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 715, Thu Nov 2 07:51:04 2000 UTC revision 1468, Fri Mar 19 18:54:52 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: Allen Leung (leunga (at) reservoir (dot) com)
17    Date: 2004/03/19 14:40:00 EST
18    Tag: leunga-20040319-cygwin-nlffi
19    Description:
20    
21    Fixed the runtime so that ml-nlffi-lib runs on the cygwin version
22    of SML/NJ.  The problem is that
23    
24       lib = dlopen(NULL, ...)
25       f   = dlsym(lib, "malloc");
26    
27    does not work on Windows unless we explicitly export symbols
28    such as 'malloc' during linking.  We fixed this by explicitly
29    exporting the required symbols with the magic gcc incantation:
30    
31        -Wl,--export-all cygwin.def
32    
33    where cygwin.def is a file containing all the symbols that
34    we wish to export.
35    
36    I suspect this is a Windows problem and we'll have to
37    do the same (somehow with windows compilers) when
38    we build the native win32 version with the system
39    calls LoadLibrary/GetProcAddress.
40    
41    ----------------------------------------------------------------------
42    Name: Matthias Blume (blume (at) tti - c (dot) org)
43    Date: 2004/03/04 16:35:00 CST
44    Tag: blume-20040304-intinf-fmt
45    Description:
46    
47    Fixed problem with IntInf.fmt (sign would show up on the right instead
48    of on the left for BIN, OCT, and HEX).
49    
50    ----------------------------------------------------------------------
51    Name: Matthias Blume (blume (at) tti - c (dot) org)
52    Date: 2004/03/04 11:25:00 CST
53    Tag: blume-20040304-symlinks
54    Description:
55    
56    Fixed problem with installer script (unix only) where bin/ml-yacc and
57    friends pointed (via symlinks) to absolute locations instead of just
58    .run-sml.  This was reported by Vesa A Norrman.
59    
60    ----------------------------------------------------------------------
61    Name: Matthias Blume (blume (at) tti - c (dot) org)
62    Date: 2004/02/13 14:50:00 CST
63    Tag: Release_110_45
64    Description:
65    
66    New working version (110.45). New bootfiles.
67    
68    ----------------------------------------------------------------------
69    Name: Matthias Blume (blume (at) tti - c (dot) org)
70    Date: 2004/01/26 15:15:15 CST
71    Tag: blume-20040126-toplevel
72    Description:
73    
74    Improved handling of exceptions at the interactive toplevel.
75    
76    ----------------------------------------------------------------------
77    Name: Matthias Blume (blume (at) tti - c (dot) org)
78    Date: 2004/01/26 11:25:00
79    Tag: blume-20040126-app
80    Description:
81    
82    Type of top-level "app" corrected.
83    Added code for setting vp_limitPtrMask to Win32-specific runtime.
84    
85    ----------------------------------------------------------------------
86    Name: Matthias Blume (blume (at) tti - c (dot) org)
87    Date: 2003/11/18 17:10 CST
88    Tag: blume-20031118-basis-fiddle
89    Description:
90    
91    - changed Timer interface to what might become the spec
92    - POSIX_FLAGS -> BIT_FLAGS according to spec
93    - some other minor discrepancies wrt. spec eliminated
94    
95    ----------------------------------------------------------------------
96    Name: Matthias Blume (blume (at) tti - c (dot) org)
97    Date: 2003/11/06 12:00:00 CST
98    Tag: Release_110_44
99    Description:
100    
101    New working version (110.44). New bootfiles.
102    
103    ----------------------------------------------------------------------
104    Name: Matthias Blume (blume (at) tti - c (dot) org)
105    Date: 2003/11/04 11:50:00 CST
106    Tag: blume-20031104-move-libraries
107    Description:
108    
109    Eliminated the "dont_move_libraries" directive in config/targets.
110    (The mechanism was broken and could not be fixed easily.  Moreover,
111    there does not seem to be any reason not to move all libraries into
112    lib during installation.  I originally implemented this directive as a
113    backward-compatibility feature when I first introduced the new CM.
114    Now that things have been stable for a long time and going back to the
115    old CM is not an option, there is no reason to keep it around.)
116    
117    ----------------------------------------------------------------------
118    Name: Matthias Blume (blume (at) tti - c (dot) org)
119    Date: 2003/11/03 16:00:00 CST
120    Tag: blume-20031103-installdir
121    Description:
122    
123    Made installer honor INSTALLDIR variable again.  (Thanks to Chris
124    Richards for pointing out the problem and providing the solution.)
125    
126    ----------------------------------------------------------------------
127    Name: Matthias Blume (blume (at) tti - c (dot) org)
128    Date: 2003/10/01 17:05:00 CDT
129    Tag: blume-20031001-lal-mlrisc
130    Description:
131    
132    MLRISC bug fix from Lal.
133    
134    ----------------------------------------------------------------------
135    Name: Matthias Blume (blume (at) tti - c (dot) org)
136    Date: 2003/09/30 16:10:00 CDT
137    Tag: blume-20030930-primio-bat
138    Description:
139    
140    1. Added openVector, nullRd, and nullWr to PRIM_IO.
141    2. Improved .bat files (for Win32 port) to make things work under Win95.
142       (thanks to Aaron S. Hawley for this one)
143    
144    ----------------------------------------------------------------------
145    Name: Matthias Blume (blume (at) tti - c (dot) org)
146    Date: 2003/09/26 16:05:00 CDT
147    Tag: blume-20030926-wrappriv
148    Description:
149    
150    Added missing wrapper for privilege "primitive" in $smlnj/viscomp/core.cm.
151    
152    ----------------------------------------------------------------------
153    Name: Matthias Blume (blume (at) tti - c (dot) org)
154    Date: 2003/09/26 15:00:00 CDT
155    Tag: blume-20030926-110_43_3
156    Description:
157    
158    - additional cleanup
159    - version number bump, NEW BOOTFILES
160    
161    ----------------------------------------------------------------------
162    Name: Matthias Blume (blume (at) tti - c (dot) org)
163    Date: 2003/09/26 12:00:00 CDT
164    Tag: blume-20030926-ppautoload
165    Description:
166    
167    I modified the read-eval-print loop so that the autoloader gets
168    invoked whenever the prettyprinter tries to look up a symbol that
169    is not currently defined in the toplevel environment but which
170    appears in CM's autoload registry.  As a result, we see far fewer of
171    those ?.Foo.Bar.xxx names in the prettyprinter's output.
172    
173    In addition to this I tried to clean up some pieces of the Basis
174    implementation (e.g., Socket, Word8Array) in order to prevent other
175    instances of these ?.Foo.Bar.xxx names from being printed.
176    
177    The mechanism that picks names for types still needs some work, though.
178    (Right now it seems that if there is a type A.t which is defined to
179    be B.u, but B is unavailable at toplevel, then A.t gets printed as
180    "?.B.u" although the perhaps more sensible solution would be to use
181    "A.t" in this case.  In other words, the prettyprinter should follow
182    a chain of DEFtycs not farther than there are corresponding toplevel
183    names in the current environment.)
184    
185    ----------------------------------------------------------------------
186    Name: Matthias Blume (blume (at) tti - c (dot) org)
187    Date: 2003/09/24 16:31:00 CDT
188    Tag: blume-20030924-installer
189    Description:
190    
191    Another installer tweak:  All the ML code for the installer is now
192    compiled during CMB.make and put into a little library called
193    $smlnj/installer.cm.  The installation then simply invokes
194    
195       sml -m $smlnj/installer.cm
196    
197    and everything happens automagically.
198    
199    Win32:  ML code senses value of environment variable SMLNJ_HOME.
200    Unix:   ML code senses values of environment variables ROOT, CONFIGDIR,
201            and BINDIR.
202    
203    The new scheme guarantees that the ML code responsible for the installation
204    is in sync with the APIs of the main system.  Also, the installer is
205    somewhat faster because the installer script is precompiled.
206    
207    ----------------------------------------------------------------------
208    Name: Matthias Blume (blume (at) tti - c (dot) org)
209    Date: 2003/09/24 15:35:00 CDT
210    Tag: blume-20030924-synsock
211    Description:
212    
213    Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
214    but excludes all non-blocking operations.  Defined SOCKET (in Basis)
215    and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
216    implementations of non-blocking operations from CML's Socket
217    structure.
218    
219    ----------------------------------------------------------------------
220    Name: Matthias Blume (blume (at) tti - c (dot) org)
221    Date: 2003/09/24 15:10:05 CDT
222    Tag: blume-20030924-sockets
223    Description:
224    
225    1. Fixed SOCKET API and implementation to match Basis spec.
226       This required changing the internal representation of sockets to one
227       that remembers (for each socket file descriptor) whether it is currently
228       blocking or non-blocking.  This state is maintained lazily (i.e., a system
229       call is made only if the state actually needs to change).
230    
231    2. OS-specific details of sockets were moved into separate files, thus
232       making it possible to unify the bulk of the socket implementations
233       between Unix and Win32.
234    
235    3. CML's socket API changed accordingly.
236       (Note that we need to remove non-blocking functions from this API
237       since they are redundant in the case of CML!)
238    
239    4. CML's socket implementation now makes use of non-blocking functions
240       provided by Basis, thus removing all OS-dependent code from this part
241       of CML.
242    
243    5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
244    
245    ----------------------------------------------------------------------
246    Name: Matthias Blume (blume (at) tti - c (dot) org)
247    Date: 2003/09/22 12:10:00 CDT
248    Tag: blume-20030922-110_43_2
249    Description:
250    
251    Made a new interim version and bootfiles for developer's bootstrapping
252    convenience.
253    
254       110.43.2 -- NEW BOOTFILES
255    
256    ----------------------------------------------------------------------
257    Name: Matthias Blume (blume (at) tti - c (dot) org)
258    Date: 2003/09/19 15:55:00 CDT
259    Tag: blume-20030919-cmdir
260    Description:
261    
262    1. new-install.sh -> install.sh
263    2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
264    3. tweaked installer so that another name instead of .cm can be chosen
265       at install time (by setting the CM_DIR_ARC environment variable
266       during installation); once installation is complete, the name is
267       fixed
268    
269    ----------------------------------------------------------------------
270    Name: Matthias Blume (blume (at) tti - c (dot) org)
271    Date: 2003/09/18 16:00:00 CDT
272    Tag: blume-20030918-110_43_1
273    Description:
274    
275    Made a new interim version and bootfiles for developer's bootstrapping
276    convenience.
277    
278       110.43.1 -- NEW BOOTFILES
279    
280    ----------------------------------------------------------------------
281    Name: Matthias Blume (blume (at) tti - c (dot) org)
282    Date: 2003/09/18 15:20:00 CDT
283    Tag: blume-20030918-misc
284    Description:
285    
286    1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
287       this seems to be controversial at the moment) and used those in
288       Posix.ProcEnv.times.
289    
290    2. Added Time.{from,to}Nanoseconds to Time.
291    
292    3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
293       For example, fromLargeInt never needs to look at more than 3 "big
294       digits" to get its 53 bits of precision.
295    
296    ----------------------------------------------------------------------
297    Name: Matthias Blume (blume (at) tti - c (dot) org)
298    Date: 2003/09/17 16:30:00 CDT
299    Tag: blume-20030917-real32-slices
300    Description:
301    
302    Added an entry to the primitive environment
303    (compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
304    added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
305    
306    Removed some of the "magic" constants in real64.sml and replaced them
307    with code that generates these values from their corresponding
308    integer counterparts.
309    
310    Made all(?) the slice-related changes to the Basis and made everything
311    compile again...
312    
313    ----------------------------------------------------------------------
314    Name: Matthias Blume (blume (at) tti - c (dot) org)
315    Date: 2003/09/15 17:45:00 CDT
316    Tag: blume-20030915-rbase
317    Description:
318    
319    Fixed bug in Real.fromLargeInt.
320    
321    ----------------------------------------------------------------------
322    Name: Matthias Blume (blume (at) tti - c (dot) org)
323    Date: 2003/09/13 18:11:00 CDT
324    Tag: blume-20030913-libinstall
325    Description:
326    
327    Minor bugfix in config/libinstall (set anchor with path to
328    standalone tool after installing it, otherwise libraries that
329    need ml-lex or ml-yacc won't compile the first time the installer
330    runs).
331    
332    ----------------------------------------------------------------------
333    Name: Matthias Blume (blume (at) tti - c (dot) org)
334    Date: 2003/09/12 11:45:00 CDT
335    Tag: blume-20030912-various
336    Description:
337    
338     - fixed bug in Real.toLargeInt
339     - fixed bug in Posix.ProcEnv.times
340     - changed inputLine functions to return an option
341     - minor installer improvements / bugfixes
342     - changed default @SMLalloc parameter for x86/celeron to 64k
343    
344    ----------------------------------------------------------------------
345    Name: Matthias Blume (blume (at) tti - c (dot) org)
346    Date: 2003/09/09 22:00:00 CDT
347    Tag: Release_110_43
348    Description:
349    
350        New working release 110.43.  New bootfiles.
351    
352    ----------------------------------------------------------------------
353    Name: Matthias Blume (blume (at) tti - c (dot) org)
354    Date: 2003/09/09 19:20:00 CDT
355    Tag: blume-20030909-installer
356    Description:
357    
358    Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
359    Modified config/install.bat to take advantage of it.  Also modified
360    config/install.sh (and called it config/new-install.sh) to take advantage
361    of it on Unix systems.  (The SML code is (supposed to be) platform-
362    independent.)
363    
364    The installer can now install everything under Win32
365    as well as under *nix as long as it compiles.
366    
367    Other changes:
368    
369      - made CML compile again under Win32
370      - made eXene compile under Win32 (by providing a fake structure UnixSock
371        and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
372      - fixed a bug in nowhere: it assumed that type OS.Process.status is the
373        same as type int;  under Win32 it isn't
374      - fixed some slice-related problems in the win32-specific parts of CML
375      - added a functor argument "sameVol" to os-path-fn.sml in the Basis
376        (under Win32, the volume name is case-insensitive, and the
377        OS.Path code compares volume names for equality)
378    
379    ----------------------------------------------------------------------
380    Name: Matthias Blume (blume (at) tti - c (dot) org)
381    Date: 2003/09/08 11:55:00 CDT
382    Tag: blume-20030908-fullpath
383    Description:
384    
385    Made Win32 version of OS.FileSys.fullPath return current directory
386    when given an empty string.  This is what the spec says, and incidentally,
387    CM depends on it.  (CM otherwise goes into an infinite loop in certain
388    cases when presented with the name of a non-existing .cm file.)
389    
390    ----------------------------------------------------------------------
391    Name: Matthias Blume (blume (at) tti - c (dot) org)
392    Date: 2003/09/04 16:30:00 CDT
393    Tag: blume-20030905-slices-etc
394    Description:
395    
396      1. Changed interface to vectors and arrays in Basis to match
397         (draft) Basis spec.
398      2. Added signatures and implementations of slices according to
399         Basis spec.
400      3. Edited source code throughout the system to make it compile again
401         under 1. and 2.  (In some cases code had to be added to have it
402         match the new signatures.)
403      4. MLRISC should be backward-compatible: the copies of the originals
404         of files that needed to change under 3. were retained, the .cm files
405         check the compiler version number and use old versions when
406         appropriate.
407      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
408         dirstream -> string option (in accordance with Basis spec).
409      6. When generating code that counts lines, ml-lex used function
410         CharVector.foldli, taking advantage of its old interface.
411         This has been replaced with the corresponding code from
412         CharVectorSlice.  (html-lex must be re-lexed!)
413      7. BitArray in smlnj-lib/Util has been extended/modified to match the
414         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
415      8. Removed temporary additions (fromInternal, toInternal) from the
416         (now obsolete) IntInf in smlnj-lib/Util.
417      9. Cleaned up structure Byte.
418     10. Added localOffset, scan, and fromString to Date (according to spec).
419         Cleaned/corrected implementation of Date.
420         (Still need to check for correctness; implement better canonicalizeDate.)
421     11. Added "scan" to signature IEEE_REAL.
422     12. Some improvements to IntInf [in particular: efficiency-hack for
423         mod and rem when second operand is 2 (for parity checks).]
424     13. Changed representation of type Time.time, using a single IntInf.int
425         value counting microseconds.  This considerably simplified the
426         implementation of structure Time.  We now support negative time
427         values; scan and fromString handle signs.
