias Blume (blume (at) tti - c (dot) org)
129    Date: 2003/11/18 17:10 CST
130    Tag: blume-20031118-basis-fiddle
131    Description:
132    
133    - changed Timer interface to what might become the spec
134    - POSIX_FLAGS -> BIT_FLAGS according to spec
135    - some other minor discrepancies wrt. spec eliminated
136    
137    ----------------------------------------------------------------------
138    Name: Matthias Blume (blume (at) tti - c (dot) org)
139    Date: 2003/11/06 12:00:00 CST
140    Tag: Release_110_44
141    Description:
142    
143    New working version (110.44). New bootfiles.
144    
145    ----------------------------------------------------------------------
146    Name: Matthias Blume (blume (at) tti - c (dot) org)
147    Date: 2003/11/04 11:50:00 CST
148    Tag: blume-20031104-move-libraries
149    Description:
150    
151    Eliminated the "dont_move_libraries" directive in config/targets.
152    (The mechanism was broken and could not be fixed easily.  Moreover,
153    there does not seem to be any reason not to move all libraries into
154    lib during installation.  I originally implemented this directive as a
155    backward-compatibility feature when I first introduced the new CM.
156    Now that things have been stable for a long time and going back to the
157    old CM is not an option, there is no reason to keep it around.)
158    
159    ----------------------------------------------------------------------
160    Name: Matthias Blume (blume (at) tti - c (dot) org)
161    Date: 2003/11/03 16:00:00 CST
162    Tag: blume-20031103-installdir
163    Description:
164    
165    Made installer honor INSTALLDIR variable again.  (Thanks to Chris
166    Richards for pointing out the problem and providing the solution.)
167    
168    ----------------------------------------------------------------------
169    Name: Matthias Blume (blume (at) tti - c (dot) org)
170    Date: 2003/10/01 17:05:00 CDT
171    Tag: blume-20031001-lal-mlrisc
172    Description:
173    
174    MLRISC bug fix from Lal.
175    
176    ----------------------------------------------------------------------
177    Name: Matthias Blume (blume (at) tti - c (dot) org)
178    Date: 2003/09/30 16:10:00 CDT
179    Tag: blume-20030930-primio-bat
180    Description:
181    
182    1. Added openVector, nullRd, and nullWr to PRIM_IO.
183    2. Improved .bat files (for Win32 port) to make things work under Win95.
184       (thanks to Aaron S. Hawley for this one)
185    
186    ----------------------------------------------------------------------
187    Name: Matthias Blume (blume (at) tti - c (dot) org)
188    Date: 2003/09/26 16:05:00 CDT
189    Tag: blume-20030926-wrappriv
190    Description:
191    
192    Added missing wrapper for privilege "primitive" in $smlnj/viscomp/core.cm.
193    
194    ----------------------------------------------------------------------
195    Name: Matthias Blume (blume (at) tti - c (dot) org)
196    Date: 2003/09/26 15:00:00 CDT
197    Tag: blume-20030926-110_43_3
198    Description:
199    
200    - additional cleanup
201    - version number bump, NEW BOOTFILES
202    
203    ----------------------------------------------------------------------
204    Name: Matthias Blume (blume (at) tti - c (dot) org)
205    Date: 2003/09/26 12:00:00 CDT
206    Tag: blume-20030926-ppautoload
207    Description:
208    
209    I modified the read-eval-print loop so that the autoloader gets
210    invoked whenever the prettyprinter tries to look up a symbol that
211    is not currently defined in the toplevel environment but which
212    appears in CM's autoload registry.  As a result, we see far fewer of
213    those ?.Foo.Bar.xxx names in the prettyprinter's output.
214    
215    In addition to this I tried to clean up some pieces of the Basis
216    implementation (e.g., Socket, Word8Array) in order to prevent other
217    instances of these ?.Foo.Bar.xxx names from being printed.
218    
219    The mechanism that picks names for types still needs some work, though.
220    (Right now it seems that if there is a type A.t which is defined to
221    be B.u, but B is unavailable at toplevel, then A.t gets printed as
222    "?.B.u" although the perhaps more sensible solution would be to use
223    "A.t" in this case.  In other words, the prettyprinter should follow
224    a chain of DEFtycs not farther than there are corresponding toplevel
225    names in the current environment.)
226    
227    ----------------------------------------------------------------------
228    Name: Matthias Blume (blume (at) tti - c (dot) org)
229    Date: 2003/09/24 16:31:00 CDT
230    Tag: blume-20030924-installer
231    Description:
232    
233    Another installer tweak:  All the ML code for the installer is now
234    compiled during CMB.make and put into a little library called
235    $smlnj/installer.cm.  The installation then simply invokes
236    
237       sml -m $smlnj/installer.cm
238    
239    and everything happens automagically.
240    
241    Win32:  ML code senses value of environment variable SMLNJ_HOME.
242    Unix:   ML code senses values of environment variables ROOT, CONFIGDIR,
243            and BINDIR.
244    
245    The new scheme guarantees that the ML code responsible for the installation
246    is in sync with the APIs of the main system.  Also, the installer is
247    somewhat faster because the installer script is precompiled.
248    
249    ----------------------------------------------------------------------
250    Name: Matthias Blume (blume (at) tti - c (dot) org)
251    Date: 2003/09/24 15:35:00 CDT
252    Tag: blume-20030924-synsock
253    Description:
254    
255    Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
256    but excludes all non-blocking operations.  Defined SOCKET (in Basis)
257    and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
258    implementations of non-blocking operations from CML's Socket
259    structure.
260    
261    ----------------------------------------------------------------------
262    Name: Matthias Blume (blume (at) tti - c (dot) org)
263    Date: 2003/09/24 15:10:05 CDT
264    Tag: blume-20030924-sockets
265    Description:
266    
267    1. Fixed SOCKET API and implementation to match Basis spec.
268       This required changing the internal representation of sockets to one
269       that remembers (for each socket file descriptor) whether it is currently
270       blocking or non-blocking.  This state is maintained lazily (i.e., a system
271       call is made only if the state actually needs to change).
272    
273    2. OS-specific details of sockets were moved into separate files, thus
274       making it possible to unify the bulk of the socket implementations
275       between Unix and Win32.
276    
277    3. CML's socket API changed accordingly.
278       (Note that we need to remove non-blocking functions from this API
279       since they are redundant in the case of CML!)
280    
281    4. CML's socket implementation now makes use of non-blocking functions
282       provided by Basis, thus removing all OS-dependent code from this part
283       of CML.
284    
285    5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
286    
287    ----------------------------------------------------------------------
288    Name: Matthias Blume (blume (at) tti - c (dot) org)
289    Date: 2003/09/22 12:10:00 CDT
290    Tag: blume-20030922-110_43_2
291    Description:
292    
293    Made a new interim version and bootfiles for developer's bootstrapping
294    convenience.
295    
296       110.43.2 -- NEW BOOTFILES
297    
298    ----------------------------------------------------------------------
299    Name: Matthias Blume (blume (at) tti - c (dot) org)
300    Date: 2003/09/19 15:55:00 CDT
301    Tag: blume-20030919-cmdir
302    Description:
303    
304    1. new-install.sh -> install.sh
305    2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
306    3. tweaked installer so that another name instead of .cm can be chosen
307       at install time (by setting the CM_DIR_ARC environment variable
308       during installation); once installation is complete, the name is
309       fixed
310    
311    ----------------------------------------------------------------------
312    Name: Matthias Blume (blume (at) tti - c (dot) org)
313    Date: 2003/09/18 16:00:00 CDT
314    Tag: blume-20030918-110_43_1
315    Description:
316    
317    Made a new interim version and bootfiles for developer's bootstrapping
318    convenience.
319    
320       110.43.1 -- NEW BOOTFILES
321    
322    ----------------------------------------------------------------------
323    Name: Matthias Blume (blume (at) tti - c (dot) org)
324    Date: 2003/09/18 15:20:00 CDT
325    Tag: blume-20030918-misc
326    Description:
327    
328    1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
329       this seems to be controversial at the moment) and used those in
330       Posix.ProcEnv.times.
331    
332    2. Added Time.{from,to}Nanoseconds to Time.
333    
334    3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
335       For example, fromLargeInt never needs to look at more than 3 "big
336       digits" to get its 53 bits of precision.
337    
338    ----------------------------------------------------------------------
339    Name: Matthias Blume (blume (at) tti - c (dot) org)
340    Date: 2003/09/17 16:30:00 CDT
341    Tag: blume-20030917-real32-slices
342    Description:
343    
344    Added an entry to the primitive environment
345    (compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
346    added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
347    
348    Removed some of the "magic" constants in real64.sml and replaced them
349    with code that generates these values from their corresponding
350    integer counterparts.
351    
352    Made all(?) the slice-related changes to the Basis and made everything
353    compile again...
354    
355    ----------------------------------------------------------------------
356    Name: Matthias Blume (blume (at) tti - c (dot) org)
357    Date: 2003/09/15 17:45:00 CDT
358    Tag: blume-20030915-rbase
359    Description:
360    
361    Fixed bug in Real.fromLargeInt.
362    
363    ----------------------------------------------------------------------
364    Name: Matthias Blume (blume (at) tti - c (dot) org)
365    Date: 2003/09/13 18:11:00 CDT
366    Tag: blume-20030913-libinstall
367    Description:
368    
369    Minor bugfix in config/libinstall (set anchor with path to
370    standalone tool after installing it, otherwise libraries that
371    need ml-lex or ml-yacc won't compile the first time the installer
372    runs).
373    
374    ----------------------------------------------------------------------
375    Name: Matthias Blume (blume (at) tti - c (dot) org)
376    Date: 2003/09/12 11:45:00 CDT
377    Tag: blume-20030912-various
378    Description:
379    
380     - fixed bug in Real.toLargeInt
381     - fixed bug in Posix.ProcEnv.times
382     - changed inputLine functions to return an option
383     - minor installer improvements / bugfixes
384     - changed default @SMLalloc parameter for x86/celeron to 64k
385    
386    ----------------------------------------------------------------------
387    Name: Matthias Blume (blume (at) tti - c (dot) org)
388    Date: 2003/09/09 22:00:00 CDT
389    Tag: Release_110_43
390    Description:
391    
392        New working release 110.43.  New bootfiles.
393    
394    ----------------------------------------------------------------------
395    Name: Matthias Blume (blume (at) tti - c (dot) org)
396    Date: 2003/09/09 19:20:00 CDT
397    Tag: blume-20030909-installer
398    Description:
399    
400    Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
401    Modified config/install.bat to take advantage of it.  Also modified
402    config/install.sh (and called it config/new-install.sh) to take advantage
403    of it on Unix systems.  (The SML code is (supposed to be) platform-
404    independent.)
405    
406    The installer can now install everything under Win32
407    as well as under *nix as long as it compiles.
408    
409    Other changes:
410    
411      - made CML compile again under Win32
412      - made eXene compile under Win32 (by providing a fake structure UnixSock
413        and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
414      - fixed a bug in nowhere: it assumed that type OS.Process.status is the
415        same as type int;  under Win32 it isn't
416      - fixed some slice-related problems in the win32-specific parts of CML
417      - added a functor argument "sameVol" to os-path-fn.sml in the Basis
418        (under Win32, the volume name is case-insensitive, and the
419        OS.Path code compares volume names for equality)
420    
421    ----------------------------------------------------------------------
422    Name: Matthias Blume (blume (at) tti - c (dot) org)
423    Date: 2003/09/08 11:55:00 CDT
424    Tag: blume-20030908-fullpath
425    Description:
426    
427    Made Win32 version of OS.FileSys.fullPath return current directory
428    when given an empty string.  This is what the spec says, and incidentally,
429    CM depends on it.  (CM otherwise goes into an infinite loop in certain
430    cases when presented with the name of a non-existing .cm file.)
431    
432    ----------------------------------------------------------------------
433    Name: Matthias Blume (blume (at) tti - c (dot) org)
434    Date: 2003/09/04 16:30:00 CDT
435    Tag: blume-20030905-slices-etc
436    Description:
437    
438      1. Changed interface to vectors and arrays in Basis to match
439         (draft) Basis spec.
440      2. Added signatures and implementations of slices according to
441         Basis spec.
442      3. Edited source code throughout the system to make it compile again
443         under 1. and 2.  (In some cases code had to be added to have it
444         match the new signatures.)
445      4. MLRISC should be backward-compatible: the copies of the originals
446         of files that needed to change under 3. were retained, the .cm files
447         check the compiler version number and use old versions when
448         appropriate.
449      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
450         dirstream -> string option (in accordance with Basis spec).
451      6. When generating code that counts lines, ml-lex used function
452         CharVector.foldli, taking advantage of its old interface.
453         This has been replaced with the corresponding code from
454         CharVectorSlice.  (html-lex must be re-lexed!)
455      7. BitArray in smlnj-lib/Util has been extended/modified to match the
456         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
457      8. Removed temporary additions (fromInternal, toInternal) from the
458         (now obsolete) IntInf in smlnj-lib/Util.
459      9. Cleaned up structure Byte.
460     10. Added localOffset, scan, and fromString to Date (according to spec).
461         Cleaned/corrected implementation of Date.
462         (Still need to check for correctness; implement better canonicalizeDate.)
463     11. Added "scan" to signature IEEE_REAL.
464     12. Some improvements to IntInf [in particular: efficiency-hack for
465         mod and rem when second operand is 2 (for parity checks).]
466     13. Changed representation of type Time.time, using a single IntInf.int
467         value counting microseconds.  This considerably simplified the
468         implementation of structure Time.  We now support negative time
469         values; scan and fromString handle signs.
470     14. Functor PrimIO now takes two additional arguments (VectorSlice and
471         ArraySlice).
472    
473    ----------------------------------------------------------------------
474    Name: Matthias Blume (blume (at) tti - c (dot) org)
475    Date: 2003/08/28 17:00:00 CDT
476    Tag: blume-20030828-intinf
477    Description:
478    
479    This is a major update which comes with a version number bump
480    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
481    and an implementation of IntInf in the Basis.
482    
483    There are a fairly large number of related changes and updates throughout
484    the system:
485    
486    Basis:
487            - Implemented IntInf.
488            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
489            - Added some missing Real64 operations, most notably Real.toLargeInt.
490            - Added FixedInt as a synonym for Int32.
491    
492    compiler:
493            * Added support for a built-in intinf type.
494              - literals
495              - pattern matching
496              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
497              - overloading on literals and operations
498    
499            This required adding a primitive type intinf, some additional
500            primops, and implementations for several non-trivial intinf
501            operations in Core.  (The intinf type is completely abstract
502            to the compiler; all operations get delegated back to the Core.)
503    
504            * Intinf equality is handled by polyequal.  However, the compiler
505            does not print its usual warning in this case (since polyequal
506            is the right thing to do there).
507    
508            * Improved the organization of structure InlineT.
509    
510            * A word about conversion primops:
511            If conversions involving intinf do not cancel out during
512            CPS contract, then the compiler must insert calls to Core functions.
513            Since all core access must be resolved already during the FLINT
514            translate phase, it would be too late a the time of CPS contract
515            to add new Core calls.  For this reason, conversion primops
516            for intinf carry two arguments:  1. the numeric argument that
517            they are supposed to convert, and 2. the Core function that
518            can help with this conversion if necessary.  If CPS contract
519            eliminates a primop, then the associated Core function becomes
520            dead and goes away.  Intinf conversion primops that do not get
521            eliminated by CPS contract get rewritten into calls of their
522            core functions by a separate, new phase.
523    
524    interactive system:
525            - Control.Print.intinfDepth controls max length of intinf constants
526            being printed. (Analogous to Control.Print.stringDepth.)
527            - Cleanup in printutil and pputil: got rid of unused stuff and
528            duplicates; replaced some of the code with code that makes better
529            use of library functionality.
530    
531    CM:
532            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
533            are no longer silent.
