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 580, Wed Mar 22 06:33:52 2000 UTC revision 1464, Thu Mar 4 17:19:17 2004 UTC
# Line 8  Line 8 
8  The form of an entry should be:  The form of an entry should be:
9    
10  Name:  Name:
11  Date:  Date: yyyy/mm/dd
12  Tag: <post-commit CVS tag>  Tag: <post-commit CVS tag>
13  Description:  Description:
14    
15    ----------------------------------------------------------------------
16    Name: Matthias Blume (blume (at) tti - c (dot) org)
17    Date: 2004/03/04 11:25:00 CST
18    Tag: blume-20040304-symlinks
19    Description:
20    
21    Fixed problem with installer script (unix only) where bin/ml-yacc and
22    friends pointed (via symlinks) to absolute locations instead of just
23    .run-sml.  This was reported by Vesa A Norrman.
24    
25    ----------------------------------------------------------------------
26    Name: Matthias Blume (blume (at) tti - c (dot) org)
27    Date: 2004/02/13 14:50:00 CST
28    Tag: Release_110_45
29    Description:
30    
31    New working version (110.45). New bootfiles.
32    
33    ----------------------------------------------------------------------
34    Name: Matthias Blume (blume (at) tti - c (dot) org)
35    Date: 2004/01/26 15:15:15 CST
36    Tag: blume-20040126-toplevel
37    Description:
38    
39    Improved handling of exceptions at the interactive toplevel.
40    
41    ----------------------------------------------------------------------
42    Name: Matthias Blume (blume (at) tti - c (dot) org)
43    Date: 2004/01/26 11:25:00
44    Tag: blume-20040126-app
45    Description:
46    
47    Type of top-level "app" corrected.
48    Added code for setting vp_limitPtrMask to Win32-specific runtime.
49    
50    ----------------------------------------------------------------------
51    Name: Matthias Blume (blume (at) tti - c (dot) org)
52    Date: 2003/11/18 17:10 CST
53    Tag: blume-20031118-basis-fiddle
54    Description:
55    
56    - changed Timer interface to what might become the spec
57    - POSIX_FLAGS -> BIT_FLAGS according to spec
58    - some other minor discrepancies wrt. spec eliminated
59    
60    ----------------------------------------------------------------------
61    Name: Matthias Blume (blume (at) tti - c (dot) org)
62    Date: 2003/11/06 12:00:00 CST
63    Tag: Release_110_44
64    Description:
65    
66    New working version (110.44). New bootfiles.
67    
68    ----------------------------------------------------------------------
69    Name: Matthias Blume (blume (at) tti - c (dot) org)
70    Date: 2003/11/04 11:50:00 CST
71    Tag: blume-20031104-move-libraries
72    Description:
73    
74    Eliminated the "dont_move_libraries" directive in config/targets.
75    (The mechanism was broken and could not be fixed easily.  Moreover,
76    there does not seem to be any reason not to move all libraries into
77    lib during installation.  I originally implemented this directive as a
78    backward-compatibility feature when I first introduced the new CM.
79    Now that things have been stable for a long time and going back to the
80    old CM is not an option, there is no reason to keep it around.)
81    
82    ----------------------------------------------------------------------
83    Name: Matthias Blume (blume (at) tti - c (dot) org)
84    Date: 2003/11/03 16:00:00 CST
85    Tag: blume-20031103-installdir
86    Description:
87    
88    Made installer honor INSTALLDIR variable again.  (Thanks to Chris
89    Richards for pointing out the problem and providing the solution.)
90    
91    ----------------------------------------------------------------------
92    Name: Matthias Blume (blume (at) tti - c (dot) org)
93    Date: 2003/10/01 17:05:00 CDT
94    Tag: blume-20031001-lal-mlrisc
95    Description:
96    
97    MLRISC bug fix from Lal.
98    
99    ----------------------------------------------------------------------
100    Name: Matthias Blume (blume (at) tti - c (dot) org)
101    Date: 2003/09/30 16:10:00 CDT
102    Tag: blume-20030930-primio-bat
103    Description:
104    
105    1. Added openVector, nullRd, and nullWr to PRIM_IO.
106    2. Improved .bat files (for Win32 port) to make things work under Win95.
107       (thanks to Aaron S. Hawley for this one)
108    
109    ----------------------------------------------------------------------
110    Name: Matthias Blume (blume (at) tti - c (dot) org)
111    Date: 2003/09/26 16:05:00 CDT
112    Tag: blume-20030926-wrappriv
113    Description:
114    
115    Added missing wrapper for privilege "primitive" in $smlnj/viscomp/core.cm.
116    
117    ----------------------------------------------------------------------
118    Name: Matthias Blume (blume (at) tti - c (dot) org)
119    Date: 2003/09/26 15:00:00 CDT
120    Tag: blume-20030926-110_43_3
121    Description:
122    
123    - additional cleanup
124    - version number bump, NEW BOOTFILES
125    
126    ----------------------------------------------------------------------
127    Name: Matthias Blume (blume (at) tti - c (dot) org)
128    Date: 2003/09/26 12:00:00 CDT
129    Tag: blume-20030926-ppautoload
130    Description:
131    
132    I modified the read-eval-print loop so that the autoloader gets
133    invoked whenever the prettyprinter tries to look up a symbol that
134    is not currently defined in the toplevel environment but which
135    appears in CM's autoload registry.  As a result, we see far fewer of
136    those ?.Foo.Bar.xxx names in the prettyprinter's output.
137    
138    In addition to this I tried to clean up some pieces of the Basis
139    implementation (e.g., Socket, Word8Array) in order to prevent other
140    instances of these ?.Foo.Bar.xxx names from being printed.
141    
142    The mechanism that picks names for types still needs some work, though.
143    (Right now it seems that if there is a type A.t which is defined to
144    be B.u, but B is unavailable at toplevel, then A.t gets printed as
145    "?.B.u" although the perhaps more sensible solution would be to use
146    "A.t" in this case.  In other words, the prettyprinter should follow
147    a chain of DEFtycs not farther than there are corresponding toplevel
148    names in the current environment.)
149    
150    ----------------------------------------------------------------------
151    Name: Matthias Blume (blume (at) tti - c (dot) org)
152    Date: 2003/09/24 16:31:00 CDT
153    Tag: blume-20030924-installer
154    Description:
155    
156    Another installer tweak:  All the ML code for the installer is now
157    compiled during CMB.make and put into a little library called
158    $smlnj/installer.cm.  The installation then simply invokes
159    
160       sml -m $smlnj/installer.cm
161    
162    and everything happens automagically.
163    
164    Win32:  ML code senses value of environment variable SMLNJ_HOME.
165    Unix:   ML code senses values of environment variables ROOT, CONFIGDIR,
166            and BINDIR.
167    
168    The new scheme guarantees that the ML code responsible for the installation
169    is in sync with the APIs of the main system.  Also, the installer is
170    somewhat faster because the installer script is precompiled.
171    
172    ----------------------------------------------------------------------
173    Name: Matthias Blume (blume (at) tti - c (dot) org)
174    Date: 2003/09/24 15:35:00 CDT
175    Tag: blume-20030924-synsock
176    Description:
177    
178    Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
179    but excludes all non-blocking operations.  Defined SOCKET (in Basis)
180    and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
181    implementations of non-blocking operations from CML's Socket
182    structure.
183    
184    ----------------------------------------------------------------------
185    Name: Matthias Blume (blume (at) tti - c (dot) org)
186    Date: 2003/09/24 15:10:05 CDT
187    Tag: blume-20030924-sockets
188    Description:
189    
190    1. Fixed SOCKET API and implementation to match Basis spec.
191       This required changing the internal representation of sockets to one
192       that remembers (for each socket file descriptor) whether it is currently
193       blocking or non-blocking.  This state is maintained lazily (i.e., a system
194       call is made only if the state actually needs to change).
195    
196    2. OS-specific details of sockets were moved into separate files, thus
197       making it possible to unify the bulk of the socket implementations
198       between Unix and Win32.
199    
200    3. CML's socket API changed accordingly.
201       (Note that we need to remove non-blocking functions from this API
202       since they are redundant in the case of CML!)
203    
204    4. CML's socket implementation now makes use of non-blocking functions
205       provided by Basis, thus removing all OS-dependent code from this part
206       of CML.
207    
208    5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
209    
210    ----------------------------------------------------------------------
211    Name: Matthias Blume (blume (at) tti - c (dot) org)
212    Date: 2003/09/22 12:10:00 CDT
213    Tag: blume-20030922-110_43_2
214    Description:
215    
216    Made a new interim version and bootfiles for developer's bootstrapping
217    convenience.
218    
219       110.43.2 -- NEW BOOTFILES
220    
221    ----------------------------------------------------------------------
222    Name: Matthias Blume (blume (at) tti - c (dot) org)
223    Date: 2003/09/19 15:55:00 CDT
224    Tag: blume-20030919-cmdir
225    Description:
226    
227    1. new-install.sh -> install.sh
228    2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
229    3. tweaked installer so that another name instead of .cm can be chosen
230       at install time (by setting the CM_DIR_ARC environment variable
231       during installation); once installation is complete, the name is
232       fixed
233    
234    ----------------------------------------------------------------------
235    Name: Matthias Blume (blume (at) tti - c (dot) org)
236    Date: 2003/09/18 16:00:00 CDT
237    Tag: blume-20030918-110_43_1
238    Description:
239    
240    Made a new interim version and bootfiles for developer's bootstrapping
241    convenience.
242    
243       110.43.1 -- NEW BOOTFILES
244    
245    ----------------------------------------------------------------------
246    Name: Matthias Blume (blume (at) tti - c (dot) org)
247    Date: 2003/09/18 15:20:00 CDT
248    Tag: blume-20030918-misc
249    Description:
250    
251    1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
252       this seems to be controversial at the moment) and used those in
253       Posix.ProcEnv.times.
254    
255    2. Added Time.{from,to}Nanoseconds to Time.
256    
257    3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
258       For example, fromLargeInt never needs to look at more than 3 "big
259       digits" to get its 53 bits of precision.
260    
261    ----------------------------------------------------------------------
262    Name: Matthias Blume (blume (at) tti - c (dot) org)
263    Date: 2003/09/17 16:30:00 CDT
264    Tag: blume-20030917-real32-slices
265    Description:
266    
267    Added an entry to the primitive environment
268    (compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
269    added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
270    
271    Removed some of the "magic" constants in real64.sml and replaced them
272    with code that generates these values from their corresponding
273    integer counterparts.
274    
275    Made all(?) the slice-related changes to the Basis and made everything
276    compile again...
277    
278    ----------------------------------------------------------------------
279    Name: Matthias Blume (blume (at) tti - c (dot) org)
280    Date: 2003/09/15 17:45:00 CDT
281    Tag: blume-20030915-rbase
282    Description:
283    
284    Fixed bug in Real.fromLargeInt.
285    
286    ----------------------------------------------------------------------
287    Name: Matthias Blume (blume (at) tti - c (dot) org)
288    Date: 2003/09/13 18:11:00 CDT
289    Tag: blume-20030913-libinstall
290    Description:
291    
292    Minor bugfix in config/libinstall (set anchor with path to
293    standalone tool after installing it, otherwise libraries that
294    need ml-lex or ml-yacc won't compile the first time the installer
295    runs).
296    
297    ----------------------------------------------------------------------
298    Name: Matthias Blume (blume (at) tti - c (dot) org)
299    Date: 2003/09/12 11:45:00 CDT
300    Tag: blume-20030912-various
301    Description:
302    
303     - fixed bug in Real.toLargeInt
304     - fixed bug in Posix.ProcEnv.times
305     - changed inputLine functions to return an option
306     - minor installer improvements / bugfixes
307     - changed default @SMLalloc parameter for x86/celeron to 64k
308    
309    ----------------------------------------------------------------------
310    Name: Matthias Blume (blume (at) tti - c (dot) org)
311    Date: 2003/09/09 22:00:00 CDT
312    Tag: Release_110_43
313    Description:
314    
315        New working release 110.43.  New bootfiles.
316    
317    ----------------------------------------------------------------------
318    Name: Matthias Blume (blume (at) tti - c (dot) org)
319    Date: 2003/09/09 19:20:00 CDT
320    Tag: blume-20030909-installer
321    Description:
322    
323    Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
324    Modified config/install.bat to take advantage of it.  Also modified
325    config/install.sh (and called it config/new-install.sh) to take advantage
326    of it on Unix systems.  (The SML code is (supposed to be) platform-
327    independent.)
328    
329    The installer can now install everything under Win32
330    as well as under *nix as long as it compiles.
331    
332    Other changes:
333    
334      - made CML compile again under Win32
335      - made eXene compile under Win32 (by providing a fake structure UnixSock
336        and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
337      - fixed a bug in nowhere: it assumed that type OS.Process.status is the
338        same as type int;  under Win32 it isn't
339      - fixed some slice-related problems in the win32-specific parts of CML
340      - added a functor argument "sameVol" to os-path-fn.sml in the Basis
341        (under Win32, the volume name is case-insensitive, and the
342        OS.Path code compares volume names for equality)
343    
344    ----------------------------------------------------------------------
345    Name: Matthias Blume (blume (at) tti - c (dot) org)
346    Date: 2003/09/08 11:55:00 CDT
347    Tag: blume-20030908-fullpath
348    Description:
349    
350    Made Win32 version of OS.FileSys.fullPath return current directory
351    when given an empty string.  This is what the spec says, and incidentally,
352    CM depends on it.  (CM otherwise goes into an infinite loop in certain
353    cases when presented with the name of a non-existing .cm file.)
354    
355    ----------------------------------------------------------------------
356    Name: Matthias Blume (blume (at) tti - c (dot) org)
357    Date: 2003/09/04 16:30:00 CDT
358    Tag: blume-20030905-slices-etc
359    Description:
360    
361      1. Changed interface to vectors and arrays in Basis to match
362         (draft) Basis spec.
363      2. Added signatures and implementations of slices according to
364         Basis spec.
365      3. Edited source code throughout the system to make it compile again
366         under 1. and 2.  (In some cases code had to be added to have it
367         match the new signatures.)
368      4. MLRISC should be backward-compatible: the copies of the originals
369         of files that needed to change under 3. were retained, the .cm files
370         check the compiler version number and use old versions when
371         appropriate.
372      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
373         dirstream -> string option (in accordance with Basis spec).
374      6. When generating code that counts lines, ml-lex used function
375         CharVector.foldli, taking advantage of its old interface.
376         This has been replaced with the corresponding code from
377         CharVectorSlice.  (html-lex must be re-lexed!)
378      7. BitArray in smlnj-lib/Util has been extended/modified to match the
379         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
380      8. Removed temporary additions (fromInternal, toInternal) from the
381         (now obsolete) IntInf in smlnj-lib/Util.
382      9. Cleaned up structure Byte.
383     10. Added localOffset, scan, and fromString to Date (according to spec).
384         Cleaned/corrected implementation of Date.
385         (Still need to check for correctness; implement better canonicalizeDate.)
386     11. Added "scan" to signature IEEE_REAL.
387     12. Some improvements to IntInf [in particular: efficiency-hack for
388         mod and rem when second operand is 2 (for parity checks).]