428     14. Functor PrimIO now takes two additional arguments (VectorSlice and
429         ArraySlice).
430    
431    ----------------------------------------------------------------------
432    Name: Matthias Blume (blume (at) tti - c (dot) org)
433    Date: 2003/08/28 17:00:00 CDT
434    Tag: blume-20030828-intinf
435    Description:
436    
437    This is a major update which comes with a version number bump
438    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
439    and an implementation of IntInf in the Basis.
440    
441    There are a fairly large number of related changes and updates throughout
442    the system:
443    
444    Basis:
445            - Implemented IntInf.
446            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
447            - Added some missing Real64 operations, most notably Real.toLargeInt.
448            - Added FixedInt as a synonym for Int32.
449    
450    compiler:
451            * Added support for a built-in intinf type.
452              - literals
453              - pattern matching
454              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
455              - overloading on literals and operations
456    
457            This required adding a primitive type intinf, some additional
458            primops, and implementations for several non-trivial intinf
459            operations in Core.  (The intinf type is completely abstract
460            to the compiler; all operations get delegated back to the Core.)
461    
462            * Intinf equality is handled by polyequal.  However, the compiler
463            does not print its usual warning in this case (since polyequal
464            is the right thing to do there).
465    
466            * Improved the organization of structure InlineT.
467    
468            * A word about conversion primops:
469            If conversions involving intinf do not cancel out during
470            CPS contract, then the compiler must insert calls to Core functions.
471            Since all core access must be resolved already during the FLINT
472            translate phase, it would be too late a the time of CPS contract
473            to add new Core calls.  For this reason, conversion primops
474            for intinf carry two arguments:  1. the numeric argument that
475            they are supposed to convert, and 2. the Core function that
476            can help with this conversion if necessary.  If CPS contract
477            eliminates a primop, then the associated Core function becomes
478            dead and goes away.  Intinf conversion primops that do not get
479            eliminated by CPS contract get rewritten into calls of their
480            core functions by a separate, new phase.
481    
482    interactive system:
483            - Control.Print.intinfDepth controls max length of intinf constants
484            being printed. (Analogous to Control.Print.stringDepth.)
485            - Cleanup in printutil and pputil: got rid of unused stuff and
486            duplicates; replaced some of the code with code that makes better
487            use of library functionality.
488    
489    CM:
490            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
491            are no longer silent.
492    
493    CKIT:
494            Fixed mismatched uses of Int32 and LargeInt.  I always decided
495            in favor of LargeInt -- which is now the same as IntInf.
496            CKIT-knowledgable people should check whether this is what's
497            intended and otherwise change things back to using Int32 or
498            FixedInt.
499    
500    Throughout the code:
501            Started using IntInf.int literals and built-in operations
502            (e.g., comparison with 0) where this seems appropriate.
503    
504    
505    ----------------------------------------------------------------------
506    Name: Dave MacQueen (dbm@cs.uchicago.edu)
507    Date: 2003/08/13 11:36:00 CDT
508    Tag: dbm-20030813-mcz-merge1
509    Description:
510    
511    Merging changes from the mcz-branch development branch into trunk.
512    These changes involve replacement of the emulated old prettyprinter
513    interface with direct use of the SML/NJ Lib PP library, and fixing
514    of a couple of bugs (895, 1186) relating to error messages.  A new
515    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
516    has been added.
517    
518    ----------------------------------------------------------------------
519    Name: Matthias Blume (blume (at) tti - c (dot) org)
520    Date: 2003/08/11 15:45:00 CDT
521    Tag: blume-20030811-windows
522    Description:
523    
524    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
525    
526        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
527    
528    This patch restores SML/NJ's ability to run under win32.  There are a
529    number of changes, including fixes for several bugs that had gone
530    unnoticed until now:
531    
532      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
533        trying to run the win32 version from within cygwin.)
534    
535      - There are a number of simple .bat scripts that substitute for their
536        corresponding Unix shell-scripts.  (See below.)
537    
538      - The internals of ml-build have been modified slightly.  The main
539        difference is that instead of calling ".link-sml" (or link-sml.bat)
540        using OS.Process.system, the ML process delegates this task back
541        to the script.  Otherwise problems arise in mixed environments such
542        as Cygwin where scripts look and work like Unix scripts, but
543        where OS.Process.system cannot run them.
544    
545      - In CM, the srcpath pickler used native pathname syntax -- which
546        is incorrect in the case of cross-compilation.  The new pickle format
547        is independent of platform-specific naming conventions.
548    
549      - Path configuration files (such as lib/pathconfig) can now choose
550        between native and standard syntax.  Placing a line of the form
551    
552            standard!
553    
554        into the file causes all subsequent paths to be interpreted using
555        CM standard pathname syntax (= Unix conventions); a line
556    
557            native!
558    
559        switches back to native style.  This was needed so that
560        path config files can be written portably, see src/system/pathconfig.
561    
562      - Runtime system:
563    
564          - win32-filesys.c:  get_file_time and set_file_time now
565            access modification time, not creation time.
566    
567          - I/O code made aware of new array representation.
568    
569          - Bug fixes in X86.prim.masm.
570    
571       - src/system/makeml made aware of win32. (For use under cygwin
572            and other Unix-environments for windows.)
573    
574       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
575         which caused BinIO.inputAll to fail consistently.
576    
577    .bat scripts:
578    
579       Windows .bat scripts assume that SMLNJ_HOME is defined.
580    
581       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
582         applications (sml, ml-yacc, ml-lex).
583       - ml-build.bat: analogous to ml-build.
584       - config\install.bat:  Analogous to config/install.sh.  This requires
585         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
586         (nmake etc. must be on the path, and vcvars32 must have been run.)
587         Moreover, sources for ml-lex and ml-yacc need to exist under src,
588         and the bootfile hierarchy must have been unpacked under
589         sml.boot.x86-win32.
590         The script is very primitive and does a poor job at error checking.
591         It only installs the base system, ml-lex, and ml-yacc.  No other
592         libraries are being installed (i.e., you get only those that
593         are part of the compiler.)
594       - link-sml.bat:  analogous to .link-sml, but not currently used
595    
596    Unrelated bug fixes:
597    
598      - ml-nlffigen now exports structures ST_* corresponding to incomplete
599        types.
600      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
601        otherwise.)
602    
603    ----------------------------------------------------------------------
604    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
605    Date: 2003/06/17
606    Tag: macqueen-20030617-bug895
607    Description:
608    
609    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
610    Tag will be used for new development branch (mcz-branch) for
611    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
612    
613    ----------------------------------------------------------------------
614    Name: Matthias Blume (blume (at) tti - c (dot) org)
615    Date: 2003/05/27 16:55:00 CDT
616    Tag: blume-20030527-polyeq
617    Description:
618    
619    Tried to eliminated most cases of polymorphic equality.
620    
621    ----------------------------------------------------------------------
622    Name: Matthias Blume (blume (at) tti - c (dot) org)
623    Date: 2003/05/21 17:45:00 CDT
624    Tag: blume-20030517-complete
625    Description:
626    
627    Two changes:
628    
629      1. Added a flag for controlling whether non-exhaustive bindings will
630         be treated as errors (default is false).
631      2. Cleaned up the *entire* source tree so that CMB.make goes through
632         without a single non-exhaustive match- or bind warning.
633    
634    ----------------------------------------------------------------------
635    Name: Matthias Blume (blume (at) tti - c (dot) org)
636    Date: 2003/05/17 10:20:00 CDT
637    Tag: blume-20030517-absyn
638    Description:
639    
640    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
641    
642       This mainly affects the quality of error messages.  However, some
643       of the code is now more straightforward than before. (Treatment of
644       the above four constructs in translate.sml is much simpler than
645       the "macro-expansion" that was going on before.  Plus, the mach-
646       compiler no longer gets invoked just to be able to compile an
647       if-expression.)
648    
649    2. The ErrorMsg.Error exception is now caught and absorbed by the
650       interactive loop.
651    
652    ----------------------------------------------------------------------
653    Name: Allen Leung
654    Date: 2003/05/16 13:05:00 CDT
655    Tag: leunga-20030516-cygwin-runtime
656    Description:
657    
658        Ported the runtime system to cygwin, which uses the unix
659    x86-unix bin files.  Missing/buggy features:
660    
661       o getnetbyname, getnetbyaddr: these functions seem to be missing in
662         the Cygwin library.
663       o Ctrl-C handling may be flaky.
664       o Windows system calls and Windows I/O are not supported.
665    
666       A new set of binfiles is located at:
667    
668         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
669    
670       This is only needed for bootstrapping the cygwin version of smlnj.
671       Other x86 versions can use the existing binfiles.
672    
673    ----------------------------------------------------------------------
674    Name: Matthias Blume
675    Date: 2003/04/08 15:42:00 CDT
676    Tag: blume-20030408-listpair
677    Description:
678    
679    1. Added a target 'mlrisc' to installer.
680    
681    2. Added missing elements to structure ListPair.
682    
683    ----------------------------------------------------------------------
684    Name: Allen Leung
685    Date: 2003/01/07 10:40:00 EST
686    Tag: leunga-20030107-int-rem
687    Description:
688    
689    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
690    The arguments to the SUBL instruction were swapped.
691    
692    ----------------------------------------------------------------------
693    Name: Matthias Blume
694    Date: 2002/12/12 16:25:00 EST
695    Tag: blume-20021212-risc-ra
696    Description:
697    
698    Fixed a serious bug in the rewrite code for FP spilling/reloading that
699    sent the RA into an infinite loop when floating point registers get
700    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
701    110.37 and 110.38.)
702    There was another set of potential problems related to the handling of
703    MLRISC annotations (but those did not yet cause real problems, apparently).
704    
705    ----------------------------------------------------------------------
706    Name: Matthias Blume
707    Date: 2002/12/06 22:40:00 EST
708    Tag: blume-20021206-cm-fileid
709    Description:
710    
711    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
712    This fixes the problem of CM getting confused by files that suddenly
713    change their identity (e.g., by getting unlinked and recreated by some
714    text editor such as vi).  There might be a better/cheaper/cleaner way
715    of doing this, but for now this will have to do.
716    
717    ----------------------------------------------------------------------
718    Name: Matthias Blume
719    Date: 2002/10/28 09:50:00 EST
720    Tag: blume-20021028-typecheck
721    Description:
722    
723    Exported structure Typecheck from $smlnj/viscomp/core.cm.
724    
725    ----------------------------------------------------------------------
726    Name: Matthias Blume
727    Date: 2002/10/17 09:10:00 EDT
728    Tag: Release_110_42
729    Description:
730    
731    In good old tradition, there has been a slight hiccup so that we have
732    to patch 110.42 after the fact.   The old release tag has been replaced
733    (see below).
734    
735    The change solves a problem with two competing approaches the
736    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
737    each other's way.
738    
739    This change only affects the runtime system code and the installer script.
740    (No new bootfiles.)
741    
742    ----------------------------------------------------------------------
743    Name: Matthias Blume
744    Date: 2002/10/16 12:00:00 EDT
745    Tag: Release_110_42_removed
746    Description:
747    
748    New working release.  New bootfiles.
749    
750    ----------------------------------------------------------------------
751    Name: Matthias Blume
752    Date: 2002/10/10 13:10:00 EDT
753    Tag: blume-20021010-ppc-divs
754    Description:
755    
756    The mltree operator DIVS must be implemented with an overflow check on
757    the PPC because the hardware indicates divide-by-zero using "overflow" as
758    well.
759    
760    ----------------------------------------------------------------------
761    Name: Matthias Blume
762    Date: 2002/07/23 11:45:00
763    Tag: blume-20020723-smlnj-home
764    Description:
765    
766    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
767    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
768    CM_PATHCONFIG is also set), the path configuration file is assumed
769    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
770    the entire tree to some other place and everything will "just work".
771    
772    (Companion commands such as ml-build and ml-makedepend also sense this
773    variable.)
774    
775    ----------------------------------------------------------------------
776    Name: Matthias Blume
777    Date: 2002/07/12 21:19:00 EDT
778    Tag: blume-20020712-liveness
779    Description:
780    
781    Exported two useful "step" functions from liveness module (MLRISC).
782    
783    ----------------------------------------------------------------------
784    Name: Matthias Blume
785    Date: 2002/07/05 16:00 EDT
786    Tag: Release_110_41
787    Description:
788    
789    New working release.  New bootfiles.
790    
791    ----------------------------------------------------------------------
792    Name: Matthias Blume
793    Date: 2002/07/05 10:25:00 EDT
794    Tag: blume-20020705-btimp
795    Description:
796    
797    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
798    other clients can set up backtracing support.
799    
800    ----------------------------------------------------------------------
801    Name: Matthias Blume
802    Date: 2002/06/25 17:23:00 EDT
803    Tag: blume-20020625-fpmax
804    Description:
805    
806    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
807    case.  (The sense of the isNaN test was reversed -- which made min and
808    max always return their first argument.)
809    
810    ----------------------------------------------------------------------
811    Name: Matthias Blume
812    Date: 2002/06/11
813    Tag: blume-20020611-unixpath
814    Description:
815    
816    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
817    
818    ----------------------------------------------------------------------
819    Name: Matthias Blume
820    Date: 2002/06/10 16:35:00 EDT
821    Tag: blume-20020610-ieeereal
822    Description:
823    
824    I back-ported my implementation of IEEEReal.fromString from the
825    idlbasis-devel branch so that we can test it.
826    
827    Another small change is that ppDec tries to give more information
828    than just "<sig>" in the case of functors.  However, this code is
829    broken in some mysterious way if the functor's body's signature
830    has not been declared by ascription but gets inferred from the
831    implementation.  This needs fixing...
832    
833    ----------------------------------------------------------------------
834    Name: Matthias Blume
835    Date: 2002/05/31
836    Tag: blume-20020531-btrace-mode
837    Description:
838    
839    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
840    the wayside when I switched over to using Controls everywhere.)
841    
842    ----------------------------------------------------------------------
843    Name: Lal George
844    Date: 2002/05/23 12:21:40 EDT
845    Tag: george-20020523-visual-labels
846    Description:
847    
848    Labels are now displayed in the graphical output to make
849    the fall-through and target blocks obvious.
850    
851    ----------------------------------------------------------------------
852    Name: Matthias Blume
853    Date: 2002/05/22 11:03:00 EDT
854    Tag: blume-20020522-shrink
855    Description:
856    
857    John tweaked yesterday's fix for 1131 to handle an out-of-memory
858    situation that comes up when allocating huge arrays.
859    
860    ----------------------------------------------------------------------
861    Name: Matthias Blume
862    Date: 2002/05/21 16:00:00 EDT
863    Tag: Release_110_40
864    Description:
865    
866    New working release (110.40).  New bootfiles.
867    
868    [Also: John Reppy fixed GC bug 1131.]
869    
870    ----------------------------------------------------------------------
871    Name: Matthias Blume
872    Date: 2002/05/21 12:35:00 EDT
873    Tag: blume-20020521-cmdoc
874    Description:
875    
876    CM documentation update.
877    
878    ----------------------------------------------------------------------
879    Name: Matthias Blume
880    Date: 2002/05/21 10:55:00 EDT
881    Tag: blume-20020521-misc
882    Description:
883    
884    - John tweaked runtime to be silent on heap export (except when
885      GC messages are on).
886    - I added a few more things (cross-compiling versions of CMB) to
887      config/preloads (as suggestions).
888    
889    ----------------------------------------------------------------------
890    Name: Matthias Blume
891    Date: 2002/05/20 22:25:00 EDT
892    Tag: blume-20020520-controls
893    Description:
894    
895    - Added ControlUtil structure to control-lib.cm.
896    - Use it throughout.
897    - Used Controls facility to define MLRISC controls (as opposed to
898      registering MLRISC control ref cells with Controls after the fact)
899    - Fixed messed-up controls priorities.
900    
901    * Removed again all the stuff from config/preloads that one wouldn't
902      be able to preload at the time the initial heap image is built.
903      (Many libraries, e.g., CML, do not exist yet at this time.  The
904       only libraries that can be preloaded via config/preloads are those
905       that come bundled with the bootfiles.)
906    
907    ----------------------------------------------------------------------
908    Name: Matthias Blume
909    Date: 2002/05/20 10:59:00 EDT
910    Tag: blume-20020520-preloads
911    Description:
912    
913    Added a lot of commented-out suggestions for things to be included
914    in config/preloads.
915    
916    ----------------------------------------------------------------------
917    Name: Allen Leung
918    Date: 2002/05/18 14:20:00 EDT
919    Tag: leunga-20020518-mdl
920    Description:
921    
922    o Made the mdl tool stuff compile and run again.