534    
535    CKIT:
536            Fixed mismatched uses of Int32 and LargeInt.  I always decided
537            in favor of LargeInt -- which is now the same as IntInf.
538            CKIT-knowledgable people should check whether this is what's
539            intended and otherwise change things back to using Int32 or
540            FixedInt.
541    
542    Throughout the code:
543            Started using IntInf.int literals and built-in operations
544            (e.g., comparison with 0) where this seems appropriate.
545    
546    
547    ----------------------------------------------------------------------
548    Name: Dave MacQueen (dbm@cs.uchicago.edu)
549    Date: 2003/08/13 11:36:00 CDT
550    Tag: dbm-20030813-mcz-merge1
551    Description:
552    
553    Merging changes from the mcz-branch development branch into trunk.
554    These changes involve replacement of the emulated old prettyprinter
555    interface with direct use of the SML/NJ Lib PP library, and fixing
556    of a couple of bugs (895, 1186) relating to error messages.  A new
557    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
558    has been added.
559    
560    ----------------------------------------------------------------------
561    Name: Matthias Blume (blume (at) tti - c (dot) org)
562    Date: 2003/08/11 15:45:00 CDT
563    Tag: blume-20030811-windows
564    Description:
565    
566    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
567    
568        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
569    
570    This patch restores SML/NJ's ability to run under win32.  There are a
571    number of changes, including fixes for several bugs that had gone
572    unnoticed until now:
573    
574      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
575        trying to run the win32 version from within cygwin.)
576    
577      - There are a number of simple .bat scripts that substitute for their
578        corresponding Unix shell-scripts.  (See below.)
579    
580      - The internals of ml-build have been modified slightly.  The main
581        difference is that instead of calling ".link-sml" (or link-sml.bat)
582        using OS.Process.system, the ML process delegates this task back
583        to the script.  Otherwise problems arise in mixed environments such
584        as Cygwin where scripts look and work like Unix scripts, but
585        where OS.Process.system cannot run them.
586    
587      - In CM, the srcpath pickler used native pathname syntax -- which
588        is incorrect in the case of cross-compilation.  The new pickle format
589        is independent of platform-specific naming conventions.
590    
591      - Path configuration files (such as lib/pathconfig) can now choose
592        between native and standard syntax.  Placing a line of the form
593    
594            standard!
595    
596        into the file causes all subsequent paths to be interpreted using
597        CM standard pathname syntax (= Unix conventions); a line
598    
599            native!
600    
601        switches back to native style.  This was needed so that
602        path config files can be written portably, see src/system/pathconfig.
603    
604      - Runtime system:
605    
606          - win32-filesys.c:  get_file_time and set_file_time now
607            access modification time, not creation time.
608    
609          - I/O code made aware of new array representation.
610    
611          - Bug fixes in X86.prim.masm.
612    
613       - src/system/makeml made aware of win32. (For use under cygwin
614            and other Unix-environments for windows.)
615    
616       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
617         which caused BinIO.inputAll to fail consistently.
618    
619    .bat scripts:
620    
621       Windows .bat scripts assume that SMLNJ_HOME is defined.
622    
623       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
624         applications (sml, ml-yacc, ml-lex).
625       - ml-build.bat: analogous to ml-build.
626       - config\install.bat:  Analogous to config/install.sh.  This requires
627         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
628         (nmake etc. must be on the path, and vcvars32 must have been run.)
629         Moreover, sources for ml-lex and ml-yacc need to exist under src,
630         and the bootfile hierarchy must have been unpacked under
631         sml.boot.x86-win32.
632         The script is very primitive and does a poor job at error checking.
633         It only installs the base system, ml-lex, and ml-yacc.  No other
634         libraries are being installed (i.e., you get only those that
635         are part of the compiler.)
636       - link-sml.bat:  analogous to .link-sml, but not currently used
637    
638    Unrelated bug fixes:
639    
640      - ml-nlffigen now exports structures ST_* corresponding to incomplete
641        types.
642      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
643        otherwise.)
644    
645    ----------------------------------------------------------------------
646    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
647    Date: 2003/06/17
648    Tag: macqueen-20030617-bug895
649    Description:
650    
651    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
652    Tag will be used for new development branch (mcz-branch) for
653    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
654    
655    ----------------------------------------------------------------------
656    Name: Matthias Blume (blume (at) tti - c (dot) org)
657    Date: 2003/05/27 16:55:00 CDT
658    Tag: blume-20030527-polyeq
659    Description:
660    
661    Tried to eliminated most cases of polymorphic equality.
662    
663    ----------------------------------------------------------------------
664    Name: Matthias Blume (blume (at) tti - c (dot) org)
665    Date: 2003/05/21 17:45:00 CDT
666    Tag: blume-20030517-complete
667    Description:
668    
669    Two changes:
670    
671      1. Added a flag for controlling whether non-exhaustive bindings will
672         be treated as errors (default is false).
673      2. Cleaned up the *entire* source tree so that CMB.make goes through
674         without a single non-exhaustive match- or bind warning.
675    
676    ----------------------------------------------------------------------
677    Name: Matthias Blume (blume (at) tti - c (dot) org)
678    Date: 2003/05/17 10:20:00 CDT
679    Tag: blume-20030517-absyn
680    Description:
681    
682    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
683    
684       This mainly affects the quality of error messages.  However, some
685       of the code is now more straightforward than before. (Treatment of
686       the above four constructs in translate.sml is much simpler than
687       the "macro-expansion" that was going on before.  Plus, the mach-
688       compiler no longer gets invoked just to be able to compile an
689       if-expression.)
690    
691    2. The ErrorMsg.Error exception is now caught and absorbed by the
692       interactive loop.
693    
694    ----------------------------------------------------------------------
695    Name: Allen Leung
696    Date: 2003/05/16 13:05:00 CDT
697    Tag: leunga-20030516-cygwin-runtime
698    Description:
699    
700        Ported the runtime system to cygwin, which uses the unix
701    x86-unix bin files.  Missing/buggy features:
702    
703       o getnetbyname, getnetbyaddr: these functions seem to be missing in
704         the Cygwin library.
705       o Ctrl-C handling may be flaky.
706       o Windows system calls and Windows I/O are not supported.
707    
708       A new set of binfiles is located at:
709    
710         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
711    
712       This is only needed for bootstrapping the cygwin version of smlnj.
713       Other x86 versions can use the existing binfiles.
714    
715    ----------------------------------------------------------------------
716    Name: Matthias Blume
717    Date: 2003/04/08 15:42:00 CDT
718    Tag: blume-20030408-listpair
719    Description:
720    
721    1. Added a target 'mlrisc' to installer.
722    
723    2. Added missing elements to structure ListPair.
724    
725    ----------------------------------------------------------------------
726    Name: Allen Leung
727    Date: 2003/01/07 10:40:00 EST
728    Tag: leunga-20030107-int-rem
729    Description:
730    
731    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
732    The arguments to the SUBL instruction were swapped.
733    
734    ----------------------------------------------------------------------
735    Name: Matthias Blume
736    Date: 2002/12/12 16:25:00 EST
737    Tag: blume-20021212-risc-ra
738    Description:
739    
740    Fixed a serious bug in the rewrite code for FP spilling/reloading that
741    sent the RA into an infinite loop when floating point registers get
742    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
743    110.37 and 110.38.)
744    There was another set of potential problems related to the handling of
745    MLRISC annotations (but those did not yet cause real problems, apparently).
746    
747    ----------------------------------------------------------------------
748    Name: Matthias Blume
749    Date: 2002/12/06 22:40:00 EST
750    Tag: blume-20021206-cm-fileid
751    Description:
752    
753    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
754    This fixes the problem of CM getting confused by files that suddenly
755    change their identity (e.g., by getting unlinked and recreated by some
756    text editor such as vi).  There might be a better/cheaper/cleaner way
757    of doing this, but for now this will have to do.
758    
759    ----------------------------------------------------------------------
760    Name: Matthias Blume
761    Date: 2002/10/28 09:50:00 EST
762    Tag: blume-20021028-typecheck
763    Description:
764    
765    Exported structure Typecheck from $smlnj/viscomp/core.cm.
766    
767    ----------------------------------------------------------------------
768    Name: Matthias Blume
769    Date: 2002/10/17 09:10:00 EDT
770    Tag: Release_110_42
771    Description:
772    
773    In good old tradition, there has been a slight hiccup so that we have
774    to patch 110.42 after the fact.   The old release tag has been replaced
775    (see below).
776    
777    The change solves a problem with two competing approaches the
778    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
779    each other's way.
780    
781    This change only affects the runtime system code and the installer script.
782    (No new bootfiles.)
783    
784    ----------------------------------------------------------------------
785    Name: Matthias Blume
786    Date: 2002/10/16 12:00:00 EDT
787    Tag: Release_110_42_removed
788    Description:
789    
790    New working release.  New bootfiles.
791    
792    ----------------------------------------------------------------------
793    Name: Matthias Blume
794    Date: 2002/10/10 13:10:00 EDT
795    Tag: blume-20021010-ppc-divs
796    Description:
797    
798    The mltree operator DIVS must be implemented with an overflow check on
799    the PPC because the hardware indicates divide-by-zero using "overflow" as
800    well.
801    
802    ----------------------------------------------------------------------
803    Name: Matthias Blume
804    Date: 2002/07/23 11:45:00
805    Tag: blume-20020723-smlnj-home
806    Description:
807    
808    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
809    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
810    CM_PATHCONFIG is also set), the path configuration file is assumed
811    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
812    the entire tree to some other place and everything will "just work".
813    
814    (Companion commands such as ml-build and ml-makedepend also sense this
815    variable.)
816    
817    ----------------------------------------------------------------------
818    Name: Matthias Blume
819    Date: 2002/07/12 21:19:00 EDT
820    Tag: blume-20020712-liveness
821    Description:
822    
823    Exported two useful "step" functions from liveness module (MLRISC).
824    
825    ----------------------------------------------------------------------
826    Name: Matthias Blume
827    Date: 2002/07/05 16:00 EDT
828    Tag: Release_110_41
829    Description:
830    
831    New working release.  New bootfiles.
832    
833    ----------------------------------------------------------------------
834    Name: Matthias Blume
835    Date: 2002/07/05 10:25:00 EDT
836    Tag: blume-20020705-btimp
837    Description:
838    
839    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
840    other clients can set up backtracing support.
841    
842    ----------------------------------------------------------------------
843    Name: Matthias Blume
844    Date: 2002/06/25 17:23:00 EDT
845    Tag: blume-20020625-fpmax
846    Description:
847    
848    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
849    case.  (The sense of the isNaN test was reversed -- which made min and
850    max always return their first argument.)
851    
852    ----------------------------------------------------------------------
853    Name: Matthias Blume
854    Date: 2002/06/11
855    Tag: blume-20020611-unixpath
856    Description:
857    
858    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
859    
860    ----------------------------------------------------------------------
861    Name: Matthias Blume
862    Date: 2002/06/10 16:35:00 EDT
863    Tag: blume-20020610-ieeereal
864    Description:
865    
866    I back-ported my implementation of IEEEReal.fromString from the
867    idlbasis-devel branch so that we can test it.
868    
869    Another small change is that ppDec tries to give more information
870    than just "<sig>" in the case of functors.  However, this code is
871    broken in some mysterious way if the functor's body's signature
872    has not been declared by ascription but gets inferred from the
873    implementation.  This needs fixing...
874    
875    ----------------------------------------------------------------------
876    Name: Matthias Blume
877    Date: 2002/05/31
878    Tag: blume-20020531-btrace-mode
879    Description:
880    
881    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
882    the wayside when I switched over to using Controls everywhere.)
883    
884    ----------------------------------------------------------------------
885    Name: Lal George
886    Date: 2002/05/23 12:21:40 EDT
887    Tag: george-20020523-visual-labels
888    Description:
889    
890    Labels are now displayed in the graphical output to make
891    the fall-through and target blocks obvious.
892    
893    ----------------------------------------------------------------------
894    Name: Matthias Blume
895    Date: 2002/05/22 11:03:00 EDT
896    Tag: blume-20020522-shrink
897    Description:
898    
899    John tweaked yesterday's fix for 1131 to handle an out-of-memory
900    situation that comes up when allocating huge arrays.
901    
902    ----------------------------------------------------------------------
903    Name: Matthias Blume
904    Date: 2002/05/21 16:00:00 EDT
905    Tag: Release_110_40
906    Description:
907    
908    New working release (110.40).  New bootfiles.
909    
910    [Also: John Reppy fixed GC bug 1131.]
911    
912    ----------------------------------------------------------------------
913    Name: Matthias Blume
914    Date: 2002/05/21 12:35:00 EDT
915    Tag: blume-20020521-cmdoc
916    Description:
917    
918    CM documentation update.
919    
920    ----------------------------------------------------------------------
921    Name: Matthias Blume
922    Date: 2002/05/21 10:55:00 EDT
923    Tag: blume-20020521-misc
924    Description:
925    
926    - John tweaked runtime to be silent on heap export (except when
927      GC messages are on).
928    - I added a few more things (cross-compiling versions of CMB) to
929      config/preloads (as suggestions).
930    
931    ----------------------------------------------------------------------
932    Name: Matthias Blume
933    Date: 2002/05/20 22:25:00 EDT
934    Tag: blume-20020520-controls
935    Description:
936    
937    - Added ControlUtil structure to control-lib.cm.
938    - Use it throughout.
939    - Used Controls facility to define MLRISC controls (as opposed to
940      registering MLRISC control ref cells with Controls after the fact)
941    - Fixed messed-up controls priorities.
942    
943    * Removed again all the stuff from config/preloads that one wouldn't
944      be able to preload at the time the initial heap image is built.
945      (Many libraries, e.g., CML, do not exist yet at this time.  The
946       only libraries that can be preloaded via config/preloads are those
947       that come bundled with the bootfiles.)
948    
949    ----------------------------------------------------------------------
950    Name: Matthias Blume
951    Date: 2002/05/20 10:59:00 EDT
952    Tag: blume-20020520-preloads
953    Description:
954    
955    Added a lot of commented-out suggestions for things to be included
956    in config/preloads.
957    
958    ----------------------------------------------------------------------
959    Name: Allen Leung
960    Date: 2002/05/18 14:20:00 EDT
961    Tag: leunga-20020518-mdl
962    Description:
963    
964    o Made the mdl tool stuff compile and run again.
965    o I've disabled all the stuff that depends on RTL specifications; they
966      are all badly broken anyway.
967    
968    ----------------------------------------------------------------------
969    Name: Matthias Blume
970    Date: 2002/05/17 16:49:00 EDT
971    Tag: blume-20020517-controls
972    Description:
973    
974    0. John Reppy made several modifications to the SML/NJ library.
975       In particular, there is a shiny new controls-lib.cm.
976    
977    1. Pushed new controls interface through compiler so that everything
978       compiles again.
979    
980    2. Added FormatComb and FORMAT_COMB to the CML version of the
981       SML/NJ library (so that CML compiles again).
982    
983    3. Modified init scripts because XXX_DEFAULT environment variables
984       are no longer with us.  (Boot-time initialization is now done
985       using the same environment variables that are also used for
986       startup-time initialization of controls.)
987    
988    ----------------------------------------------------------------------
989    Name: Lal George
990    Date: 2002/05/15 09:20:10 EDT
991    Tag: george-20020515-pseudo-op-decls
992    Description:
993    
994            All pseudo-ops emitted before the first segment declaration
995    such as TEXT, DATA, and BSS directives are assumed to be global
996    declarations and are emitted first in the assembly file. This is
997    useful in a number of situations where one has pseudo-ops that are not
998    specific to any segment, and also works around the constraint that one
999    cannot have client pseudo-ops in the TEXT segment.
1000    
1001            Because no segment is associated with these declarations it is
1002    an error to allocate any space or objects before the first segment
1003    directive and an exception will be raised. However, we cannot make
1004    this check for client pseudo-ops.