389     13. Changed representation of type Time.time, using a single IntInf.int
390         value counting microseconds.  This considerably simplified the
391         implementation of structure Time.  We now support negative time
392         values; scan and fromString handle signs.
393     14. Functor PrimIO now takes two additional arguments (VectorSlice and
394         ArraySlice).
395    
396    ----------------------------------------------------------------------
397    Name: Matthias Blume (blume (at) tti - c (dot) org)
398    Date: 2003/08/28 17:00:00 CDT
399    Tag: blume-20030828-intinf
400    Description:
401    
402    This is a major update which comes with a version number bump
403    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
404    and an implementation of IntInf in the Basis.
405    
406    There are a fairly large number of related changes and updates throughout
407    the system:
408    
409    Basis:
410            - Implemented IntInf.
411            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
412            - Added some missing Real64 operations, most notably Real.toLargeInt.
413            - Added FixedInt as a synonym for Int32.
414    
415    compiler:
416            * Added support for a built-in intinf type.
417              - literals
418              - pattern matching
419              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
420              - overloading on literals and operations
421    
422            This required adding a primitive type intinf, some additional
423            primops, and implementations for several non-trivial intinf
424            operations in Core.  (The intinf type is completely abstract
425            to the compiler; all operations get delegated back to the Core.)
426    
427            * Intinf equality is handled by polyequal.  However, the compiler
428            does not print its usual warning in this case (since polyequal
429            is the right thing to do there).
430    
431            * Improved the organization of structure InlineT.
432    
433            * A word about conversion primops:
434            If conversions involving intinf do not cancel out during
435            CPS contract, then the compiler must insert calls to Core functions.
436            Since all core access must be resolved already during the FLINT
437            translate phase, it would be too late a the time of CPS contract
438            to add new Core calls.  For this reason, conversion primops
439            for intinf carry two arguments:  1. the numeric argument that
440            they are supposed to convert, and 2. the Core function that
441            can help with this conversion if necessary.  If CPS contract
442            eliminates a primop, then the associated Core function becomes
443            dead and goes away.  Intinf conversion primops that do not get
444            eliminated by CPS contract get rewritten into calls of their
445            core functions by a separate, new phase.
446    
447    interactive system:
448            - Control.Print.intinfDepth controls max length of intinf constants
449            being printed. (Analogous to Control.Print.stringDepth.)
450            - Cleanup in printutil and pputil: got rid of unused stuff and
451            duplicates; replaced some of the code with code that makes better
452            use of library functionality.
453    
454    CM:
455            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
456            are no longer silent.
457    
458    CKIT:
459            Fixed mismatched uses of Int32 and LargeInt.  I always decided
460            in favor of LargeInt -- which is now the same as IntInf.
461            CKIT-knowledgable people should check whether this is what's
462            intended and otherwise change things back to using Int32 or
463            FixedInt.
464    
465    Throughout the code:
466            Started using IntInf.int literals and built-in operations
467            (e.g., comparison with 0) where this seems appropriate.
468    
469    
470    ----------------------------------------------------------------------
471    Name: Dave MacQueen (dbm@cs.uchicago.edu)
472    Date: 2003/08/13 11:36:00 CDT
473    Tag: dbm-20030813-mcz-merge1
474    Description:
475    
476    Merging changes from the mcz-branch development branch into trunk.
477    These changes involve replacement of the emulated old prettyprinter
478    interface with direct use of the SML/NJ Lib PP library, and fixing
479    of a couple of bugs (895, 1186) relating to error messages.  A new
480    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
481    has been added.
482    
483    ----------------------------------------------------------------------
484    Name: Matthias Blume (blume (at) tti - c (dot) org)
485    Date: 2003/08/11 15:45:00 CDT
486    Tag: blume-20030811-windows
487    Description:
488    
489    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
490    
491        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
492    
493    This patch restores SML/NJ's ability to run under win32.  There are a
494    number of changes, including fixes for several bugs that had gone
495    unnoticed until now:
496    
497      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
498        trying to run the win32 version from within cygwin.)
499    
500      - There are a number of simple .bat scripts that substitute for their
501        corresponding Unix shell-scripts.  (See below.)
502    
503      - The internals of ml-build have been modified slightly.  The main
504        difference is that instead of calling ".link-sml" (or link-sml.bat)
505        using OS.Process.system, the ML process delegates this task back
506        to the script.  Otherwise problems arise in mixed environments such
507        as Cygwin where scripts look and work like Unix scripts, but
508        where OS.Process.system cannot run them.
509    
510      - In CM, the srcpath pickler used native pathname syntax -- which
511        is incorrect in the case of cross-compilation.  The new pickle format
512        is independent of platform-specific naming conventions.
513    
514      - Path configuration files (such as lib/pathconfig) can now choose
515        between native and standard syntax.  Placing a line of the form
516    
517            standard!
518    
519        into the file causes all subsequent paths to be interpreted using
520        CM standard pathname syntax (= Unix conventions); a line
521    
522            native!
523    
524        switches back to native style.  This was needed so that
525        path config files can be written portably, see src/system/pathconfig.
526    
527      - Runtime system:
528    
529          - win32-filesys.c:  get_file_time and set_file_time now
530            access modification time, not creation time.
531    
532          - I/O code made aware of new array representation.
533    
534          - Bug fixes in X86.prim.masm.
535    
536       - src/system/makeml made aware of win32. (For use under cygwin
537            and other Unix-environments for windows.)
538    
539       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
540         which caused BinIO.inputAll to fail consistently.
541    
542    .bat scripts:
543    
544       Windows .bat scripts assume that SMLNJ_HOME is defined.
545    
546       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
547         applications (sml, ml-yacc, ml-lex).
548       - ml-build.bat: analogous to ml-build.
549       - config\install.bat:  Analogous to config/install.sh.  This requires
550         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
551         (nmake etc. must be on the path, and vcvars32 must have been run.)
552         Moreover, sources for ml-lex and ml-yacc need to exist under src,
553         and the bootfile hierarchy must have been unpacked under
554         sml.boot.x86-win32.
555         The script is very primitive and does a poor job at error checking.
556         It only installs the base system, ml-lex, and ml-yacc.  No other
557         libraries are being installed (i.e., you get only those that
558         are part of the compiler.)
559       - link-sml.bat:  analogous to .link-sml, but not currently used
560    
561    Unrelated bug fixes:
562    
563      - ml-nlffigen now exports structures ST_* corresponding to incomplete
564        types.
565      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
566        otherwise.)
567    
568    ----------------------------------------------------------------------
569    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
570    Date: 2003/06/17
571    Tag: macqueen-20030617-bug895
572    Description:
573    
574    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
575    Tag will be used for new development branch (mcz-branch) for
576    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
577    
578    ----------------------------------------------------------------------
579    Name: Matthias Blume (blume (at) tti - c (dot) org)
580    Date: 2003/05/27 16:55:00 CDT
581    Tag: blume-20030527-polyeq
582    Description:
583    
584    Tried to eliminated most cases of polymorphic equality.
585    
586    ----------------------------------------------------------------------
587    Name: Matthias Blume (blume (at) tti - c (dot) org)
588    Date: 2003/05/21 17:45:00 CDT
589    Tag: blume-20030517-complete
590    Description:
591    
592    Two changes:
593    
594      1. Added a flag for controlling whether non-exhaustive bindings will
595         be treated as errors (default is false).
596      2. Cleaned up the *entire* source tree so that CMB.make goes through
597         without a single non-exhaustive match- or bind warning.
598    
599    ----------------------------------------------------------------------
600    Name: Matthias Blume (blume (at) tti - c (dot) org)
601    Date: 2003/05/17 10:20:00 CDT
602    Tag: blume-20030517-absyn
603    Description:
604    
605    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
606    
607       This mainly affects the quality of error messages.  However, some
608       of the code is now more straightforward than before. (Treatment of
609       the above four constructs in translate.sml is much simpler than
610       the "macro-expansion" that was going on before.  Plus, the mach-
611       compiler no longer gets invoked just to be able to compile an
612       if-expression.)
613    
614    2. The ErrorMsg.Error exception is now caught and absorbed by the
615       interactive loop.
616    
617    ----------------------------------------------------------------------
618    Name: Allen Leung
619    Date: 2003/05/16 13:05:00 CDT
620    Tag: leunga-20030516-cygwin-runtime
621    Description:
622    
623        Ported the runtime system to cygwin, which uses the unix
624    x86-unix bin files.  Missing/buggy features:
625    
626       o getnetbyname, getnetbyaddr: these functions seem to be missing in
627         the Cygwin library.
628       o Ctrl-C handling may be flaky.
629       o Windows system calls and Windows I/O are not supported.
630    
631       A new set of binfiles is located at:
632    
633         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
634    
635       This is only needed for bootstrapping the cygwin version of smlnj.
636       Other x86 versions can use the existing binfiles.
637    
638    ----------------------------------------------------------------------
639    Name: Matthias Blume
640    Date: 2003/04/08 15:42:00 CDT
641    Tag: blume-20030408-listpair
642    Description:
643    
644    1. Added a target 'mlrisc' to installer.
645    
646    2. Added missing elements to structure ListPair.
647    
648    ----------------------------------------------------------------------
649    Name: Allen Leung
650    Date: 2003/01/07 10:40:00 EST
651    Tag: leunga-20030107-int-rem
652    Description:
653    
654    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
655    The arguments to the SUBL instruction were swapped.
656    
657    ----------------------------------------------------------------------
658    Name: Matthias Blume
659    Date: 2002/12/12 16:25:00 EST
660    Tag: blume-20021212-risc-ra
661    Description:
662    
663    Fixed a serious bug in the rewrite code for FP spilling/reloading that
664    sent the RA into an infinite loop when floating point registers get
665    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
666    110.37 and 110.38.)
667    There was another set of potential problems related to the handling of
668    MLRISC annotations (but those did not yet cause real problems, apparently).
669    
670    ----------------------------------------------------------------------
671    Name: Matthias Blume
672    Date: 2002/12/06 22:40:00 EST
673    Tag: blume-20021206-cm-fileid
674    Description:
675    
676    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
677    This fixes the problem of CM getting confused by files that suddenly
678    change their identity (e.g., by getting unlinked and recreated by some
679    text editor such as vi).  There might be a better/cheaper/cleaner way
680    of doing this, but for now this will have to do.
681    
682    ----------------------------------------------------------------------
683    Name: Matthias Blume
684    Date: 2002/10/28 09:50:00 EST
685    Tag: blume-20021028-typecheck
686    Description:
687    
688    Exported structure Typecheck from $smlnj/viscomp/core.cm.
689    
690    ----------------------------------------------------------------------
691    Name: Matthias Blume
692    Date: 2002/10/17 09:10:00 EDT
693    Tag: Release_110_42
694    Description:
695    
696    In good old tradition, there has been a slight hiccup so that we have
697    to patch 110.42 after the fact.   The old release tag has been replaced
698    (see below).
699    
700    The change solves a problem with two competing approaches the
701    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
702    each other's way.
703    
704    This change only affects the runtime system code and the installer script.
705    (No new bootfiles.)
706    
707    ----------------------------------------------------------------------
708    Name: Matthias Blume
709    Date: 2002/10/16 12:00:00 EDT
710    Tag: Release_110_42_removed
711    Description:
712    
713    New working release.  New bootfiles.
714    
715    ----------------------------------------------------------------------
716    Name: Matthias Blume
717    Date: 2002/10/10 13:10:00 EDT
718    Tag: blume-20021010-ppc-divs
719    Description:
720    
721    The mltree operator DIVS must be implemented with an overflow check on
722    the PPC because the hardware indicates divide-by-zero using "overflow" as
723    well.
724    
725    ----------------------------------------------------------------------
726    Name: Matthias Blume
727    Date: 2002/07/23 11:45:00
728    Tag: blume-20020723-smlnj-home
729    Description:
730    
731    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
732    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
733    CM_PATHCONFIG is also set), the path configuration file is assumed
734    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
735    the entire tree to some other place and everything will "just work".
736    
737    (Companion commands such as ml-build and ml-makedepend also sense this
738    variable.)
739    
740    ----------------------------------------------------------------------
741    Name: Matthias Blume
742    Date: 2002/07/12 21:19:00 EDT
743    Tag: blume-20020712-liveness
744    Description:
745    
746    Exported two useful "step" functions from liveness module (MLRISC).
747    
748    ----------------------------------------------------------------------
749    Name: Matthias Blume
750    Date: 2002/07/05 16:00 EDT
751    Tag: Release_110_41
752    Description:
753    
754    New working release.  New bootfiles.
755    
756    ----------------------------------------------------------------------
757    Name: Matthias Blume
758    Date: 2002/07/05 10:25:00 EDT
759    Tag: blume-20020705-btimp
760    Description:
761    
762    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
763    other clients can set up backtracing support.
764    
765    ----------------------------------------------------------------------
766    Name: Matthias Blume
767    Date: 2002/06/25 17:23:00 EDT
768    Tag: blume-20020625-fpmax
769    Description:
770    
771    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
772    case.  (The sense of the isNaN test was reversed -- which made min and
773    max always return their first argument.)
774    
775    ----------------------------------------------------------------------
776    Name: Matthias Blume
777    Date: 2002/06/11
778    Tag: blume-20020611-unixpath
779    Description:
780    
781    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
782    
783    ----------------------------------------------------------------------
784    Name: Matthias Blume
785    Date: 2002/06/10 16:35:00 EDT
786    Tag: blume-20020610-ieeereal
787    Description:
788    
789    I back-ported my implementation of IEEEReal.fromString from the
790    idlbasis-devel branch so that we can test it.
791    
792    Another small change is that ppDec tries to give more information
793    than just "<sig>" in the case of functors.  However, this code is
794    broken in some mysterious way if the functor's body's signature
795    has not been declared by ascription but gets inferred from the
796    implementation.  This needs fixing...
797    
798    ----------------------------------------------------------------------
799    Name: Matthias Blume
800    Date: 2002/05/31
801    Tag: blume-20020531-btrace-mode
802    Description:
803    
804    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
805    the wayside when I switched over to using Controls everywhere.)
806    
807    ----------------------------------------------------------------------
808    Name: Lal George
809    Date: 2002/05/23 12:21:40 EDT
810    Tag: george-20020523-visual-labels
811    Description:
812    
813    Labels are now displayed in the graphical output to make
814    the fall-through and target blocks obvious.
815    
816    ----------------------------------------------------------------------
817    Name: Matthias Blume
818    Date: 2002/05/22 11:03:00 EDT
819    Tag: blume-20020522-shrink
820    Description:
821    
822    John tweaked yesterday's fix for 1131 to handle an out-of-memory
823    situation that comes up when allocating huge arrays.
824    
825    ----------------------------------------------------------------------
826    Name: Matthias Blume
827    Date: 2002/05/21 16:00:00 EDT
828    Tag: Release_110_40
829    Description:
830    
831    New working release (110.40).  New bootfiles.
832    
833    [Also: John Reppy fixed GC bug 1131.]
834    
835    ----------------------------------------------------------------------
836    Name: Matthias Blume
837    Date: 2002/05/21 12:35:00 EDT
838    Tag: blume-20020521-cmdoc
839    Description:
840    
841    CM documentation update.
842    
843    ----------------------------------------------------------------------
844    Name: Matthias Blume
845    Date: 2002/05/21 10:55:00 EDT
846    Tag: blume-20020521-misc
847    Description:
848    
849    - John tweaked runtime to be silent on heap export (except when
850      GC messages are on).
851    - I added a few more things (cross-compiling versions of CMB) to
852      config/preloads (as suggestions).