923    o I've disabled all the stuff that depends on RTL specifications; they
924      are all badly broken anyway.
925    
926    ----------------------------------------------------------------------
927    Name: Matthias Blume
928    Date: 2002/05/17 16:49:00 EDT
929    Tag: blume-20020517-controls
930    Description:
931    
932    0. John Reppy made several modifications to the SML/NJ library.
933       In particular, there is a shiny new controls-lib.cm.
934    
935    1. Pushed new controls interface through compiler so that everything
936       compiles again.
937    
938    2. Added FormatComb and FORMAT_COMB to the CML version of the
939       SML/NJ library (so that CML compiles again).
940    
941    3. Modified init scripts because XXX_DEFAULT environment variables
942       are no longer with us.  (Boot-time initialization is now done
943       using the same environment variables that are also used for
944       startup-time initialization of controls.)
945    
946    ----------------------------------------------------------------------
947    Name: Lal George
948    Date: 2002/05/15 09:20:10 EDT
949    Tag: george-20020515-pseudo-op-decls
950    Description:
951    
952            All pseudo-ops emitted before the first segment declaration
953    such as TEXT, DATA, and BSS directives are assumed to be global
954    declarations and are emitted first in the assembly file. This is
955    useful in a number of situations where one has pseudo-ops that are not
956    specific to any segment, and also works around the constraint that one
957    cannot have client pseudo-ops in the TEXT segment.
958    
959            Because no segment is associated with these declarations it is
960    an error to allocate any space or objects before the first segment
961    directive and an exception will be raised. However, we cannot make
962    this check for client pseudo-ops.
963    
964            These top level declarations are a field in the CFG graph_info.
965    In theory you can continue to add to this field after the CFG has been
966    built -- provided you know what you are doing;-)
967    
968    ----------------------------------------------------------------------
969    Name: Matthias Blume
970    Date: 2002/05/13 16:40:00 EDT
971    Tag: blume-20020513-pp-etc
972    Description:
973    
974    A few minor bugfixes:
975    
976      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
977        (Bogus printouts for redefined bindings still occur.  Compiler
978        bug should no longer occur now.  We need to redo the prettyprinter
979        from scratch.)
980    
981      - CM pathname printer now also adds escape sequences for ( and )
982    
983      - commend and docu fixes for ml-nlffi
984    
985    ----------------------------------------------------------------------
986    Name: Matthias Blume
987    Date: 2002/05/10 16:40:00 EDT
988    Tag: blume-20020510-erg-textio
989    Description:
990    
991    Applied the following bugfix provided by Emden Gansner:
992    
993        Output is corrupted when outputSubstr is used rather than output.
994    
995        The problem occurs when a substring
996    
997            ss = (s, dataStart, dataLen)
998    
999        where dataStart > 0, fills a stream buffer with avail bytes left.
1000        avail bytes of s, starting at index dataStart, are copied into the
1001        buffer, the buffer is flushed, and then the remaining dataLen-avail
1002        bytes of ss are copied into the beginning of the buffer. Instead of
1003        starting this copy at index dataStart+avail in s, the current code
1004        starts the copy at index avail.
1005    
1006        Fix:
1007        In text-io-fn.sml, change line 695 from
1008             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
1009        to
1010             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
1011    
1012    ----------------------------------------------------------------------
1013    Name: Matthias Blume
1014    Date: 2002/04/12 13:55:00 EDT
1015    Tag: blume-20020412-assyntax
1016    Description:
1017    
1018    1. Grabbed newer assyntax.h from the XFree86 project.
1019    2. Fiddled with how to compile X86.prim.asm without warnings.
1020    3. (Very) Minor cleanup in CM.
1021    
1022    ----------------------------------------------------------------------
1023    Name: Matthias Blume
1024    Date: 2002/04/01 (no joke!) 17:07:00 EST
1025    Tag: blume-20020401-x86div
1026    Description:
1027    
1028    Added full support for div/mod/rem/quot on the x86, using the machine
1029    instruction's two results (without clumsily recomputing the remainder)
1030    directly where appropriate.
1031    
1032    Some more extensive power-of-two support was added to the x86 instruction
1033    selector (avoiding expensive divs, mods, and muls where they can be
1034    replaced with cheaper shifts and masks).  However, this sort of thing
1035    ought to be done earlier, e.g., within the CPS optimizer so that
1036    all architectures benefit from it.
1037    
1038    The compiler compiles to a fixed point, but changes might be somewhat
1039    fragile nevertheless.  Please, report any strange things that you might
1040    see wrt. div/mod/quot/rem...
1041    
1042    ----------------------------------------------------------------------
1043    Name: Matthias Blume
1044    Date: 2002/03/29 17:22:00
1045    Tag: blume-20020329-div
1046    Description:
1047    
1048    Fixed my broken div/mod logic.  Unfortunately, this means that the
1049    inline code for div/mod now has one more comparison than before.
1050    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
1051    The problem was with quotient = 0, because that alone does not tell
1052    us which way the rounding went.  One then has to look at whether
1053    remainder and divisor have the same sign...  :(
1054    
1055    Anyway, I replaced the bootfiles with fresh ones...
1056    
1057    ----------------------------------------------------------------------
1058    Name: Matthias Blume
1059    Date: 2002/03/29 14:10:00 EST
1060    Tag: blume-20020329-inlprims
1061    Description:
1062    
1063    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
1064    
1065    Primops have changed. This means that the bin/boot-file formats have
1066    changed as well.
1067    
1068    To make sure that there is no confusion, I made a new version.
1069    
1070    
1071    CHANGES:
1072    
1073    * removed REMT from mltree (remainder should never overflow).
1074    
1075    * added primops to deal with divisions of all flavors to the frontend
1076    
1077    * handled these primops all the way through so they map to their respective
1078      MLRISC support
1079    
1080    * used these primops in the implementation of Int, Int32, Word, Word32
1081    
1082    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
1083    
1084    * parameterized INLMIN, INLMAX, and INLABS by a numkind
1085    
1086    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
1087      floating point
1088    
1089    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
1090      and Real (but Real.abs maps to a separate floating-point-only primop)
1091    
1092    
1093    TODO items:
1094    
1095    * Hacked Alpha32 instruction selection, disabling the selection of REMx
1096      instructions because the machine instruction encoder cannot handle
1097      them.  (Hppa, PPC, and Sparc instruction selection did not handle
1098      REM in the first place, and REM is supported by the x86 machine coder.)
1099    
1100    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
1101      selection phase.  (The two can be streamlined because the hardware
1102      delivers both quotient and remainder at the same time anyway.)
1103    
1104    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
1105      (Currently the behavior is inconsistent both across architectures and
1106      wrt. the draft Basis spec.)
1107    
1108    * Word8 should eventually be handled natively, too.
1109    
1110    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
1111      as if there currently is no execution path that could trigger it in
1112      SML/NJ.  (The assumptions underlying functions arith and promotable do not
1113      hold for things like multiplication and division.)
1114    
1115    ----------------------------------------------------------------------
1116    Name: Matthias Blume
1117    Date: 2002/03/27 16:27:00 EST
1118    Tag: blume-20020327-mlrisc-divisions
1119    Description:
1120    
1121    Added support for all four division operations (ML's div, mod, quot,
1122    and rem) to MLRISC.  In the course of doing so, I also rationalized
1123    the naming (no more annoying switch-around of DIV and QUOT), by
1124    parameterizing the operation by div_rounding_mode (which can be either
1125    DIV_TO_ZERO or DIV_TO_NEGINF).
1126    
1127    The generic MLTreeGen functor takes care of compiling all four
1128    operations down to only round-to-zero div.
1129    
1130    Missing pieces:
1131    
1132      * Doing something smarter than relying on MLTreeGen on architectures
1133        like, e.g., the x86 where hardware division delivers both quotient and
1134        remainder at the same time.  With this, the implementation of the
1135        round-to-neginf operations could be further streamlined.
1136    
1137      * Remove inlining support for div/mod/rem from the frontend and replace it
1138        with primops that get carried through to the backend.  Do this for all
1139        int and word types.
1140    
1141    ----------------------------------------------------------------------
1142    Name: Matthias Blume
1143    Date: 2002/03/25 17:25:00 EST
1144    Tag: blume-20020325-divmod
1145    Description:
1146    
1147    I improved (hopefully without breaking them) the implementation of Int.div,
1148    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
1149    advantage of the following observations:
1150    
1151      Let  q = x quot y      r = x rem y
1152           d = x div  y      m = x mod y
1153    
1154    where "quot" is the round-to-zero version of integer division that
1155    hardware usually provides.  Then we have:
1156    
1157         r = x - q * y        where neither the * nor the - will overflow
1158         d = if q >= 0 orelse x = q * y then q else q - 1
1159                              where neither the * nor the - will overflow
1160         m = if q >= 0 orelse r = 0 then r else r + y
1161                              where the + will not overflow
1162    
1163    This results in substantial simplification of the generated code.
1164    The following table shows the number of CFG nodes and edges generated
1165    for
1166            fun f (x, y) = x OPER y
1167            (* with OPER \in div, mod, quot, rem *)
1168    
1169    
1170        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
1171        --------------------------------------------------------
1172         div |         24 |         39 |         12 |         16
1173         mod |         41 |         71 |         12 |         16
1174        quot |          8 |         10 |          8 |         10
1175         rem |         10 |         14 |          8 |         10
1176    
1177    
1178    ----------------------------------------------------------------------
1179    Name: Matthias Blume
1180    Date: 2002/03/25 22:06:00 EST
1181    Tag: blume-20020325-cprotobug
1182    Description:
1183    
1184    Fixed a bug in cproto (c prototype decoder).
1185    
1186    ----------------------------------------------------------------------
1187    Name: Matthias Blume
1188    Date: 2002/03/25 16:00:00 EST
1189    Tag: blume-20020325-raw-primops
1190    Description:
1191    
1192    I did some cleanup to Allen's new primop code and
1193    replaced yesterday's bootfiles with new ones.
1194    (But they are stored in the same place.)
1195    
1196    ----------------------------------------------------------------------
1197    Name: Matthias Blume
1198    Date: 2002/03/24 22:40:00 EST
1199    Tag: blume-20020324-bootfiles
1200    Description:
1201    
1202    Made the bootfiles that Allen asked for.
1203    
1204    ----------------------------------------------------------------------
1205    Name: Allen Leung
1206    Date: 2002/03/23 15:50:00 EST
1207    Tag: leunga-20020323-flint-cps-rcc-primops
1208    Description:
1209    
1210      1. Changes to FLINT primops:
1211    
1212        (* make a call to a C-function;
1213         * The primop carries C function prototype information and specifies
1214         * which of its (ML-) arguments are floating point. C prototype
1215         * information is for use by the backend, ML information is for
1216         * use by the CPS converter. *)
1217      | RAW_CCALL of { c_proto: CTypes.c_proto,
1218                       ml_args: ccall_type list,
1219                       ml_res_opt: ccall_type option,
1220                       reentrant : bool
1221                     } option
1222       (* Allocate uninitialized storage on the heap.
1223        * The record is meant to hold short-lived C objects, i.e., they
1224        * are not ML pointers.  With the tag, the representation is
1225        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
1226        *)
1227      | RAW_RECORD of {tag:bool,sz:int}
1228      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
1229    
1230      2.  These CPS primops are now overloaded:
1231    
1232           rawload of {kind:numkind}
1233           rawstore of {kind:numkind}
1234    
1235          The one argument form is:
1236    
1237             rawload {kind} address
1238    
1239          The two argument form is:
1240    
1241             rawload {kind} [ml object, byte-offset]
1242    
1243      3. RAW_CCALL/RCC now takes two extra arguments:
1244    
1245         a. The first is whether the C call is reentrant, i.e., whether
1246            ML state should be saved and restored.
1247         b. The second argument is a string argument specifying the name of
1248            library and the C function.
1249    
1250         These things are currently not handled in the code generator, yet.
1251    
1252      4. In CProto,
1253    
1254         An encoding type of "bool" means "ml object" and is mapped into
1255         C prototype of PTR.  Note that "bool" is different than "string",
1256         even though "string" is also mapped into PTR, because "bool"
1257         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
1258    
1259      5. Pickler/unpicker
1260    
1261         Changed to handle RAW_RECORD and newest RAW_CCALL
1262    
1263      6. MLRiscGen,
1264    
1265         1. Changed to handle the new rawload/rawstore/rawrecord operators.
1266         2. Code for handling C Calls has been moved to a new module CPSCCalls,
1267            in the file CodeGen/cpscompile/cps-c-calls.sml
1268    
1269      7. Added the conditional move operator
1270    
1271             condmove of branch
1272    
1273         to cps.  Generation of this is still buggy so it is currently
1274         disabled.
1275    
1276    ----------------------------------------------------------------------
1277    Name: Lal George
1278    Date: 2002/03/22 14:18:25 EST
1279    Tag: george-20020322-cps-branch-prob
1280    Description:
1281    
1282    Implemented the Ball-Larus branch prediction-heuristics, and
1283    incorporated graphical viewers for control flow graphs.
1284    
1285    Ball-Larus Heuristics:
1286    ---------------------
1287    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1288    
1289    By design it uses the Dempster-Shafer theory for combining
1290    probabilities.  For example, in the function:
1291    
1292        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1293    
1294    the ball-larus heuristics predicts that the n=0 is unlikely
1295    (OH-heuristic), and the 'then' branch is unlikely because of the
1296    RH-heuristic -- giving the 'then' branch an even lower combined
1297    probability using the Dempster-Shafer theory.
1298    
1299    Finally, John Reppy's loop analysis in MLRISC, further lowers the
1300    probability of the 'then' branch because of the loop in the else
1301    branch.
1302    
1303    
1304    Graphical Viewing:
1305    ------------------
1306    I merely plugged in Allen's graphical viewers into the compiler. The
1307    additional code is not much. At the top level, saying:
1308    
1309            Control.MLRISC.getFlag "cfg-graphical-view" := true;
1310    
1311    will display the graphical view of the control flow graph just before
1312    back-patching.  daVinci must be in your path for this to work. If
1313    daVinci is not available, then the default viewer can be changed
1314    using:
1315    
1316            Control.MLRISC.getString "viewer"
1317    
1318    which can be set to "dot" or "vcg" for the corresponding viewers. Of
1319    course, these viewers must be in your path.
1320    
1321    The above will display the compilation unit at the level of clusters,
1322    many of which are small, boring, and un-interesting. Also setting:
1323    
1324            Control.MLRISC.getInt "cfg-graphical-view_size"
1325    
1326    will display clusters that are larger than the value set by the above.
1327    
1328    
1329    ----------------------------------------------------------------------
1330    Name: Matthias Blume
1331    Date: 2002/03/21 22:20:00 EST
1332    Tag: blume-20020321-kmp-bugfix
1333    Description:
1334    
1335    Changed the interface to the KMP routine in PreString and fixed
1336    a minor bug in one place where it was used.
1337    
1338    ----------------------------------------------------------------------
1339    Name: Allen Leung
1340    Date: 2002/03/21 20:30:00 EST
1341    Tag: leunga-20020321-cfg
1342    Description:
1343    
1344      Fixed a potential problem in cfg edge splitting.
1345    
1346    ----------------------------------------------------------------------
1347    Name: Allen Leung
1348    Date: 2002/03/21 17:15:00 EST
1349    Tag: leunga-20020321-x86-fp-cfg
1350    Description:
1351    
1352      1. Recoded the buggy parts of x86-fp.
1353    
1354         a. All the block reordering code has been removed.
1355            We now depend on the block placement phases to do this work.
1356    
1357         b. Critical edge splitting code has been simplified and moved into the
1358            CFG modules, as where they belong.
1359    
1360         Both of these were quite buggy and complex.  The code is now much, much
1361         simpler.
1362    
1363      2. X86 backend.
1364    
1365         a. Added instructions for 64-bit support.  Instruction selection for
1366            64-bit has not been committed, however, since that
1367            requires changes to MLTREE which haven't been approved by
1368            Lal and John.
1369    
1370         b. Added support for FUCOMI and FUCOMIP when generating code for
1371            PentiumPro and above.  We only generate these instructions in
1372            the fast-fp mode.
1373    
1374         c. Added cases for JP and JNP in X86FreqProps.
1375    
1376      3. CFG
1377    
1378         CFG now has a bunch of methods for edge splitting and merging.
1379    
1380      4. Machine description.