1005    
1006            These top level declarations are a field in the CFG graph_info.
1007    In theory you can continue to add to this field after the CFG has been
1008    built -- provided you know what you are doing;-)
1009    
1010    ----------------------------------------------------------------------
1011    Name: Matthias Blume
1012    Date: 2002/05/13 16:40:00 EDT
1013    Tag: blume-20020513-pp-etc
1014    Description:
1015    
1016    A few minor bugfixes:
1017    
1018      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
1019        (Bogus printouts for redefined bindings still occur.  Compiler
1020        bug should no longer occur now.  We need to redo the prettyprinter
1021        from scratch.)
1022    
1023      - CM pathname printer now also adds escape sequences for ( and )
1024    
1025      - commend and docu fixes for ml-nlffi
1026    
1027    ----------------------------------------------------------------------
1028    Name: Matthias Blume
1029    Date: 2002/05/10 16:40:00 EDT
1030    Tag: blume-20020510-erg-textio
1031    Description:
1032    
1033    Applied the following bugfix provided by Emden Gansner:
1034    
1035        Output is corrupted when outputSubstr is used rather than output.
1036    
1037        The problem occurs when a substring
1038    
1039            ss = (s, dataStart, dataLen)
1040    
1041        where dataStart > 0, fills a stream buffer with avail bytes left.
1042        avail bytes of s, starting at index dataStart, are copied into the
1043        buffer, the buffer is flushed, and then the remaining dataLen-avail
1044        bytes of ss are copied into the beginning of the buffer. Instead of
1045        starting this copy at index dataStart+avail in s, the current code
1046        starts the copy at index avail.
1047    
1048        Fix:
1049        In text-io-fn.sml, change line 695 from
1050             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
1051        to
1052             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
1053    
1054    ----------------------------------------------------------------------
1055    Name: Matthias Blume
1056    Date: 2002/04/12 13:55:00 EDT
1057    Tag: blume-20020412-assyntax
1058    Description:
1059    
1060    1. Grabbed newer assyntax.h from the XFree86 project.
1061    2. Fiddled with how to compile X86.prim.asm without warnings.
1062    3. (Very) Minor cleanup in CM.
1063    
1064    ----------------------------------------------------------------------
1065    Name: Matthias Blume
1066    Date: 2002/04/01 (no joke!) 17:07:00 EST
1067    Tag: blume-20020401-x86div
1068    Description:
1069    
1070    Added full support for div/mod/rem/quot on the x86, using the machine
1071    instruction's two results (without clumsily recomputing the remainder)
1072    directly where appropriate.
1073    
1074    Some more extensive power-of-two support was added to the x86 instruction
1075    selector (avoiding expensive divs, mods, and muls where they can be
1076    replaced with cheaper shifts and masks).  However, this sort of thing
1077    ought to be done earlier, e.g., within the CPS optimizer so that
1078    all architectures benefit from it.
1079    
1080    The compiler compiles to a fixed point, but changes might be somewhat
1081    fragile nevertheless.  Please, report any strange things that you might
1082    see wrt. div/mod/quot/rem...
1083    
1084    ----------------------------------------------------------------------
1085    Name: Matthias Blume
1086    Date: 2002/03/29 17:22:00
1087    Tag: blume-20020329-div
1088    Description:
1089    
1090    Fixed my broken div/mod logic.  Unfortunately, this means that the
1091    inline code for div/mod now has one more comparison than before.
1092    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
1093    The problem was with quotient = 0, because that alone does not tell
1094    us which way the rounding went.  One then has to look at whether
1095    remainder and divisor have the same sign...  :(
1096    
1097    Anyway, I replaced the bootfiles with fresh ones...
1098    
1099    ----------------------------------------------------------------------
1100    Name: Matthias Blume
1101    Date: 2002/03/29 14:10:00 EST
1102    Tag: blume-20020329-inlprims
1103    Description:
1104    
1105    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
1106    
1107    Primops have changed. This means that the bin/boot-file formats have
1108    changed as well.
1109    
1110    To make sure that there is no confusion, I made a new version.
1111    
1112    
1113    CHANGES:
1114    
1115    * removed REMT from mltree (remainder should never overflow).
1116    
1117    * added primops to deal with divisions of all flavors to the frontend
1118    
1119    * handled these primops all the way through so they map to their respective
1120      MLRISC support
1121    
1122    * used these primops in the implementation of Int, Int32, Word, Word32
1123    
1124    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
1125    
1126    * parameterized INLMIN, INLMAX, and INLABS by a numkind
1127    
1128    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
1129      floating point
1130    
1131    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
1132      and Real (but Real.abs maps to a separate floating-point-only primop)
1133    
1134    
1135    TODO items:
1136    
1137    * Hacked Alpha32 instruction selection, disabling the selection of REMx
1138      instructions because the machine instruction encoder cannot handle
1139      them.  (Hppa, PPC, and Sparc instruction selection did not handle
1140      REM in the first place, and REM is supported by the x86 machine coder.)
1141    
1142    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
1143      selection phase.  (The two can be streamlined because the hardware
1144      delivers both quotient and remainder at the same time anyway.)
1145    
1146    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
1147      (Currently the behavior is inconsistent both across architectures and
1148      wrt. the draft Basis spec.)
1149    
1150    * Word8 should eventually be handled natively, too.
1151    
1152    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
1153      as if there currently is no execution path that could trigger it in
1154      SML/NJ.  (The assumptions underlying functions arith and promotable do not
1155      hold for things like multiplication and division.)
1156    
1157    ----------------------------------------------------------------------
1158    Name: Matthias Blume
1159    Date: 2002/03/27 16:27:00 EST
1160    Tag: blume-20020327-mlrisc-divisions
1161    Description:
1162    
1163    Added support for all four division operations (ML's div, mod, quot,
1164    and rem) to MLRISC.  In the course of doing so, I also rationalized
1165    the naming (no more annoying switch-around of DIV and QUOT), by
1166    parameterizing the operation by div_rounding_mode (which can be either
1167    DIV_TO_ZERO or DIV_TO_NEGINF).
1168    
1169    The generic MLTreeGen functor takes care of compiling all four
1170    operations down to only round-to-zero div.
1171    
1172    Missing pieces:
1173    
1174      * Doing something smarter than relying on MLTreeGen on architectures
1175        like, e.g., the x86 where hardware division delivers both quotient and
1176        remainder at the same time.  With this, the implementation of the
1177        round-to-neginf operations could be further streamlined.
1178    
1179      * Remove inlining support for div/mod/rem from the frontend and replace it
1180        with primops that get carried through to the backend.  Do this for all
1181        int and word types.
1182    
1183    ----------------------------------------------------------------------
1184    Name: Matthias Blume
1185    Date: 2002/03/25 17:25:00 EST
1186    Tag: blume-20020325-divmod
1187    Description:
1188    
1189    I improved (hopefully without breaking them) the implementation of Int.div,
1190    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
1191    advantage of the following observations:
1192    
1193      Let  q = x quot y      r = x rem y
1194           d = x div  y      m = x mod y
1195    
1196    where "quot" is the round-to-zero version of integer division that
1197    hardware usually provides.  Then we have:
1198    
1199         r = x - q * y        where neither the * nor the - will overflow
1200         d = if q >= 0 orelse x = q * y then q else q - 1
1201                              where neither the * nor the - will overflow
1202         m = if q >= 0 orelse r = 0 then r else r + y
1203                              where the + will not overflow
1204    
1205    This results in substantial simplification of the generated code.
1206    The following table shows the number of CFG nodes and edges generated
1207    for
1208            fun f (x, y) = x OPER y
1209            (* with OPER \in div, mod, quot, rem *)
1210    
1211    
1212        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
1213        --------------------------------------------------------
1214         div |         24 |         39 |         12 |         16
1215         mod |         41 |         71 |         12 |         16
1216        quot |          8 |         10 |          8 |         10
1217         rem |         10 |         14 |          8 |         10
1218    
1219    
1220    ----------------------------------------------------------------------
1221    Name: Matthias Blume
1222    Date: 2002/03/25 22:06:00 EST
1223    Tag: blume-20020325-cprotobug
1224    Description:
1225    
1226    Fixed a bug in cproto (c prototype decoder).
1227    
1228    ----------------------------------------------------------------------
1229    Name: Matthias Blume
1230    Date: 2002/03/25 16:00:00 EST
1231    Tag: blume-20020325-raw-primops
1232    Description:
1233    
1234    I did some cleanup to Allen's new primop code and
1235    replaced yesterday's bootfiles with new ones.
1236    (But they are stored in the same place.)
1237    
1238    ----------------------------------------------------------------------
1239    Name: Matthias Blume
1240    Date: 2002/03/24 22:40:00 EST
1241    Tag: blume-20020324-bootfiles
1242    Description:
1243    
1244    Made the bootfiles that Allen asked for.
1245    
1246    ----------------------------------------------------------------------
1247    Name: Allen Leung
1248    Date: 2002/03/23 15:50:00 EST
1249    Tag: leunga-20020323-flint-cps-rcc-primops
1250    Description:
1251    
1252      1. Changes to FLINT primops:
1253    
1254        (* make a call to a C-function;
1255         * The primop carries C function prototype information and specifies
1256         * which of its (ML-) arguments are floating point. C prototype
1257         * information is for use by the backend, ML information is for
1258         * use by the CPS converter. *)
1259      | RAW_CCALL of { c_proto: CTypes.c_proto,
1260                       ml_args: ccall_type list,
1261                       ml_res_opt: ccall_type option,
1262                       reentrant : bool
1263                     } option
1264       (* Allocate uninitialized storage on the heap.
1265        * The record is meant to hold short-lived C objects, i.e., they
1266        * are not ML pointers.  With the tag, the representation is
1267        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
1268        *)
1269      | RAW_RECORD of {tag:bool,sz:int}
1270      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
1271    
1272      2.  These CPS primops are now overloaded:
1273    
1274           rawload of {kind:numkind}
1275           rawstore of {kind:numkind}
1276    
1277          The one argument form is:
1278    
1279             rawload {kind} address
1280    
1281          The two argument form is:
1282    
1283             rawload {kind} [ml object, byte-offset]
1284    
1285      3. RAW_CCALL/RCC now takes two extra arguments:
1286    
1287         a. The first is whether the C call is reentrant, i.e., whether
1288            ML state should be saved and restored.
1289         b. The second argument is a string argument specifying the name of
1290            library and the C function.
1291    
1292         These things are currently not handled in the code generator, yet.
1293    
1294      4. In CProto,
1295    
1296         An encoding type of "bool" means "ml object" and is mapped into
1297         C prototype of PTR.  Note that "bool" is different than "string",
1298         even though "string" is also mapped into PTR, because "bool"
1299         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
1300    
1301      5. Pickler/unpicker
1302    
1303         Changed to handle RAW_RECORD and newest RAW_CCALL
1304    
1305      6. MLRiscGen,
1306    
1307         1. Changed to handle the new rawload/rawstore/rawrecord operators.
1308         2. Code for handling C Calls has been moved to a new module CPSCCalls,
1309            in the file CodeGen/cpscompile/cps-c-calls.sml
1310    
1311      7. Added the conditional move operator
1312    
1313             condmove of branch
1314    
1315         to cps.  Generation of this is still buggy so it is currently
1316         disabled.
1317    
1318    ----------------------------------------------------------------------
1319    Name: Lal George
1320    Date: 2002/03/22 14:18:25 EST
1321    Tag: george-20020322-cps-branch-prob
1322    Description:
1323    
1324    Implemented the Ball-Larus branch prediction-heuristics, and
1325    incorporated graphical viewers for control flow graphs.
1326    
1327    Ball-Larus Heuristics:
1328    ---------------------
1329    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1330    
1331    By design it uses the Dempster-Shafer theory for combining
1332    probabilities.  For example, in the function:
1333    
1334        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1335    
1336    the ball-larus heuristics predicts that the n=0 is unlikely
1337    (OH-heuristic), and the 'then' branch is unlikely because of the
1338    RH-heuristic -- giving the 'then' branch an even lower combined
1339    probability using the Dempster-Shafer theory.
1340    
1341    Finally, John Reppy's loop analysis in MLRISC, further lowers the
1342    probability of the 'then' branch because of the loop in the else
1343    branch.
1344    
1345    
1346    Graphical Viewing:
1347    ------------------
1348    I merely plugged in Allen's graphical viewers into the compiler. The
1349    additional code is not much. At the top level, saying:
1350    
1351            Control.MLRISC.getFlag "cfg-graphical-view" := true;
1352    
1353    will display the graphical view of the control flow graph just before
1354    back-patching.  daVinci must be in your path for this to work. If
1355    daVinci is not available, then the default viewer can be changed
1356    using:
1357    
1358            Control.MLRISC.getString "viewer"
1359    
1360    which can be set to "dot" or "vcg" for the corresponding viewers. Of
1361    course, these viewers must be in your path.
1362    
1363    The above will display the compilation unit at the level of clusters,
1364    many of which are small, boring, and un-interesting. Also setting:
1365    
1366            Control.MLRISC.getInt "cfg-graphical-view_size"
1367    
1368    will display clusters that are larger than the value set by the above.
1369    
1370    
1371    ----------------------------------------------------------------------
1372    Name: Matthias Blume
1373    Date: 2002/03/21 22:20:00 EST
1374    Tag: blume-20020321-kmp-bugfix
1375    Description:
1376    
1377    Changed the interface to the KMP routine in PreString and fixed
1378    a minor bug in one place where it was used.
1379    
1380    ----------------------------------------------------------------------
1381    Name: Allen Leung
1382    Date: 2002/03/21 20:30:00 EST
1383    Tag: leunga-20020321-cfg
1384    Description:
1385    
1386      Fixed a potential problem in cfg edge splitting.
1387    
1388    ----------------------------------------------------------------------
1389    Name: Allen Leung
1390    Date: 2002/03/21 17:15:00 EST
1391    Tag: leunga-20020321-x86-fp-cfg
1392    Description:
1393    
1394      1. Recoded the buggy parts of x86-fp.
1395    
1396         a. All the block reordering code has been removed.
1397            We now depend on the block placement phases to do this work.
1398    
1399         b. Critical edge splitting code has been simplified and moved into the
1400            CFG modules, as where they belong.
1401    
1402         Both of these were quite buggy and complex.  The code is now much, much
1403         simpler.
1404    
1405      2. X86 backend.
1406    
1407         a. Added instructions for 64-bit support.  Instruction selection for
1408            64-bit has not been committed, however, since that
1409            requires changes to MLTREE which haven't been approved by
1410            Lal and John.
1411    
1412         b. Added support for FUCOMI and FUCOMIP when generating code for
1413            PentiumPro and above.  We only generate these instructions in
1414            the fast-fp mode.
1415    
1416         c. Added cases for JP and JNP in X86FreqProps.
1417    
1418      3. CFG
1419    
1420         CFG now has a bunch of methods for edge splitting and merging.
1421    
1422      4. Machine description.
1423    
1424         John's simplification of MLTREE_BASIS.fcond broke a few machine
1425         description things:
1426    
1427         rtl-build.{sig,sml} and hppa.mdl fixed.
1428    
1429         NOTE: the machine description stuff in the repository is still broken.
1430               Again, I can't put my fixes in because that involves
1431               changes to MLTREE.
1432    
1433    ----------------------------------------------------------------------
1434    Name: Matthias Blume
1435    Date: 2002/03/20 15:55:00 EST
1436    Tag: blume-20020320-kmp
1437    Description:
1438    
1439    Implemented Knuth-Morris-Pratt string matching in PreString and used
1440    it for String.isSubstring, Substring.isSubstring, and
1441    Substring.position.
1442    
1443    (Might need some stress-testing.  Simple examples worked fine.)
1444    
1445    ----------------------------------------------------------------------
1446    Name: Matthias Blume
1447    Date: 2002/03/19 16:37:00 EST
1448    Tag: blume-20020319-witnesses
1449    Description:
1450    
1451    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1452    
1453    This implements a generic mechanism for changing constness qualifiers
1454    anywhere within big C types without resorting to outright "casts".