853    
854    ----------------------------------------------------------------------
855    Name: Matthias Blume
856    Date: 2002/05/20 22:25:00 EDT
857    Tag: blume-20020520-controls
858    Description:
859    
860    - Added ControlUtil structure to control-lib.cm.
861    - Use it throughout.
862    - Used Controls facility to define MLRISC controls (as opposed to
863      registering MLRISC control ref cells with Controls after the fact)
864    - Fixed messed-up controls priorities.
865    
866    * Removed again all the stuff from config/preloads that one wouldn't
867      be able to preload at the time the initial heap image is built.
868      (Many libraries, e.g., CML, do not exist yet at this time.  The
869       only libraries that can be preloaded via config/preloads are those
870       that come bundled with the bootfiles.)
871    
872    ----------------------------------------------------------------------
873    Name: Matthias Blume
874    Date: 2002/05/20 10:59:00 EDT
875    Tag: blume-20020520-preloads
876    Description:
877    
878    Added a lot of commented-out suggestions for things to be included
879    in config/preloads.
880    
881    ----------------------------------------------------------------------
882    Name: Allen Leung
883    Date: 2002/05/18 14:20:00 EDT
884    Tag: leunga-20020518-mdl
885    Description:
886    
887    o Made the mdl tool stuff compile and run again.
888    o I've disabled all the stuff that depends on RTL specifications; they
889      are all badly broken anyway.
890    
891    ----------------------------------------------------------------------
892    Name: Matthias Blume
893    Date: 2002/05/17 16:49:00 EDT
894    Tag: blume-20020517-controls
895    Description:
896    
897    0. John Reppy made several modifications to the SML/NJ library.
898       In particular, there is a shiny new controls-lib.cm.
899    
900    1. Pushed new controls interface through compiler so that everything
901       compiles again.
902    
903    2. Added FormatComb and FORMAT_COMB to the CML version of the
904       SML/NJ library (so that CML compiles again).
905    
906    3. Modified init scripts because XXX_DEFAULT environment variables
907       are no longer with us.  (Boot-time initialization is now done
908       using the same environment variables that are also used for
909       startup-time initialization of controls.)
910    
911    ----------------------------------------------------------------------
912    Name: Lal George
913    Date: 2002/05/15 09:20:10 EDT
914    Tag: george-20020515-pseudo-op-decls
915    Description:
916    
917            All pseudo-ops emitted before the first segment declaration
918    such as TEXT, DATA, and BSS directives are assumed to be global
919    declarations and are emitted first in the assembly file. This is
920    useful in a number of situations where one has pseudo-ops that are not
921    specific to any segment, and also works around the constraint that one
922    cannot have client pseudo-ops in the TEXT segment.
923    
924            Because no segment is associated with these declarations it is
925    an error to allocate any space or objects before the first segment
926    directive and an exception will be raised. However, we cannot make
927    this check for client pseudo-ops.
928    
929            These top level declarations are a field in the CFG graph_info.
930    In theory you can continue to add to this field after the CFG has been
931    built -- provided you know what you are doing;-)
932    
933    ----------------------------------------------------------------------
934    Name: Matthias Blume
935    Date: 2002/05/13 16:40:00 EDT
936    Tag: blume-20020513-pp-etc
937    Description:
938    
939    A few minor bugfixes:
940    
941      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
942        (Bogus printouts for redefined bindings still occur.  Compiler
943        bug should no longer occur now.  We need to redo the prettyprinter
944        from scratch.)
945    
946      - CM pathname printer now also adds escape sequences for ( and )
947    
948      - commend and docu fixes for ml-nlffi
949    
950    ----------------------------------------------------------------------
951    Name: Matthias Blume
952    Date: 2002/05/10 16:40:00 EDT
953    Tag: blume-20020510-erg-textio
954    Description:
955    
956    Applied the following bugfix provided by Emden Gansner:
957    
958        Output is corrupted when outputSubstr is used rather than output.
959    
960        The problem occurs when a substring
961    
962            ss = (s, dataStart, dataLen)
963    
964        where dataStart > 0, fills a stream buffer with avail bytes left.
965        avail bytes of s, starting at index dataStart, are copied into the
966        buffer, the buffer is flushed, and then the remaining dataLen-avail
967        bytes of ss are copied into the beginning of the buffer. Instead of
968        starting this copy at index dataStart+avail in s, the current code
969        starts the copy at index avail.
970    
971        Fix:
972        In text-io-fn.sml, change line 695 from
973             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
974        to
975             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
976    
977    ----------------------------------------------------------------------
978    Name: Matthias Blume
979    Date: 2002/04/12 13:55:00 EDT
980    Tag: blume-20020412-assyntax
981    Description:
982    
983    1. Grabbed newer assyntax.h from the XFree86 project.
984    2. Fiddled with how to compile X86.prim.asm without warnings.
985    3. (Very) Minor cleanup in CM.
986    
987    ----------------------------------------------------------------------
988    Name: Matthias Blume
989    Date: 2002/04/01 (no joke!) 17:07:00 EST
990    Tag: blume-20020401-x86div
991    Description:
992    
993    Added full support for div/mod/rem/quot on the x86, using the machine
994    instruction's two results (without clumsily recomputing the remainder)
995    directly where appropriate.
996    
997    Some more extensive power-of-two support was added to the x86 instruction
998    selector (avoiding expensive divs, mods, and muls where they can be
999    replaced with cheaper shifts and masks).  However, this sort of thing
1000    ought to be done earlier, e.g., within the CPS optimizer so that
1001    all architectures benefit from it.
1002    
1003    The compiler compiles to a fixed point, but changes might be somewhat
1004    fragile nevertheless.  Please, report any strange things that you might
1005    see wrt. div/mod/quot/rem...
1006    
1007    ----------------------------------------------------------------------
1008    Name: Matthias Blume
1009    Date: 2002/03/29 17:22:00
1010    Tag: blume-20020329-div
1011    Description:
1012    
1013    Fixed my broken div/mod logic.  Unfortunately, this means that the
1014    inline code for div/mod now has one more comparison than before.
1015    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
1016    The problem was with quotient = 0, because that alone does not tell
1017    us which way the rounding went.  One then has to look at whether
1018    remainder and divisor have the same sign...  :(
1019    
1020    Anyway, I replaced the bootfiles with fresh ones...
1021    
1022    ----------------------------------------------------------------------
1023    Name: Matthias Blume
1024    Date: 2002/03/29 14:10:00 EST
1025    Tag: blume-20020329-inlprims
1026    Description:
1027    
1028    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
1029    
1030    Primops have changed. This means that the bin/boot-file formats have
1031    changed as well.
1032    
1033    To make sure that there is no confusion, I made a new version.
1034    
1035    
1036    CHANGES:
1037    
1038    * removed REMT from mltree (remainder should never overflow).
1039    
1040    * added primops to deal with divisions of all flavors to the frontend
1041    
1042    * handled these primops all the way through so they map to their respective
1043      MLRISC support
1044    
1045    * used these primops in the implementation of Int, Int32, Word, Word32
1046    
1047    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
1048    
1049    * parameterized INLMIN, INLMAX, and INLABS by a numkind
1050    
1051    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
1052      floating point
1053    
1054    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
1055      and Real (but Real.abs maps to a separate floating-point-only primop)
1056    
1057    
1058    TODO items:
1059    
1060    * Hacked Alpha32 instruction selection, disabling the selection of REMx
1061      instructions because the machine instruction encoder cannot handle
1062      them.  (Hppa, PPC, and Sparc instruction selection did not handle
1063      REM in the first place, and REM is supported by the x86 machine coder.)
1064    
1065    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
1066      selection phase.  (The two can be streamlined because the hardware
1067      delivers both quotient and remainder at the same time anyway.)
1068    
1069    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
1070      (Currently the behavior is inconsistent both across architectures and
1071      wrt. the draft Basis spec.)
1072    
1073    * Word8 should eventually be handled natively, too.
1074    
1075    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
1076      as if there currently is no execution path that could trigger it in
1077      SML/NJ.  (The assumptions underlying functions arith and promotable do not
1078      hold for things like multiplication and division.)
1079    
1080    ----------------------------------------------------------------------
1081    Name: Matthias Blume
1082    Date: 2002/03/27 16:27:00 EST
1083    Tag: blume-20020327-mlrisc-divisions
1084    Description:
1085    
1086    Added support for all four division operations (ML's div, mod, quot,
1087    and rem) to MLRISC.  In the course of doing so, I also rationalized
1088    the naming (no more annoying switch-around of DIV and QUOT), by
1089    parameterizing the operation by div_rounding_mode (which can be either
1090    DIV_TO_ZERO or DIV_TO_NEGINF).
1091    
1092    The generic MLTreeGen functor takes care of compiling all four
1093    operations down to only round-to-zero div.
1094    
1095    Missing pieces:
1096    
1097      * Doing something smarter than relying on MLTreeGen on architectures
1098        like, e.g., the x86 where hardware division delivers both quotient and
1099        remainder at the same time.  With this, the implementation of the
1100        round-to-neginf operations could be further streamlined.
1101    
1102      * Remove inlining support for div/mod/rem from the frontend and replace it
1103        with primops that get carried through to the backend.  Do this for all
1104        int and word types.
1105    
1106    ----------------------------------------------------------------------
1107    Name: Matthias Blume
1108    Date: 2002/03/25 17:25:00 EST
1109    Tag: blume-20020325-divmod
1110    Description:
1111    
1112    I improved (hopefully without breaking them) the implementation of Int.div,
1113    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
1114    advantage of the following observations:
1115    
1116      Let  q = x quot y      r = x rem y
1117           d = x div  y      m = x mod y
1118    
1119    where "quot" is the round-to-zero version of integer division that
1120    hardware usually provides.  Then we have:
1121    
1122         r = x - q * y        where neither the * nor the - will overflow
1123         d = if q >= 0 orelse x = q * y then q else q - 1
1124                              where neither the * nor the - will overflow
1125         m = if q >= 0 orelse r = 0 then r else r + y
1126                              where the + will not overflow
1127    
1128    This results in substantial simplification of the generated code.
1129    The following table shows the number of CFG nodes and edges generated
1130    for
1131            fun f (x, y) = x OPER y
1132            (* with OPER \in div, mod, quot, rem *)
1133    
1134    
1135        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
1136        --------------------------------------------------------
1137         div |         24 |         39 |         12 |         16
1138         mod |         41 |         71 |         12 |         16
1139        quot |          8 |         10 |          8 |         10
1140         rem |         10 |         14 |          8 |         10
1141    
1142    
1143    ----------------------------------------------------------------------
1144    Name: Matthias Blume
1145    Date: 2002/03/25 22:06:00 EST
1146    Tag: blume-20020325-cprotobug
1147    Description:
1148    
1149    Fixed a bug in cproto (c prototype decoder).
1150    
1151    ----------------------------------------------------------------------
1152    Name: Matthias Blume
1153    Date: 2002/03/25 16:00:00 EST
1154    Tag: blume-20020325-raw-primops
1155    Description:
1156    
1157    I did some cleanup to Allen's new primop code and
1158    replaced yesterday's bootfiles with new ones.
1159    (But they are stored in the same place.)
1160    
1161    ----------------------------------------------------------------------
1162    Name: Matthias Blume
1163    Date: 2002/03/24 22:40:00 EST
1164    Tag: blume-20020324-bootfiles
1165    Description:
1166    
1167    Made the bootfiles that Allen asked for.
1168    
1169    ----------------------------------------------------------------------
1170    Name: Allen Leung
1171    Date: 2002/03/23 15:50:00 EST
1172    Tag: leunga-20020323-flint-cps-rcc-primops
1173    Description:
1174    
1175      1. Changes to FLINT primops:
1176    
1177        (* make a call to a C-function;
1178         * The primop carries C function prototype information and specifies
1179         * which of its (ML-) arguments are floating point. C prototype
1180         * information is for use by the backend, ML information is for
1181         * use by the CPS converter. *)
1182      | RAW_CCALL of { c_proto: CTypes.c_proto,
1183                       ml_args: ccall_type list,
1184                       ml_res_opt: ccall_type option,
1185                       reentrant : bool
1186                     } option
1187       (* Allocate uninitialized storage on the heap.
1188        * The record is meant to hold short-lived C objects, i.e., they
1189        * are not ML pointers.  With the tag, the representation is
1190        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
1191        *)
1192      | RAW_RECORD of {tag:bool,sz:int}
1193      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
1194    
1195      2.  These CPS primops are now overloaded:
1196    
1197           rawload of {kind:numkind}
1198           rawstore of {kind:numkind}
1199    
1200          The one argument form is:
1201    
1202             rawload {kind} address
1203    
1204          The two argument form is:
1205    
1206             rawload {kind} [ml object, byte-offset]
1207    
1208      3. RAW_CCALL/RCC now takes two extra arguments:
1209    
1210         a. The first is whether the C call is reentrant, i.e., whether
1211            ML state should be saved and restored.
1212         b. The second argument is a string argument specifying the name of
1213            library and the C function.
1214    
1215         These things are currently not handled in the code generator, yet.
1216    
1217      4. In CProto,
1218    
1219         An encoding type of "bool" means "ml object" and is mapped into
1220         C prototype of PTR.  Note that "bool" is different than "string",
1221         even though "string" is also mapped into PTR, because "bool"
1222         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
1223    
1224      5. Pickler/unpicker
1225    
1226         Changed to handle RAW_RECORD and newest RAW_CCALL
1227    
1228      6. MLRiscGen,
1229    
1230         1. Changed to handle the new rawload/rawstore/rawrecord operators.
1231         2. Code for handling C Calls has been moved to a new module CPSCCalls,
1232            in the file CodeGen/cpscompile/cps-c-calls.sml
1233    
1234      7. Added the conditional move operator
1235    
1236             condmove of branch
1237    
1238         to cps.  Generation of this is still buggy so it is currently
1239         disabled.
1240    
1241    ----------------------------------------------------------------------
1242    Name: Lal George
1243    Date: 2002/03/22 14:18:25 EST
1244    Tag: george-20020322-cps-branch-prob
1245    Description:
1246    
1247    Implemented the Ball-Larus branch prediction-heuristics, and
1248    incorporated graphical viewers for control flow graphs.
1249    
1250    Ball-Larus Heuristics:
1251    ---------------------
1252    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1253    
1254    By design it uses the Dempster-Shafer theory for combining
1255    probabilities.  For example, in the function:
1256    
1257        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1258    
1259    the ball-larus heuristics predicts that the n=0 is unlikely
1260    (OH-heuristic), and the 'then' branch is unlikely because of the
1261    RH-heuristic -- giving the 'then' branch an even lower combined
1262    probability using the Dempster-Shafer theory.
1263    
1264    Finally, John Reppy's loop analysis in MLRISC, further lowers the
1265    probability of the 'then' branch because of the loop in the else
1266    branch.