1381    
1382         John's simplification of MLTREE_BASIS.fcond broke a few machine
1383         description things:
1384    
1385         rtl-build.{sig,sml} and hppa.mdl fixed.
1386    
1387         NOTE: the machine description stuff in the repository is still broken.
1388               Again, I can't put my fixes in because that involves
1389               changes to MLTREE.
1390    
1391    ----------------------------------------------------------------------
1392    Name: Matthias Blume
1393    Date: 2002/03/20 15:55:00 EST
1394    Tag: blume-20020320-kmp
1395    Description:
1396    
1397    Implemented Knuth-Morris-Pratt string matching in PreString and used
1398    it for String.isSubstring, Substring.isSubstring, and
1399    Substring.position.
1400    
1401    (Might need some stress-testing.  Simple examples worked fine.)
1402    
1403    ----------------------------------------------------------------------
1404    Name: Matthias Blume
1405    Date: 2002/03/19 16:37:00 EST
1406    Tag: blume-20020319-witnesses
1407    Description:
1408    
1409    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1410    
1411    This implements a generic mechanism for changing constness qualifiers
1412    anywhere within big C types without resorting to outright "casts".
1413    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1414    modify the constness at the outermost level.)
1415    The implementation of "convert" is based on the idea of "witness"
1416    values -- values that are not used by the operation but whose types
1417    "testify" to their applicability.  On the implementation side, "convert"
1418    is simply a projection (returning its second curried argument).  With
1419    cross-module inlining, it should not result in any machine code being
1420    generated.
1421    
1422    ----------------------------------------------------------------------
1423    Name: Matthias Blume
1424    Date: 2002/03/15 16:40:00 EST
1425    Tag: blume-20020315-basis
1426    Description:
1427    
1428    Provided (preliminary?) implementations for
1429    
1430      {String,Substring}.{concatWith,isSuffix,isSubstring}
1431    
1432    and
1433    
1434      Substring.full
1435    
1436    Those are in the Basis spec but they were missing in SML/NJ.
1437    
1438    ----------------------------------------------------------------------
1439    Name: Matthias Blume
1440    Date: 2002/03/14 21:30:00 EST
1441    Tag: blume-20020314-controls
1442    Description:
1443    
1444    Controls:
1445    ---------
1446    
1447    1. Factored out the recently-added Controls : CONTROLS stuff and put
1448       it into its own library $/controls-lib.cm.  The source tree for
1449       this is under src/smlnj-lib/Controls.
1450    
1451    2. Changed the names of types and functions in this interface, so they
1452       make a bit more "sense":
1453    
1454          module -> registry
1455          'a registry -> 'a group
1456    
1457    3. The interface now deals in ref cells only.  The getter/setter interface
1458       is (mostly) gone.
1459    
1460    4. Added a function that lets one register an already-existing ref cell.
1461    
1462    5. Made the corresponding modifications to the rest of the code so that
1463       everything compiles again.
1464    
1465    6. Changed the implementation of Controls.MLRISC back to something closer
1466       to the original.  In particular, this module (and therefore MLRISC)
1467       does not depend on Controls.  There now is some link-time code in
1468       int-sys.sml that registers the MLRISC controls with the Controls
1469       module.
1470    
1471    CM:
1472    ---
1473    
1474      * One can now specify the lambda-split aggressiveness in init.cmi.
1475    
1476    ----------------------------------------------------------------------
1477    Name: Allen Leung
1478    Date: 2002/03/13 17:30:00 EST
1479    Tag: leunga-20020313-x86-fp-unary
1480    Description:
1481    
1482    Bug fix for:
1483    
1484    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1485    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1486    > - fun f(x,(y,z)) = Real.~ y;
1487    > [autoloading]
1488    > [autoloading done]
1489    >       fchsl   (%eax), 184(%esp)
1490    > Error: MLRisc bug: X86MCEmitter.emitInstr
1491    >
1492    > uncaught exception Error
1493    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1494    
1495    The problem was that the code generator did not generate any fp registers
1496    in this case, and the ra didn't know that it needed to run the X86FP phase to
1497    translate the pseudo fp instruction.   This only happened with unary fp
1498    operators in certain situations.
1499    
1500    ----------------------------------------------------------------------
1501    Name: Matthias Blume
1502    Date: 2002/03/13 14:00:00 EST
1503    Tag: blume-20020313-overload-etc
1504    Description:
1505    
1506    1. Added _overload as a synonym for overload for backward compatibility.
1507       (Control.overloadKW must be true for either version to be accepted.)
1508    
1509    2. Fixed bug in install script that caused more things to be installed
1510       than what was requested in config/targets.
1511    
1512    3. Made CM aware of the (_)overload construct so that autoloading
1513       works.
1514    
1515    ----------------------------------------------------------------------
1516    Name: Matthias Blume
1517    Date: 2002/03/12 22:03:00 EST
1518    Tag: blume-20020312-url
1519    Description:
1520    
1521    Forgot to update BOOT and srcarchiveurl.
1522    
1523    ----------------------------------------------------------------------
1524    Name: Matthias Blume
1525    Date: 2002/03/12 17:30:00 EST
1526    Tag: blume-20020312-version110392
1527    Description:
1528    
1529    Yet another version number bump (because of small changes to the
1530    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1531    
1532    Changes:
1533    
1534      The new pid generation scheme described a few weeks ago was overly
1535      complicated.  I implemented a new mechanism that is simpler and
1536      provides a bit more "stability":  Once CM has seen a compilation
1537      unit, it keeps its identity constant (as long as you do not delete
1538      those crucial CM/GUID/* files).  This means that when you change
1539      an interface, compile, then go back to the old interface, and
1540      compile again, you arrive at the original pid.
1541    
1542      There now also is a mechanism that instructs CM to use the plain
1543      environment hash as a module's pid (effectively making its GUID
1544      the empty string).  For this, "noguid" must be specified as an
1545      option to the .sml file in question within its .cm file.
1546      This is most useful for code that is being generated by tools such
1547      as ml-nlffigen (because during development programmers tend to
1548      erase the tool's entire output directory tree including CM's cached
1549      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1550      revert to the old, broken behavior of SML/NJ, but in specific cases
1551      where there is no danger of interface confusion, its use is ok
1552      (I think).
1553    
1554      ml-nlffigen by default generates "noguid" annotations.  They can be
1555      turned off by specifying -guid in its command line.
1556    
1557    ----------------------------------------------------------------------
1558    Name: Lal George
1559    Date: 2002/03/12 12 14:42:36 EST
1560    Tag: george-20020312-frequency-computation
1561    Description:
1562    
1563    Integrated jump chaining and static block frequency into the
1564    compiler. More details and numbers later.
1565    
1566    ----------------------------------------------------------------------
1567    Name: Lal George
1568    Date: 2002/03/11 11 22:38:53 EST
1569    Tag: george-20020311-jump-chain-elim
1570    Description:
1571    
1572    Tested the jump chain elimination on all architectures (except the
1573    hppa).  This is on by default right now and is profitable for the
1574    alpha and x86, however, it may not be profitable for the sparc and ppc
1575    when compiling the compiler.
1576    
1577    The gc test will typically jump to a label at the end of the cluster,
1578    where there is another jump to an external cluster containing the actual
1579    code to invoke gc. This is to allow factoring of common gc invocation
1580    sequences. That is to say, we generate:
1581    
1582            f:
1583               testgc
1584               ja   L1      % jump if above to L1
1585    
1586            L1:
1587               jmp L2
1588    
1589    
1590    After jump chain elimination the 'ja L1' instructions is converted to
1591    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1592    up being implemented in their long form (if L2 is far away) using:
1593    
1594            jbe     L3      % jump if below or equal to L3
1595            jmp     L2
1596         L3:
1597            ...
1598    
1599    
1600    For large compilation units L2  may be far away.
1601    
1602    
1603    ----------------------------------------------------------------------
1604    Name: Matthias Blume
1605    Date: 2002/03/11 13:30:00 EST
1606    Tag: blume-20020311-mltreeeval
1607    Description:
1608    
1609    A functor parameter was missing.
1610    
1611    ----------------------------------------------------------------------
1612    Name: Allen Leung
1613    Date: 2002/03/11 10:30:00 EST
1614    Tag: leunga-20020311-runtime-string0
1615    Description:
1616    
1617       The representation of the empty string now points to a
1618    legal null terminated C string instead of unit.  It is now possible
1619    to convert an ML string into C string with InlineT.CharVector.getData.
1620    This compiles into one single machine instruction.
1621    
1622    ----------------------------------------------------------------------
1623    Name: Allen Leung
1624    Date: 2002/03/10 23:55:00 EST
1625    Tag: leunga-20020310-x86-call
1626    Description:
1627    
1628       Added machine generation for CALL instruction (relative displacement mode)
1629    
1630    ----------------------------------------------------------------------
1631    Name: Matthias Blume
1632    Date: 2002/03/08 16:05:00
1633    Tag: blume-20020308-entrypoints
1634    Description:
1635    
1636    Version number bumped to 110.39.1.  NEW BOOTFILES!
1637    
1638    Entrypoints: non-zero offset into a code object where execution should begin.
1639    
1640    - Added the notion of an entrypoint to CodeObj.
1641    - Added reading/writing of entrypoint info to Binfile.
1642    - Made runtime system bootloader aware of entrypoints.
1643    - Use the address of the label of the first function given to mlriscGen
1644      as the entrypoint.  This address is currently always 0, but it will
1645      not be 0 once we turn on block placement.
1646    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1647      with entry points) from mlriscGen.
1648    
1649    ----------------------------------------------------------------------
1650    Name: Allen Leung
1651    Date: 2002/03/07 20:45:00 EST
1652    Tag: leunga-20020307-x86-cmov
1653    Description:
1654    
1655       Bug fixes for CMOVcc on x86.
1656    
1657       1. Added machine code generation for CMOVcc
1658       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1659       3. CMOVcc cannot have an immediate operand as argument.
1660    
1661    ----------------------------------------------------------------------
1662    Name: Matthias Blume
1663    Date: 2002/03/07 16:15:00 EST
1664    Tag: blume-20020307-controls
1665    Description:
1666    
1667    This is a very large but mostly boring patch which makes (almost)
1668    every tuneable compiler knob (i.e., pretty much everything under
1669    Control.* plus a few other things) configurable via both the command
1670    line and environment variables in the style CM did its configuration
1671    until now.
1672    
1673    Try starting sml with '-h' (or, if you are brave, '-H')
1674    
1675    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1676    implements the underlying generic mechanism.
1677    
1678    The interface to some of the existing such facilities has changed somewhat.
1679    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1680    (The getFoo interface is still there for backward-compatibility, but its
1681    use is deprecated.)
1682    
1683    The ml-build script passes -Cxxx=yyy command-line arguments through so
1684    that one can now twiddle the compiler settings when using this "batch"
1685    compiler.
1686    
1687    TODO items:
1688    
1689    We should go through and throw out all controls that are no longer
1690    connected to anything.  Moreover, we should go through and provide
1691    meaningful (and correct!) documentation strings for those controls
1692    that still are connected.
1693    
1694    Currently, multiple calls to Controls.new are accepted (only the first
1695    has any effect).  Eventually we should make sure that every control
1696    is being made (via Controls.new) exactly once.  Future access can then
1697    be done using Controls.acc.
1698    
1699    Finally, it would probably be a good idea to use the getter-setter
1700    interface to controls rather than ref cells.  For the time being, both
1701    styles are provided by the Controls module, but getter-setter pairs are
1702    better if thread-safety is of any concern because they can be wrapped.
1703    
1704    *****************************************
1705    
1706    One bug fix: The function blockPlacement in three of the MLRISC
1707    backpatch files used to be hard-wired to one of two possibilities at
1708    link time (according to the value of the placementFlag).  But (I
1709    think) it should rather sense the flag every time.
1710    
1711    *****************************************
1712    
1713    Other assorted changes (by other people who did not supply a HISTORY entry):
1714    
1715    1. the cross-module inliner now works much better (Monnier)
1716    2. representation of weights, frequencies, and probabilities in MLRISC
1717       changed in preparation of using those for weighted block placement
1718       (Reppy, George)
1719    
1720    ----------------------------------------------------------------------
1721    Name: Lal George
1722    Date: 2002/03/07 14:44:24 EST 2002
1723    Tag: george-20020307-weighted-block-placement
1724    
1725    Tested the weighted block placement optimization on all architectures
1726    (except the hppa) using AMPL to generate the block and edge frequencies.
1727    Changes were required in the machine properties to correctly
1728    categorize trap instructions. There is an MLRISC flag
1729    "weighted-block-placement" that can be used to enable weighted block
1730    placement, but this will be ineffective without block/edge
1731    frequencies (coming soon).
1732    
1733    
1734    ----------------------------------------------------------------------
1735    Name: Lal George
1736    Date: 2002/03/05 17:24:48 EST
1737    Tag: george-20020305-linkage-cluster
1738    
1739    In order to support the block placement optimization, a new cluster
1740    is generated as the very first cluster (called the linkage cluster).
1741    It contains a single jump to the 'real' entry point for the compilation
1742    unit. Block placement has no effect on the linkage cluster itself, but
1743    all the other clusters  have full freedom in the manner in which they
1744    reorder blocks or functions.
1745    
1746    On the x86 the typical linkage code that is generated is:
1747       ----------------------
1748            .align 2
1749       L0:
1750            addl    $L1-L0, 72(%esp)
1751            jmp     L1
1752    
1753    
1754            .align  2
1755       L1:
1756       ----------------------
1757    
1758    72(%esp) is the memory location for the stdlink register. This
1759    must contain the address of the CPS function being called. In the
1760    above example, it contains the address of  L0; before
1761    calling L1 (the real entry point for the compilation unit), it
1762    must contain the address for L1, and hence
1763    
1764            addl $L1-L0, 72(%esp)
1765    
1766    I have tested this on all architectures except the hppa.The increase
1767    in code size is of course negligible
1768    
1769    ----------------------------------------------------------------------
1770    Name: Allen Leung
1771    Date: 2002/03/03 13:20:00 EST
1772    Tag: leunga-20020303-mlrisc-tools
1773    
1774      Added #[ ... ] expressions to mlrisc tools
1775    
1776    ----------------------------------------------------------------------
1777    Name: Matthias Blume
1778    Date: 2002/02/27 12:29:00 EST
1779    Tag: blume-20020227-cdebug
1780    Description:
1781    
1782    - made types in structure C and C_Debug to be equal
1783    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1784    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1785    
1786    ----------------------------------------------------------------------
1787    Name: Matthias Blume
1788    Date: 2002/02/26 12:00:00 EST
1789    Tag: blume-20020226-ffi
1790    Description:
1791    
1792    1. Fixed a minor bug in CM's "noweb" tool:
1793       If numbering is turned off, then truly don't number (i.e., do not
1794       supply the -L option to noweb).  The previous behavior was to supply
1795       -L'' -- which caused noweb to use the "default" line numbering scheme.
1796       Thanks to Chris Richards for pointing this out (and supplying the fix).
1797    
1798    2. Once again, I reworked some aspects of the FFI:
1799    
1800       A. The incomplete/complete type business:
1801    
1802       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1803         gone!
1804       - ML types representing an incomplete type are now *equal* to
1805         ML types representing their corresponding complete types (just like
1806         in C).  This is still safe because ml-nlffigen will not generate
1807         RTTI for incomplete types, nor will it generate functions that
1808         require access to such RTTI.   But when ML code generated from both
1809         incomplete and complete versions of the C type meet, the ML types
1810         are trivially interoperable.
1811    
1812         NOTE:  These changes restore the full generality of the translation
1813         (which was previously lost when I eliminated functorization)!
1814    
1815       B. Enum types:
1816    
1817       - Structure C now has a type constructor "enum" that is similar to
1818         how the "su" constructor works.  However, "enum" is not a phantom
1819         type because each "T enum" has values (and is isomorphic to
1820         MLRep.Signed.int).
1821       - There are generic access operations for enum objects (using
1822         MLRep.Signed.int).
1823       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1824         * The structure contains the definition of type "mlrep" (the ML-side
1825         representation type of the enum).  Normally, mlrep is the same
1826         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1827         then mlrep will be defined as a datatype -- thus facilitating
1828         pattern matching on mlrep values.
1829         ("-ec" will be suppressed if there are duplicate values in an
1830          enumeration.)
1831         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1832         will be generated for each C enum constant xxx.