1455    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1456    modify the constness at the outermost level.)
1457    The implementation of "convert" is based on the idea of "witness"
1458    values -- values that are not used by the operation but whose types
1459    "testify" to their applicability.  On the implementation side, "convert"
1460    is simply a projection (returning its second curried argument).  With
1461    cross-module inlining, it should not result in any machine code being
1462    generated.
1463    
1464    ----------------------------------------------------------------------
1465    Name: Matthias Blume
1466    Date: 2002/03/15 16:40:00 EST
1467    Tag: blume-20020315-basis
1468    Description:
1469    
1470    Provided (preliminary?) implementations for
1471    
1472      {String,Substring}.{concatWith,isSuffix,isSubstring}
1473    
1474    and
1475    
1476      Substring.full
1477    
1478    Those are in the Basis spec but they were missing in SML/NJ.
1479    
1480    ----------------------------------------------------------------------
1481    Name: Matthias Blume
1482    Date: 2002/03/14 21:30:00 EST
1483    Tag: blume-20020314-controls
1484    Description:
1485    
1486    Controls:
1487    ---------
1488    
1489    1. Factored out the recently-added Controls : CONTROLS stuff and put
1490       it into its own library $/controls-lib.cm.  The source tree for
1491       this is under src/smlnj-lib/Controls.
1492    
1493    2. Changed the names of types and functions in this interface, so they
1494       make a bit more "sense":
1495    
1496          module -> registry
1497          'a registry -> 'a group
1498    
1499    3. The interface now deals in ref cells only.  The getter/setter interface
1500       is (mostly) gone.
1501    
1502    4. Added a function that lets one register an already-existing ref cell.
1503    
1504    5. Made the corresponding modifications to the rest of the code so that
1505       everything compiles again.
1506    
1507    6. Changed the implementation of Controls.MLRISC back to something closer
1508       to the original.  In particular, this module (and therefore MLRISC)
1509       does not depend on Controls.  There now is some link-time code in
1510       int-sys.sml that registers the MLRISC controls with the Controls
1511       module.
1512    
1513    CM:
1514    ---
1515    
1516      * One can now specify the lambda-split aggressiveness in init.cmi.
1517    
1518    ----------------------------------------------------------------------
1519    Name: Allen Leung
1520    Date: 2002/03/13 17:30:00 EST
1521    Tag: leunga-20020313-x86-fp-unary
1522    Description:
1523    
1524    Bug fix for:
1525    
1526    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1527    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1528    > - fun f(x,(y,z)) = Real.~ y;
1529    > [autoloading]
1530    > [autoloading done]
1531    >       fchsl   (%eax), 184(%esp)
1532    > Error: MLRisc bug: X86MCEmitter.emitInstr
1533    >
1534    > uncaught exception Error
1535    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1536    
1537    The problem was that the code generator did not generate any fp registers
1538    in this case, and the ra didn't know that it needed to run the X86FP phase to
1539    translate the pseudo fp instruction.   This only happened with unary fp
1540    operators in certain situations.
1541    
1542    ----------------------------------------------------------------------
1543    Name: Matthias Blume
1544    Date: 2002/03/13 14:00:00 EST
1545    Tag: blume-20020313-overload-etc
1546    Description:
1547    
1548    1. Added _overload as a synonym for overload for backward compatibility.
1549       (Control.overloadKW must be true for either version to be accepted.)
1550    
1551    2. Fixed bug in install script that caused more things to be installed
1552       than what was requested in config/targets.
1553    
1554    3. Made CM aware of the (_)overload construct so that autoloading
1555       works.
1556    
1557    ----------------------------------------------------------------------
1558    Name: Matthias Blume
1559    Date: 2002/03/12 22:03:00 EST
1560    Tag: blume-20020312-url
1561    Description:
1562    
1563    Forgot to update BOOT and srcarchiveurl.
1564    
1565    ----------------------------------------------------------------------
1566    Name: Matthias Blume
1567    Date: 2002/03/12 17:30:00 EST
1568    Tag: blume-20020312-version110392
1569    Description:
1570    
1571    Yet another version number bump (because of small changes to the
1572    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1573    
1574    Changes:
1575    
1576      The new pid generation scheme described a few weeks ago was overly
1577      complicated.  I implemented a new mechanism that is simpler and
1578      provides a bit more "stability":  Once CM has seen a compilation
1579      unit, it keeps its identity constant (as long as you do not delete
1580      those crucial CM/GUID/* files).  This means that when you change
1581      an interface, compile, then go back to the old interface, and
1582      compile again, you arrive at the original pid.
1583    
1584      There now also is a mechanism that instructs CM to use the plain
1585      environment hash as a module's pid (effectively making its GUID
1586      the empty string).  For this, "noguid" must be specified as an
1587      option to the .sml file in question within its .cm file.
1588      This is most useful for code that is being generated by tools such
1589      as ml-nlffigen (because during development programmers tend to
1590      erase the tool's entire output directory tree including CM's cached
1591      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1592      revert to the old, broken behavior of SML/NJ, but in specific cases
1593      where there is no danger of interface confusion, its use is ok
1594      (I think).
1595    
1596      ml-nlffigen by default generates "noguid" annotations.  They can be
1597      turned off by specifying -guid in its command line.
1598    
1599    ----------------------------------------------------------------------
1600    Name: Lal George
1601    Date: 2002/03/12 12 14:42:36 EST
1602    Tag: george-20020312-frequency-computation
1603    Description:
1604    
1605    Integrated jump chaining and static block frequency into the
1606    compiler. More details and numbers later.
1607    
1608    ----------------------------------------------------------------------
1609    Name: Lal George
1610    Date: 2002/03/11 11 22:38:53 EST
1611    Tag: george-20020311-jump-chain-elim
1612    Description:
1613    
1614    Tested the jump chain elimination on all architectures (except the
1615    hppa).  This is on by default right now and is profitable for the
1616    alpha and x86, however, it may not be profitable for the sparc and ppc
1617    when compiling the compiler.
1618    
1619    The gc test will typically jump to a label at the end of the cluster,
1620    where there is another jump to an external cluster containing the actual
1621    code to invoke gc. This is to allow factoring of common gc invocation
1622    sequences. That is to say, we generate:
1623    
1624            f:
1625               testgc
1626               ja   L1      % jump if above to L1
1627    
1628            L1:
1629               jmp L2
1630    
1631    
1632    After jump chain elimination the 'ja L1' instructions is converted to
1633    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1634    up being implemented in their long form (if L2 is far away) using:
1635    
1636            jbe     L3      % jump if below or equal to L3
1637            jmp     L2
1638         L3:
1639            ...
1640    
1641    
1642    For large compilation units L2  may be far away.
1643    
1644    
1645    ----------------------------------------------------------------------
1646    Name: Matthias Blume
1647    Date: 2002/03/11 13:30:00 EST
1648    Tag: blume-20020311-mltreeeval
1649    Description:
1650    
1651    A functor parameter was missing.
1652    
1653    ----------------------------------------------------------------------
1654    Name: Allen Leung
1655    Date: 2002/03/11 10:30:00 EST
1656    Tag: leunga-20020311-runtime-string0
1657    Description:
1658    
1659       The representation of the empty string now points to a
1660    legal null terminated C string instead of unit.  It is now possible
1661    to convert an ML string into C string with InlineT.CharVector.getData.
1662    This compiles into one single machine instruction.
1663    
1664    ----------------------------------------------------------------------
1665    Name: Allen Leung
1666    Date: 2002/03/10 23:55:00 EST
1667    Tag: leunga-20020310-x86-call
1668    Description:
1669    
1670       Added machine generation for CALL instruction (relative displacement mode)
1671    
1672    ----------------------------------------------------------------------
1673    Name: Matthias Blume
1674    Date: 2002/03/08 16:05:00
1675    Tag: blume-20020308-entrypoints
1676    Description:
1677    
1678    Version number bumped to 110.39.1.  NEW BOOTFILES!
1679    
1680    Entrypoints: non-zero offset into a code object where execution should begin.
1681    
1682    - Added the notion of an entrypoint to CodeObj.
1683    - Added reading/writing of entrypoint info to Binfile.
1684    - Made runtime system bootloader aware of entrypoints.
1685    - Use the address of the label of the first function given to mlriscGen
1686      as the entrypoint.  This address is currently always 0, but it will
1687      not be 0 once we turn on block placement.
1688    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1689      with entry points) from mlriscGen.
1690    
1691    ----------------------------------------------------------------------
1692    Name: Allen Leung
1693    Date: 2002/03/07 20:45:00 EST
1694    Tag: leunga-20020307-x86-cmov
1695    Description:
1696    
1697       Bug fixes for CMOVcc on x86.
1698    
1699       1. Added machine code generation for CMOVcc
1700       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1701       3. CMOVcc cannot have an immediate operand as argument.
1702    
1703    ----------------------------------------------------------------------
1704    Name: Matthias Blume
1705    Date: 2002/03/07 16:15:00 EST
1706    Tag: blume-20020307-controls
1707    Description:
1708    
1709    This is a very large but mostly boring patch which makes (almost)
1710    every tuneable compiler knob (i.e., pretty much everything under
1711    Control.* plus a few other things) configurable via both the command
1712    line and environment variables in the style CM did its configuration
1713    until now.
1714    
1715    Try starting sml with '-h' (or, if you are brave, '-H')
1716    
1717    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1718    implements the underlying generic mechanism.
1719    
1720    The interface to some of the existing such facilities has changed somewhat.
1721    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1722    (The getFoo interface is still there for backward-compatibility, but its
1723    use is deprecated.)
1724    
1725    The ml-build script passes -Cxxx=yyy command-line arguments through so
1726    that one can now twiddle the compiler settings when using this "batch"
1727    compiler.
1728    
1729    TODO items:
1730    
1731    We should go through and throw out all controls that are no longer
1732    connected to anything.  Moreover, we should go through and provide
1733    meaningful (and correct!) documentation strings for those controls
1734    that still are connected.
1735    
1736    Currently, multiple calls to Controls.new are accepted (only the first
1737    has any effect).  Eventually we should make sure that every control
1738    is being made (via Controls.new) exactly once.  Future access can then
1739    be done using Controls.acc.
1740    
1741    Finally, it would probably be a good idea to use the getter-setter
1742    interface to controls rather than ref cells.  For the time being, both
1743    styles are provided by the Controls module, but getter-setter pairs are
1744    better if thread-safety is of any concern because they can be wrapped.
1745    
1746    *****************************************
1747    
1748    One bug fix: The function blockPlacement in three of the MLRISC
1749    backpatch files used to be hard-wired to one of two possibilities at
1750    link time (according to the value of the placementFlag).  But (I
1751    think) it should rather sense the flag every time.
1752    
1753    *****************************************
1754    
1755    Other assorted changes (by other people who did not supply a HISTORY entry):
1756    
1757    1. the cross-module inliner now works much better (Monnier)
1758    2. representation of weights, frequencies, and probabilities in MLRISC
1759       changed in preparation of using those for weighted block placement
1760       (Reppy, George)
1761    
1762    ----------------------------------------------------------------------
1763    Name: Lal George
1764    Date: 2002/03/07 14:44:24 EST 2002
1765    Tag: george-20020307-weighted-block-placement
1766    
1767    Tested the weighted block placement optimization on all architectures
1768    (except the hppa) using AMPL to generate the block and edge frequencies.
1769    Changes were required in the machine properties to correctly
1770    categorize trap instructions. There is an MLRISC flag
1771    "weighted-block-placement" that can be used to enable weighted block
1772    placement, but this will be ineffective without block/edge
1773    frequencies (coming soon).
1774    
1775    
1776    ----------------------------------------------------------------------
1777    Name: Lal George
1778    Date: 2002/03/05 17:24:48 EST
1779    Tag: george-20020305-linkage-cluster
1780    
1781    In order to support the block placement optimization, a new cluster
1782    is generated as the very first cluster (called the linkage cluster).
1783    It contains a single jump to the 'real' entry point for the compilation
1784    unit. Block placement has no effect on the linkage cluster itself, but
1785    all the other clusters  have full freedom in the manner in which they
1786    reorder blocks or functions.
1787    
1788    On the x86 the typical linkage code that is generated is:
1789       ----------------------
1790            .align 2
1791       L0:
1792            addl    $L1-L0, 72(%esp)
1793            jmp     L1
1794    
1795    
1796            .align  2
1797       L1:
1798       ----------------------
1799    
1800    72(%esp) is the memory location for the stdlink register. This
1801    must contain the address of the CPS function being called. In the
1802    above example, it contains the address of  L0; before
1803    calling L1 (the real entry point for the compilation unit), it
1804    must contain the address for L1, and hence
1805    
1806            addl $L1-L0, 72(%esp)
1807    
1808    I have tested this on all architectures except the hppa.The increase
1809    in code size is of course negligible
1810    
1811    ----------------------------------------------------------------------
1812    Name: Allen Leung
1813    Date: 2002/03/03 13:20:00 EST
1814    Tag: leunga-20020303-mlrisc-tools
1815    
1816      Added #[ ... ] expressions to mlrisc tools
1817    
1818    ----------------------------------------------------------------------
1819    Name: Matthias Blume
1820    Date: 2002/02/27 12:29:00 EST
1821    Tag: blume-20020227-cdebug
1822    Description:
1823    
1824    - made types in structure C and C_Debug to be equal
1825    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1826    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1827    
1828    ----------------------------------------------------------------------
1829    Name: Matthias Blume
1830    Date: 2002/02/26 12:00:00 EST
1831    Tag: blume-20020226-ffi
1832    Description:
1833    
1834    1. Fixed a minor bug in CM's "noweb" tool:
1835       If numbering is turned off, then truly don't number (i.e., do not
1836       supply the -L option to noweb).  The previous behavior was to supply
1837       -L'' -- which caused noweb to use the "default" line numbering scheme.
1838       Thanks to Chris Richards for pointing this out (and supplying the fix).
1839    
1840    2. Once again, I reworked some aspects of the FFI:
1841    
1842       A. The incomplete/complete type business:
1843    
1844       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1845         gone!
1846       - ML types representing an incomplete type are now *equal* to
1847         ML types representing their corresponding complete types (just like
1848         in C).  This is still safe because ml-nlffigen will not generate
1849         RTTI for incomplete types, nor will it generate functions that
1850         require access to such RTTI.   But when ML code generated from both
1851         incomplete and complete versions of the C type meet, the ML types
1852         are trivially interoperable.
1853    
1854         NOTE:  These changes restore the full generality of the translation
1855         (which was previously lost when I eliminated functorization)!
1856    
1857       B. Enum types:
1858    
1859       - Structure C now has a type constructor "enum" that is similar to
1860         how the "su" constructor works.  However, "enum" is not a phantom
1861         type because each "T enum" has values (and is isomorphic to
1862         MLRep.Signed.int).
1863       - There are generic access operations for enum objects (using
1864         MLRep.Signed.int).
1865       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1866         * The structure contains the definition of type "mlrep" (the ML-side
1867         representation type of the enum).  Normally, mlrep is the same
1868         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1869         then mlrep will be defined as a datatype -- thus facilitating
1870         pattern matching on mlrep values.
1871         ("-ec" will be suppressed if there are duplicate values in an
1872          enumeration.)
1873         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1874         will be generated for each C enum constant xxx.
1875         * Conversion functions m2i and i2m convert between mlrep and
1876         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1877         * Coversion functions c and ml convert between mlrep and "tag enum".
1878         * Access functions (get/set) fetch and store mlrep values.
1879       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1880         enumerations are merged into one single enumeration represented by
1881         structure E_'.