1267    
1268    
1269    Graphical Viewing:
1270    ------------------
1271    I merely plugged in Allen's graphical viewers into the compiler. The
1272    additional code is not much. At the top level, saying:
1273    
1274            Control.MLRISC.getFlag "cfg-graphical-view" := true;
1275    
1276    will display the graphical view of the control flow graph just before
1277    back-patching.  daVinci must be in your path for this to work. If
1278    daVinci is not available, then the default viewer can be changed
1279    using:
1280    
1281            Control.MLRISC.getString "viewer"
1282    
1283    which can be set to "dot" or "vcg" for the corresponding viewers. Of
1284    course, these viewers must be in your path.
1285    
1286    The above will display the compilation unit at the level of clusters,
1287    many of which are small, boring, and un-interesting. Also setting:
1288    
1289            Control.MLRISC.getInt "cfg-graphical-view_size"
1290    
1291    will display clusters that are larger than the value set by the above.
1292    
1293    
1294    ----------------------------------------------------------------------
1295    Name: Matthias Blume
1296    Date: 2002/03/21 22:20:00 EST
1297    Tag: blume-20020321-kmp-bugfix
1298    Description:
1299    
1300    Changed the interface to the KMP routine in PreString and fixed
1301    a minor bug in one place where it was used.
1302    
1303    ----------------------------------------------------------------------
1304    Name: Allen Leung
1305    Date: 2002/03/21 20:30:00 EST
1306    Tag: leunga-20020321-cfg
1307    Description:
1308    
1309      Fixed a potential problem in cfg edge splitting.
1310    
1311    ----------------------------------------------------------------------
1312    Name: Allen Leung
1313    Date: 2002/03/21 17:15:00 EST
1314    Tag: leunga-20020321-x86-fp-cfg
1315    Description:
1316    
1317      1. Recoded the buggy parts of x86-fp.
1318    
1319         a. All the block reordering code has been removed.
1320            We now depend on the block placement phases to do this work.
1321    
1322         b. Critical edge splitting code has been simplified and moved into the
1323            CFG modules, as where they belong.
1324    
1325         Both of these were quite buggy and complex.  The code is now much, much
1326         simpler.
1327    
1328      2. X86 backend.
1329    
1330         a. Added instructions for 64-bit support.  Instruction selection for
1331            64-bit has not been committed, however, since that
1332            requires changes to MLTREE which haven't been approved by
1333            Lal and John.
1334    
1335         b. Added support for FUCOMI and FUCOMIP when generating code for
1336            PentiumPro and above.  We only generate these instructions in
1337            the fast-fp mode.
1338    
1339         c. Added cases for JP and JNP in X86FreqProps.
1340    
1341      3. CFG
1342    
1343         CFG now has a bunch of methods for edge splitting and merging.
1344    
1345      4. Machine description.
1346    
1347         John's simplification of MLTREE_BASIS.fcond broke a few machine
1348         description things:
1349    
1350         rtl-build.{sig,sml} and hppa.mdl fixed.
1351    
1352         NOTE: the machine description stuff in the repository is still broken.
1353               Again, I can't put my fixes in because that involves
1354               changes to MLTREE.
1355    
1356    ----------------------------------------------------------------------
1357    Name: Matthias Blume
1358    Date: 2002/03/20 15:55:00 EST
1359    Tag: blume-20020320-kmp
1360    Description:
1361    
1362    Implemented Knuth-Morris-Pratt string matching in PreString and used
1363    it for String.isSubstring, Substring.isSubstring, and
1364    Substring.position.
1365    
1366    (Might need some stress-testing.  Simple examples worked fine.)
1367    
1368    ----------------------------------------------------------------------
1369    Name: Matthias Blume
1370    Date: 2002/03/19 16:37:00 EST
1371    Tag: blume-20020319-witnesses
1372    Description:
1373    
1374    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1375    
1376    This implements a generic mechanism for changing constness qualifiers
1377    anywhere within big C types without resorting to outright "casts".
1378    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1379    modify the constness at the outermost level.)
1380    The implementation of "convert" is based on the idea of "witness"
1381    values -- values that are not used by the operation but whose types
1382    "testify" to their applicability.  On the implementation side, "convert"
1383    is simply a projection (returning its second curried argument).  With
1384    cross-module inlining, it should not result in any machine code being
1385    generated.
1386    
1387    ----------------------------------------------------------------------
1388    Name: Matthias Blume
1389    Date: 2002/03/15 16:40:00 EST
1390    Tag: blume-20020315-basis
1391    Description:
1392    
1393    Provided (preliminary?) implementations for
1394    
1395      {String,Substring}.{concatWith,isSuffix,isSubstring}
1396    
1397    and
1398    
1399      Substring.full
1400    
1401    Those are in the Basis spec but they were missing in SML/NJ.
1402    
1403    ----------------------------------------------------------------------
1404    Name: Matthias Blume
1405    Date: 2002/03/14 21:30:00 EST
1406    Tag: blume-20020314-controls
1407    Description:
1408    
1409    Controls:
1410    ---------
1411    
1412    1. Factored out the recently-added Controls : CONTROLS stuff and put
1413       it into its own library $/controls-lib.cm.  The source tree for
1414       this is under src/smlnj-lib/Controls.
1415    
1416    2. Changed the names of types and functions in this interface, so they
1417       make a bit more "sense":
1418    
1419          module -> registry
1420          'a registry -> 'a group
1421    
1422    3. The interface now deals in ref cells only.  The getter/setter interface
1423       is (mostly) gone.
1424    
1425    4. Added a function that lets one register an already-existing ref cell.
1426    
1427    5. Made the corresponding modifications to the rest of the code so that
1428       everything compiles again.
1429    
1430    6. Changed the implementation of Controls.MLRISC back to something closer
1431       to the original.  In particular, this module (and therefore MLRISC)
1432       does not depend on Controls.  There now is some link-time code in
1433       int-sys.sml that registers the MLRISC controls with the Controls
1434       module.
1435    
1436    CM:
1437    ---
1438    
1439      * One can now specify the lambda-split aggressiveness in init.cmi.
1440    
1441    ----------------------------------------------------------------------
1442    Name: Allen Leung
1443    Date: 2002/03/13 17:30:00 EST
1444    Tag: leunga-20020313-x86-fp-unary
1445    Description:
1446    
1447    Bug fix for:
1448    
1449    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1450    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1451    > - fun f(x,(y,z)) = Real.~ y;
1452    > [autoloading]
1453    > [autoloading done]
1454    >       fchsl   (%eax), 184(%esp)
1455    > Error: MLRisc bug: X86MCEmitter.emitInstr
1456    >
1457    > uncaught exception Error
1458    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1459    
1460    The problem was that the code generator did not generate any fp registers
1461    in this case, and the ra didn't know that it needed to run the X86FP phase to
1462    translate the pseudo fp instruction.   This only happened with unary fp
1463    operators in certain situations.
1464    
1465    ----------------------------------------------------------------------
1466    Name: Matthias Blume
1467    Date: 2002/03/13 14:00:00 EST
1468    Tag: blume-20020313-overload-etc
1469    Description:
1470    
1471    1. Added _overload as a synonym for overload for backward compatibility.
1472       (Control.overloadKW must be true for either version to be accepted.)
1473    
1474    2. Fixed bug in install script that caused more things to be installed
1475       than what was requested in config/targets.
1476    
1477    3. Made CM aware of the (_)overload construct so that autoloading
1478       works.
1479    
1480    ----------------------------------------------------------------------
1481    Name: Matthias Blume
1482    Date: 2002/03/12 22:03:00 EST
1483    Tag: blume-20020312-url
1484    Description:
1485    
1486    Forgot to update BOOT and srcarchiveurl.
1487    
1488    ----------------------------------------------------------------------
1489    Name: Matthias Blume
1490    Date: 2002/03/12 17:30:00 EST
1491    Tag: blume-20020312-version110392
1492    Description:
1493    
1494    Yet another version number bump (because of small changes to the
1495    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1496    
1497    Changes:
1498    
1499      The new pid generation scheme described a few weeks ago was overly
1500      complicated.  I implemented a new mechanism that is simpler and
1501      provides a bit more "stability":  Once CM has seen a compilation
1502      unit, it keeps its identity constant (as long as you do not delete
1503      those crucial CM/GUID/* files).  This means that when you change
1504      an interface, compile, then go back to the old interface, and
1505      compile again, you arrive at the original pid.
1506    
1507      There now also is a mechanism that instructs CM to use the plain
1508      environment hash as a module's pid (effectively making its GUID
1509      the empty string).  For this, "noguid" must be specified as an
1510      option to the .sml file in question within its .cm file.
1511      This is most useful for code that is being generated by tools such
1512      as ml-nlffigen (because during development programmers tend to
1513      erase the tool's entire output directory tree including CM's cached
1514      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1515      revert to the old, broken behavior of SML/NJ, but in specific cases
1516      where there is no danger of interface confusion, its use is ok
1517      (I think).
1518    
1519      ml-nlffigen by default generates "noguid" annotations.  They can be
1520      turned off by specifying -guid in its command line.
1521    
1522    ----------------------------------------------------------------------
1523    Name: Lal George
1524    Date: 2002/03/12 12 14:42:36 EST
1525    Tag: george-20020312-frequency-computation
1526    Description:
1527    
1528    Integrated jump chaining and static block frequency into the
1529    compiler. More details and numbers later.
1530    
1531    ----------------------------------------------------------------------
1532    Name: Lal George
1533    Date: 2002/03/11 11 22:38:53 EST
1534    Tag: george-20020311-jump-chain-elim
1535    Description:
1536    
1537    Tested the jump chain elimination on all architectures (except the
1538    hppa).  This is on by default right now and is profitable for the
1539    alpha and x86, however, it may not be profitable for the sparc and ppc
1540    when compiling the compiler.
1541    
1542    The gc test will typically jump to a label at the end of the cluster,
1543    where there is another jump to an external cluster containing the actual
1544    code to invoke gc. This is to allow factoring of common gc invocation
1545    sequences. That is to say, we generate:
1546    
1547            f:
1548               testgc
1549               ja   L1      % jump if above to L1
1550    
1551            L1:
1552               jmp L2
1553    
1554    
1555    After jump chain elimination the 'ja L1' instructions is converted to
1556    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1557    up being implemented in their long form (if L2 is far away) using:
1558    
1559            jbe     L3      % jump if below or equal to L3
1560            jmp     L2
1561         L3:
1562            ...
1563    
1564    
1565    For large compilation units L2  may be far away.
1566    
1567    
1568    ----------------------------------------------------------------------
1569    Name: Matthias Blume
1570    Date: 2002/03/11 13:30:00 EST
1571    Tag: blume-20020311-mltreeeval
1572    Description:
1573    
1574    A functor parameter was missing.
1575    
1576    ----------------------------------------------------------------------
1577    Name: Allen Leung
1578    Date: 2002/03/11 10:30:00 EST
1579    Tag: leunga-20020311-runtime-string0
1580    Description:
1581    
1582       The representation of the empty string now points to a
1583    legal null terminated C string instead of unit.  It is now possible
1584    to convert an ML string into C string with InlineT.CharVector.getData.
1585    This compiles into one single machine instruction.
1586    
1587    ----------------------------------------------------------------------
1588    Name: Allen Leung
1589    Date: 2002/03/10 23:55:00 EST
1590    Tag: leunga-20020310-x86-call
1591    Description:
1592    
1593       Added machine generation for CALL instruction (relative displacement mode)
1594    
1595    ----------------------------------------------------------------------
1596    Name: Matthias Blume
1597    Date: 2002/03/08 16:05:00
1598    Tag: blume-20020308-entrypoints
1599    Description:
1600    
1601    Version number bumped to 110.39.1.  NEW BOOTFILES!
1602    
1603    Entrypoints: non-zero offset into a code object where execution should begin.
1604    
1605    - Added the notion of an entrypoint to CodeObj.
1606    - Added reading/writing of entrypoint info to Binfile.
1607    - Made runtime system bootloader aware of entrypoints.
1608    - Use the address of the label of the first function given to mlriscGen
1609      as the entrypoint.  This address is currently always 0, but it will
1610      not be 0 once we turn on block placement.
1611    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1612      with entry points) from mlriscGen.
1613    
1614    ----------------------------------------------------------------------
1615    Name: Allen Leung
1616    Date: 2002/03/07 20:45:00 EST
1617    Tag: leunga-20020307-x86-cmov
1618    Description:
1619    
1620       Bug fixes for CMOVcc on x86.
1621    
1622       1. Added machine code generation for CMOVcc
1623       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1624       3. CMOVcc cannot have an immediate operand as argument.
1625    
1626    ----------------------------------------------------------------------
1627    Name: Matthias Blume
1628    Date: 2002/03/07 16:15:00 EST
1629    Tag: blume-20020307-controls
1630    Description:
1631    
1632    This is a very large but mostly boring patch which makes (almost)
1633    every tuneable compiler knob (i.e., pretty much everything under
1634    Control.* plus a few other things) configurable via both the command
1635    line and environment variables in the style CM did its configuration
1636    until now.
1637    
1638    Try starting sml with '-h' (or, if you are brave, '-H')
1639    
1640    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1641    implements the underlying generic mechanism.
1642    
1643    The interface to some of the existing such facilities has changed somewhat.
1644    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1645    (The getFoo interface is still there for backward-compatibility, but its
1646    use is deprecated.)
1647    
1648    The ml-build script passes -Cxxx=yyy command-line arguments through so
1649    that one can now twiddle the compiler settings when using this "batch"
1650    compiler.
1651    
1652    TODO items:
1653    
1654    We should go through and throw out all controls that are no longer
1655    connected to anything.  Moreover, we should go through and provide
1656    meaningful (and correct!) documentation strings for those controls
1657    that still are connected.
1658    
1659    Currently, multiple calls to Controls.new are accepted (only the first
1660    has any effect).  Eventually we should make sure that every control
1661    is being made (via Controls.new) exactly once.  Future access can then
1662    be done using Controls.acc.
1663    
1664    Finally, it would probably be a good idea to use the getter-setter
1665    interface to controls rather than ref cells.  For the time being, both
1666    styles are provided by the Controls module, but getter-setter pairs are
1667    better if thread-safety is of any concern because they can be wrapped.
1668    
1669    *****************************************
1670    
1671    One bug fix: The function blockPlacement in three of the MLRISC
1672    backpatch files used to be hard-wired to one of two possibilities at
1673    link time (according to the value of the placementFlag).  But (I
1674    think) it should rather sense the flag every time.
1675    
1676    *****************************************
1677    
1678    Other assorted changes (by other people who did not supply a HISTORY entry):
1679    
1680    1. the cross-module inliner now works much better (Monnier)
1681    2. representation of weights, frequencies, and probabilities in MLRISC
1682       changed in preparation of using those for weighted block placement
1683       (Reppy, George)
1684    
1685    ----------------------------------------------------------------------
1686    Name: Lal George
1687    Date: 2002/03/07 14:44:24 EST 2002
1688    Tag: george-20020307-weighted-block-placement
1689    
1690    Tested the weighted block placement optimization on all architectures
1691    (except the hppa) using AMPL to generate the block and edge frequencies.
1692    Changes were required in the machine properties to correctly
1693    categorize trap instructions. There is an MLRISC flag
1694    "weighted-block-placement" that can be used to enable weighted block
1695    placement, but this will be ineffective without block/edge
1696    frequencies (coming soon).