1833         * Conversion functions m2i and i2m convert between mlrep and
1834         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1835         * Coversion functions c and ml convert between mlrep and "tag enum".
1836         * Access functions (get/set) fetch and store mlrep values.
1837       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1838         enumerations are merged into one single enumeration represented by
1839         structure E_'.
1840    
1841    ----------------------------------------------------------------------
1842    Name: Allen Leung
1843    Date: 2002/02/25 04:45:00 EST
1844    Tag: leunga-20020225-cps-spill
1845    
1846    This is a new implementation of the CPS spill phase.
1847    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1848    In case of problems, replace it with the old file spill.sml
1849    
1850    The current compiler runs into some serious performance problems when
1851    constructing a large record.  This can happen when we try to compile a
1852    structure with many items.  Even a very simple structure like the following
1853    makes the compiler slow down.
1854    
1855        structure Foo = struct
1856           val x_1 = 0w1 : Word32.int
1857           val x_2 = 0w2 : Word32.int
1858           val x_3 = 0w3 : Word32.int
1859           ...
1860           val x_N = 0wN : Word32.int
1861        end
1862    
1863    The following table shows the compile time, from N=1000 to N=4000,
1864    with the old compiler:
1865    
1866    N
1867    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1868           MLRISC ra                               0.06u  0.00s  0.05g
1869              (spills = 0 reloads = 0)
1870           TOTAL                                   0.63u  0.07s  0.21g
1871    
1872    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1873           MLRISC ra                               5.68u  0.59s  3.93g
1874              (spills = 0 reloads = 0)
1875           TOTAL                                   14.71u  0.99s  4.81g
1876    
1877    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1878           MLRISC ra                               5.54u  0.65s  3.91g
1879              (spills = 543 reloads = 1082)
1880           TOTAL                                   65.40u  3.13s  6.00g
1881    
1882    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1883           MLRISC ra                               0.80u  0.10s  0.55g
1884              (spills = 42 reloads = 84)
1885           TOTAL                                   129.42u  5.10s  4.13g
1886    
1887    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1888           MLRISC ra                               2.69u  0.27s  1.38g
1889              (spills = 62 reloads = 124)
1890           TOTAL                                   682.48u  19.61s  13.99g
1891    
1892    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1893           MLRISC ra                               4.96u  0.27s  2.72g
1894              (spills = 85 reloads = 170)
1895           TOTAL                                   2375.26u  57.21s  48.00g
1896    
1897    As you can see the old cps spill module suffers from some serious
1898    performance problem.  But since I cannot decipher the old code fully,
1899    instead of patching the problems up, I'm reimplementing it
1900    with a different algorithm.  The new code is more modular,
1901    smaller when compiled, and substantially faster
1902    (O(n log n) time and O(n) space).  Timing of the new spill module:
1903    
1904    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1905          MLRISC ra                               0.25u  0.02s  0.15g
1906             (spills=1 reloads=3)
1907          TOTAL                                   7.74u  0.34s  1.62g
1908    
1909    Implementation details:
1910    
1911    As far as I can tell, the purpose of the CPS spill module is to make sure the
1912    number of live variables at any program point (the bandwidth)
1913    does not exceed a certain limit, which is determined by the
1914    size of the spill area.
1915    
1916    When the bandwidth is too large, we decrease the register pressure by
1917    packing live variables into spill records.  How we achieve this is
1918    completely different than what we did in the old code.
1919    
1920    First, there is something about the MLRiscGen code generator
1921    that we should be aware of:
1922    
1923    o MLRiscGen performs code motion!
1924    
1925       In particular, it will move floating point computations and
1926       address computations involving only the heap pointer to
1927       their use sites (if there is only a single use).
1928       What this means is that if we have a CPS record construction
1929       statement
1930    
1931           RECORD(k,vl,w,e)
1932    
1933       we should never count the new record address w as live if w
1934       has only one use (which is often the case).
1935    
1936       We should do something similar to floating point, but the transformation
1937       there is much more complex, so I won't deal with that.
1938    
1939    Secondly, there are now two new cps primops at our disposal:
1940    
1941     1. rawrecord of record_kind option
1942        This pure operator allocates some uninitialized storage from the heap.
1943        There are two forms:
1944    
1945         rawrecord NONE [INT n]  allocates a tagless record of length n
1946         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1947                                     and initializes the tag.
1948    
1949     2. rawupdate of cty
1950          rawupdate cty (v,i,x)
1951          Assigns to x to the ith component of record v.
1952          The storelist is not updated.
1953    
1954    We use these new primops for both spilling and increment record construction.
1955    
1956     1. Spilling.
1957    
1958        This is implemented with a linear scan algorithm (but generalized
1959        to trees).  The algorithm will create a single spill record at the
1960        beginning of the cps function and use rawupdate to spill to it,
1961        and SELECT or SELp to reload from it.  So both spills and reloads
1962        are fine-grain operations.  In contrast, in the old algorithm
1963        "spills" have to be bundled together in records.
1964    
1965        Ideally, we should sink the spill record construction to where
1966        it is needed.  We can even split the spill record into multiple ones
1967        at the places where they are needed.  But CPS is not a good
1968        representation for global code motion, so I'll keep it simple and
1969        am not attempting this.
1970    
1971     2. Incremental record construction (aka record splitting).
1972    
1973        Long records with many component values which are simulatenously live
1974        (recall that single use record addresses are not considered to
1975         be live) are constructed with rawrecord and rawupdate.
1976        We allocate space on the heap with rawrecord first, then gradually
1977        fill it in with rawupdate.  This is the technique suggested to me
1978        by Matthias.
1979    
1980        Some restrictions on when this is applicable:
1981        1. It is not a VECTOR record.  The code generator currently does not handle
1982           this case. VECTOR record uses double indirection like arrays.
1983        2. All the record component values are defined in the same "basic block"
1984           as the record constructor.  This is to prevent speculative
1985           record construction.
1986    
1987    ----------------------------------------------------------------------
1988    Name: Allen Leung
1989    Date: 2002/02/22 01:02:00 EST
1990    Tag: leunga-20020222-mlrisc-tools
1991    
1992    Minor bug fixes in the parser and rewriter
1993    
1994    ----------------------------------------------------------------------
1995    Name: Allen Leung
1996    Date: 2002/02/21 20:20:00 EST
1997    Tag: leunga-20020221-peephole
1998    
1999    Regenerated the peephole files.  Some contained typos in the specification
2000    and some didn't compile because of pretty printing bugs in the old version
2001    of 'nowhere'.
2002    
2003    ----------------------------------------------------------------------
2004    Name: Allen Leung
2005    Date: 2002/02/19 20:20:00 EST
2006    Tag: leunga-20020219-mlrisc-tools
2007    Description:
2008    
2009       Minor bug fixes to the mlrisc-tools library:
2010    
2011       1.  Fixed up parsing colon suffixed keywords
2012       2.  Added the ability to shut the error messages up
2013       3.  Reimplemented the pretty printer and fixed up/improved
2014           the pretty printing of handle and -> types.
2015       4.  Fixed up generation of literal symbols in the nowhere tool.
2016       5.  Added some SML keywords to to sml.sty
2017    
2018    ----------------------------------------------------------------------
2019    Name: Matthias Blume
2020    Date: 2002/02/19 16:20:00 EST
2021    Tag: blume-20020219-cmffi
2022    Description:
2023    
2024    A wild mix of changes, some minor, some major:
2025    
2026    * All C FFI-related libraries are now anchored under $c:
2027        $/c.cm      --> $c/c.cm
2028        $/c-int.cm  --> $c/internals/c-int.cm
2029        $/memory.cm --> $c/memory/memory.cm
2030    
2031    * "make" tool (in CM) now treats its argument pathname slightly
2032      differently:
2033        1. If the native expansion is an absolute name, then before invoking
2034           the "make" command on it, CM will apply OS.Path.mkRelative
2035           (with relativeTo = OS.FileSys.getDir()) to it.
2036        2. The argument will be passed through to subsequent phases of CM
2037           processing without "going native".  In particular, if the argument
2038           was an anchored path, then "make" will not lose track of that anchor.
2039    
2040    * Compiler backends now "know" their respective C calling conventions
2041      instead of having to be told about it by ml-nlffigen.  This relieves
2042      ml-nlffigen from one of its burdens.
2043    
2044    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
2045    
2046    * Export C_DEBUG and C_Debug from $c/c.cm.
2047    
2048    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
2049      subtyping relationship between incomplete pointers and their complete
2050      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
2051      with the convention of instantiating 'o with ('t, 'c) obj whenever
2052      the pointer target type is complete.  In the incomplete case, 'o
2053      will be instantiated with some "'c iobj" -- a type obtained by
2054      using one of the functors PointerToIncompleteType or PointerToCompleteType.
2055    
2056      Operations that work on both incomplete and complete pointer types are
2057      typed as taking an 'o ptr while operations that require the target to
2058      be known are typed as taking some ('t, 'c) obj ptr.
2059    
2060      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
2061      where void is an eqtype without any values.  This makes it possible
2062      to work on voidptr values using functions meant to operate on light
2063      incomplete pointers.
2064    
2065    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
2066      been vastly simplified.
2067    
2068    ----------------------------------------------------------------------
2069    Name: Matthias Blume
2070    Date: 2002/02/19 10:48:00 EST
2071    Tag: blume-20020219-pqfix
2072    Description:
2073    
2074    Applied Chris Okasaki's bug fix for priority queues.
2075    
2076    ----------------------------------------------------------------------
2077    Name: Matthias Blume
2078    Date: 2002/02/15 17:05:00
2079    Tag: Release_110_39
2080    Description:
2081    
2082    Last-minute retagging is becoming a tradition... :-(
2083    
2084    This is the working release 110.39.
2085    
2086    ----------------------------------------------------------------------
2087    Name: Matthias Blume
2088    Date: 2002/02/15 16:00:00 EST
2089    Tag: Release_110_39-orig
2090    Description:
2091    
2092    Working release 110.39.  New bootfiles.
2093    
2094    (Update: There was a small bug in the installer so it wouldn't work
2095    with all shells.  So I retagged. -Matthias)
2096    
2097    ----------------------------------------------------------------------
2098    Name: Matthias Blume
2099    Date: 2002/02/15 14:17:00 EST
2100    Tag: blume-20020215-showbindings
2101    Description:
2102    
2103    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
2104    the latter can be useful for exploring what bindings are available at
2105    the interactive prompt.  (The first function returns only the list
2106    of symbols that are really bound, the second prints those but also the
2107    ones that CM's autoloading mechanism knows about.)
2108    
2109    ----------------------------------------------------------------------
2110    Name: Matthias Blume
2111    Date: 2002/02/15 12:08:00 EST
2112    Tag: blume-20020215-iptrs
2113    Description:
2114    
2115    Two improvements to ml-nlffigen:
2116    
2117      1. Write files only if they do not exist or if their current contents
2118         do not coincide with what's being written.  (That is, avoid messing
2119         with the time stamps unless absolutely necessary.)
2120    
2121      2. Implement a "repository" mechanism for generated files related
2122         to "incomplete pointer types".   See the README file for details.
2123    
2124    ----------------------------------------------------------------------
2125    Name: Matthias Blume
2126    Date: 2002/02/14 11:50:00 EST
2127    Tag: blume-20020214-quote
2128    Description:
2129    
2130    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
2131    because of the new and improved tag generation scheme.  (Thanks to Allen
2132    Leung for pointing it out.)
2133    
2134    ----------------------------------------------------------------------
2135    Name: Lal George
2136    Date: 2002/02/14 09:55:27 EST 2002
2137    Tag: george-20020214-isabelle-bug
2138    Description:
2139    
2140    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
2141    of Isabelle on the x86.
2142    
2143    From Allen:
2144    -----------
2145     I've found the problem:
2146    
2147         in ra-core.sml, I use the counter "blocked" to keep track of the
2148         true number of elements in the freeze queue.  When the counter goes
2149         to zero, I skip examining the queue.  But I've messed up the
2150         bookkeeping in combine():
2151    
2152             else ();
2153             case !ucol of
2154               PSEUDO => (if !cntv > 0 then
2155                     (if !cntu > 0 then blocked := !blocked - 1 else ();
2156                                        ^^^^^^^^^^^^^^^^^^^^^^^
2157                      moveu := mergeMoveList(!movev, !moveu)
2158                     )
2159                  else ();
2160    
2161         combine() is called to coalesce two nodes u and v.
2162         I think I was thinking that if the move counts of u and v are both
2163         greater than zero then after they are coalesced then one node is
2164         removed from the freeze queue.  Apparently I was thinking that
2165         both u and v are of low degree, but that's clearly not necessarily true.
2166    
2167    
2168    02/12/2002:
2169        Here's the patch.  HOL now compiles.
2170    
2171        I don't know how this impact on performance (compile
2172        time or runtime).  This bug caused the RA (especially on the x86)
2173        to go thru the potential spill phase when there are still nodes on the
2174        freeze queue.
2175    
2176    
2177    
2178    
2179    ----------------------------------------------------------------------
2180    Name: Matthias Blume
2181    Date: 2002/02/13 22:40:00 EST
2182    Tag: blume-20020213-fptr-rtti
2183    Description:
2184    
2185    Fixed a bug in ml-nlffigen that was introduced with one of the previous
2186    updates.
2187    
2188    ----------------------------------------------------------------------
2189    Name: Matthias Blume
2190    Date: 2002/02/13 16:41:00 EST
2191    Tag: blume-20020213-cmlpq
2192    Description:
2193    
2194    Added new priority queue export symbols (which have just been added to
2195    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
2196    not compile and the installer would choke.)
2197    
2198    ----------------------------------------------------------------------
2199    Name: Matthias Blume
2200    Date: 2002/02/13 16:15:00 EST
2201    Tag: blume-20020213-various
2202    Description:
2203    
2204    1. More tweaks to ml-nlffigen:
2205    
2206       - better internal datastructures (resulting in slight speedup)
2207       - "-match" option requires exact match
2208       - "localized" gensym counters (untagged structs/unions nested within
2209         other structs/unions or within typedefs get a fresh counter; their
2210         tag will be prefixed by a concatenation of their parents' tags)
2211       - bug fixes (related to calculation of transitive closure of types
2212         to be included in the output)
2213    
2214    2. Minor Basis updates:
2215    
2216       - added implementations for List.collate and Option.app
2217    
2218    ----------------------------------------------------------------------
2219    Name: Matthias Blume
2220    Date: 2002/02/11 15:55:00 EST
2221    Tag: blume-20020211-gensym
2222    Description:
2223    
2224    Added a "-gensym" option to command line of ml-nlffigen.  This can be
2225    used to specify a "stem" -- a string that is inserted in all "gensym'd"
2226    names (ML structure names that correspond to unnamed C structs, unions,
2227    and enums), so that separate runs of ml-nlffigen do not clash.
2228    
2229    ----------------------------------------------------------------------
2230    Name: Matthias Blume
2231    Date: 2002/02/11 12:05:00 EST
2232    Tag: blume-20020211-gensml
2233    Description:
2234    
2235    A quick fix for a problem with GenSML (in the pgraph-util library):
2236    Make generation of toplevel "local" optional.  (Strictly speaking,
2237    signature definitions within "local" are not legal SML.)
2238    
2239    Other than that: updates to INSTALL and cm/TODO.
2240    
2241    ----------------------------------------------------------------------
2242    Name: Matthias Blume
2243    Date: 2002/02/08 15:00:00 EST
2244    Tag: blume-20020208-uniquepid
2245    Description:
2246    
2247    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
2248    
2249    1. The installer (config/install.sh) has gotten smarter:
2250    
2251         - Configuration options are a bit easier to specify now
2252           (in config/targets).
2253         - Bug in recognizing .tar.bz2 files fixed.
2254         - Installer automatically resolves dependencies between
2255           configuration options (e.g., if you ask for eXene, you will
2256           also get cml -- regardless whether you asked for it or not).
2257         - Installer can run in "quieter mode" by setting the environment
2258           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
2259           "completely silent", though.
2260         - Build HashCons library as part of smlnj-lib.
2261    
2262    2. A new scheme for assigning persistent identifiers to compilation
2263       units (and, by extension, to types etc.) has been put into place.
2264       This fixes a long-standing bug where types and even dynamic values
2265       can get internally confused, thereby compromising type safety
2266       (abstraction) and dynamic correctness.  See
2267    
2268         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
2269    
2270       for an example of how things could go wrong until now.