1882    
1883    ----------------------------------------------------------------------
1884    Name: Allen Leung
1885    Date: 2002/02/25 04:45:00 EST
1886    Tag: leunga-20020225-cps-spill
1887    
1888    This is a new implementation of the CPS spill phase.
1889    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1890    In case of problems, replace it with the old file spill.sml
1891    
1892    The current compiler runs into some serious performance problems when
1893    constructing a large record.  This can happen when we try to compile a
1894    structure with many items.  Even a very simple structure like the following
1895    makes the compiler slow down.
1896    
1897        structure Foo = struct
1898           val x_1 = 0w1 : Word32.int
1899           val x_2 = 0w2 : Word32.int
1900           val x_3 = 0w3 : Word32.int
1901           ...
1902           val x_N = 0wN : Word32.int
1903        end
1904    
1905    The following table shows the compile time, from N=1000 to N=4000,
1906    with the old compiler:
1907    
1908    N
1909    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1910           MLRISC ra                               0.06u  0.00s  0.05g
1911              (spills = 0 reloads = 0)
1912           TOTAL                                   0.63u  0.07s  0.21g
1913    
1914    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1915           MLRISC ra                               5.68u  0.59s  3.93g
1916              (spills = 0 reloads = 0)
1917           TOTAL                                   14.71u  0.99s  4.81g
1918    
1919    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1920           MLRISC ra                               5.54u  0.65s  3.91g
1921              (spills = 543 reloads = 1082)
1922           TOTAL                                   65.40u  3.13s  6.00g
1923    
1924    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1925           MLRISC ra                               0.80u  0.10s  0.55g
1926              (spills = 42 reloads = 84)
1927           TOTAL                                   129.42u  5.10s  4.13g
1928    
1929    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1930           MLRISC ra                               2.69u  0.27s  1.38g
1931              (spills = 62 reloads = 124)
1932           TOTAL                                   682.48u  19.61s  13.99g
1933    
1934    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1935           MLRISC ra                               4.96u  0.27s  2.72g
1936              (spills = 85 reloads = 170)
1937           TOTAL                                   2375.26u  57.21s  48.00g
1938    
1939    As you can see the old cps spill module suffers from some serious
1940    performance problem.  But since I cannot decipher the old code fully,
1941    instead of patching the problems up, I'm reimplementing it
1942    with a different algorithm.  The new code is more modular,
1943    smaller when compiled, and substantially faster
1944    (O(n log n) time and O(n) space).  Timing of the new spill module:
1945    
1946    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1947          MLRISC ra                               0.25u  0.02s  0.15g
1948             (spills=1 reloads=3)
1949          TOTAL                                   7.74u  0.34s  1.62g
1950    
1951    Implementation details:
1952    
1953    As far as I can tell, the purpose of the CPS spill module is to make sure the
1954    number of live variables at any program point (the bandwidth)
1955    does not exceed a certain limit, which is determined by the
1956    size of the spill area.
1957    
1958    When the bandwidth is too large, we decrease the register pressure by
1959    packing live variables into spill records.  How we achieve this is
1960    completely different than what we did in the old code.
1961    
1962    First, there is something about the MLRiscGen code generator
1963    that we should be aware of:
1964    
1965    o MLRiscGen performs code motion!
1966    
1967       In particular, it will move floating point computations and
1968       address computations involving only the heap pointer to
1969       their use sites (if there is only a single use).
1970       What this means is that if we have a CPS record construction
1971       statement
1972    
1973           RECORD(k,vl,w,e)
1974    
1975       we should never count the new record address w as live if w
1976       has only one use (which is often the case).
1977    
1978       We should do something similar to floating point, but the transformation
1979       there is much more complex, so I won't deal with that.
1980    
1981    Secondly, there are now two new cps primops at our disposal:
1982    
1983     1. rawrecord of record_kind option
1984        This pure operator allocates some uninitialized storage from the heap.
1985        There are two forms:
1986    
1987         rawrecord NONE [INT n]  allocates a tagless record of length n
1988         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1989                                     and initializes the tag.
1990    
1991     2. rawupdate of cty
1992          rawupdate cty (v,i,x)
1993          Assigns to x to the ith component of record v.
1994          The storelist is not updated.
1995    
1996    We use these new primops for both spilling and increment record construction.
1997    
1998     1. Spilling.
1999    
2000        This is implemented with a linear scan algorithm (but generalized
2001        to trees).  The algorithm will create a single spill record at the
2002        beginning of the cps function and use rawupdate to spill to it,
2003        and SELECT or SELp to reload from it.  So both spills and reloads
2004        are fine-grain operations.  In contrast, in the old algorithm
2005        "spills" have to be bundled together in records.
2006    
2007        Ideally, we should sink the spill record construction to where
2008        it is needed.  We can even split the spill record into multiple ones
2009        at the places where they are needed.  But CPS is not a good
2010        representation for global code motion, so I'll keep it simple and
2011        am not attempting this.
2012    
2013     2. Incremental record construction (aka record splitting).
2014    
2015        Long records with many component values which are simulatenously live
2016        (recall that single use record addresses are not considered to
2017         be live) are constructed with rawrecord and rawupdate.
2018        We allocate space on the heap with rawrecord first, then gradually
2019        fill it in with rawupdate.  This is the technique suggested to me
2020        by Matthias.
2021    
2022        Some restrictions on when this is applicable:
2023        1. It is not a VECTOR record.  The code generator currently does not handle
2024           this case. VECTOR record uses double indirection like arrays.
2025        2. All the record component values are defined in the same "basic block"
2026           as the record constructor.  This is to prevent speculative
2027           record construction.
2028    
2029    ----------------------------------------------------------------------
2030    Name: Allen Leung
2031    Date: 2002/02/22 01:02:00 EST
2032    Tag: leunga-20020222-mlrisc-tools
2033    
2034    Minor bug fixes in the parser and rewriter
2035    
2036    ----------------------------------------------------------------------
2037    Name: Allen Leung
2038    Date: 2002/02/21 20:20:00 EST
2039    Tag: leunga-20020221-peephole
2040    
2041    Regenerated the peephole files.  Some contained typos in the specification
2042    and some didn't compile because of pretty printing bugs in the old version
2043    of 'nowhere'.
2044    
2045    ----------------------------------------------------------------------
2046    Name: Allen Leung
2047    Date: 2002/02/19 20:20:00 EST
2048    Tag: leunga-20020219-mlrisc-tools
2049    Description:
2050    
2051       Minor bug fixes to the mlrisc-tools library:
2052    
2053       1.  Fixed up parsing colon suffixed keywords
2054       2.  Added the ability to shut the error messages up
2055       3.  Reimplemented the pretty printer and fixed up/improved
2056           the pretty printing of handle and -> types.
2057       4.  Fixed up generation of literal symbols in the nowhere tool.
2058       5.  Added some SML keywords to to sml.sty
2059    
2060    ----------------------------------------------------------------------
2061    Name: Matthias Blume
2062    Date: 2002/02/19 16:20:00 EST
2063    Tag: blume-20020219-cmffi
2064    Description:
2065    
2066    A wild mix of changes, some minor, some major:
2067    
2068    * All C FFI-related libraries are now anchored under $c:
2069        $/c.cm      --> $c/c.cm
2070        $/c-int.cm  --> $c/internals/c-int.cm
2071        $/memory.cm --> $c/memory/memory.cm
2072    
2073    * "make" tool (in CM) now treats its argument pathname slightly
2074      differently:
2075        1. If the native expansion is an absolute name, then before invoking
2076           the "make" command on it, CM will apply OS.Path.mkRelative
2077           (with relativeTo = OS.FileSys.getDir()) to it.
2078        2. The argument will be passed through to subsequent phases of CM
2079           processing without "going native".  In particular, if the argument
2080           was an anchored path, then "make" will not lose track of that anchor.
2081    
2082    * Compiler backends now "know" their respective C calling conventions
2083      instead of having to be told about it by ml-nlffigen.  This relieves
2084      ml-nlffigen from one of its burdens.
2085    
2086    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
2087    
2088    * Export C_DEBUG and C_Debug from $c/c.cm.
2089    
2090    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
2091      subtyping relationship between incomplete pointers and their complete
2092      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
2093      with the convention of instantiating 'o with ('t, 'c) obj whenever
2094      the pointer target type is complete.  In the incomplete case, 'o
2095      will be instantiated with some "'c iobj" -- a type obtained by
2096      using one of the functors PointerToIncompleteType or PointerToCompleteType.
2097    
2098      Operations that work on both incomplete and complete pointer types are
2099      typed as taking an 'o ptr while operations that require the target to
2100      be known are typed as taking some ('t, 'c) obj ptr.
2101    
2102      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
2103      where void is an eqtype without any values.  This makes it possible
2104      to work on voidptr values using functions meant to operate on light
2105      incomplete pointers.
2106    
2107    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
2108      been vastly simplified.
2109    
2110    ----------------------------------------------------------------------
2111    Name: Matthias Blume
2112    Date: 2002/02/19 10:48:00 EST
2113    Tag: blume-20020219-pqfix
2114    Description:
2115    
2116    Applied Chris Okasaki's bug fix for priority queues.
2117    
2118    ----------------------------------------------------------------------
2119    Name: Matthias Blume
2120    Date: 2002/02/15 17:05:00
2121    Tag: Release_110_39
2122    Description:
2123    
2124    Last-minute retagging is becoming a tradition... :-(
2125    
2126    This is the working release 110.39.
2127    
2128    ----------------------------------------------------------------------
2129    Name: Matthias Blume
2130    Date: 2002/02/15 16:00:00 EST
2131    Tag: Release_110_39-orig
2132    Description:
2133    
2134    Working release 110.39.  New bootfiles.
2135    
2136    (Update: There was a small bug in the installer so it wouldn't work
2137    with all shells.  So I retagged. -Matthias)
2138    
2139    ----------------------------------------------------------------------
2140    Name: Matthias Blume
2141    Date: 2002/02/15 14:17:00 EST
2142    Tag: blume-20020215-showbindings
2143    Description:
2144    
2145    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
2146    the latter can be useful for exploring what bindings are available at
2147    the interactive prompt.  (The first function returns only the list
2148    of symbols that are really bound, the second prints those but also the
2149    ones that CM's autoloading mechanism knows about.)
2150    
2151    ----------------------------------------------------------------------
2152    Name: Matthias Blume
2153    Date: 2002/02/15 12:08:00 EST
2154    Tag: blume-20020215-iptrs
2155    Description:
2156    
2157    Two improvements to ml-nlffigen:
2158    
2159      1. Write files only if they do not exist or if their current contents
2160         do not coincide with what's being written.  (That is, avoid messing
2161         with the time stamps unless absolutely necessary.)
2162    
2163      2. Implement a "repository" mechanism for generated files related
2164         to "incomplete pointer types".   See the README file for details.
2165    
2166    ----------------------------------------------------------------------
2167    Name: Matthias Blume
2168    Date: 2002/02/14 11:50:00 EST
2169    Tag: blume-20020214-quote
2170    Description:
2171    
2172    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
2173    because of the new and improved tag generation scheme.  (Thanks to Allen
2174    Leung for pointing it out.)
2175    
2176    ----------------------------------------------------------------------
2177    Name: Lal George
2178    Date: 2002/02/14 09:55:27 EST 2002
2179    Tag: george-20020214-isabelle-bug
2180    Description:
2181    
2182    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
2183    of Isabelle on the x86.
2184    
2185    From Allen:
2186    -----------
2187     I've found the problem:
2188    
2189         in ra-core.sml, I use the counter "blocked" to keep track of the
2190         true number of elements in the freeze queue.  When the counter goes
2191         to zero, I skip examining the queue.  But I've messed up the
2192         bookkeeping in combine():
2193    
2194             else ();
2195             case !ucol of
2196               PSEUDO => (if !cntv > 0 then
2197                     (if !cntu > 0 then blocked := !blocked - 1 else ();
2198                                        ^^^^^^^^^^^^^^^^^^^^^^^
2199                      moveu := mergeMoveList(!movev, !moveu)
2200                     )
2201                  else ();
2202    
2203         combine() is called to coalesce two nodes u and v.
2204         I think I was thinking that if the move counts of u and v are both
2205         greater than zero then after they are coalesced then one node is
2206         removed from the freeze queue.  Apparently I was thinking that
2207         both u and v are of low degree, but that's clearly not necessarily true.
2208    
2209    
2210    02/12/2002:
2211        Here's the patch.  HOL now compiles.
2212    
2213        I don't know how this impact on performance (compile
2214        time or runtime).  This bug caused the RA (especially on the x86)
2215        to go thru the potential spill phase when there are still nodes on the
2216        freeze queue.
2217    
2218    
2219    
2220    
2221    ----------------------------------------------------------------------
2222    Name: Matthias Blume
2223    Date: 2002/02/13 22:40:00 EST
2224    Tag: blume-20020213-fptr-rtti
2225    Description:
2226    
2227    Fixed a bug in ml-nlffigen that was introduced with one of the previous
2228    updates.
2229    
2230    ----------------------------------------------------------------------
2231    Name: Matthias Blume
2232    Date: 2002/02/13 16:41:00 EST
2233    Tag: blume-20020213-cmlpq
2234    Description:
2235    
2236    Added new priority queue export symbols (which have just been added to
2237    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
2238    not compile and the installer would choke.)
2239    
2240    ----------------------------------------------------------------------
2241    Name: Matthias Blume
2242    Date: 2002/02/13 16:15:00 EST
2243    Tag: blume-20020213-various
2244    Description:
2245    
2246    1. More tweaks to ml-nlffigen:
2247    
2248       - better internal datastructures (resulting in slight speedup)
2249       - "-match" option requires exact match
2250       - "localized" gensym counters (untagged structs/unions nested within
2251         other structs/unions or within typedefs get a fresh counter; their
2252         tag will be prefixed by a concatenation of their parents' tags)
2253       - bug fixes (related to calculation of transitive closure of types
2254         to be included in the output)
2255    
2256    2. Minor Basis updates:
2257    
2258       - added implementations for List.collate and Option.app
2259    
2260    ----------------------------------------------------------------------
2261    Name: Matthias Blume
2262    Date: 2002/02/11 15:55:00 EST
2263    Tag: blume-20020211-gensym
2264    Description:
2265    
2266    Added a "-gensym" option to command line of ml-nlffigen.  This can be
2267    used to specify a "stem" -- a string that is inserted in all "gensym'd"
2268    names (ML structure names that correspond to unnamed C structs, unions,
2269    and enums), so that separate runs of ml-nlffigen do not clash.
2270    
2271    ----------------------------------------------------------------------
2272    Name: Matthias Blume
2273    Date: 2002/02/11 12:05:00 EST
2274    Tag: blume-20020211-gensml
2275    Description:
2276    
2277    A quick fix for a problem with GenSML (in the pgraph-util library):
2278    Make generation of toplevel "local" optional.  (Strictly speaking,
2279    signature definitions within "local" are not legal SML.)
2280    
2281    Other than that: updates to INSTALL and cm/TODO.
2282    
2283    ----------------------------------------------------------------------
2284    Name: Matthias Blume
2285    Date: 2002/02/08 15:00:00 EST
2286    Tag: blume-20020208-uniquepid
2287    Description:
2288    
2289    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
2290    
2291    1. The installer (config/install.sh) has gotten smarter:
2292    
2293         - Configuration options are a bit easier to specify now
2294           (in config/targets).
2295         - Bug in recognizing .tar.bz2 files fixed.
2296         - Installer automatically resolves dependencies between
2297           configuration options (e.g., if you ask for eXene, you will
2298           also get cml -- regardless whether you asked for it or not).
2299         - Installer can run in "quieter mode" by setting the environment
2300           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
2301           "completely silent", though.
2302         - Build HashCons library as part of smlnj-lib.
2303    
2304    2. A new scheme for assigning persistent identifiers to compilation
2305       units (and, by extension, to types etc.) has been put into place.
2306       This fixes a long-standing bug where types and even dynamic values
2307       can get internally confused, thereby compromising type safety
2308       (abstraction) and dynamic correctness.  See
2309    
2310         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
2311    
2312       for an example of how things could go wrong until now.