1697    
1698    
1699    ----------------------------------------------------------------------
1700    Name: Lal George
1701    Date: 2002/03/05 17:24:48 EST
1702    Tag: george-20020305-linkage-cluster
1703    
1704    In order to support the block placement optimization, a new cluster
1705    is generated as the very first cluster (called the linkage cluster).
1706    It contains a single jump to the 'real' entry point for the compilation
1707    unit. Block placement has no effect on the linkage cluster itself, but
1708    all the other clusters  have full freedom in the manner in which they
1709    reorder blocks or functions.
1710    
1711    On the x86 the typical linkage code that is generated is:
1712       ----------------------
1713            .align 2
1714       L0:
1715            addl    $L1-L0, 72(%esp)
1716            jmp     L1
1717    
1718    
1719            .align  2
1720       L1:
1721       ----------------------
1722    
1723    72(%esp) is the memory location for the stdlink register. This
1724    must contain the address of the CPS function being called. In the
1725    above example, it contains the address of  L0; before
1726    calling L1 (the real entry point for the compilation unit), it
1727    must contain the address for L1, and hence
1728    
1729            addl $L1-L0, 72(%esp)
1730    
1731    I have tested this on all architectures except the hppa.The increase
1732    in code size is of course negligible
1733    
1734    ----------------------------------------------------------------------
1735    Name: Allen Leung
1736    Date: 2002/03/03 13:20:00 EST
1737    Tag: leunga-20020303-mlrisc-tools
1738    
1739      Added #[ ... ] expressions to mlrisc tools
1740    
1741    ----------------------------------------------------------------------
1742    Name: Matthias Blume
1743    Date: 2002/02/27 12:29:00 EST
1744    Tag: blume-20020227-cdebug
1745    Description:
1746    
1747    - made types in structure C and C_Debug to be equal
1748    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1749    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1750    
1751    ----------------------------------------------------------------------
1752    Name: Matthias Blume
1753    Date: 2002/02/26 12:00:00 EST
1754    Tag: blume-20020226-ffi
1755    Description:
1756    
1757    1. Fixed a minor bug in CM's "noweb" tool:
1758       If numbering is turned off, then truly don't number (i.e., do not
1759       supply the -L option to noweb).  The previous behavior was to supply
1760       -L'' -- which caused noweb to use the "default" line numbering scheme.
1761       Thanks to Chris Richards for pointing this out (and supplying the fix).
1762    
1763    2. Once again, I reworked some aspects of the FFI:
1764    
1765       A. The incomplete/complete type business:
1766    
1767       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1768         gone!
1769       - ML types representing an incomplete type are now *equal* to
1770         ML types representing their corresponding complete types (just like
1771         in C).  This is still safe because ml-nlffigen will not generate
1772         RTTI for incomplete types, nor will it generate functions that
1773         require access to such RTTI.   But when ML code generated from both
1774         incomplete and complete versions of the C type meet, the ML types
1775         are trivially interoperable.
1776    
1777         NOTE:  These changes restore the full generality of the translation
1778         (which was previously lost when I eliminated functorization)!
1779    
1780       B. Enum types:
1781    
1782       - Structure C now has a type constructor "enum" that is similar to
1783         how the "su" constructor works.  However, "enum" is not a phantom
1784         type because each "T enum" has values (and is isomorphic to
1785         MLRep.Signed.int).
1786       - There are generic access operations for enum objects (using
1787         MLRep.Signed.int).
1788       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1789         * The structure contains the definition of type "mlrep" (the ML-side
1790         representation type of the enum).  Normally, mlrep is the same
1791         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1792         then mlrep will be defined as a datatype -- thus facilitating
1793         pattern matching on mlrep values.
1794         ("-ec" will be suppressed if there are duplicate values in an
1795          enumeration.)
1796         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1797         will be generated for each C enum constant xxx.
1798         * Conversion functions m2i and i2m convert between mlrep and
1799         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1800         * Coversion functions c and ml convert between mlrep and "tag enum".
1801         * Access functions (get/set) fetch and store mlrep values.
1802       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1803         enumerations are merged into one single enumeration represented by
1804         structure E_'.
1805    
1806    ----------------------------------------------------------------------
1807    Name: Allen Leung
1808    Date: 2002/02/25 04:45:00 EST
1809    Tag: leunga-20020225-cps-spill
1810    
1811    This is a new implementation of the CPS spill phase.
1812    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1813    In case of problems, replace it with the old file spill.sml
1814    
1815    The current compiler runs into some serious performance problems when
1816    constructing a large record.  This can happen when we try to compile a
1817    structure with many items.  Even a very simple structure like the following
1818    makes the compiler slow down.
1819    
1820        structure Foo = struct
1821           val x_1 = 0w1 : Word32.int
1822           val x_2 = 0w2 : Word32.int
1823           val x_3 = 0w3 : Word32.int
1824           ...
1825           val x_N = 0wN : Word32.int
1826        end
1827    
1828    The following table shows the compile time, from N=1000 to N=4000,
1829    with the old compiler:
1830    
1831    N
1832    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1833           MLRISC ra                               0.06u  0.00s  0.05g
1834              (spills = 0 reloads = 0)
1835           TOTAL                                   0.63u  0.07s  0.21g
1836    
1837    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1838           MLRISC ra                               5.68u  0.59s  3.93g
1839              (spills = 0 reloads = 0)
1840           TOTAL                                   14.71u  0.99s  4.81g
1841    
1842    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1843           MLRISC ra                               5.54u  0.65s  3.91g
1844              (spills = 543 reloads = 1082)
1845           TOTAL                                   65.40u  3.13s  6.00g
1846    
1847    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1848           MLRISC ra                               0.80u  0.10s  0.55g
1849              (spills = 42 reloads = 84)
1850           TOTAL                                   129.42u  5.10s  4.13g
1851    
1852    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1853           MLRISC ra                               2.69u  0.27s  1.38g
1854              (spills = 62 reloads = 124)
1855           TOTAL                                   682.48u  19.61s  13.99g
1856    
1857    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1858           MLRISC ra                               4.96u  0.27s  2.72g
1859              (spills = 85 reloads = 170)
1860           TOTAL                                   2375.26u  57.21s  48.00g
1861    
1862    As you can see the old cps spill module suffers from some serious
1863    performance problem.  But since I cannot decipher the old code fully,
1864    instead of patching the problems up, I'm reimplementing it
1865    with a different algorithm.  The new code is more modular,
1866    smaller when compiled, and substantially faster
1867    (O(n log n) time and O(n) space).  Timing of the new spill module:
1868    
1869    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1870          MLRISC ra                               0.25u  0.02s  0.15g
1871             (spills=1 reloads=3)
1872          TOTAL                                   7.74u  0.34s  1.62g
1873    
1874    Implementation details:
1875    
1876    As far as I can tell, the purpose of the CPS spill module is to make sure the
1877    number of live variables at any program point (the bandwidth)
1878    does not exceed a certain limit, which is determined by the
1879    size of the spill area.
1880    
1881    When the bandwidth is too large, we decrease the register pressure by
1882    packing live variables into spill records.  How we achieve this is
1883    completely different than what we did in the old code.
1884    
1885    First, there is something about the MLRiscGen code generator
1886    that we should be aware of:
1887    
1888    o MLRiscGen performs code motion!
1889    
1890       In particular, it will move floating point computations and
1891       address computations involving only the heap pointer to
1892       their use sites (if there is only a single use).
1893       What this means is that if we have a CPS record construction
1894       statement
1895    
1896           RECORD(k,vl,w,e)
1897    
1898       we should never count the new record address w as live if w
1899       has only one use (which is often the case).
1900    
1901       We should do something similar to floating point, but the transformation
1902       there is much more complex, so I won't deal with that.
1903    
1904    Secondly, there are now two new cps primops at our disposal:
1905    
1906     1. rawrecord of record_kind option
1907        This pure operator allocates some uninitialized storage from the heap.
1908        There are two forms:
1909    
1910         rawrecord NONE [INT n]  allocates a tagless record of length n
1911         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1912                                     and initializes the tag.
1913    
1914     2. rawupdate of cty
1915          rawupdate cty (v,i,x)
1916          Assigns to x to the ith component of record v.
1917          The storelist is not updated.
1918    
1919    We use these new primops for both spilling and increment record construction.
1920    
1921     1. Spilling.
1922    
1923        This is implemented with a linear scan algorithm (but generalized
1924        to trees).  The algorithm will create a single spill record at the
1925        beginning of the cps function and use rawupdate to spill to it,
1926        and SELECT or SELp to reload from it.  So both spills and reloads
1927        are fine-grain operations.  In contrast, in the old algorithm
1928        "spills" have to be bundled together in records.
1929    
1930        Ideally, we should sink the spill record construction to where
1931        it is needed.  We can even split the spill record into multiple ones
1932        at the places where they are needed.  But CPS is not a good
1933        representation for global code motion, so I'll keep it simple and
1934        am not attempting this.
1935    
1936     2. Incremental record construction (aka record splitting).
1937    
1938        Long records with many component values which are simulatenously live
1939        (recall that single use record addresses are not considered to
1940         be live) are constructed with rawrecord and rawupdate.
1941        We allocate space on the heap with rawrecord first, then gradually
1942        fill it in with rawupdate.  This is the technique suggested to me
1943        by Matthias.
1944    
1945        Some restrictions on when this is applicable:
1946        1. It is not a VECTOR record.  The code generator currently does not handle
1947           this case. VECTOR record uses double indirection like arrays.
1948        2. All the record component values are defined in the same "basic block"
1949           as the record constructor.  This is to prevent speculative
1950           record construction.
1951    
1952    ----------------------------------------------------------------------
1953    Name: Allen Leung
1954    Date: 2002/02/22 01:02:00 EST
1955    Tag: leunga-20020222-mlrisc-tools
1956    
1957    Minor bug fixes in the parser and rewriter
1958    
1959    ----------------------------------------------------------------------
1960    Name: Allen Leung
1961    Date: 2002/02/21 20:20:00 EST
1962    Tag: leunga-20020221-peephole
1963    
1964    Regenerated the peephole files.  Some contained typos in the specification
1965    and some didn't compile because of pretty printing bugs in the old version
1966    of 'nowhere'.
1967    
1968    ----------------------------------------------------------------------
1969    Name: Allen Leung
1970    Date: 2002/02/19 20:20:00 EST
1971    Tag: leunga-20020219-mlrisc-tools
1972    Description:
1973    
1974       Minor bug fixes to the mlrisc-tools library:
1975    
1976       1.  Fixed up parsing colon suffixed keywords
1977       2.  Added the ability to shut the error messages up
1978       3.  Reimplemented the pretty printer and fixed up/improved
1979           the pretty printing of handle and -> types.
1980       4.  Fixed up generation of literal symbols in the nowhere tool.
1981       5.  Added some SML keywords to to sml.sty
1982    
1983    ----------------------------------------------------------------------
1984    Name: Matthias Blume
1985    Date: 2002/02/19 16:20:00 EST
1986    Tag: blume-20020219-cmffi
1987    Description:
1988    
1989    A wild mix of changes, some minor, some major:
1990    
1991    * All C FFI-related libraries are now anchored under $c:
1992        $/c.cm      --> $c/c.cm
1993        $/c-int.cm  --> $c/internals/c-int.cm
1994        $/memory.cm --> $c/memory/memory.cm
1995    
1996    * "make" tool (in CM) now treats its argument pathname slightly
1997      differently:
1998        1. If the native expansion is an absolute name, then before invoking
1999           the "make" command on it, CM will apply OS.Path.mkRelative
2000           (with relativeTo = OS.FileSys.getDir()) to it.
2001        2. The argument will be passed through to subsequent phases of CM
2002           processing without "going native".  In particular, if the argument
2003           was an anchored path, then "make" will not lose track of that anchor.
2004    
2005    * Compiler backends now "know" their respective C calling conventions
2006      instead of having to be told about it by ml-nlffigen.  This relieves
2007      ml-nlffigen from one of its burdens.
2008    
2009    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
2010    
2011    * Export C_DEBUG and C_Debug from $c/c.cm.
2012    
2013    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
2014      subtyping relationship between incomplete pointers and their complete
2015      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
2016      with the convention of instantiating 'o with ('t, 'c) obj whenever
2017      the pointer target type is complete.  In the incomplete case, 'o
2018      will be instantiated with some "'c iobj" -- a type obtained by
2019      using one of the functors PointerToIncompleteType or PointerToCompleteType.
2020    
2021      Operations that work on both incomplete and complete pointer types are
2022      typed as taking an 'o ptr while operations that require the target to
2023      be known are typed as taking some ('t, 'c) obj ptr.
2024    
2025      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
2026      where void is an eqtype without any values.  This makes it possible
2027      to work on voidptr values using functions meant to operate on light
2028      incomplete pointers.
2029    
2030    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
2031      been vastly simplified.
2032    
2033    ----------------------------------------------------------------------
2034    Name: Matthias Blume
2035    Date: 2002/02/19 10:48:00 EST
2036    Tag: blume-20020219-pqfix
2037    Description:
2038    
2039    Applied Chris Okasaki's bug fix for priority queues.
2040    
2041    ----------------------------------------------------------------------
2042    Name: Matthias Blume
2043    Date: 2002/02/15 17:05:00
2044    Tag: Release_110_39
2045    Description:
2046    
2047    Last-minute retagging is becoming a tradition... :-(
2048    
2049    This is the working release 110.39.
2050    
2051    ----------------------------------------------------------------------
2052    Name: Matthias Blume
2053    Date: 2002/02/15 16:00:00 EST
2054    Tag: Release_110_39-orig
2055    Description:
2056    
2057    Working release 110.39.  New bootfiles.
2058    
2059    (Update: There was a small bug in the installer so it wouldn't work
2060    with all shells.  So I retagged. -Matthias)
2061    
2062    ----------------------------------------------------------------------
2063    Name: Matthias Blume
2064    Date: 2002/02/15 14:17:00 EST
2065    Tag: blume-20020215-showbindings
2066    Description:
2067    
2068    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
2069    the latter can be useful for exploring what bindings are available at
2070    the interactive prompt.  (The first function returns only the list
2071    of symbols that are really bound, the second prints those but also the
2072    ones that CM's autoloading mechanism knows about.)
2073    
2074    ----------------------------------------------------------------------
2075    Name: Matthias Blume
2076    Date: 2002/02/15 12:08:00 EST
2077    Tag: blume-20020215-iptrs
2078    Description:
2079    
2080    Two improvements to ml-nlffigen:
2081    
2082      1. Write files only if they do not exist or if their current contents
2083         do not coincide with what's being written.  (That is, avoid messing
2084         with the time stamps unless absolutely necessary.)
2085    
2086      2. Implement a "repository" mechanism for generated files related
2087         to "incomplete pointer types".   See the README file for details.
2088    
2089    ----------------------------------------------------------------------
2090    Name: Matthias Blume
2091    Date: 2002/02/14 11:50:00 EST
2092    Tag: blume-20020214-quote
2093    Description:
2094    
2095    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
2096    because of the new and improved tag generation scheme.  (Thanks to Allen
2097    Leung for pointing it out.)
2098    
2099    ----------------------------------------------------------------------
2100    Name: Lal George
2101    Date: 2002/02/14 09:55:27 EST 2002
2102    Tag: george-20020214-isabelle-bug
2103    Description:
2104    
2105    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
2106    of Isabelle on the x86.