2271    
2272       The downside of the new scheme is that pids are not quite as
2273       persistent as they used to be: CM will generate a fresh pid
2274       for every compilation unit that it thinks it sees for the first
2275       time.  That means that if you compile starting from a clean, fresh
2276       source tree at two different times, you end up with different
2277       binaries.
2278    
2279       Cutoff recompilation, however, has not been compromised because
2280       CM keeps pid information in special caches between runs.
2281    
2282    ----------------------------------------------------------------------
2283    Name: Lal George
2284    Date: 2002/02/07 15:34:13 EST 2002
2285    Tag: <none>
2286    Description:
2287    
2288    Compilers that generate assembly code may produce  global labels
2289    whose value is resolved at link time. The various peephole optimization
2290    modules did not take this in account.
2291    
2292    TODO. The Labels.addrOf function should really return an option
2293    type so that clients are forced to deal with this issue, rather
2294    than an exception being raised.
2295    
2296    ----------------------------------------------------------------------
2297    Name: Lal George
2298    Date: 2002/02/06 13:55:02 EST
2299    Tag: george-20020206-ra-breakup
2300    Description:
2301    
2302    1. A bug fix from Allen.
2303    
2304        A typo causes extra fstp %st(0)'s to be generated at compensation
2305        edges, which might cause stack underflow traps at runtime.  This
2306        occurs in fft where there are extraneous fstps right before the 'into'
2307        trap instruction (in this case they are harmless since none of the
2308        integers overflow.)
2309    
2310    2. Pulled out various utility modules that were embedded in the modules
2311       of the register allocator. I need these modules for other purposes, but
2312       they are not complete enough to put into a library (just yet).
2313    ----------------------------------------------------------------------
2314    Name: Matthias Blume
2315    Date: 2002/01/31 16:05:00 EST
2316    Tag: blume-20020131-sparc-ccalls
2317    Description:
2318    
2319    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
2320       of extra stack space by mistake.  Fixed.
2321    
2322    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
2323    
2324    ----------------------------------------------------------------------
2325    Name: Allen Leung
2326    Date: 2002/01/30
2327    Tag: leunga-20020130-nowhere-bug-fix
2328    Description:
2329    
2330       MLRISC bug fixes:
2331       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
2332       2. MachineInt.fromString was negating its value.
2333    
2334    ----------------------------------------------------------------------
2335    Name: Matthias Blume
2336    Date: 2002/01/29
2337    Tag: blume-20020129-INSTALL
2338    Description:
2339    
2340    - Added somewhat detailed installation instructions (file INSTALL).
2341    - Fixed curl-detection bug in config/install.sh.
2342    - It is now possible to select the URL getter using the URLGETTER
2343      environment variable:
2344    
2345          not set / "unknown"      --> automatic detection (script tries wget,
2346                                       curl, and lynx)
2347          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
2348                                       how to properly invoke them)
2349          other                    --> use $URLGETTER directly, it must take
2350                                       precisely two command-line arguments
2351                                       (source URL and destination file name)
2352    
2353    ----------------------------------------------------------------------
2354    Name: Matthias Blume
2355    Date: 2002/01/28
2356    Tag: blume-20020128-sparc-ccalls
2357    Description:
2358    
2359    - Fixed problem with calculation of "used" registers in sparc-c-calls.
2360    - Make use of the allocParam argument in sparc-c-calls.
2361    
2362    ----------------------------------------------------------------------
2363    Name: Matthias Blume
2364    Date: 2002/01/28
2365    Tag: blume-20020128-allocParam
2366    Description:
2367    
2368    John Reppy:  Changes c-calls API to accept client-callback for
2369    allocating extra stack space.
2370    me: Corresponding changes to mlriscGen (using a dummy argument that
2371        does not change the current behavior).
2372    
2373    ----------------------------------------------------------------------
2374    Name: Matthias Blume
2375    Date: 2002/01/28 12:00:00
2376    Tag: Release_110_38
2377    Description:
2378    
2379    This time for real!!!
2380    
2381    ----------------------------------------------------------------------
2382    Name: Matthias Blume
2383    Date: 2002/01/28 10:56:00 EST
2384    Tag: blume-20020128-retraction
2385    Description:
2386    
2387    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
2388       with blume-Release_110_38-retracted.)
2389    
2390    1. Fixed a problem with incorrect rounding modes in real64.sml.
2391       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
2392    
2393    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
2394       and enums fixed.  The naming of corresponding ML identifiers should
2395       now be consistent again.
2396    
2397    ----------------------------------------------------------------------
2398    Name: Allen Leung
2399    Date: 2002/01/27
2400    Tag: leunga-20020127-nowhere
2401    Description:
2402    
2403       Added a target called nowhere in the configuration scripts.
2404       Enabling this will build the MLRISC 'nowhere' tool (for translating
2405       programs with where-clauses into legal SML code) during installation.
2406    
2407    ----------------------------------------------------------------------
2408    Name: Matthias Blume
2409    Date: 2002/01/25 21:27:00 EST
2410    Tag: blume-Release_110_38-retracted
2411    Description:
2412    
2413    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
2414    
2415    README will be added later.
2416    
2417    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
2418    (Release_110_38) removed.  Reason: Last-minute bug fixes.
2419    
2420    ----------------------------------------------------------------------
2421    Name: Matthias Blume
2422    Date: 2002/01/25
2423    Tag: blume-20020125-ffi
2424    Description:
2425    
2426    A large number of tweaks and improvements to ml-nlffi-lib and
2427    ml-nlffigen:
2428    
2429       - ML represenation types have been streamlined
2430       - getter and setter functions work with concrete values, not abstract
2431         ones where possible
2432       - ml-nlffigen command line more flexible (see README file there)
2433       - some bugs have been fixed (hopefully)
2434    
2435    ----------------------------------------------------------------------
2436    Name: Lal George
2437    Date: 2002/01/24
2438    Tag: george-20020124-risc-ra-interface
2439    Description:
2440    
2441       There is a dramatic simplification in the interface to the
2442       register allocator for RISC architectures as a result of making
2443       parallel copy instructions explicit.
2444    
2445    ----------------------------------------------------------------------
2446    Name: Matthias Blume
2447    Date: 2002/01/22
2448    Tag: blume-20020122-x86-ccalls
2449    Description:
2450    
2451    Bug fix for c-calls on x86 (having to do with how char- and
2452    short-arguments are being handled).
2453    
2454    ----------------------------------------------------------------------
2455    Name: Matthias Blume
2456    Date: 2002/01/21
2457    Tag: blume-20020121-ff
2458    Description:
2459    
2460    Another day of fiddling with the FFI...
2461    
2462    1. Bug fix/workaround:  CKIT does not complain about negative array
2463       dimensions, so ml-nlffigen has to guard itself against this possibility.
2464       (Otherwise a negative dimension would send it into an infinite loop.)
2465    
2466    2. Some of the abstract types (light objects, light pointers, most "base"
2467       types) in structure C are now eqtypes.
2468    
2469    3. Added constructors and test functions for NULL function pointers.
2470    
2471    ----------------------------------------------------------------------
2472    Name: Matthias Blume
2473    Date: 2002/01/18
2474    Tag: blume-20020118-ready-for-new-release
2475    Description:
2476    
2477    Made config/srcarchiveurl point to a new place.  (Will provide boot
2478    files shortly.)
2479    
2480    Maybe we christen this to be 110.38?
2481    
2482    ----------------------------------------------------------------------
2483    Name: Matthias Blume
2484    Date: 2002/01/18
2485    Tag: blume-20020118-more-ffifiddle
2486    Description:
2487    
2488    Today's FFI fiddling:
2489    
2490      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
2491        C.Set.* that use concrete (MLRep.*) arguments and results instead
2492        of abstract ones.
2493    
2494      - Provided word-style bit operations etc. for "int" representation
2495        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
2496        and Long.
2497    
2498    ----------------------------------------------------------------------
2499    Name: Matthias Blume
2500    Date: 2002/01/18
2501    Tag: blume-20020118-use-x86-fp
2502    Description:
2503    
2504    Now that x86-fast-fp seems to be working, I turned it back on again
2505    by default.  (Seems to work fine now, even with the FFI.)
2506    
2507    Other than that, I added some documentation about the FFI to
2508    src/ml-nlffigen/README and updated the FFI test examples in
2509    src/ml-nlffi-lib/Tests/*.
2510    
2511    ----------------------------------------------------------------------
2512    Name: Allen Leung
2513    Date: 2002/01/17
2514    Tag: leunga-20020117-x86-fast-fp-call
2515    Description:
2516    
2517       1. Fixed a problem with handling return fp values when x86's fast fp
2518          mode is turned on.
2519    
2520       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
2521          of %f32.
2522    
2523       3. Added a constructor INT32lit to the ast of MLRISC tools.
2524    
2525    ----------------------------------------------------------------------
2526    Name: Matthias Blume
2527    Date: 2002/01/16
2528    Tag: blume-20020116-ffifiddle
2529    Description:
2530    
2531    More fiddling with the FFI interface:
2532    
2533     - Make constness 'c instead of rw wherever possible.  This eliminates
2534       the need for certain explicit coercions.  (However, due to ML's
2535       value polymorphism, there will still be many cases where explicit
2536       coercions are necessary.  Phantom types are not the whole answer
2537       to modeling a subtyping relationship in ML.)
2538    
2539     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
2540    
2541     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
2542       clumsy workaround for varargs problem.)
2543    
2544    ----------------------------------------------------------------------
2545    Name: Lal George
2546    Date: 2002/01/15
2547    Tag: <none>
2548    Description:
2549    
2550    1. Since COPY instructions are no longer native to the architecture,
2551       a generic functor can be used to implement the expandCopies function.
2552    
2553    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
2554       TEXT segment.
2555    
2556    ----------------------------------------------------------------------
2557    Name: Matthias Blume
2558    Date: 2002/01/15
2559    Tag: blume-20020115-ffiupdates
2560    Description:
2561    
2562    1. Fix for bug resulting in single-precision float values being returned
2563       incorrectly from FFI calls.
2564    
2565    2. Small modifications to C FFI API:
2566    
2567        - memory-allocation routines return straight objects (no options)
2568          and raise an exception in out-of-memory situations
2569        - unsafe extensions to cast between function pointers and pointers
2570          from/to ints
2571        - added structure C_Debug as an alternative to structure C where
2572          pointer-dereferencing (|*| and |*!) always check for null-pointers
2573        - added open_lib' to DynLinkage;  open_lib' works like open_lib
2574          but also takes a (possibly empty) list of existing library handles
2575          that the current library depends on
2576    
2577    ----------------------------------------------------------------------
2578    Name: Matthias Blume
2579    Date: 2002/01/10
2580    Tag: blume-20020110-newffigen
2581    Description:
2582    
2583    1. Updates to portable graph code.
2584    
2585    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
2586       more scalable now so that even huge interfaces such as the one
2587       for GTK compile in finite time and space. :-)
2588       See src/ml-nlffigen/README for details on what's new.
2589    
2590    ----------------------------------------------------------------------
2591    Name: Lal George
2592    Date: 2001/01/09 14:31:35 EST 2002
2593    Tag: george-20011206-rm-native-copy
2594    Description:
2595    
2596            Removed the native COPY and FCOPY instructions
2597            from all the architectures and replaced it with the
2598            explicit COPY instruction from the previous commit.
2599    
2600            It is now possible to simplify many of the optimizations
2601            modules that manipulate copies. This has not been
2602            done in this change.
2603    
2604    ----------------------------------------------------------------------
2605    Name: Lal George
2606    Date: 2001/12/06 16:50:13 EST 2001
2607    Tag: george-20011206-mlrisc-instruction
2608    Description:
2609    
2610    Changed the representation of instructions from being fully abstract
2611    to being partially concrete. That is to say:
2612    
2613      from
2614            type instruction
2615    
2616      to
2617            type instr                              (* machine instruction *)
2618    
2619            datatype instruction =
2620                LIVE of {regs: C.cellset, spilled: C.cellset}
2621              | KILL of {regs: C.cellset, spilled: C.cellset}
2622              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
2623              | ANNOTATION of {i: instruction, a: Annotations.annotation}
2624              | INSTR of instr
2625    
2626    This makes the handling of certain special instructions that appear on
2627    all architectures easier and uniform.
2628    
2629    LIVE and KILL say that a list of registers are live or killed at the
2630    program point where they appear. No spill code is generated when an
2631    element of the 'regs' field is spilled, but the register is moved to
2632    the 'spilled' (which is present, more for debugging than anything else).
2633    
2634    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
2635    We used to generate:
2636    
2637            DEFFREG f1
2638            f1 := f2 + f3
2639            trapb
2640    
2641    but now generate:
2642    
2643            f1 := f2 + f3
2644            trapb
2645            LIVE {regs=[f1,f2,f3], spilled=[]}
2646    
2647    Furthermore, the DEFFREG (hack) required that all floating point instruction
2648    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
2649    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
2650    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
2651    this architecture).
2652    
2653    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
2654    all the architectures. This will result in further simplification of the
2655    register allocator that must be aware of them for coalescing purposes, and
2656    will also simplify certain aspects of the machine description that provides
2657    callbacks related to parallel copies.
2658    
2659    ANNOTATION should be obvious, and now INSTR represents the honest to God
2660    machine instruction set!
2661    
2662    The <arch>/instructions/<arch>Instr.sml files define certain utility
2663    functions for making porting easier -- essentially converting upper case
2664    to lower case. All machine instructions (of type instr) are in upper case,
2665    and the lower case form generates an MLRISC instruction. For example on
2666    the alpha we have:
2667    
2668      datatype instr =
2669         LDA of {r:cell, b:cell, d:operand}
2670       | ...
2671    
2672      val lda : {r:cell, b:cell, d:operand} -> instruction
2673        ...
2674    
2675    where lda is just (INSTR o LDA), etc.
2676    
2677    ----------------------------------------------------------------------
2678    Name: Matthias Blume
2679    Date: 2001/11/22 21:40:00 EST
2680    Tag: Release_110_37
2681    Description:
2682    
2683    Release 110.37.  This time for real.
2684    
2685    ----------------------------------------------------------------------
2686    Name: Matthias Blume
2687    Date: 2001/11/21 16:35:00 EST
2688    Tag: blume-20011121-foot-in-mouth
2689    Description:
2690    
2691    Removed the "Release_110_37" tag because of a serious bug.
2692    This will be re-tagged once the bug is fixed.
2693    
2694    ----------------------------------------------------------------------
2695    Name: Matthias Blume
2696    Date: 2001/11/21 16:14:00 EST
2697    Tag: blume-20011121-forgottenfile
2698    Description:
2699    
2700    Forgot to add a file.  (Just a .tex-file -- part of
2701    the CM manual source.)
2702    
2703    ----------------------------------------------------------------------
2704    Name: Matthias Blume
2705    Date: 2001/11/21 16:10:00 EST
2706    Tag: blume-20011121-invalid_110_37
2707    Description:
2708    
2709    Note: I removed the original tag "Release_110_37" from this commit
2710          because we found a serious bug in all non-x86 backends.
2711          - Matthias
2712    
2713    1. Modifications to the SML/NJ code generator and to the runtime system
2714       so that code object name strings are directly inserted into code
2715       objects at code generation time.  The only business the runtime system
2716       has with this is now to read the name strings on occasions.
2717       (The encoding of the name string has also changed somewhat.)
2718    
2719    2. CM now implements a simple "set calculus" for specifying export lists.
2720       In particular, it is now possible to refer to the export lists of
2721       other libraries/groups/sources and form unions as well as differences.
2722       See the latest CM manual for details.
2723    
2724    3. An separate notion of "proxy" libraries has again be eliminated from
2725       CM's model.  (Proxy libraries are now simply a special case of using
2726       the export list calculus.)
2727    
2728    4. Some of the existing libraries now take advantage of the new set
2729       calculus.
2730       (Notice that not all libraries have been converted because some
2731       of the existing .cm-files are supposed to be backward compatible
2732       with 110.0.x.)
2733    
2734    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2735       "General.exnMessage"!  The former relies on a certain hook to be
2736       initialized, and that often does not happen in the stand-alone case.)
2737    
2738    ----------------------------------------------------------------------
2739    Name: Lal George
2740    Date: 2001/11/21  13:56:18 EST
2741    Tag: george-2001121-pseudo-ops
2742    Description:
2743    
2744      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2745      ought to never be any question of incompatabilities with
2746      pseudo-op syntax expected by host assemblers.