2313    
2314       The downside of the new scheme is that pids are not quite as
2315       persistent as they used to be: CM will generate a fresh pid
2316       for every compilation unit that it thinks it sees for the first
2317       time.  That means that if you compile starting from a clean, fresh
2318       source tree at two different times, you end up with different
2319       binaries.
2320    
2321       Cutoff recompilation, however, has not been compromised because
2322       CM keeps pid information in special caches between runs.
2323    
2324    ----------------------------------------------------------------------
2325    Name: Lal George
2326    Date: 2002/02/07 15:34:13 EST 2002
2327    Tag: <none>
2328    Description:
2329    
2330    Compilers that generate assembly code may produce  global labels
2331    whose value is resolved at link time. The various peephole optimization
2332    modules did not take this in account.
2333    
2334    TODO. The Labels.addrOf function should really return an option
2335    type so that clients are forced to deal with this issue, rather
2336    than an exception being raised.
2337    
2338    ----------------------------------------------------------------------
2339    Name: Lal George
2340    Date: 2002/02/06 13:55:02 EST
2341    Tag: george-20020206-ra-breakup
2342    Description:
2343    
2344    1. A bug fix from Allen.
2345    
2346        A typo causes extra fstp %st(0)'s to be generated at compensation
2347        edges, which might cause stack underflow traps at runtime.  This
2348        occurs in fft where there are extraneous fstps right before the 'into'
2349        trap instruction (in this case they are harmless since none of the
2350        integers overflow.)
2351    
2352    2. Pulled out various utility modules that were embedded in the modules
2353       of the register allocator. I need these modules for other purposes, but
2354       they are not complete enough to put into a library (just yet).
2355    ----------------------------------------------------------------------
2356    Name: Matthias Blume
2357    Date: 2002/01/31 16:05:00 EST
2358    Tag: blume-20020131-sparc-ccalls
2359    Description:
2360    
2361    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
2362       of extra stack space by mistake.  Fixed.
2363    
2364    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
2365    
2366    ----------------------------------------------------------------------
2367    Name: Allen Leung
2368    Date: 2002/01/30
2369    Tag: leunga-20020130-nowhere-bug-fix
2370    Description:
2371    
2372       MLRISC bug fixes:
2373       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
2374       2. MachineInt.fromString was negating its value.
2375    
2376    ----------------------------------------------------------------------
2377    Name: Matthias Blume
2378    Date: 2002/01/29
2379    Tag: blume-20020129-INSTALL
2380    Description:
2381    
2382    - Added somewhat detailed installation instructions (file INSTALL).
2383    - Fixed curl-detection bug in config/install.sh.
2384    - It is now possible to select the URL getter using the URLGETTER
2385      environment variable:
2386    
2387          not set / "unknown"      --> automatic detection (script tries wget,
2388                                       curl, and lynx)
2389          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
2390                                       how to properly invoke them)
2391          other                    --> use $URLGETTER directly, it must take
2392                                       precisely two command-line arguments
2393                                       (source URL and destination file name)
2394    
2395    ----------------------------------------------------------------------
2396    Name: Matthias Blume
2397    Date: 2002/01/28
2398    Tag: blume-20020128-sparc-ccalls
2399    Description:
2400    
2401    - Fixed problem with calculation of "used" registers in sparc-c-calls.
2402    - Make use of the allocParam argument in sparc-c-calls.
2403    
2404    ----------------------------------------------------------------------
2405    Name: Matthias Blume
2406    Date: 2002/01/28
2407    Tag: blume-20020128-allocParam
2408    Description:
2409    
2410    John Reppy:  Changes c-calls API to accept client-callback for
2411    allocating extra stack space.
2412    me: Corresponding changes to mlriscGen (using a dummy argument that
2413        does not change the current behavior).
2414    
2415    ----------------------------------------------------------------------
2416    Name: Matthias Blume
2417    Date: 2002/01/28 12:00:00
2418    Tag: Release_110_38
2419    Description:
2420    
2421    This time for real!!!
2422    
2423    ----------------------------------------------------------------------
2424    Name: Matthias Blume
2425    Date: 2002/01/28 10:56:00 EST
2426    Tag: blume-20020128-retraction
2427    Description:
2428    
2429    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
2430       with blume-Release_110_38-retracted.)
2431    
2432    1. Fixed a problem with incorrect rounding modes in real64.sml.
2433       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
2434    
2435    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
2436       and enums fixed.  The naming of corresponding ML identifiers should
2437       now be consistent again.
2438    
2439    ----------------------------------------------------------------------
2440    Name: Allen Leung
2441    Date: 2002/01/27
2442    Tag: leunga-20020127-nowhere
2443    Description:
2444    
2445       Added a target called nowhere in the configuration scripts.
2446       Enabling this will build the MLRISC 'nowhere' tool (for translating
2447       programs with where-clauses into legal SML code) during installation.
2448    
2449    ----------------------------------------------------------------------
2450    Name: Matthias Blume
2451    Date: 2002/01/25 21:27:00 EST
2452    Tag: blume-Release_110_38-retracted
2453    Description:
2454    
2455    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
2456    
2457    README will be added later.
2458    
2459    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
2460    (Release_110_38) removed.  Reason: Last-minute bug fixes.
2461    
2462    ----------------------------------------------------------------------
2463    Name: Matthias Blume
2464    Date: 2002/01/25
2465    Tag: blume-20020125-ffi
2466    Description:
2467    
2468    A large number of tweaks and improvements to ml-nlffi-lib and
2469    ml-nlffigen:
2470    
2471       - ML represenation types have been streamlined
2472       - getter and setter functions work with concrete values, not abstract
2473         ones where possible
2474       - ml-nlffigen command line more flexible (see README file there)
2475       - some bugs have been fixed (hopefully)
2476    
2477    ----------------------------------------------------------------------
2478    Name: Lal George
2479    Date: 2002/01/24
2480    Tag: george-20020124-risc-ra-interface
2481    Description:
2482    
2483       There is a dramatic simplification in the interface to the
2484       register allocator for RISC architectures as a result of making
2485       parallel copy instructions explicit.
2486    
2487    ----------------------------------------------------------------------
2488    Name: Matthias Blume
2489    Date: 2002/01/22
2490    Tag: blume-20020122-x86-ccalls
2491    Description:
2492    
2493    Bug fix for c-calls on x86 (having to do with how char- and
2494    short-arguments are being handled).
2495    
2496    ----------------------------------------------------------------------
2497    Name: Matthias Blume
2498    Date: 2002/01/21
2499    Tag: blume-20020121-ff
2500    Description:
2501    
2502    Another day of fiddling with the FFI...
2503    
2504    1. Bug fix/workaround:  CKIT does not complain about negative array
2505       dimensions, so ml-nlffigen has to guard itself against this possibility.
2506       (Otherwise a negative dimension would send it into an infinite loop.)
2507    
2508    2. Some of the abstract types (light objects, light pointers, most "base"
2509       types) in structure C are now eqtypes.
2510    
2511    3. Added constructors and test functions for NULL function pointers.
2512    
2513    ----------------------------------------------------------------------
2514    Name: Matthias Blume
2515    Date: 2002/01/18
2516    Tag: blume-20020118-ready-for-new-release
2517    Description:
2518    
2519    Made config/srcarchiveurl point to a new place.  (Will provide boot
2520    files shortly.)
2521    
2522    Maybe we christen this to be 110.38?
2523    
2524    ----------------------------------------------------------------------
2525    Name: Matthias Blume
2526    Date: 2002/01/18
2527    Tag: blume-20020118-more-ffifiddle
2528    Description:
2529    
2530    Today's FFI fiddling:
2531    
2532      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
2533        C.Set.* that use concrete (MLRep.*) arguments and results instead
2534        of abstract ones.
2535    
2536      - Provided word-style bit operations etc. for "int" representation
2537        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
2538        and Long.
2539    
2540    ----------------------------------------------------------------------
2541    Name: Matthias Blume
2542    Date: 2002/01/18
2543    Tag: blume-20020118-use-x86-fp
2544    Description:
2545    
2546    Now that x86-fast-fp seems to be working, I turned it back on again
2547    by default.  (Seems to work fine now, even with the FFI.)
2548    
2549    Other than that, I added some documentation about the FFI to
2550    src/ml-nlffigen/README and updated the FFI test examples in
2551    src/ml-nlffi-lib/Tests/*.
2552    
2553    ----------------------------------------------------------------------
2554    Name: Allen Leung
2555    Date: 2002/01/17
2556    Tag: leunga-20020117-x86-fast-fp-call
2557    Description:
2558    
2559       1. Fixed a problem with handling return fp values when x86's fast fp
2560          mode is turned on.
2561    
2562       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
2563          of %f32.
2564    
2565       3. Added a constructor INT32lit to the ast of MLRISC tools.
2566    
2567    ----------------------------------------------------------------------
2568    Name: Matthias Blume
2569    Date: 2002/01/16
2570    Tag: blume-20020116-ffifiddle
2571    Description:
2572    
2573    More fiddling with the FFI interface:
2574    
2575     - Make constness 'c instead of rw wherever possible.  This eliminates
2576       the need for certain explicit coercions.  (However, due to ML's
2577       value polymorphism, there will still be many cases where explicit
2578       coercions are necessary.  Phantom types are not the whole answer
2579       to modeling a subtyping relationship in ML.)
2580    
2581     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
2582    
2583     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
2584       clumsy workaround for varargs problem.)
2585    
2586    ----------------------------------------------------------------------
2587    Name: Lal George
2588    Date: 2002/01/15
2589    Tag: <none>
2590    Description:
2591    
2592    1. Since COPY instructions are no longer native to the architecture,
2593       a generic functor can be used to implement the expandCopies function.
2594    
2595    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
2596       TEXT segment.
2597    
2598    ----------------------------------------------------------------------
2599    Name: Matthias Blume
2600    Date: 2002/01/15
2601    Tag: blume-20020115-ffiupdates
2602    Description:
2603    
2604    1. Fix for bug resulting in single-precision float values being returned
2605       incorrectly from FFI calls.
2606    
2607    2. Small modifications to C FFI API:
2608    
2609        - memory-allocation routines return straight objects (no options)
2610          and raise an exception in out-of-memory situations
2611        - unsafe extensions to cast between function pointers and pointers
2612          from/to ints
2613        - added structure C_Debug as an alternative to structure C where
2614          pointer-dereferencing (|*| and |*!) always check for null-pointers
2615        - added open_lib' to DynLinkage;  open_lib' works like open_lib
2616          but also takes a (possibly empty) list of existing library handles
2617          that the current library depends on
2618    
2619    ----------------------------------------------------------------------
2620    Name: Matthias Blume
2621    Date: 2002/01/10
2622    Tag: blume-20020110-newffigen
2623    Description:
2624    
2625    1. Updates to portable graph code.
2626    
2627    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
2628       more scalable now so that even huge interfaces such as the one
2629       for GTK compile in finite time and space. :-)
2630       See src/ml-nlffigen/README for details on what's new.
2631    
2632    ----------------------------------------------------------------------
2633    Name: Lal George
2634    Date: 2001/01/09 14:31:35 EST 2002
2635    Tag: george-20011206-rm-native-copy
2636    Description:
2637    
2638            Removed the native COPY and FCOPY instructions
2639            from all the architectures and replaced it with the
2640            explicit COPY instruction from the previous commit.
2641    
2642            It is now possible to simplify many of the optimizations
2643            modules that manipulate copies. This has not been
2644            done in this change.
2645    
2646    ----------------------------------------------------------------------
2647    Name: Lal George
2648    Date: 2001/12/06 16:50:13 EST 2001
2649    Tag: george-20011206-mlrisc-instruction
2650    Description:
2651    
2652    Changed the representation of instructions from being fully abstract
2653    to being partially concrete. That is to say:
2654    
2655      from
2656            type instruction
2657    
2658      to
2659            type instr                              (* machine instruction *)
2660    
2661            datatype instruction =
2662                LIVE of {regs: C.cellset, spilled: C.cellset}
2663              | KILL of {regs: C.cellset, spilled: C.cellset}
2664              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
2665              | ANNOTATION of {i: instruction, a: Annotations.annotation}
2666              | INSTR of instr
2667    
2668    This makes the handling of certain special instructions that appear on
2669    all architectures easier and uniform.
2670    
2671    LIVE and KILL say that a list of registers are live or killed at the
2672    program point where they appear. No spill code is generated when an
2673    element of the 'regs' field is spilled, but the register is moved to
2674    the 'spilled' (which is present, more for debugging than anything else).
2675    
2676    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
2677    We used to generate:
2678    
2679            DEFFREG f1
2680            f1 := f2 + f3
2681            trapb
2682    
2683    but now generate:
2684    
2685            f1 := f2 + f3
2686            trapb
2687            LIVE {regs=[f1,f2,f3], spilled=[]}
2688    
2689    Furthermore, the DEFFREG (hack) required that all floating point instruction
2690    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
2691    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
2692    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
2693    this architecture).
2694    
2695    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
2696    all the architectures. This will result in further simplification of the
2697    register allocator that must be aware of them for coalescing purposes, and
2698    will also simplify certain aspects of the machine description that provides
2699    callbacks related to parallel copies.
2700    
2701    ANNOTATION should be obvious, and now INSTR represents the honest to God
2702    machine instruction set!
2703    
2704    The <arch>/instructions/<arch>Instr.sml files define certain utility
2705    functions for making porting easier -- essentially converting upper case
2706    to lower case. All machine instructions (of type instr) are in upper case,
2707    and the lower case form generates an MLRISC instruction. For example on
2708    the alpha we have:
2709    
2710      datatype instr =
2711         LDA of {r:cell, b:cell, d:operand}
2712       | ...
2713    
2714      val lda : {r:cell, b:cell, d:operand} -> instruction
2715        ...
2716    
2717    where lda is just (INSTR o LDA), etc.
2718    
2719    ----------------------------------------------------------------------
2720    Name: Matthias Blume
2721    Date: 2001/11/22 21:40:00 EST
2722    Tag: Release_110_37
2723    Description:
2724    
2725    Release 110.37.  This time for real.
2726    
2727    ----------------------------------------------------------------------
2728    Name: Matthias Blume
2729    Date: 2001/11/21 16:35:00 EST
2730    Tag: blume-20011121-foot-in-mouth
2731    Description:
2732    
2733    Removed the "Release_110_37" tag because of a serious bug.
2734    This will be re-tagged once the bug is fixed.
2735    
2736    ----------------------------------------------------------------------
2737    Name: Matthias Blume
2738    Date: 2001/11/21 16:14:00 EST
2739    Tag: blume-20011121-forgottenfile
2740    Description:
2741    
2742    Forgot to add a file.  (Just a .tex-file -- part of
2743    the CM manual source.)
2744    
2745    ----------------------------------------------------------------------
2746    Name: Matthias Blume
2747    Date: 2001/11/21 16:10:00 EST
2748    Tag: blume-20011121-invalid_110_37
2749    Description:
2750    
2751    Note: I removed the original tag "Release_110_37" from this commit
2752          because we found a serious bug in all non-x86 backends.
2753          - Matthias
2754    
2755    1. Modifications to the SML/NJ code generator and to the runtime system
2756       so that code object name strings are directly inserted into code
2757       objects at code generation time.  The only business the runtime system
2758       has with this is now to read the name strings on occasions.
2759       (The encoding of the name string has also changed somewhat.)
2760    
2761    2. CM now implements a simple "set calculus" for specifying export lists.
2762       In particular, it is now possible to refer to the export lists of
2763       other libraries/groups/sources and form unions as well as differences.
2764       See the latest CM manual for details.
2765    
2766    3. An separate notion of "proxy" libraries has again be eliminated from
2767       CM's model.  (Proxy libraries are now simply a special case of using
2768       the export list calculus.)
2769    
2770    4. Some of the existing libraries now take advantage of the new set
2771       calculus.