2107    
2108    From Allen:
2109    -----------
2110     I've found the problem:
2111    
2112         in ra-core.sml, I use the counter "blocked" to keep track of the
2113         true number of elements in the freeze queue.  When the counter goes
2114         to zero, I skip examining the queue.  But I've messed up the
2115         bookkeeping in combine():
2116    
2117             else ();
2118             case !ucol of
2119               PSEUDO => (if !cntv > 0 then
2120                     (if !cntu > 0 then blocked := !blocked - 1 else ();
2121                                        ^^^^^^^^^^^^^^^^^^^^^^^
2122                      moveu := mergeMoveList(!movev, !moveu)
2123                     )
2124                  else ();
2125    
2126         combine() is called to coalesce two nodes u and v.
2127         I think I was thinking that if the move counts of u and v are both
2128         greater than zero then after they are coalesced then one node is
2129         removed from the freeze queue.  Apparently I was thinking that
2130         both u and v are of low degree, but that's clearly not necessarily true.
2131    
2132    
2133    02/12/2002:
2134        Here's the patch.  HOL now compiles.
2135    
2136        I don't know how this impact on performance (compile
2137        time or runtime).  This bug caused the RA (especially on the x86)
2138        to go thru the potential spill phase when there are still nodes on the
2139        freeze queue.
2140    
2141    
2142    
2143    
2144    ----------------------------------------------------------------------
2145    Name: Matthias Blume
2146    Date: 2002/02/13 22:40:00 EST
2147    Tag: blume-20020213-fptr-rtti
2148    Description:
2149    
2150    Fixed a bug in ml-nlffigen that was introduced with one of the previous
2151    updates.
2152    
2153    ----------------------------------------------------------------------
2154    Name: Matthias Blume
2155    Date: 2002/02/13 16:41:00 EST
2156    Tag: blume-20020213-cmlpq
2157    Description:
2158    
2159    Added new priority queue export symbols (which have just been added to
2160    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
2161    not compile and the installer would choke.)
2162    
2163    ----------------------------------------------------------------------
2164    Name: Matthias Blume
2165    Date: 2002/02/13 16:15:00 EST
2166    Tag: blume-20020213-various
2167    Description:
2168    
2169    1. More tweaks to ml-nlffigen:
2170    
2171       - better internal datastructures (resulting in slight speedup)
2172       - "-match" option requires exact match
2173       - "localized" gensym counters (untagged structs/unions nested within
2174         other structs/unions or within typedefs get a fresh counter; their
2175         tag will be prefixed by a concatenation of their parents' tags)
2176       - bug fixes (related to calculation of transitive closure of types
2177         to be included in the output)
2178    
2179    2. Minor Basis updates:
2180    
2181       - added implementations for List.collate and Option.app
2182    
2183    ----------------------------------------------------------------------
2184    Name: Matthias Blume
2185    Date: 2002/02/11 15:55:00 EST
2186    Tag: blume-20020211-gensym
2187    Description:
2188    
2189    Added a "-gensym" option to command line of ml-nlffigen.  This can be
2190    used to specify a "stem" -- a string that is inserted in all "gensym'd"
2191    names (ML structure names that correspond to unnamed C structs, unions,
2192    and enums), so that separate runs of ml-nlffigen do not clash.
2193    
2194    ----------------------------------------------------------------------
2195    Name: Matthias Blume
2196    Date: 2002/02/11 12:05:00 EST
2197    Tag: blume-20020211-gensml
2198    Description:
2199    
2200    A quick fix for a problem with GenSML (in the pgraph-util library):
2201    Make generation of toplevel "local" optional.  (Strictly speaking,
2202    signature definitions within "local" are not legal SML.)
2203    
2204    Other than that: updates to INSTALL and cm/TODO.
2205    
2206    ----------------------------------------------------------------------
2207    Name: Matthias Blume
2208    Date: 2002/02/08 15:00:00 EST
2209    Tag: blume-20020208-uniquepid
2210    Description:
2211    
2212    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
2213    
2214    1. The installer (config/install.sh) has gotten smarter:
2215    
2216         - Configuration options are a bit easier to specify now
2217           (in config/targets).
2218         - Bug in recognizing .tar.bz2 files fixed.
2219         - Installer automatically resolves dependencies between
2220           configuration options (e.g., if you ask for eXene, you will
2221           also get cml -- regardless whether you asked for it or not).
2222         - Installer can run in "quieter mode" by setting the environment
2223           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
2224           "completely silent", though.
2225         - Build HashCons library as part of smlnj-lib.
2226    
2227    2. A new scheme for assigning persistent identifiers to compilation
2228       units (and, by extension, to types etc.) has been put into place.
2229       This fixes a long-standing bug where types and even dynamic values
2230       can get internally confused, thereby compromising type safety
2231       (abstraction) and dynamic correctness.  See
2232    
2233         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
2234    
2235       for an example of how things could go wrong until now.
2236    
2237       The downside of the new scheme is that pids are not quite as
2238       persistent as they used to be: CM will generate a fresh pid
2239       for every compilation unit that it thinks it sees for the first
2240       time.  That means that if you compile starting from a clean, fresh
2241       source tree at two different times, you end up with different
2242       binaries.
2243    
2244       Cutoff recompilation, however, has not been compromised because
2245       CM keeps pid information in special caches between runs.
2246    
2247    ----------------------------------------------------------------------
2248    Name: Lal George
2249    Date: 2002/02/07 15:34:13 EST 2002
2250    Tag: <none>
2251    Description:
2252    
2253    Compilers that generate assembly code may produce  global labels
2254    whose value is resolved at link time. The various peephole optimization
2255    modules did not take this in account.
2256    
2257    TODO. The Labels.addrOf function should really return an option
2258    type so that clients are forced to deal with this issue, rather
2259    than an exception being raised.
2260    
2261    ----------------------------------------------------------------------
2262    Name: Lal George
2263    Date: 2002/02/06 13:55:02 EST
2264    Tag: george-20020206-ra-breakup
2265    Description:
2266    
2267    1. A bug fix from Allen.
2268    
2269        A typo causes extra fstp %st(0)'s to be generated at compensation
2270        edges, which might cause stack underflow traps at runtime.  This
2271        occurs in fft where there are extraneous fstps right before the 'into'
2272        trap instruction (in this case they are harmless since none of the
2273        integers overflow.)
2274    
2275    2. Pulled out various utility modules that were embedded in the modules
2276       of the register allocator. I need these modules for other purposes, but
2277       they are not complete enough to put into a library (just yet).
2278    ----------------------------------------------------------------------
2279    Name: Matthias Blume
2280    Date: 2002/01/31 16:05:00 EST
2281    Tag: blume-20020131-sparc-ccalls
2282    Description:
2283    
2284    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
2285       of extra stack space by mistake.  Fixed.
2286    
2287    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
2288    
2289    ----------------------------------------------------------------------
2290    Name: Allen Leung
2291    Date: 2002/01/30
2292    Tag: leunga-20020130-nowhere-bug-fix
2293    Description:
2294    
2295       MLRISC bug fixes:
2296       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
2297       2. MachineInt.fromString was negating its value.
2298    
2299    ----------------------------------------------------------------------
2300    Name: Matthias Blume
2301    Date: 2002/01/29
2302    Tag: blume-20020129-INSTALL
2303    Description:
2304    
2305    - Added somewhat detailed installation instructions (file INSTALL).
2306    - Fixed curl-detection bug in config/install.sh.
2307    - It is now possible to select the URL getter using the URLGETTER
2308      environment variable:
2309    
2310          not set / "unknown"      --> automatic detection (script tries wget,
2311                                       curl, and lynx)
2312          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
2313                                       how to properly invoke them)
2314          other                    --> use $URLGETTER directly, it must take
2315                                       precisely two command-line arguments
2316                                       (source URL and destination file name)
2317    
2318    ----------------------------------------------------------------------
2319    Name: Matthias Blume
2320    Date: 2002/01/28
2321    Tag: blume-20020128-sparc-ccalls
2322    Description:
2323    
2324    - Fixed problem with calculation of "used" registers in sparc-c-calls.
2325    - Make use of the allocParam argument in sparc-c-calls.
2326    
2327    ----------------------------------------------------------------------
2328    Name: Matthias Blume
2329    Date: 2002/01/28
2330    Tag: blume-20020128-allocParam
2331    Description:
2332    
2333    John Reppy:  Changes c-calls API to accept client-callback for
2334    allocating extra stack space.
2335    me: Corresponding changes to mlriscGen (using a dummy argument that
2336        does not change the current behavior).
2337    
2338    ----------------------------------------------------------------------
2339    Name: Matthias Blume
2340    Date: 2002/01/28 12:00:00
2341    Tag: Release_110_38
2342    Description:
2343    
2344    This time for real!!!
2345    
2346    ----------------------------------------------------------------------
2347    Name: Matthias Blume
2348    Date: 2002/01/28 10:56:00 EST
2349    Tag: blume-20020128-retraction
2350    Description:
2351    
2352    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
2353       with blume-Release_110_38-retracted.)
2354    
2355    1. Fixed a problem with incorrect rounding modes in real64.sml.
2356       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
2357    
2358    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
2359       and enums fixed.  The naming of corresponding ML identifiers should
2360       now be consistent again.
2361    
2362    ----------------------------------------------------------------------
2363    Name: Allen Leung
2364    Date: 2002/01/27
2365    Tag: leunga-20020127-nowhere
2366    Description:
2367    
2368       Added a target called nowhere in the configuration scripts.
2369       Enabling this will build the MLRISC 'nowhere' tool (for translating
2370       programs with where-clauses into legal SML code) during installation.
2371    
2372    ----------------------------------------------------------------------
2373    Name: Matthias Blume
2374    Date: 2002/01/25 21:27:00 EST
2375    Tag: blume-Release_110_38-retracted
2376    Description:
2377    
2378    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
2379    
2380    README will be added later.
2381    
2382    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
2383    (Release_110_38) removed.  Reason: Last-minute bug fixes.
2384    
2385    ----------------------------------------------------------------------
2386    Name: Matthias Blume
2387    Date: 2002/01/25
2388    Tag: blume-20020125-ffi
2389    Description:
2390    
2391    A large number of tweaks and improvements to ml-nlffi-lib and
2392    ml-nlffigen:
2393    
2394       - ML represenation types have been streamlined
2395       - getter and setter functions work with concrete values, not abstract
2396         ones where possible
2397       - ml-nlffigen command line more flexible (see README file there)
2398       - some bugs have been fixed (hopefully)
2399    
2400    ----------------------------------------------------------------------
2401    Name: Lal George
2402    Date: 2002/01/24
2403    Tag: george-20020124-risc-ra-interface
2404    Description:
2405    
2406       There is a dramatic simplification in the interface to the
2407       register allocator for RISC architectures as a result of making
2408       parallel copy instructions explicit.
2409    
2410    ----------------------------------------------------------------------
2411    Name: Matthias Blume
2412    Date: 2002/01/22
2413    Tag: blume-20020122-x86-ccalls
2414    Description:
2415    
2416    Bug fix for c-calls on x86 (having to do with how char- and
2417    short-arguments are being handled).
2418    
2419    ----------------------------------------------------------------------
2420    Name: Matthias Blume
2421    Date: 2002/01/21
2422    Tag: blume-20020121-ff
2423    Description:
2424    
2425    Another day of fiddling with the FFI...
2426    
2427    1. Bug fix/workaround:  CKIT does not complain about negative array
2428       dimensions, so ml-nlffigen has to guard itself against this possibility.
2429       (Otherwise a negative dimension would send it into an infinite loop.)
2430    
2431    2. Some of the abstract types (light objects, light pointers, most "base"
2432       types) in structure C are now eqtypes.
2433    
2434    3. Added constructors and test functions for NULL function pointers.
2435    
2436    ----------------------------------------------------------------------
2437    Name: Matthias Blume
2438    Date: 2002/01/18
2439    Tag: blume-20020118-ready-for-new-release
2440    Description:
2441    
2442    Made config/srcarchiveurl point to a new place.  (Will provide boot
2443    files shortly.)
2444    
2445    Maybe we christen this to be 110.38?
2446    
2447    ----------------------------------------------------------------------
2448    Name: Matthias Blume
2449    Date: 2002/01/18
2450    Tag: blume-20020118-more-ffifiddle
2451    Description:
2452    
2453    Today's FFI fiddling:
2454    
2455      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
2456        C.Set.* that use concrete (MLRep.*) arguments and results instead
2457        of abstract ones.
2458    
2459      - Provided word-style bit operations etc. for "int" representation
2460        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
2461        and Long.
2462    
2463    ----------------------------------------------------------------------
2464    Name: Matthias Blume
2465    Date: 2002/01/18
2466    Tag: blume-20020118-use-x86-fp
2467    Description:
2468    
2469    Now that x86-fast-fp seems to be working, I turned it back on again
2470    by default.  (Seems to work fine now, even with the FFI.)
2471    
2472    Other than that, I added some documentation about the FFI to
2473    src/ml-nlffigen/README and updated the FFI test examples in
2474    src/ml-nlffi-lib/Tests/*.
2475    
2476    ----------------------------------------------------------------------
2477    Name: Allen Leung
2478    Date: 2002/01/17
2479    Tag: leunga-20020117-x86-fast-fp-call
2480    Description:
2481    
2482       1. Fixed a problem with handling return fp values when x86's fast fp
2483          mode is turned on.
2484    
2485       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
2486          of %f32.
2487    
2488       3. Added a constructor INT32lit to the ast of MLRISC tools.
2489    
2490    ----------------------------------------------------------------------
2491    Name: Matthias Blume
2492    Date: 2002/01/16
2493    Tag: blume-20020116-ffifiddle
2494    Description:
2495    
2496    More fiddling with the FFI interface:
2497    
2498     - Make constness 'c instead of rw wherever possible.  This eliminates
2499       the need for certain explicit coercions.  (However, due to ML's
2500       value polymorphism, there will still be many cases where explicit
2501       coercions are necessary.  Phantom types are not the whole answer
2502       to modeling a subtyping relationship in ML.)
2503    
2504     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
2505    
2506     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
2507       clumsy workaround for varargs problem.)
2508    
2509    ----------------------------------------------------------------------
2510    Name: Lal George
2511    Date: 2002/01/15
2512    Tag: <none>
2513    Description:
2514    
2515    1. Since COPY instructions are no longer native to the architecture,
2516       a generic functor can be used to implement the expandCopies function.
2517    
2518    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
2519       TEXT segment.
2520    
2521    ----------------------------------------------------------------------
2522    Name: Matthias Blume
2523    Date: 2002/01/15
2524    Tag: blume-20020115-ffiupdates
2525    Description:
2526    
2527    1. Fix for bug resulting in single-precision float values being returned
2528       incorrectly from FFI calls.