2747    
2748      For now, only modules supporting GAS syntax are implemented
2749      but more should follow, such as MASM, and vendor assembler
2750      syntax, e.g. IBM as, Sun as, etc.
2751    
2752    ----------------------------------------------------------------------
2753    Name: Matthias Blume
2754    Date: 2001/11/14 11:52:00 EST
2755    Tag: blume-20011114-srcname
2756    Description:
2757    
2758    1. Routed the name of the current source file to mlriscgen where it
2759       should be directly emitted into the code object.  (This last part
2760       is yet to be done.)
2761    
2762    2. Some cleanup of the pgraph code to make it match the proposal that
2763       I put out the other day.  (The proposal notwithstanding, things are
2764       still in flux here.)
2765    
2766    ----------------------------------------------------------------------
2767    Name: Lal George
2768    Date: 2001/11/14 09:44:04 EST
2769    Tag:
2770    Description:
2771    
2772      Fix for a backpatching bug reported by Allen.
2773    
2774      Because the boundary between short and long span-dependent
2775      instructions is +/- 128, there are an astounding number of
2776      span-dependent instructions whose size is over estimated.
2777    
2778      Allen came up with the idea of letting the size of span
2779      dependent instructions be non-monotonic, for a maxIter
2780      number of times, after which the size must be monotonically
2781      increasing.
2782    
2783      This table shows the number of span-dependent instructions
2784      whose size was over-estimated as a function of maxIter, for the
2785      file Parse/parse/ml.grm.sml:
2786    
2787         maxIter            # of instructions:
2788            10                      687
2789            20                      438
2790            30                      198
2791            40                        0
2792    
2793      In compiling the compiler, there is no significant difference in
2794      compilation speed between maxIter=10 and maxIter=40. Actually,
2795      my measurements showed that maxIter=40 was a tad faster than
2796      maxIter=10! Also 96% of the  files in the compiler reach a fix
2797      point within 13 iterations, so fixing maxIter at 40, while high,
2798      is okay.
2799    
2800    ----------------------------------------------------------------------
2801    Name: Matthias Blume
2802    Date: 2001/10/31 15:25:00 EST
2803    Tag: blume-20011031-pgraph
2804    Description:
2805    
2806    CKIT:
2807    * Changed the "Function" constructor of type Ast.ctype to carry optional
2808      argument identifiers.
2809    * Changed the return type of TypeUtil.getFunction accordingly.
2810    * Type equality ignores the argument names.
2811    * TypeUtil.composite tries to preserve argument names but gives up quickly
2812      if there is a mismatch.
2813    
2814    installation script:
2815    * attempts to use "curl" if available (unless "wget" is available as well)
2816    
2817    CM:
2818    * has an experimental implementation of "portable graphs" which I will
2819      soon propose as an implementation-independent library format
2820    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2821    
2822    NLFFI-LIB:
2823    * some cleanup (all cosmetic)
2824    
2825    NLFFIGEN:
2826    * temporarily disabled the mechanism that suppresses ML output for
2827      C definitions whose identifiers start with an underscore character
2828    * generate val bindings for enum constants
2829    * user can request that only one style (light or heavy) is being used;
2830      default is to use both (command-line arguments: -heavy and -light)
2831    * fixed bug in handling of function types involving incomplete pointers
2832    * generate ML entry points that take record arguments (i.e., using
2833      named arguments) for C functions that have a prototype with named
2834      arguments
2835      (see changes to CKIT)
2836    
2837    ----------------------------------------------------------------------
2838    Name: Allen Leung
2839    Date: 2001/10/27 20:34:00 EDT
2840    Tag: leunga-20011027-x86-fast-fp-call
2841    Description:
2842    
2843       Fixed the bug described in blume-20010920-slowfp.
2844    
2845       The fix involves
2846          1. generating FCOPYs in FSTP in ia32-svid
2847          2. marking a CALL with the appropriate annotation
2848    
2849    ----------------------------------------------------------------------
2850    Name: Matthias Blume
2851    Date: 2001/10/16 11:32:00 EDT
2852    Tag: blume-20011016-netbsd
2853    Description:
2854    
2855    Underscore patch from Chris Richards (fixing problem with compiling
2856    runtime system under recent NetBSD).
2857    
2858    ----------------------------------------------------------------------
2859    Name: Allen Leung
2860    Date: 2001/10/12 17:18:32 EDT 2001
2861    Tag: leung-20011012-x86-printflowgraph
2862    Description:
2863    
2864    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2865    
2866    ----------------------------------------------------------------------
2867    Name: Lal George
2868    Date: 2001/10/11 23:51:34 EDT
2869    Tag: george-20011011-too-many-instrs
2870    Description:
2871    
2872    The representation of a program point never expected to see more
2873    than 65536 instructions in a basic block!
2874    
2875    ----------------------------------------------------------------------
2876    Name: Lal George
2877    Date: 2001/10/09 09:41:37 EDT
2878    Tag: george-20011008-mlrisc-labels
2879    Description:
2880    
2881    Changed the machine description files to support printing of
2882    local and global labels in assembly code, based on host assembler
2883    conventions.
2884    
2885    ----------------------------------------------------------------------
2886    Name: Matthias Blume
2887    Date: 2001/09/25 15:25:00 EDT
2888    Tag: blume-20010925-exninfo
2889    Description:
2890    
2891    I provided a non-hook implementation of exnName (at the toplevel) and
2892    made the "dummy" implementation of exnMessage (at the toplevel) more
2893    useful: if nothing gets "hooked in", then at least you are going to
2894    see the exception name and a message indicating why you don't see more.
2895    
2896    [For the time being, programs that need exnMessage and want to use
2897    ml-build should either use General.exnMessage (strongly recommended) or
2898    refer to structure General at some other point so that CM sees a
2899    static dependency.]
2900    
2901    [Similar remarks go for "print" and "use":  If you want to use their
2902    functionality in stand-alone programs generated by ml-build, then use
2903    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2904    
2905    ----------------------------------------------------------------------
2906    Name: Matthias Blume
2907    Date: 2001/09/20 17:28:00 EDT
2908    Tag: blume-20010920-slowfp
2909    Description:
2910    
2911    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2912    
2913    ----------------------------------------------------------------------
2914    Name: Matthias Blume
2915    Date: 2001/09/20 17:20:00 EDT
2916    Tag: blume-20010920-canonicalpaths
2917    Description:
2918    
2919    0. Updated the BOOT file (something that I forgot to do earlier).
2920    
2921    1. Small internal change to CM so that it avoids "/../" in filenames
2922       as much as possible (but only where it is safe).
2923    
2924    2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2925       that arguments that contain delimiters are passed through correctly.
2926       This change also means that all "special" arguments of the form
2927       @SMLxxx... must come first.
2928    
2929    3. Changed install script to put relative anchor names for tool commands
2930       into pathconfig.
2931    
2932    ----------------------------------------------------------------------
2933    Name: Matthias Blume
2934    Date: 2001/09/18 15:35:00 EDT
2935    Tag: blume-20010918-readme11036
2936    Description:
2937    
2938    Added README files.
2939    
2940    ----------------------------------------------------------------------
2941    Name: Matthias Blume
2942    Date: 2001/09/18 11:45:00 EDT
2943    Tag: Release_110_36 (retag)
2944    Description:
2945    
2946    Fixed mistake in config/preloads. Retagged as 110.36.
2947    
2948    ----------------------------------------------------------------------
2949    Name: Matthias Blume
2950    Date: 2001/09/18 09:40:00 EDT
2951    Tag: Release_110_36_orig (tag changed)
2952    Description:
2953    
2954    New version (110.36).  New bootfiles.
2955    
2956    ----------------------------------------------------------------------
2957    Name: Matthias Blume
2958    Date: 2001/09/14 16:15:00 EDT
2959    Tag: blume-20010914-x86fastfp
2960    Description:
2961    
2962    John committed some changes that Allen made, in particular a (hopefully)
2963    correctly working version of the x86-fp module.
2964    
2965    I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
2966    flag to "true".  Everything seems to compile to a fixpoint ok, and
2967    "mandelbrot" speeds up by about 15%.
2968    
2969    ----------------------------------------------------------------------
2970    Name: Matthias Blume
2971    Date: 2001/09/13 11:20:00 EDT
2972    Tag: blume-20010913-minimal
2973    Description:
2974    
2975    1. Stefan Monnier's patch to fix a miscompilation problem that
2976       was brought to light by John Reppy's work on Moby.
2977    
2978    2. Implemented a minimal "structure Compiler" that contains just
2979       "version" and "architecture".  The minimal version will be
2980       available when the full version is not.  This is for backward-
2981       compatibility with code that wants to test Compiler.version.
2982    
2983    ----------------------------------------------------------------------
2984    Name: Matthias Blume
2985    Date: 2001/08/28 14:03:00 EDT
2986    Tag: blume-20010828-ml-lex
2987    Description:
2988    
2989    Fix for bug 1581, received from Neophytos Michael.
2990    
2991    ----------------------------------------------------------------------
2992    Name: Matthias Blume
2993    Date: 2001/08/27 11:20:00 EDT
2994    Tag: blume-20010827-readme11035
2995    Description:
2996    
2997    Fleshed out the README file for 110.35.
2998    
2999    ----------------------------------------------------------------------
3000    Name: Matthias Blume
3001    Date: 2001/08/24 17:10:00 EDT
3002    Tag: Release_110_35
3003    Description:
3004    
3005    New version number (110.35).  New bootfiles.
3006    
3007    ----------------------------------------------------------------------
3008    Name: Lal George
3009    Date: 2001/08/24 13:47:18 EDT 2001
3010    Tag: george-20010824-MLRISC-graphs
3011    Description:
3012    
3013     removed clusters from MLRISC completely and replaced with graphs.
3014    
3015    ----------------------------------------------------------------------
3016    Name: Matthias Blume
3017    Date: 2001/08/23 17:50:00 EDT
3018    Tag: blume-20010823-toplevel
3019    Description:
3020    
3021    - some reorganization of the code that implements various kinds of
3022      environments in the compiler (static, dynamic, symbolic, combined)
3023    - re-implemented the EnvRef module so that evalStream works properly
3024      (if the stream contains references to "use", "CM.make", etc.)
3025    - cleaned up evalloop.sml and interact.sml (but they need more cleaning)
3026    
3027    ----------------------------------------------------------------------
3028    Name: Matthias Blume
3029    Date: 2001/08/20 15:50 EDT
3030    Tag: blume20010820-slipup
3031    Description:
3032    
3033    I forgot to commit a few files.  Here they are...
3034    
3035    ----------------------------------------------------------------------
3036    Name: Matthias Blume
3037    Date: 2001/08/20 15:35:00 EDT
3038    Tag: blume-20010820-debugprof
3039    Description:
3040    
3041    !!!! NEW BOOTFILES !!!!
3042    
3043    This is another round of reorganizing the compiler sources.  This
3044    time the main goal was to factor out all the "instrumentation"
3045    passes (for profiling and backtracing) into their own library.
3046    The difficulty was to do it in such a way that it does not depend
3047    on elaborate.cm but only on elabdata.cm.
3048    
3049    Therefore there have been further changes to both elaborate.cm and
3050    elabdata.cm -- more "generic" things have been moved from the former
3051    to the latter.  As a result, I was forced to split the assignment
3052    of numbers indicating "primtyc"s into two portions: SML-generic and
3053    SML/NJ-specific.  Since it would have been awkward to maintain,
3054    I bit the bullet and actually _changed_ the mapping between these
3055    numbers and primtycs.  The bottom line of this is that you need
3056    a new set of bin- and bootfiles.
3057    
3058    I have built new bootfiles for all architectures, so doing a fresh
3059    checkout and config/install.sh should be all you need.
3060    
3061    The newly created library's name is
3062    
3063        $smlnj/viscomp/debugprof.cm
3064    
3065    and its sources live under
3066    
3067        src/compiler/DebugProf
3068    
3069    ----------------------------------------------------------------------
3070    Name: Matthias Blume
3071    Date: 2001/08/15 17:15:00 EDT
3072    Tag: blume-20010815-compreorg
3073    Description:
3074    
3075    This is a first cut at reorganizing the CM libraries that make up the
3076    core of the compiler.  The idea is to separate out pieces that could
3077    be used independently by tools, e.g., the parser, the typechecker, etc.
3078    
3079    The current status is a step in this direction, but it is not quite
3080    satisfactory yet.  Expect more changes in the future.
3081    
3082    Here is the current (new) organization...
3083    
3084        What used to be $smlnj/viscomp/core.cm is now divided into
3085        six CM libraries:
3086    
3087             $smlnj/viscomp/basics.cm
3088                           /parser.cm
3089                           /elabdata.cm
3090                           /elaborate.cm
3091                           /execute.cm
3092                           /core.cm
3093    
3094        The CM files for these libraries live under src/system/smlnj/viscomp.
3095        All these libraries are proxy libraries that contain precisely
3096        one CM library component.  Here are the locations of the components
3097        (all within the src/compiler tree):
3098    
3099             Basics/basics.cm
3100             Parse/parser.cm
3101             ElabData/elabdata.cm
3102             Elaborator/elaborate.cm
3103             Execution/execute.cm
3104             core.cm
3105    
3106         [This organization is the same that has been used already
3107         for a while for the architecture-specific parts of the visible
3108         compiler and for the old version of core.cm.]
3109    
3110         As you will notice, many source files have been moved from their
3111         respective original locations to a new home in one of the above
3112         subtrees.
3113    
3114         The division of labor between the new libraries is the following:
3115    
3116             basics.cm:
3117                - Simple, basic definitions that pertain to many (or all) of
3118                  the other libraries.
3119             parser.cm:
3120                - The SML parser, producing output of type Ast.dec.
3121                - The type family for Ast is also defined and exported here.
3122             elabdata.cm:
3123                - The datatypes that describe input and output of the elaborator.
3124                  This includes types, absyn, and static environments.
3125             elaborator.cm:
3126                - The SML/NJ type checker and elaborator.
3127                  This maps an Ast.dec (with a given static environment) to
3128                  an Absyn.dec (with a new static environment).
3129                - This libraries implements certain modules that used to be
3130                  structures as functors (to remove dependencies on FLINT).
3131             execute.cm:
3132                - Everything having to do with executing binary code objects.
3133                - Dynamic environments.
3134             core.cm:
3135                - SML/NJ-specific instantiations of the elaborator and MLRISC.
3136                - Top-level modules.
3137                - FLINT (this should eventually become its own library)
3138    
3139    Notes:
3140    
3141    I am not 100% happy with the way I separated the elaborator (and its
3142    data structures) from FLINT.  Two instances of the same problem:
3143    
3144        1. Data structures contain certain fields that carry FLINT-specific
3145           information.  I hacked around this using exn and the property list
3146           module from smlnj-lib.  But the fact that there are middle-end
3147           specific fields around at all is a bit annoying.
3148    
3149        2. The elaborator calculates certain FLINT-related information.  I tried
3150           to make this as abstract as I could using functorization, but, again,
3151           the fact that the elaborator has to perform calculations on behalf
3152           of the middle-end at all is not nice.
3153    
3154        3. Having to used exn and property lists is unfortunate because it
3155           weakens type checking.  The other alternative (parameterizing
3156           nearly *everything*) is not appealing, though.
3157    
3158    I removed the "rebinding =" warning hack because due to the new organization
3159    it was awkward to maintain it.  As a result, the compiler now issues some of
3160    these warnings when compiling init.cmi during bootstrap compilation. On
3161    the plus side, you also get a warning when you do, for example:
3162       val op = = Int32.+
3163    which was not the case up to now.
3164    
3165    I placed "assign" and "deref" into the _Core structure so that the
3166    code that deals with the "lazy" keyword can find them there.  This
3167    removes the need for having access to the primitive environment
3168    during elaboration.
3169    
3170    ----------------------------------------------------------------------
3171    Name: Matthias Blume
3172    Date: 2001/08/13
3173    Tag: blume-20010813-closures
3174    Description:
3175    
3176    This fix was sent to us by Zhong Shao.  It is supposed to improve the
3177    performance of certain loops by avoiding needless closure allocation.
3178    
3179    ----------------------------------------------------------------------
3180    Name: Lal George
3181    Date: 2001/07/31 10:03:23 EDT 2001
3182    Tag: george-20010731-x86-fmalloc
3183    Description: Fixed bug in x86 calls
3184    
3185        There was a bug where call instructions would mysteriously
3186        vanish. The call instruction had to be one that returned
3187        a floating point value.