2772       (Notice that not all libraries have been converted because some
2773       of the existing .cm-files are supposed to be backward compatible
2774       with 110.0.x.)
2775    
2776    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2777       "General.exnMessage"!  The former relies on a certain hook to be
2778       initialized, and that often does not happen in the stand-alone case.)
2779    
2780    ----------------------------------------------------------------------
2781    Name: Lal George
2782    Date: 2001/11/21  13:56:18 EST
2783    Tag: george-2001121-pseudo-ops
2784    Description:
2785    
2786      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2787      ought to never be any question of incompatabilities with
2788      pseudo-op syntax expected by host assemblers.
2789    
2790      For now, only modules supporting GAS syntax are implemented
2791      but more should follow, such as MASM, and vendor assembler
2792      syntax, e.g. IBM as, Sun as, etc.
2793    
2794    ----------------------------------------------------------------------
2795    Name: Matthias Blume
2796    Date: 2001/11/14 11:52:00 EST
2797    Tag: blume-20011114-srcname
2798    Description:
2799    
2800    1. Routed the name of the current source file to mlriscgen where it
2801       should be directly emitted into the code object.  (This last part
2802       is yet to be done.)
2803    
2804    2. Some cleanup of the pgraph code to make it match the proposal that
2805       I put out the other day.  (The proposal notwithstanding, things are
2806       still in flux here.)
2807    
2808    ----------------------------------------------------------------------
2809    Name: Lal George
2810    Date: 2001/11/14 09:44:04 EST
2811    Tag:
2812    Description:
2813    
2814      Fix for a backpatching bug reported by Allen.
2815    
2816      Because the boundary between short and long span-dependent
2817      instructions is +/- 128, there are an astounding number of
2818      span-dependent instructions whose size is over estimated.
2819    
2820      Allen came up with the idea of letting the size of span
2821      dependent instructions be non-monotonic, for a maxIter
2822      number of times, after which the size must be monotonically
2823      increasing.
2824    
2825      This table shows the number of span-dependent instructions
2826      whose size was over-estimated as a function of maxIter, for the
2827      file Parse/parse/ml.grm.sml:
2828    
2829         maxIter            # of instructions:
2830            10                      687
2831            20                      438
2832            30                      198
2833            40                        0
2834    
2835      In compiling the compiler, there is no significant difference in
2836      compilation speed between maxIter=10 and maxIter=40. Actually,
2837      my measurements showed that maxIter=40 was a tad faster than
2838      maxIter=10! Also 96% of the  files in the compiler reach a fix
2839      point within 13 iterations, so fixing maxIter at 40, while high,
2840      is okay.
2841    
2842    ----------------------------------------------------------------------
2843    Name: Matthias Blume
2844    Date: 2001/10/31 15:25:00 EST
2845    Tag: blume-20011031-pgraph
2846    Description:
2847    
2848    CKIT:
2849    * Changed the "Function" constructor of type Ast.ctype to carry optional
2850      argument identifiers.
2851    * Changed the return type of TypeUtil.getFunction accordingly.
2852    * Type equality ignores the argument names.
2853    * TypeUtil.composite tries to preserve argument names but gives up quickly
2854      if there is a mismatch.
2855    
2856    installation script:
2857    * attempts to use "curl" if available (unless "wget" is available as well)
2858    
2859    CM:
2860    * has an experimental implementation of "portable graphs" which I will
2861      soon propose as an implementation-independent library format
2862    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2863    
2864    NLFFI-LIB:
2865    * some cleanup (all cosmetic)
2866    
2867    NLFFIGEN:
2868    * temporarily disabled the mechanism that suppresses ML output for
2869      C definitions whose identifiers start with an underscore character
2870    * generate val bindings for enum constants
2871    * user can request that only one style (light or heavy) is being used;
2872      default is to use both (command-line arguments: -heavy and -light)
2873    * fixed bug in handling of function types involving incomplete pointers
2874    * generate ML entry points that take record arguments (i.e., using
2875      named arguments) for C functions that have a prototype with named
2876      arguments
2877      (see changes to CKIT)
2878    
2879    ----------------------------------------------------------------------
2880    Name: Allen Leung
2881    Date: 2001/10/27 20:34:00 EDT
2882    Tag: leunga-20011027-x86-fast-fp-call
2883    Description:
2884    
2885       Fixed the bug described in blume-20010920-slowfp.
2886    
2887       The fix involves
2888          1. generating FCOPYs in FSTP in ia32-svid
2889          2. marking a CALL with the appropriate annotation
2890    
2891    ----------------------------------------------------------------------
2892    Name: Matthias Blume
2893    Date: 2001/10/16 11:32:00 EDT
2894    Tag: blume-20011016-netbsd
2895    Description:
2896    
2897    Underscore patch from Chris Richards (fixing problem with compiling
2898    runtime system under recent NetBSD).
2899    
2900    ----------------------------------------------------------------------
2901    Name: Allen Leung
2902    Date: 2001/10/12 17:18:32 EDT 2001
2903    Tag: leung-20011012-x86-printflowgraph
2904    Description:
2905    
2906    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2907    
2908    ----------------------------------------------------------------------
2909    Name: Lal George
2910    Date: 2001/10/11 23:51:34 EDT
2911    Tag: george-20011011-too-many-instrs
2912    Description:
2913    
2914    The representation of a program point never expected to see more
2915    than 65536 instructions in a basic block!
2916    
2917    ----------------------------------------------------------------------
2918    Name: Lal George
2919    Date: 2001/10/09 09:41:37 EDT
2920    Tag: george-20011008-mlrisc-labels
2921    Description:
2922    
2923    Changed the machine description files to support printing of
2924    local and global labels in assembly code, based on host assembler
2925    conventions.
2926    
2927    ----------------------------------------------------------------------
2928    Name: Matthias Blume
2929    Date: 2001/09/25 15:25:00 EDT
2930    Tag: blume-20010925-exninfo
2931    Description:
2932    
2933    I provided a non-hook implementation of exnName (at the toplevel) and
2934    made the "dummy" implementation of exnMessage (at the toplevel) more
2935    useful: if nothing gets "hooked in", then at least you are going to
2936    see the exception name and a message indicating why you don't see more.
2937    
2938    [For the time being, programs that need exnMessage and want to use
2939    ml-build should either use General.exnMessage (strongly recommended) or
2940    refer to structure General at some other point so that CM sees a
2941    static dependency.]
2942    
2943    [Similar remarks go for "print" and "use":  If you want to use their
2944    functionality in stand-alone programs generated by ml-build, then use
2945    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2946    
2947    ----------------------------------------------------------------------
2948    Name: Matthias Blume
2949    Date: 2001/09/20 17:28:00 EDT
2950    Tag: blume-20010920-slowfp
2951    Description:
2952    
2953    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2954    
2955    ----------------------------------------------------------------------
2956    Name: Matthias Blume
2957    Date: 2001/09/20 17:20:00 EDT
2958    Tag: blume-20010920-canonicalpaths
2959    Description:
2960    
2961    0. Updated the BOOT file (something that I forgot to do earlier).
2962    
2963    1. Small internal change to CM so that it avoids "/../" in filenames
2964       as much as possible (but only where it is safe).
2965    
2966    2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2967       that arguments that contain delimiters are passed through correctly.
2968       This change also means that all "special" arguments of the form
2969       @SMLxxx... must come first.
2970    
2971    3. Changed install script to put relative anchor names for tool commands
2972       into pathconfig.
2973    
2974    ----------------------------------------------------------------------
2975    Name: Matthias Blume
2976    Date: 2001/09/18 15:35:00 EDT
2977    Tag: blume-20010918-readme11036
2978    Description:
2979    
2980    Added README files.
2981    
2982    ----------------------------------------------------------------------
2983    Name: Matthias Blume
2984    Date: 2001/09/18 11:45:00 EDT
2985    Tag: Release_110_36 (retag)
2986    Description:
2987    
2988    Fixed mistake in config/preloads. Retagged as 110.36.
2989    
2990    ----------------------------------------------------------------------
2991    Name: Matthias Blume
2992    Date: 2001/09/18 09:40:00 EDT
2993    Tag: Release_110_36_orig (tag changed)
2994    Description:
2995    
2996    New version (110.36).  New bootfiles.
2997    
2998    ----------------------------------------------------------------------
2999    Name: Matthias Blume
3000    Date: 2001/09/14 16:15:00 EDT
3001    Tag: blume-20010914-x86fastfp
3002    Description:
3003    
3004    John committed some changes that Allen made, in particular a (hopefully)
3005    correctly working version of the x86-fp module.
3006    
3007    I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
3008    flag to "true".  Everything seems to compile to a fixpoint ok, and
3009    "mandelbrot" speeds up by about 15%.
3010    
3011    ----------------------------------------------------------------------
3012    Name: Matthias Blume
3013    Date: 2001/09/13 11:20:00 EDT
3014    Tag: blume-20010913-minimal
3015    Description:
3016    
3017    1. Stefan Monnier's patch to fix a miscompilation problem that
3018       was brought to light by John Reppy's work on Moby.
3019    
3020    2. Implemented a minimal "structure Compiler" that contains just
3021       "version" and "architecture".  The minimal version will be
3022       available when the full version is not.  This is for backward-
3023       compatibility with code that wants to test Compiler.version.
3024    
3025    ----------------------------------------------------------------------
3026    Name: Matthias Blume
3027    Date: 2001/08/28 14:03:00 EDT
3028    Tag: blume-20010828-ml-lex
3029    Description:
3030    
3031    Fix for bug 1581, received from Neophytos Michael.
3032    
3033    ----------------------------------------------------------------------
3034    Name: Matthias Blume
3035    Date: 2001/08/27 11:20:00 EDT
3036    Tag: blume-20010827-readme11035
3037    Description:
3038    
3039    Fleshed out the README file for 110.35.
3040    
3041    ----------------------------------------------------------------------
3042    Name: Matthias Blume
3043    Date: 2001/08/24 17:10:00 EDT
3044    Tag: Release_110_35
3045    Description:
3046    
3047    New version number (110.35).  New bootfiles.
3048    
3049    ----------------------------------------------------------------------
3050    Name: Lal George
3051    Date: 2001/08/24 13:47:18 EDT 2001
3052    Tag: george-20010824-MLRISC-graphs
3053    Description:
3054    
3055     removed clusters from MLRISC completely and replaced with graphs.
3056    
3057    ----------------------------------------------------------------------
3058    Name: Matthias Blume
3059    Date: 2001/08/23 17:50:00 EDT
3060    Tag: blume-20010823-toplevel
3061    Description:
3062    
3063    - some reorganization of the code that implements various kinds of
3064      environments in the compiler (static, dynamic, symbolic, combined)
3065    - re-implemented the EnvRef module so that evalStream works properly
3066      (if the stream contains references to "use", "CM.make", etc.)
3067    - cleaned up evalloop.sml and interact.sml (but they need more cleaning)
3068    
3069    ----------------------------------------------------------------------
3070    Name: Matthias Blume
3071    Date: 2001/08/20 15:50 EDT
3072    Tag: blume20010820-slipup
3073    Description:
3074    
3075    I forgot to commit a few files.  Here they are...
3076    
3077    ----------------------------------------------------------------------
3078    Name: Matthias Blume
3079    Date: 2001/08/20 15:35:00 EDT
3080    Tag: blume-20010820-debugprof
3081    Description:
3082    
3083    !!!! NEW BOOTFILES !!!!
3084    
3085    This is another round of reorganizing the compiler sources.  This
3086    time the main goal was to factor out all the "instrumentation"
3087    passes (for profiling and backtracing) into their own library.
3088    The difficulty was to do it in such a way that it does not depend
3089    on elaborate.cm but only on elabdata.cm.
3090    
3091    Therefore there have been further changes to both elaborate.cm and
3092    elabdata.cm -- more "generic" things have been moved from the former
3093    to the latter.  As a result, I was forced to split the assignment
3094    of numbers indicating "primtyc"s into two portions: SML-generic and
3095    SML/NJ-specific.  Since it would have been awkward to maintain,
3096    I bit the bullet and actually _changed_ the mapping between these
3097    numbers and primtycs.  The bottom line of this is that you need
3098    a new set of bin- and bootfiles.
3099    
3100    I have built new bootfiles for all architectures, so doing a fresh
3101    checkout and config/install.sh should be all you need.
3102    
3103    The newly created library's name is
3104    
3105        $smlnj/viscomp/debugprof.cm
3106    
3107    and its sources live under
3108    
3109        src/compiler/DebugProf
3110    
3111    ----------------------------------------------------------------------
3112    Name: Matthias Blume
3113    Date: 2001/08/15 17:15:00 EDT
3114    Tag: blume-20010815-compreorg
3115    Description:
3116    
3117    This is a first cut at reorganizing the CM libraries that make up the
3118    core of the compiler.  The idea is to separate out pieces that could
3119    be used independently by tools, e.g., the parser, the typechecker, etc.
3120    
3121    The current status is a step in this direction, but it is not quite
3122    satisfactory yet.  Expect more changes in the future.
3123    
3124    Here is the current (new) organization...
3125    
3126        What used to be $smlnj/viscomp/core.cm is now divided into
3127        six CM libraries:
3128    
3129             $smlnj/viscomp/basics.cm
3130                           /parser.cm
3131                           /elabdata.cm
3132                           /elaborate.cm
3133                           /execute.cm
3134                           /core.cm
3135    
3136        The CM files for these libraries live under src/system/smlnj/viscomp.
3137        All these libraries are proxy libraries that contain precisely
3138        one CM library component.  Here are the locations of the components
3139        (all within the src/compiler tree):
3140    
3141             Basics/basics.cm
3142             Parse/parser.cm
3143             ElabData/elabdata.cm
3144             Elaborator/elaborate.cm
3145             Execution/execute.cm
3146             core.cm
3147    
3148         [This organization is the same that has been used already
3149         for a while for the architecture-specific parts of the visible
3150         compiler and for the old version of core.cm.]
3151    
3152         As you will notice, many source files have been moved from their
3153         respective original locations to a new home in one of the above
3154         subtrees.
3155    
3156         The division of labor between the new libraries is the following:
3157    
3158             basics.cm:
3159                - Simple, basic definitions that pertain to many (or all) of
3160                  the other libraries.
3161             parser.cm:
3162                - The SML parser, producing output of type Ast.dec.
3163                - The type family for Ast is also defined and exported here.
3164             elabdata.cm:
3165                - The datatypes that describe input and output of the elaborator.
3166                  This includes types, absyn, and static environments.
3167             elaborator.cm:
3168                - The SML/NJ type checker and elaborator.
3169                  This maps an Ast.dec (with a given static environment) to
3170                  an Absyn.dec (with a new static environment).
3171                - This libraries implements certain modules that used to be
3172                  structures as functors (to remove dependencies on FLINT).
3173             execute.cm:
3174                - Everything having to do with executing binary code objects.
3175                - Dynamic environments.
3176             core.cm:
3177                - SML/NJ-specific instantiations of the elaborator and MLRISC.
3178                - Top-level modules.
3179                - FLINT (this should eventually become its own library)
3180    
3181    Notes:
3182    
3183    I am not 100% happy with the way I separated the elaborator (and its
3184    data structures) from FLINT.  Two instances of the same problem:
3185    
3186        1. Data structures contain certain fields that carry FLINT-specific
3187           information.  I hacked around this using exn and the property list
3188           module from smlnj-lib.  But the fact that there are middle-end
3189           specific fields around at all is a bit annoying.
3190    
3191        2. The elaborator calculates certain FLINT-related information.  I tried
3192           to make this as abstract as I could using functorization, but, again,
3193           the fact that the elaborator has to perform calculations on behalf
3194           of the middle-end at all is not nice.
3195    
3196        3. Having to used exn and property lists is unfortunate because it
3197           weakens type checking.  The other alternative (parameterizing
3198           nearly *everything*) is not appealing, though.