2529    
2530    2. Small modifications to C FFI API:
2531    
2532        - memory-allocation routines return straight objects (no options)
2533          and raise an exception in out-of-memory situations
2534        - unsafe extensions to cast between function pointers and pointers
2535          from/to ints
2536        - added structure C_Debug as an alternative to structure C where
2537          pointer-dereferencing (|*| and |*!) always check for null-pointers
2538        - added open_lib' to DynLinkage;  open_lib' works like open_lib
2539          but also takes a (possibly empty) list of existing library handles
2540          that the current library depends on
2541    
2542    ----------------------------------------------------------------------
2543    Name: Matthias Blume
2544    Date: 2002/01/10
2545    Tag: blume-20020110-newffigen
2546    Description:
2547    
2548    1. Updates to portable graph code.
2549    
2550    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
2551       more scalable now so that even huge interfaces such as the one
2552       for GTK compile in finite time and space. :-)
2553       See src/ml-nlffigen/README for details on what's new.
2554    
2555    ----------------------------------------------------------------------
2556    Name: Lal George
2557    Date: 2001/01/09 14:31:35 EST 2002
2558    Tag: george-20011206-rm-native-copy
2559    Description:
2560    
2561            Removed the native COPY and FCOPY instructions
2562            from all the architectures and replaced it with the
2563            explicit COPY instruction from the previous commit.
2564    
2565            It is now possible to simplify many of the optimizations
2566            modules that manipulate copies. This has not been
2567            done in this change.
2568    
2569    ----------------------------------------------------------------------
2570    Name: Lal George
2571    Date: 2001/12/06 16:50:13 EST 2001
2572    Tag: george-20011206-mlrisc-instruction
2573    Description:
2574    
2575    Changed the representation of instructions from being fully abstract
2576    to being partially concrete. That is to say:
2577    
2578      from
2579            type instruction
2580    
2581      to
2582            type instr                              (* machine instruction *)
2583    
2584            datatype instruction =
2585                LIVE of {regs: C.cellset, spilled: C.cellset}
2586              | KILL of {regs: C.cellset, spilled: C.cellset}
2587              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
2588              | ANNOTATION of {i: instruction, a: Annotations.annotation}
2589              | INSTR of instr
2590    
2591    This makes the handling of certain special instructions that appear on
2592    all architectures easier and uniform.
2593    
2594    LIVE and KILL say that a list of registers are live or killed at the
2595    program point where they appear. No spill code is generated when an
2596    element of the 'regs' field is spilled, but the register is moved to
2597    the 'spilled' (which is present, more for debugging than anything else).
2598    
2599    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
2600    We used to generate:
2601    
2602            DEFFREG f1
2603            f1 := f2 + f3
2604            trapb
2605    
2606    but now generate:
2607    
2608            f1 := f2 + f3
2609            trapb
2610            LIVE {regs=[f1,f2,f3], spilled=[]}
2611    
2612    Furthermore, the DEFFREG (hack) required that all floating point instruction
2613    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
2614    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
2615    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
2616    this architecture).
2617    
2618    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
2619    all the architectures. This will result in further simplification of the
2620    register allocator that must be aware of them for coalescing purposes, and
2621    will also simplify certain aspects of the machine description that provides
2622    callbacks related to parallel copies.
2623    
2624    ANNOTATION should be obvious, and now INSTR represents the honest to God
2625    machine instruction set!
2626    
2627    The <arch>/instructions/<arch>Instr.sml files define certain utility
2628    functions for making porting easier -- essentially converting upper case
2629    to lower case. All machine instructions (of type instr) are in upper case,
2630    and the lower case form generates an MLRISC instruction. For example on
2631    the alpha we have:
2632    
2633      datatype instr =
2634         LDA of {r:cell, b:cell, d:operand}
2635       | ...
2636    
2637      val lda : {r:cell, b:cell, d:operand} -> instruction
2638        ...
2639    
2640    where lda is just (INSTR o LDA), etc.
2641    
2642    ----------------------------------------------------------------------
2643    Name: Matthias Blume
2644    Date: 2001/11/22 21:40:00 EST
2645    Tag: Release_110_37
2646    Description:
2647    
2648    Release 110.37.  This time for real.
2649    
2650    ----------------------------------------------------------------------
2651    Name: Matthias Blume
2652    Date: 2001/11/21 16:35:00 EST
2653    Tag: blume-20011121-foot-in-mouth
2654    Description:
2655    
2656    Removed the "Release_110_37" tag because of a serious bug.
2657    This will be re-tagged once the bug is fixed.
2658    
2659    ----------------------------------------------------------------------
2660    Name: Matthias Blume
2661    Date: 2001/11/21 16:14:00 EST
2662    Tag: blume-20011121-forgottenfile
2663    Description:
2664    
2665    Forgot to add a file.  (Just a .tex-file -- part of
2666    the CM manual source.)
2667    
2668    ----------------------------------------------------------------------
2669    Name: Matthias Blume
2670    Date: 2001/11/21 16:10:00 EST
2671    Tag: blume-20011121-invalid_110_37
2672    Description:
2673    
2674    Note: I removed the original tag "Release_110_37" from this commit
2675          because we found a serious bug in all non-x86 backends.
2676          - Matthias
2677    
2678    1. Modifications to the SML/NJ code generator and to the runtime system
2679       so that code object name strings are directly inserted into code
2680       objects at code generation time.  The only business the runtime system
2681       has with this is now to read the name strings on occasions.
2682       (The encoding of the name string has also changed somewhat.)
2683    
2684    2. CM now implements a simple "set calculus" for specifying export lists.
2685       In particular, it is now possible to refer to the export lists of
2686       other libraries/groups/sources and form unions as well as differences.
2687       See the latest CM manual for details.
2688    
2689    3. An separate notion of "proxy" libraries has again be eliminated from
2690       CM's model.  (Proxy libraries are now simply a special case of using
2691       the export list calculus.)
2692    
2693    4. Some of the existing libraries now take advantage of the new set
2694       calculus.
2695       (Notice that not all libraries have been converted because some
2696       of the existing .cm-files are supposed to be backward compatible
2697       with 110.0.x.)
2698    
2699    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2700       "General.exnMessage"!  The former relies on a certain hook to be
2701       initialized, and that often does not happen in the stand-alone case.)
2702    
2703    ----------------------------------------------------------------------
2704    Name: Lal George
2705    Date: 2001/11/21  13:56:18 EST
2706    Tag: george-2001121-pseudo-ops
2707    Description:
2708    
2709      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2710      ought to never be any question of incompatabilities with
2711      pseudo-op syntax expected by host assemblers.
2712    
2713      For now, only modules supporting GAS syntax are implemented
2714      but more should follow, such as MASM, and vendor assembler
2715      syntax, e.g. IBM as, Sun as, etc.
2716    
2717    ----------------------------------------------------------------------
2718    Name: Matthias Blume
2719    Date: 2001/11/14 11:52:00 EST
2720    Tag: blume-20011114-srcname
2721    Description:
2722    
2723    1. Routed the name of the current source file to mlriscgen where it
2724       should be directly emitted into the code object.  (This last part
2725       is yet to be done.)
2726    
2727    2. Some cleanup of the pgraph code to make it match the proposal that
2728       I put out the other day.  (The proposal notwithstanding, things are
2729       still in flux here.)
2730    
2731    ----------------------------------------------------------------------
2732    Name: Lal George
2733    Date: 2001/11/14 09:44:04 EST
2734    Tag:
2735    Description:
2736    
2737      Fix for a backpatching bug reported by Allen.
2738    
2739      Because the boundary between short and long span-dependent
2740      instructions is +/- 128, there are an astounding number of
2741      span-dependent instructions whose size is over estimated.
2742    
2743      Allen came up with the idea of letting the size of span
2744      dependent instructions be non-monotonic, for a maxIter
2745      number of times, after which the size must be monotonically
2746      increasing.
2747    
2748      This table shows the number of span-dependent instructions
2749      whose size was over-estimated as a function of maxIter, for the
2750      file Parse/parse/ml.grm.sml:
2751    
2752         maxIter            # of instructions:
2753            10                      687
2754            20                      438
2755            30                      198
2756            40                        0
2757    
2758      In compiling the compiler, there is no significant difference in
2759      compilation speed between maxIter=10 and maxIter=40. Actually,
2760      my measurements showed that maxIter=40 was a tad faster than
2761      maxIter=10! Also 96% of the  files in the compiler reach a fix
2762      point within 13 iterations, so fixing maxIter at 40, while high,
2763      is okay.
2764    
2765    ----------------------------------------------------------------------
2766    Name: Matthias Blume
2767    Date: 2001/10/31 15:25:00 EST
2768    Tag: blume-20011031-pgraph
2769    Description:
2770    
2771    CKIT:
2772    * Changed the "Function" constructor of type Ast.ctype to carry optional
2773      argument identifiers.
2774    * Changed the return type of TypeUtil.getFunction accordingly.
2775    * Type equality ignores the argument names.
2776    * TypeUtil.composite tries to preserve argument names but gives up quickly
2777      if there is a mismatch.
2778    
2779    installation script:
2780    * attempts to use "curl" if available (unless "wget" is available as well)
2781    
2782    CM:
2783    * has an experimental implementation of "portable graphs" which I will
2784      soon propose as an implementation-independent library format
2785    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2786    
2787    NLFFI-LIB:
2788    * some cleanup (all cosmetic)
2789    
2790    NLFFIGEN:
2791    * temporarily disabled the mechanism that suppresses ML output for
2792      C definitions whose identifiers start with an underscore character
2793    * generate val bindings for enum constants
2794    * user can request that only one style (light or heavy) is being used;
2795      default is to use both (command-line arguments: -heavy and -light)
2796    * fixed bug in handling of function types involving incomplete pointers
2797    * generate ML entry points that take record arguments (i.e., using
2798      named arguments) for C functions that have a prototype with named
2799      arguments
2800      (see changes to CKIT)
2801    
2802    ----------------------------------------------------------------------
2803    Name: Allen Leung
2804    Date: 2001/10/27 20:34:00 EDT
2805    Tag: leunga-20011027-x86-fast-fp-call
2806    Description:
2807    
2808       Fixed the bug described in blume-20010920-slowfp.
2809    
2810       The fix involves
2811          1. generating FCOPYs in FSTP in ia32-svid
2812          2. marking a CALL with the appropriate annotation
2813    
2814    ----------------------------------------------------------------------
2815    Name: Matthias Blume
2816    Date: 2001/10/16 11:32:00 EDT
2817    Tag: blume-20011016-netbsd
2818    Description:
2819    
2820    Underscore patch from Chris Richards (fixing problem with compiling
2821    runtime system under recent NetBSD).
2822    
2823    ----------------------------------------------------------------------
2824    Name: Allen Leung
2825    Date: 2001/10/12 17:18:32 EDT 2001
2826    Tag: leung-20011012-x86-printflowgraph
2827    Description:
2828    
2829    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2830    
2831    ----------------------------------------------------------------------
2832    Name: Lal George
2833    Date: 2001/10/11 23:51:34 EDT
2834    Tag: george-20011011-too-many-instrs
2835    Description:
2836    
2837    The representation of a program point never expected to see more
2838    than 65536 instructions in a basic block!
2839    
2840    ----------------------------------------------------------------------
2841    Name: Lal George
2842    Date: 2001/10/09 09:41:37 EDT
2843    Tag: george-20011008-mlrisc-labels
2844    Description:
2845    
2846    Changed the machine description files to support printing of
2847    local and global labels in assembly code, based on host assembler
2848    conventions.
2849    
2850    ----------------------------------------------------------------------
2851    Name: Matthias Blume
2852    Date: 2001/09/25 15:25:00 EDT
2853    Tag: blume-20010925-exninfo
2854    Description:
2855    
2856    I provided a non-hook implementation of exnName (at the toplevel) and
2857    made the "dummy" implementation of exnMessage (at the toplevel) more
2858    useful: if nothing gets "hooked in", then at least you are going to
2859    see the exception name and a message indicating why you don't see more.
2860    
2861    [For the time being, programs that need exnMessage and want to use
2862    ml-build should either use General.exnMessage (strongly recommended) or
2863    refer to structure General at some other point so that CM sees a
2864    static dependency.]
2865    
2866    [Similar remarks go for "print" and "use":  If you want to use their
2867    functionality in stand-alone programs generated by ml-build, then use
2868    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2869    
2870    ----------------------------------------------------------------------
2871    Name: Matthias Blume
2872    Date: 2001/09/20 17:28:00 EDT
2873    Tag: blume-20010920-slowfp
2874    Description:
2875    
2876    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2877    
2878    ----------------------------------------------------------------------
2879    Name: Matthias Blume
2880    Date: 2001/09/20 17:20:00 EDT
2881    Tag: blume-20010920-canonicalpaths
2882    Description:
2883    
2884    0. Updated the BOOT file (something that I forgot to do earlier).
2885    
2886    1. Small internal change to CM so that it avoids "/../" in filenames
2887       as much as possible (but only where it is safe).
2888    
2889    2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2890       that arguments that contain delimiters are passed through correctly.
2891       This change also means that all "special" arguments of the form
2892       @SMLxxx... must come first.
2893    
2894    3. Changed install script to put relative anchor names for tool commands
2895       into pathconfig.
2896    
2897    ----------------------------------------------------------------------
2898    Name: Matthias Blume
2899    Date: 2001/09/18 15:35:00 EDT
2900    Tag: blume-20010918-readme11036
2901    Description:
2902    
2903    Added README files.
2904    
2905    ----------------------------------------------------------------------
2906    Name: Matthias Blume
2907    Date: 2001/09/18 11:45:00 EDT
2908    Tag: Release_110_36 (retag)
2909    Description:
2910    
2911    Fixed mistake in config/preloads. Retagged as 110.36.
2912    
2913    ----------------------------------------------------------------------
2914    Name: Matthias Blume
2915    Date: 2001/09/18 09:40:00 EDT
2916    Tag: Release_110_36_orig (tag changed)
2917    Description:
2918    
2919    New version (110.36).  New bootfiles.
2920    
2921    ----------------------------------------------------------------------
2922    Name: Matthias Blume
2923    Date: 2001/09/14 16:15:00 EDT
2924    Tag: blume-20010914-x86fastfp
2925    Description:
2926    
2927    John committed some changes that Allen made, in particular a (hopefully)
2928    correctly working version of the x86-fp module.
2929    
2930    I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
2931    flag to "true".  Everything seems to compile to a fixpoint ok, and
2932    "mandelbrot" speeds up by about 15%.
2933    
2934    ----------------------------------------------------------------------
2935    Name: Matthias Blume
2936    Date: 2001/09/13 11:20:00 EDT
2937    Tag: blume-20010913-minimal
2938    Description:
2939    
2940    1. Stefan Monnier's patch to fix a miscompilation problem that
2941       was brought to light by John Reppy's work on Moby.
2942    
2943    2. Implemented a minimal "structure Compiler" that contains just
2944       "version" and "architecture".  The minimal version will be
2945       available when the full version is not.  This is for backward-
2946       compatibility with code that wants to test Compiler.version.