3188    
3189    ----------------------------------------------------------------------
3190    Name: Lal George
3191    Date: 2001/07/19 16:36:29 EDT 2001
3192    Tag: george-20010719-simple-cells
3193    Description:
3194    
3195    I have dramatically simplified the interface for CELLS in MLRISC.
3196    
3197    In summary, the cells interface is broken up into three parts:
3198    
3199      1. CellsBasis : CELLS_BASIS
3200    
3201            CellsBasis is a top level structure and common for all
3202            architectures.  it contains the definitions of basic datatypes
3203            and utility  functions over these types.
3204    
3205      2. functor Cells() : CELLS
3206    
3207            Cells generates an interface for CELLS that incorporates the
3208            specific resources on the target architecture, such as the
3209            presence of special register classes, their number and size,
3210            and various useful substructures.
3211    
3212      3. <ARCH>CELLS
3213    
3214            e.g. SparcCells: SPARCCELLS
3215    
3216            <ARCH>CELLS usually contains additional bindings for special
3217            registers  on the architecture, such as:
3218    
3219                    val r0 : cell           (* register zero *)
3220                    val y : cell            (* Y register *)
3221                    val psr : cell          (* processor status register *)
3222                    ...
3223    
3224            The structure returned by applying the Cells functor is opened
3225            in this interface.
3226    
3227    The main implication of all this is that the datatypes for cells is
3228    split between CellsBasis and CELLS -- a fairly simple change for user
3229    code.
3230    
3231    In the old scheme the CELLS interface had a definitional binding of
3232    the form:
3233    
3234            signature CELLS = sig
3235    
3236               structure CellsBasis = CellsBasis
3237    
3238               ...
3239    
3240            end
3241    
3242    With all the sharing constraints that goes on in MLRISC, this old
3243    design  quickly leads to errors such as:
3244    
3245            "structure definition spec inside of sharing ... "
3246    
3247    
3248    and appears to require an unacceptable amount of sharing and where
3249    constraint hackery.
3250    
3251    I think this error message (the interaction of definitional specs and
3252    sharing) requires more explanation on our web page.
3253    
3254    ----------------------------------------------------------------------
3255    Name: Matthias Blume
3256    Date: 2001/07/19 15:00:00 EDT
3257    Tag: blume-20010719-libreorg
3258    Description:
3259    
3260    This update puts together a fairly extensive but straightforward change
3261    to the way the libraries that implement the interactive system are
3262    organized:
3263    
3264       The biggest change is the elimination of structure Compiler.  As a
3265       replacement for this structure, there is now a CM library
3266       (known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
3267       that exports all the substructures of the original structure Compiler
3268       directly.  So instead of saying Compiler.Foo.bar one now simply
3269       says Foo.bar.  (The CM libraries actually export a collection of
3270       structures that is richer than the collection of substructures of
3271       structure Compiler.)
3272    
3273       To make the transition smooth, there is a separate library called
3274       $smlnj/compiler/compiler.cm which puts together and exports the
3275       original structure Compiler (or at least something very close to it).
3276    
3277       There are five members of the original structure Compiler
3278       that are not exported directly but which instead became members
3279       of a new structure Backend (described by signature BACKEND).  These are:
3280       structure Profile (: PROFILE), structure Compile (: COMPILE), structure
3281       Interact (: INTERACT), structure Machine (: MACHINE), and val
3282       architecture (: string).
3283    
3284       Structure Compiler.Version has become structure CompilerVersion.
3285    
3286       Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
3287       by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
3288       or x86, respectively.
3289       Each of these exports the same frontend structures that
3290       $smlnj/compiler.cm exports.  But they do not have a structure Backend
3291       and instead export some structure <Arch>Backend where <Arch> is Alpha32,
3292       Hppa, PPC, Sparc, or X86, respectively.
3293    
3294       Library $smlnj/compiler/all.cm exports the union of the exports of
3295       $smlnj/compiler/<arch>.cm
3296    
3297       There are no structures <Arch>Compiler anymore, use
3298       $smlnj/compiler/<arch>.cm instead.
3299    
3300       Library host-compiler-0.cm is gone.  Instead, the internal library
3301       that instantiates CM is now called cm0.cm.  Selection of the host
3302       compiler (backend) is no longer done here but. (Responsibility for it
3303       now lies with $smlnj/compiler/current.cm.  This seems to be more
3304       logical.)
3305    
3306       Many individual files have been moved or renamed.  Some files have
3307       been split into multiple files, and some "dead" files have been deleted.
3308    
3309    Aside from these changes to library organization, there are also changes
3310    to the way the code itself is organized:
3311    
3312       Structure Binfile has been re-implemented in such a way that it no
3313       longer needs any knowledge of the compiler.  It exclusively deals
3314       with the details of binfile layout.  It no longer invokes the
3315       compiler (for the purpose of creating new prospective binfile
3316       content), and it no longer has any knowledge of how to interpret
3317       pickles.
3318    
3319       Structure Compile (: COMPILE) has been stripped down to the bare
3320       essentials of compilation.  It no longer deals with linking/execution.
3321       The interface has been cleaned up considerably.
3322    
3323       Utility routines for dealing with linking and execution have been
3324       moved into their own substructures.
3325    
3326       (The ultimate goal of these changes is to provide a light-weight
3327       binfile loader/linker (at least for, e.g., stable libraries) that
3328       does not require CM or the compiler to be present.)
3329    
3330    CM documentation has been updated to reflect the changes to library
3331    organization.
3332    
3333    ----------------------------------------------------------------------
3334    Name: Matthias Blume
3335    Date: 2001/07/10 17:30:00 EDT
3336    Tag: Release_110_34
3337    Description:
3338    
3339    Minor tweak to 110.34 (re-tagged):
3340    
3341      - README.html file added to CVS repository
3342      - runtime compiles properly under FreeBSD 3.X and 4.X
3343    
3344    ----------------------------------------------------------------------
3345    Name: Matthias Blume
3346    Date: 2001/07/10 17:30:00 EDT
3347    Tag: Release_110_34
3348    Description:
3349    
3350    New version number (110.34). New bootfiles.
3351    
3352    ----------------------------------------------------------------------
3353    Name: Matthias Blume
3354    Date: 2001/07/09 16:00:00 EDT
3355    Tag: blume-20010709-more-varargs
3356    Description:
3357    
3358    I changed the handling of varargs in ml-nlffigen again:
3359    The ellipsis ... will now simply be ignored (with an accompanying warning).
3360    
3361    The immediate effect is that you can actually call a varargs function
3362    from ML -- but you can't actually supply any arguments beyond the ones
3363    specified explicitly.  (For example, you can call printf with its format
3364    string, but you cannot pass additional arguments.)
3365    
3366    This behavior is only marginally more useful than the one before, but
3367    it has the advantage that a function or, more importantly, a function
3368    type never gets dropped on the floor, thus avoiding follow-up problems with
3369    other types that refer to the offending one.
3370    
3371    ----------------------------------------------------------------------
3372    Name: Matthias Blume
3373    Date: 2001/07/09 11:25:00 EDT
3374    Tag: blume-20010709-varargs
3375    Description:
3376    
3377    1. ckit-lib.cm now exports structure Error
3378    2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
3379       with a warning accompanied by a source location.  Moreover, it
3380       merely skips the offending function or type and proceeds with the
3381       rest of its work.u  As a result, one can safely feed C code containing
3382       "..." to ml-nlffigen.
3383    3. There are some internal improvements to CM, providing slightly
3384       more general string substitutions in the tools subsystem.
3385    
3386    ----------------------------------------------------------------------
3387    Name: Matthias Blume
3388    Date: 2001/06/27 15:10:00 EDT
3389    Tag: blume-20010627-concur
3390    Description:
3391    
3392    Fixed a small bug in CM's handling of parallel compilation.
3393    (You could observe the bug by Control-C-interrupting an ordinary
3394    CMB.make or CM.stabilize and then attaching some compile servers.
3395    The result was that all of a sudden the previously interrupted
3396    compilation would continue on its own.  This was because of
3397    an over-optimization: CM did not bother to clean out certain queues
3398    when no servers were attached "anyway", resulting in the contents
3399    of these queues to grab control when new servers did get attached.)
3400    
3401    There is also another minor update to the CM manual.
3402    
3403    ----------------------------------------------------------------------
3404    Name: Matthias Blume
3405    Date: 2001/06/26 16:15:00 EDT
3406    Tag: blume-20010626-cmdoc
3407    Description:
3408    
3409    Minor typo fixed in CM manual (syntax diagram for libraries).
3410    
3411    ----------------------------------------------------------------------
3412    Name: Matthias Blume
3413    Date: 2001/06/25 22:55:00 EDT
3414    Tag: blume-20010625-x86pc
3415    Description:
3416    
3417    Fixed a nasty bug in the X86 assembly code that caused signal
3418    handlers to fail (crash) randomly.
3419    
3420    ----------------------------------------------------------------------
3421    Name: Matthias Blume
3422    Date: 2001/06/25 12:05:00 EDT
3423    Tag: blume-20010625-nlffigen
3424    Description:
3425    
3426    This update fixes a number of minor bugs in ml-nlffigen as reported by
3427    Nick Carter <nbc@andrew.cmu.edu>.
3428    
3429      1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
3430      2. Default names for generated files are now derived from the name of
3431         the C file *without its directory*.  In particular, this causes generated
3432         files to be placed locally even if the C file is in some system directory.
3433      3. Default names for generated signatures and structures are also derived
3434         from the C file name without its directory.  This avoids silly things
3435         like "structure GL/GL".
3436         (Other silly names are still possible because ml-nlffigen does not do
3437          a thorough check of whether generated names are legal ML identifiers.
3438          When in doubt, use command line arguments to force particular names.)
3439    
3440    ----------------------------------------------------------------------
3441    Name: Matthias Blume
3442    Date: 2001/06/21 12:25:00 EDT
3443    Tag: blume-20010621-eXene
3444    Description:
3445    
3446    eXene now compiles and (sort of) works again.
3447    
3448    The library name (for version > 110.33) is $/eXene.cm.
3449    
3450    I also added an new example in src/eXene/examples/nbody.  See the
3451    README file there for details.
3452    
3453    ----------------------------------------------------------------------
3454    Name: Matthias Blume
3455    Date: 2001/06/20 16:40:00 EDT
3456    Tag: blume-20010620-cml
3457    Description:
3458    
3459    CML now compiles and works again.
3460    
3461    Libraries (for version > 110.33):
3462    
3463      $cml/cml.cm            Main CML library.
3464      $cml/basis.cm          CML's version of $/basis.cm.
3465      $cml/cml-internal.cm   Internal helper library.
3466      $cml/core-cml.cm       Internal helper library.
3467      $cml-lib/trace-cml.cm  Tracing facility.
3468      $cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
3469    
3470    The installer (config/install.sh) has been taught how to properly
3471    install this stuff.
3472    
3473    ----------------------------------------------------------------------
3474    Name: Matthias Blume
3475    Date: 2001/06/19 17:55:00 EDT
3476    Tag: blume-20010619-instantiate
3477    Description:
3478    
3479    This un-breaks the fix for bug 1432.
3480    (The bug was originally fixed in 110.9 but I broke it again some
3481    time after that.)
3482    
3483    ----------------------------------------------------------------------
3484    Name: Matthias Blume
3485    Date: 2001/06/19 17:25:00 EDT
3486    Tag: blume-20010619-signals
3487    Description:
3488    
3489    This should (hopefully) fix the long-standing signal handling bug.
3490    (The runtime system was constructing a continuation record with an
3491    incorrect descriptor which would cause the GC to drop data on the floor...)
3492    
3493    ----------------------------------------------------------------------
3494    Name: Matthias Blume
3495    Date: 2001/06/15 15:05:00 EDT
3496    Tag: blume-20010615-moresparc
3497    Description:
3498    
3499    Here is a short late-hour update related to Sparc c-calls:
3500    
3501     -- made handling of double-word arguments a bit smarter
3502    
3503     -- instruction selection phase tries to collapse certain clumsily
3504        constructed ML-Trees; typical example:
3505    
3506            ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
3507    
3508        This currently has no further impact on SML/NJ since mlriscGen does
3509        not seem to generate such patterns in the first place, and c-calls
3510        (which did generate them in the beginning) has meanwhile been fixed
3511        so as to avoid them as well.
3512    
3513    ----------------------------------------------------------------------
3514    Name: Matthias Blume
3515    Date: 2001/06/15 15:05:00 EDT
3516    Tag: blume-20010615-sparc
3517    Description:
3518    
3519    The purpose of this update is to provide an implementation of NLFFI
3520    on Sparc machines.
3521    
3522    Here are the changes in detail:
3523    
3524       * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
3525       the Sparc implementation of the c-calls API.
3526       * The Sparc backend of SML/NJ has been modified to uniformely use %fp
3527       for accessing the ML frame.  Thus, we have a real frame pointer and
3528       can freely modify %sp without need for an omit-frame-ptr phase.
3529       The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
3530       this case.
3531       * ml-nlffigen has been taught to produce code for different architectures
3532       and calling conventions.
3533       * In a way similar to what was done in the x86 case, the Sparc
3534       backend uses its own specific extension to mltree.  (For example,
3535       it needs to be able to generate UNIMP instructions which are part
3536       of the calling convention.)
3537       * ml-nlffi-lib was reorganized to make it more modular (in particular,
3538       to make it easier to plug in new machine- and os-dependent parts).
3539    
3540    There are some other fairly unrelated bug fixes and cleanups as well:
3541    
3542       * I further hacked the .cm files for MLRISC tools (like MDLGen) so
3543       that they properly share their libraries with existing SML/NJ libraries.
3544       * I fixed a minor cosmetic bug in CM, supressing certain spurious
3545       follow-up error messages.
3546       * Updates to CM/CMB documentation.
3547    
3548    TODO items:
3549    
3550       * MLRISC should use a different register as its asmTemp on the Sparc.
3551         (The current %o2 is a really bad choice because it is part of the
3552         calling conventions, so things might interfere in unexpected ways.)
3553    
3554    ----------------------------------------------------------------------
3555    Name: Matthias Blume
3556    Date: 2001/06/07
3557    Tag: blume-20010607-calls
3558    Description:
3559    
3560    A number of internal changes related to C calls and calling conventions:
3561    
3562    1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
3563       number of bytes popped implicitly (by the callee).  In most cases
3564       this field is 0 but on x86/win32 it is some non-zero value.  This
3565       is information provided for the benefit of the "omit-frameptr" pass.
3566    2. The CALL instruction on the x86 carries a similar "pops" field.
3567       The instruction selection phase copies its value from the ML-Tree
3568       CALL statement.
3569    3. On all other architectures, the instruction selection phase checks
3570       whether "pops=0" and complains if not.
3571    4. The c-calls implementation for x86 now accepts two calling conventions:
3572       "ccall" and "stdcall".  When "ccall" is selected, the caller cleans
3573       up after the call and pops is set to 0.  For "stdcall", the caller
3574       does nothing, leaving the cleanup to the callee; pops is set to
3575       the number of bytes that were pushed onto the stack.
3576    5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
3577       distinguish between "ccall" and "stdcall".
3578    6. The UNIMP instruction has been added to the supported Sparc instruction
3579       set. (This is needed for implementing the official C calling convention
3580       on this architecture.)
3581    7. I fixed some of the .cm files under src/MLRISC/Tools to make them
3582       work with the latest CM.
3583    
3584    ----------------------------------------------------------------------
3585    Name: Matthias Blume
3586    Date: 2001/06/05 15:10:00 EDT
3587    Tag: blume-20010605-cm-index
3588    Description:
3589    
3590    0. The "lambdasplit" parameter for class "sml" in CM has been documented.
3591    
3592    1. CM can now generate "index files".  These are human-readable files
3593       that list on a per-.cm-file basis each toplevel symbol defined or
3594       imported.  The location of the index file for
3595       <p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
3596       To enable index-file generation, set CM.Control.generate_index to true
3597       or export an environment-symbol: export CM_GENERATE_INDEX=true.
3598    
3599       The CM manual has been updated accordingly.
3600    
3601    2. I made some slight modifications to the c-calls API in MLRISC.
3602    
3603         a) There is now a callback to support saving/restoring of
3604            dedicated but caller-save registers around the actual call
3605            instruction.
3606         b) One can optionally specify a comment-annotation for the
3607            call instruction.
3608    
3609    3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
3610