3199    
3200    I removed the "rebinding =" warning hack because due to the new organization
3201    it was awkward to maintain it.  As a result, the compiler now issues some of
3202    these warnings when compiling init.cmi during bootstrap compilation. On
3203    the plus side, you also get a warning when you do, for example:
3204       val op = = Int32.+
3205    which was not the case up to now.
3206    
3207    I placed "assign" and "deref" into the _Core structure so that the
3208    code that deals with the "lazy" keyword can find them there.  This
3209    removes the need for having access to the primitive environment
3210    during elaboration.
3211    
3212    ----------------------------------------------------------------------
3213    Name: Matthias Blume
3214    Date: 2001/08/13
3215    Tag: blume-20010813-closures
3216    Description:
3217    
3218    This fix was sent to us by Zhong Shao.  It is supposed to improve the
3219    performance of certain loops by avoiding needless closure allocation.
3220    
3221    ----------------------------------------------------------------------
3222    Name: Lal George
3223    Date: 2001/07/31 10:03:23 EDT 2001
3224    Tag: george-20010731-x86-fmalloc
3225    Description: Fixed bug in x86 calls
3226    
3227        There was a bug where call instructions would mysteriously
3228        vanish. The call instruction had to be one that returned
3229        a floating point value.
3230    
3231    ----------------------------------------------------------------------
3232    Name: Lal George
3233    Date: 2001/07/19 16:36:29 EDT 2001
3234    Tag: george-20010719-simple-cells
3235    Description:
3236    
3237    I have dramatically simplified the interface for CELLS in MLRISC.
3238    
3239    In summary, the cells interface is broken up into three parts:
3240    
3241      1. CellsBasis : CELLS_BASIS
3242    
3243            CellsBasis is a top level structure and common for all
3244            architectures.  it contains the definitions of basic datatypes
3245            and utility  functions over these types.
3246    
3247      2. functor Cells() : CELLS
3248    
3249            Cells generates an interface for CELLS that incorporates the
3250            specific resources on the target architecture, such as the
3251            presence of special register classes, their number and size,
3252            and various useful substructures.
3253    
3254      3. <ARCH>CELLS
3255    
3256            e.g. SparcCells: SPARCCELLS
3257    
3258            <ARCH>CELLS usually contains additional bindings for special
3259            registers  on the architecture, such as:
3260    
3261                    val r0 : cell           (* register zero *)
3262                    val y : cell            (* Y register *)
3263                    val psr : cell          (* processor status register *)
3264                    ...
3265    
3266            The structure returned by applying the Cells functor is opened
3267            in this interface.
3268    
3269    The main implication of all this is that the datatypes for cells is
3270    split between CellsBasis and CELLS -- a fairly simple change for user
3271    code.
3272    
3273    In the old scheme the CELLS interface had a definitional binding of
3274    the form:
3275    
3276            signature CELLS = sig
3277    
3278               structure CellsBasis = CellsBasis
3279    
3280               ...
3281    
3282            end
3283    
3284    With all the sharing constraints that goes on in MLRISC, this old
3285    design  quickly leads to errors such as:
3286    
3287            "structure definition spec inside of sharing ... "
3288    
3289    
3290    and appears to require an unacceptable amount of sharing and where
3291    constraint hackery.
3292    
3293    I think this error message (the interaction of definitional specs and
3294    sharing) requires more explanation on our web page.
3295    
3296    ----------------------------------------------------------------------
3297    Name: Matthias Blume
3298    Date: 2001/07/19 15:00:00 EDT
3299    Tag: blume-20010719-libreorg
3300    Description:
3301    
3302    This update puts together a fairly extensive but straightforward change
3303    to the way the libraries that implement the interactive system are
3304    organized:
3305    
3306       The biggest change is the elimination of structure Compiler.  As a
3307       replacement for this structure, there is now a CM library
3308       (known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
3309       that exports all the substructures of the original structure Compiler
3310       directly.  So instead of saying Compiler.Foo.bar one now simply
3311       says Foo.bar.  (The CM libraries actually export a collection of
3312       structures that is richer than the collection of substructures of
3313       structure Compiler.)
3314    
3315       To make the transition smooth, there is a separate library called
3316       $smlnj/compiler/compiler.cm which puts together and exports the
3317       original structure Compiler (or at least something very close to it).
3318    
3319       There are five members of the original structure Compiler
3320       that are not exported directly but which instead became members
3321       of a new structure Backend (described by signature BACKEND).  These are:
3322       structure Profile (: PROFILE), structure Compile (: COMPILE), structure
3323       Interact (: INTERACT), structure Machine (: MACHINE), and val
3324       architecture (: string).
3325    
3326       Structure Compiler.Version has become structure CompilerVersion.
3327    
3328       Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
3329       by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
3330       or x86, respectively.
3331       Each of these exports the same frontend structures that
3332       $smlnj/compiler.cm exports.  But they do not have a structure Backend
3333       and instead export some structure <Arch>Backend where <Arch> is Alpha32,
3334       Hppa, PPC, Sparc, or X86, respectively.
3335    
3336       Library $smlnj/compiler/all.cm exports the union of the exports of
3337       $smlnj/compiler/<arch>.cm
3338    
3339       There are no structures <Arch>Compiler anymore, use
3340       $smlnj/compiler/<arch>.cm instead.
3341    
3342       Library host-compiler-0.cm is gone.  Instead, the internal library
3343       that instantiates CM is now called cm0.cm.  Selection of the host
3344       compiler (backend) is no longer done here but. (Responsibility for it
3345       now lies with $smlnj/compiler/current.cm.  This seems to be more
3346       logical.)
3347    
3348       Many individual files have been moved or renamed.  Some files have
3349       been split into multiple files, and some "dead" files have been deleted.
3350    
3351    Aside from these changes to library organization, there are also changes
3352    to the way the code itself is organized:
3353    
3354       Structure Binfile has been re-implemented in such a way that it no
3355       longer needs any knowledge of the compiler.  It exclusively deals
3356       with the details of binfile layout.  It no longer invokes the
3357       compiler (for the purpose of creating new prospective binfile
3358       content), and it no longer has any knowledge of how to interpret
3359       pickles.
3360    
3361       Structure Compile (: COMPILE) has been stripped down to the bare
3362       essentials of compilation.  It no longer deals with linking/execution.
3363       The interface has been cleaned up considerably.
3364    
3365       Utility routines for dealing with linking and execution have been
3366       moved into their own substructures.
3367    
3368       (The ultimate goal of these changes is to provide a light-weight
3369       binfile loader/linker (at least for, e.g., stable libraries) that
3370       does not require CM or the compiler to be present.)
3371    
3372    CM documentation has been updated to reflect the changes to library
3373    organization.
3374    
3375    ----------------------------------------------------------------------
3376    Name: Matthias Blume
3377    Date: 2001/07/10 17:30:00 EDT
3378    Tag: Release_110_34
3379    Description:
3380    
3381    Minor tweak to 110.34 (re-tagged):
3382    
3383      - README.html file added to CVS repository
3384      - runtime compiles properly under FreeBSD 3.X and 4.X
3385    
3386    ----------------------------------------------------------------------
3387    Name: Matthias Blume
3388    Date: 2001/07/10 17:30:00 EDT
3389    Tag: Release_110_34
3390    Description:
3391    
3392    New version number (110.34). New bootfiles.
3393    
3394    ----------------------------------------------------------------------
3395    Name: Matthias Blume
3396    Date: 2001/07/09 16:00:00 EDT
3397    Tag: blume-20010709-more-varargs
3398    Description:
3399    
3400    I changed the handling of varargs in ml-nlffigen again:
3401    The ellipsis ... will now simply be ignored (with an accompanying warning).
3402    
3403    The immediate effect is that you can actually call a varargs function
3404    from ML -- but you can't actually supply any arguments beyond the ones
3405    specified explicitly.  (For example, you can call printf with its format
3406    string, but you cannot pass additional arguments.)
3407    
3408    This behavior is only marginally more useful than the one before, but
3409    it has the advantage that a function or, more importantly, a function
3410    type never gets dropped on the floor, thus avoiding follow-up problems with
3411    other types that refer to the offending one.
3412    
3413    ----------------------------------------------------------------------
3414    Name: Matthias Blume
3415    Date: 2001/07/09 11:25:00 EDT
3416    Tag: blume-20010709-varargs
3417    Description:
3418    
3419    1. ckit-lib.cm now exports structure Error
3420    2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
3421       with a warning accompanied by a source location.  Moreover, it
3422       merely skips the offending function or type and proceeds with the
3423       rest of its work.u  As a result, one can safely feed C code containing
3424       "..." to ml-nlffigen.
3425    3. There are some internal improvements to CM, providing slightly
3426       more general string substitutions in the tools subsystem.
3427    
3428    ----------------------------------------------------------------------
3429    Name: Matthias Blume
3430    Date: 2001/06/27 15:10:00 EDT
3431    Tag: blume-20010627-concur
3432    Description:
3433    
3434    Fixed a small bug in CM's handling of parallel compilation.
3435    (You could observe the bug by Control-C-interrupting an ordinary
3436    CMB.make or CM.stabilize and then attaching some compile servers.
3437    The result was that all of a sudden the previously interrupted
3438    compilation would continue on its own.  This was because of
3439    an over-optimization: CM did not bother to clean out certain queues
3440    when no servers were attached "anyway", resulting in the contents
3441    of these queues to grab control when new servers did get attached.)
3442    
3443    There is also another minor update to the CM manual.
3444    
3445    ----------------------------------------------------------------------
3446    Name: Matthias Blume
3447    Date: 2001/06/26 16:15:00 EDT
3448    Tag: blume-20010626-cmdoc
3449    Description:
3450    
3451    Minor typo fixed in CM manual (syntax diagram for libraries).
3452    
3453    ----------------------------------------------------------------------
3454    Name: Matthias Blume
3455    Date: 2001/06/25 22:55:00 EDT
3456    Tag: blume-20010625-x86pc
3457    Description:
3458    
3459    Fixed a nasty bug in the X86 assembly code that caused signal
3460    handlers to fail (crash) randomly.
3461    
3462    ----------------------------------------------------------------------
3463    Name: Matthias Blume
3464    Date: 2001/06/25 12:05:00 EDT
3465    Tag: blume-20010625-nlffigen
3466    Description:
3467    
3468    This update fixes a number of minor bugs in ml-nlffigen as reported by
3469    Nick Carter <nbc@andrew.cmu.edu>.
3470    
3471      1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
3472      2. Default names for generated files are now derived from the name of
3473         the C file *without its directory*.  In particular, this causes generated
3474         files to be placed locally even if the C file is in some system directory.
3475      3. Default names for generated signatures and structures are also derived
3476         from the C file name without its directory.  This avoids silly things
3477         like "structure GL/GL".
3478         (Other silly names are still possible because ml-nlffigen does not do
3479          a thorough check of whether generated names are legal ML identifiers.
3480          When in doubt, use command line arguments to force particular names.)
3481    
3482    ----------------------------------------------------------------------
3483    Name: Matthias Blume
3484    Date: 2001/06/21 12:25:00 EDT
3485    Tag: blume-20010621-eXene
3486    Description:
3487    
3488    eXene now compiles and (sort of) works again.
3489    
3490    The library name (for version > 110.33) is $/eXene.cm.
3491    
3492    I also added an new example in src/eXene/examples/nbody.  See the
3493    README file there for details.
3494    
3495    ----------------------------------------------------------------------
3496    Name: Matthias Blume
3497    Date: 2001/06/20 16:40:00 EDT
3498    Tag: blume-20010620-cml
3499    Description:
3500    
3501    CML now compiles and works again.
3502    
3503    Libraries (for version > 110.33):
3504    
3505      $cml/cml.cm            Main CML library.
3506      $cml/basis.cm          CML's version of $/basis.cm.
3507      $cml/cml-internal.cm   Internal helper library.
3508      $cml/core-cml.cm       Internal helper library.
3509      $cml-lib/trace-cml.cm  Tracing facility.
3510      $cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
3511    
3512    The installer (config/install.sh) has been taught how to properly
3513    install this stuff.
3514    
3515    ----------------------------------------------------------------------
3516    Name: Matthias Blume
3517    Date: 2001/06/19 17:55:00 EDT
3518    Tag: blume-20010619-instantiate
3519    Description:
3520    
3521    This un-breaks the fix for bug 1432.
3522    (The bug was originally fixed in 110.9 but I broke it again some
3523    time after that.)
3524    
3525    ----------------------------------------------------------------------
3526    Name: Matthias Blume
3527    Date: 2001/06/19 17:25:00 EDT
3528    Tag: blume-20010619-signals
3529    Description:
3530    
3531    This should (hopefully) fix the long-standing signal handling bug.
3532    (The runtime system was constructing a continuation record with an
3533    incorrect descriptor which would cause the GC to drop data on the floor...)
3534    
3535    ----------------------------------------------------------------------
3536    Name: Matthias Blume
3537    Date: 2001/06/15 15:05:00 EDT
3538    Tag: blume-20010615-moresparc
3539    Description:
3540    
3541    Here is a short late-hour update related to Sparc c-calls:
3542    
3543     -- made handling of double-word arguments a bit smarter
3544    
3545     -- instruction selection phase tries to collapse certain clumsily
3546        constructed ML-Trees; typical example:
3547    
3548            ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
3549    
3550        This currently has no further impact on SML/NJ since mlriscGen does
3551        not seem to generate such patterns in the first place, and c-calls
3552        (which did generate them in the beginning) has meanwhile been fixed
3553        so as to avoid them as well.
3554    
3555    ----------------------------------------------------------------------
3556    Name: Matthias Blume
3557    Date: 2001/06/15 15:05:00 EDT
3558    Tag: blume-20010615-sparc
3559    Description:
3560    
3561    The purpose of this update is to provide an implementation of NLFFI
3562    on Sparc machines.
3563    
3564    Here are the changes in detail:
3565    
3566       * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
3567       the Sparc implementation of the c-calls API.
3568       * The Sparc backend of SML/NJ has been modified to uniformely use %fp
3569       for accessing the ML frame.  Thus, we have a real frame pointer and
3570       can freely modify %sp without need for an omit-frame-ptr phase.
3571       The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
3572       this case.
3573       * ml-nlffigen has been taught to produce code for different architectures
3574       and calling conventions.
3575       * In a way similar to what was done in the x86 case, the Sparc
3576       backend uses its own specific extension to mltree.  (For example,
3577       it needs to be able to generate UNIMP instructions which are part
3578       of the calling convention.)
3579       * ml-nlffi-lib was reorganized to make it more modular (in particular,
3580       to make it easier to plug in new machine- and os-dependent parts).
3581    
3582    There are some other fairly unrelated bug fixes and cleanups as well:
3583    
3584       * I further hacked the .cm files for MLRISC tools (like MDLGen) so
3585       that they properly share their libraries with existing SML/NJ libraries.
3586       * I fixed a minor cosmetic bug in CM, supressing certain spurious
3587       follow-up error messages.
3588       * Updates to CM/CMB documentation.
3589    
3590    TODO items:
3591    
3592       * MLRISC should use a different register as its asmTemp on the Sparc.
3593         (The current %o2 is a really bad choice because it is part of the
3594         calling conventions, so things might interfere in unexpected ways.)
3595    
3596    ----------------------------------------------------------------------
3597    Name: Matthias Blume
3598    Date: 2001/06/07
3599    Tag: blume-20010607-calls
3600    Description:
3601    
3602    A number of internal changes related to C calls and calling conventions:
3603    
3604    1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
3605       number of bytes popped implicitly (by the callee).  In most cases
3606       this field is 0 but on x86/win32 it is some non-zero value.  This
3607       is information provided for the benefit of the "omit-frameptr" pass.
3608    2. The CALL instruction on the x86 carries a similar "pops" field.
3609       The instruction selection phase copies its value from the ML-Tree
3610       CALL statement.
3611    3. On all other architectures, the instruction selection phase checks
3612 &n