2947    
2948    ----------------------------------------------------------------------
2949    Name: Matthias Blume
2950    Date: 2001/08/28 14:03:00 EDT
2951    Tag: blume-20010828-ml-lex
2952    Description:
2953    
2954    Fix for bug 1581, received from Neophytos Michael.
2955    
2956    ----------------------------------------------------------------------
2957    Name: Matthias Blume
2958    Date: 2001/08/27 11:20:00 EDT
2959    Tag: blume-20010827-readme11035
2960    Description:
2961    
2962    Fleshed out the README file for 110.35.
2963    
2964    ----------------------------------------------------------------------
2965    Name: Matthias Blume
2966    Date: 2001/08/24 17:10:00 EDT
2967    Tag: Release_110_35
2968    Description:
2969    
2970    New version number (110.35).  New bootfiles.
2971    
2972    ----------------------------------------------------------------------
2973    Name: Lal George
2974    Date: 2001/08/24 13:47:18 EDT 2001
2975    Tag: george-20010824-MLRISC-graphs
2976    Description:
2977    
2978     removed clusters from MLRISC completely and replaced with graphs.
2979    
2980    ----------------------------------------------------------------------
2981    Name: Matthias Blume
2982    Date: 2001/08/23 17:50:00 EDT
2983    Tag: blume-20010823-toplevel
2984    Description:
2985    
2986    - some reorganization of the code that implements various kinds of
2987      environments in the compiler (static, dynamic, symbolic, combined)
2988    - re-implemented the EnvRef module so that evalStream works properly
2989      (if the stream contains references to "use", "CM.make", etc.)
2990    - cleaned up evalloop.sml and interact.sml (but they need more cleaning)
2991    
2992    ----------------------------------------------------------------------
2993    Name: Matthias Blume
2994    Date: 2001/08/20 15:50 EDT
2995    Tag: blume20010820-slipup
2996    Description:
2997    
2998    I forgot to commit a few files.  Here they are...
2999    
3000    ----------------------------------------------------------------------
3001    Name: Matthias Blume
3002    Date: 2001/08/20 15:35:00 EDT
3003    Tag: blume-20010820-debugprof
3004    Description:
3005    
3006    !!!! NEW BOOTFILES !!!!
3007    
3008    This is another round of reorganizing the compiler sources.  This
3009    time the main goal was to factor out all the "instrumentation"
3010    passes (for profiling and backtracing) into their own library.
3011    The difficulty was to do it in such a way that it does not depend
3012    on elaborate.cm but only on elabdata.cm.
3013    
3014    Therefore there have been further changes to both elaborate.cm and
3015    elabdata.cm -- more "generic" things have been moved from the former
3016    to the latter.  As a result, I was forced to split the assignment
3017    of numbers indicating "primtyc"s into two portions: SML-generic and
3018    SML/NJ-specific.  Since it would have been awkward to maintain,
3019    I bit the bullet and actually _changed_ the mapping between these
3020    numbers and primtycs.  The bottom line of this is that you need
3021    a new set of bin- and bootfiles.
3022    
3023    I have built new bootfiles for all architectures, so doing a fresh
3024    checkout and config/install.sh should be all you need.
3025    
3026    The newly created library's name is
3027    
3028        $smlnj/viscomp/debugprof.cm
3029    
3030    and its sources live under
3031    
3032        src/compiler/DebugProf
3033    
3034    ----------------------------------------------------------------------
3035    Name: Matthias Blume
3036    Date: 2001/08/15 17:15:00 EDT
3037    Tag: blume-20010815-compreorg
3038    Description:
3039    
3040    This is a first cut at reorganizing the CM libraries that make up the
3041    core of the compiler.  The idea is to separate out pieces that could
3042    be used independently by tools, e.g., the parser, the typechecker, etc.
3043    
3044    The current status is a step in this direction, but it is not quite
3045    satisfactory yet.  Expect more changes in the future.
3046    
3047    Here is the current (new) organization...
3048    
3049        What used to be $smlnj/viscomp/core.cm is now divided into
3050        six CM libraries:
3051    
3052             $smlnj/viscomp/basics.cm
3053                           /parser.cm
3054                           /elabdata.cm
3055                           /elaborate.cm
3056                           /execute.cm
3057                           /core.cm
3058    
3059        The CM files for these libraries live under src/system/smlnj/viscomp.
3060        All these libraries are proxy libraries that contain precisely
3061        one CM library component.  Here are the locations of the components
3062        (all within the src/compiler tree):
3063    
3064             Basics/basics.cm
3065             Parse/parser.cm
3066             ElabData/elabdata.cm
3067             Elaborator/elaborate.cm
3068             Execution/execute.cm
3069             core.cm
3070    
3071         [This organization is the same that has been used already
3072         for a while for the architecture-specific parts of the visible
3073         compiler and for the old version of core.cm.]
3074    
3075         As you will notice, many source files have been moved from their
3076         respective original locations to a new home in one of the above
3077         subtrees.
3078    
3079         The division of labor between the new libraries is the following:
3080    
3081             basics.cm:
3082                - Simple, basic definitions that pertain to many (or all) of
3083                  the other libraries.
3084             parser.cm:
3085                - The SML parser, producing output of type Ast.dec.
3086                - The type family for Ast is also defined and exported here.
3087             elabdata.cm:
3088                - The datatypes that describe input and output of the elaborator.
3089                  This includes types, absyn, and static environments.
3090             elaborator.cm:
3091                - The SML/NJ type checker and elaborator.
3092                  This maps an Ast.dec (with a given static environment) to
3093                  an Absyn.dec (with a new static environment).
3094                - This libraries implements certain modules that used to be
3095                  structures as functors (to remove dependencies on FLINT).
3096             execute.cm:
3097                - Everything having to do with executing binary code objects.
3098                - Dynamic environments.
3099             core.cm:
3100                - SML/NJ-specific instantiations of the elaborator and MLRISC.
3101                - Top-level modules.
3102                - FLINT (this should eventually become its own library)
3103    
3104    Notes:
3105    
3106    I am not 100% happy with the way I separated the elaborator (and its
3107    data structures) from FLINT.  Two instances of the same problem:
3108    
3109        1. Data structures contain certain fields that carry FLINT-specific
3110           information.  I hacked around this using exn and the property list
3111           module from smlnj-lib.  But the fact that there are middle-end
3112           specific fields around at all is a bit annoying.
3113    
3114        2. The elaborator calculates certain FLINT-related information.  I tried
3115           to make this as abstract as I could using functorization, but, again,
3116           the fact that the elaborator has to perform calculations on behalf
3117           of the middle-end at all is not nice.
3118    
3119        3. Having to used exn and property lists is unfortunate because it
3120           weakens type checking.  The other alternative (parameterizing
3121           nearly *everything*) is not appealing, though.
3122    
3123    I removed the "rebinding =" warning hack because due to the new organization
3124    it was awkward to maintain it.  As a result, the compiler now issues some of
3125    these warnings when compiling init.cmi during bootstrap compilation. On
3126    the plus side, you also get a warning when you do, for example:
3127       val op = = Int32.+
3128    which was not the case up to now.
3129    
3130    I placed "assign" and "deref" into the _Core structure so that the
3131    code that deals with the "lazy" keyword can find them there.  This
3132    removes the need for having access to the primitive environment
3133    during elaboration.
3134    
3135    ----------------------------------------------------------------------
3136    Name: Matthias Blume
3137    Date: 2001/08/13
3138    Tag: blume-20010813-closures
3139    Description:
3140    
3141    This fix was sent to us by Zhong Shao.  It is supposed to improve the
3142    performance of certain loops by avoiding needless closure allocation.
3143    
3144    ----------------------------------------------------------------------
3145    Name: Lal George
3146    Date: 2001/07/31 10:03:23 EDT 2001
3147    Tag: george-20010731-x86-fmalloc
3148    Description: Fixed bug in x86 calls
3149    
3150        There was a bug where call instructions would mysteriously
3151        vanish. The call instruction had to be one that returned
3152        a floating point value.
3153    
3154    ----------------------------------------------------------------------
3155    Name: Lal George
3156    Date: 2001/07/19 16:36:29 EDT 2001
3157    Tag: george-20010719-simple-cells
3158    Description:
3159    
3160    I have dramatically simplified the interface for CELLS in MLRISC.
3161    
3162    In summary, the cells interface is broken up into three parts:
3163    
3164      1. CellsBasis : CELLS_BASIS
3165    
3166            CellsBasis is a top level structure and common for all
3167            architectures.  it contains the definitions of basic datatypes
3168            and utility  functions over these types.
3169    
3170      2. functor Cells() : CELLS
3171    
3172            Cells generates an interface for CELLS that incorporates the
3173            specific resources on the target architecture, such as the
3174            presence of special register classes, their number and size,
3175            and various useful substructures.
3176    
3177      3. <ARCH>CELLS
3178    
3179            e.g. SparcCells: SPARCCELLS
3180    
3181            <ARCH>CELLS usually contains additional bindings for special
3182            registers  on the architecture, such as:
3183    
3184                    val r0 : cell           (* register zero *)
3185                    val y : cell            (* Y register *)
3186                    val psr : cell          (* processor status register *)
3187                    ...
3188    
3189            The structure returned by applying the Cells functor is opened
3190            in this interface.
3191    
3192    The main implication of all this is that the datatypes for cells is
3193    split between CellsBasis and CELLS -- a fairly simple change for user
3194    code.
3195    
3196    In the old scheme the CELLS interface had a definitional binding of
3197    the form:
3198    
3199            signature CELLS = sig
3200    
3201               structure CellsBasis = CellsBasis
3202    
3203               ...
3204    
3205            end
3206    
3207    With all the sharing constraints that goes on in MLRISC, this old
3208    design  quickly leads to errors such as:
3209    
3210            "structure definition spec inside of sharing ... "
3211    
3212    
3213    and appears to require an unacceptable amount of sharing and where
3214    constraint hackery.
3215    
3216    I think this error message (the interaction of definitional specs and
3217    sharing) requires more explanation on our web page.
3218    
3219    ----------------------------------------------------------------------
3220    Name: Matthias Blume
3221    Date: 2001/07/19 15:00:00 EDT
3222    Tag: blume-20010719-libreorg
3223    Description:
3224    
3225    This update puts together a fairly extensive but straightforward change
3226    to the way the libraries that implement the interactive system are
3227    organized:
3228    
3229       The biggest change is the elimination of structure Compiler.  As a
3230       replacement for this structure, there is now a CM library
3231       (known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
3232       that exports all the substructures of the original structure Compiler
3233       directly.  So instead of saying Compiler.Foo.bar one now simply
3234       says Foo.bar.  (The CM libraries actually export a collection of
3235       structures that is richer than the collection of substructures of
3236       structure Compiler.)
3237    
3238       To make the transition smooth, there is a separate library called
3239       $smlnj/compiler/compiler.cm which puts together and exports the
3240       original structure Compiler (or at least something very close to it).
3241    
3242       There are five members of the original structure Compiler
3243       that are not exported directly but which instead became members
3244       of a new structure Backend (described by signature BACKEND).  These are:
3245       structure Profile (: PROFILE), structure Compile (: COMPILE), structure
3246       Interact (: INTERACT), structure Machine (: MACHINE), and val
3247       architecture (: string).
3248    
3249       Structure Compiler.Version has become structure CompilerVersion.
3250    
3251       Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
3252       by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
3253       or x86, respectively.
3254       Each of these exports the same frontend structures that
3255       $smlnj/compiler.cm exports.  But they do not have a structure Backend
3256       and instead export some structure <Arch>Backend where <Arch> is Alpha32,
3257       Hppa, PPC, Sparc, or X86, respectively.
3258    
3259       Library $smlnj/compiler/all.cm exports the union of the exports of
3260       $smlnj/compiler/<arch>.cm
3261    
3262       There are no structures <Arch>Compiler anymore, use
3263       $smlnj/compiler/<arch>.cm instead.
3264    
3265       Library host-compiler-0.cm is gone.  Instead, the internal library
3266       that instantiates CM is now called cm0.cm.  Selection of the host
3267       compiler (backend) is no longer done here but. (Responsibility for it
3268       now lies with $smlnj/compiler/current.cm.  This seems to be more
3269       logical.)
3270    
3271       Many individual files have been moved or renamed.  Some files have
3272       been split into multiple files, and some "dead" files have been deleted.
3273    
3274    Aside from these changes to library organization, there are also changes
3275    to the way the code itself is organized:
3276    
3277       Structure Binfile has been re-implemented in such a way that it no
3278       longer needs any knowledge of the compiler.  It exclusively deals
3279       with the details of binfile layout.  It no longer invokes the
3280       compiler (for the purpose of creating new prospective binfile
3281       content), and it no longer has any knowledge of how to interpret
3282       pickles.
3283    
3284       Structure Compile (: COMPILE) has been stripped down to the bare
3285       essentials of compilation.  It no longer deals with linking/execution.
3286       The interface has been cleaned up considerably.
3287    
3288       Utility routines for dealing with linking and execution have been
3289       moved into their own substructures.
3290    
3291       (The ultimate goal of these changes is to provide a light-weight
3292       binfile loader/linker (at least for, e.g., stable libraries) that
3293       does not require CM or the compiler to be present.)
3294    
3295    CM documentation has been updated to reflect the changes to library
3296    organization.
3297    
3298    ----------------------------------------------------------------------
3299    Name: Matthias Blume
3300    Date: 2001/07/10 17:30:00 EDT
3301    Tag: Release_110_34
3302    Description:
3303    
3304    Minor tweak to 110.34 (re-tagged):
3305    
3306      - README.html file added to CVS repository
3307      - runtime compiles properly under FreeBSD 3.X and 4.X
3308    
3309    ----------------------------------------------------------------------
3310    Name: Matthias Blume
3311    Date: 2001/07/10 17:30:00 EDT
3312    Tag: Release_110_34
3313    Description:
3314    
3315    New version number (110.34). New bootfiles.
3316    
3317    ----------------------------------------------------------------------
3318    Name: Matthias Blume
3319    Date: 2001/07/09 16:00:00 EDT
3320    Tag: blume-20010709-more-varargs
3321    Description:
3322    
3323    I changed the handling of varargs in ml-nlffigen again:
3324    The ellipsis ... will now simply be ignored (with an accompanying warning).
3325    
3326    The immediate effect is that you can actually call a varargs function
3327    from ML -- but you can't actually supply any arguments beyond the ones
3328    specified explicitly.  (For example, you can call printf with its format
3329    string, but you cannot pass additional arguments.)
3330    
3331    This behavior is only marginally more useful than the one before, but
3332    it has the advantage that a function or, more importantly, a function
3333    type never gets dropped on the floor, thus avoiding follow-up problems with
3334    other types that refer to the offending one.
3335    
3336    ----------------------------------------------------------------------
3337    Name: Matthias Blume
3338    Date: 2001/07/09 11:25:00 EDT
3339    Tag: blume-20010709-varargs
3340    Description:
3341    
3342    1. ckit-lib.cm now exports structure Error
3343    2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
3344       with a warning accompanied by a source location.  Moreover, it
3345       merely skips the offending function or type and proceeds with the
3346       rest of its work.u  As a result, one can safely feed C code containing
3347       "..." to ml-nlffigen.
3348    3. There are some internal improvements to CM, providing slightly
3349       more general string substitutions in the tools subsystem.
3350    
3351    ----------------------------------------------------------------------
3352    Name: Matthias Blume
3353    Date: 2001/06/27 15:10:00 EDT
3354    Tag: blume-20010627-concur
3355    Description:
3356    
3357    Fixed a small bug in CM's handling of parallel compilation.
3358    (You could observe the bug by Control-C-interrupting an ordinary
3359    CMB.make or CM.stabilize and then attaching some compile servers.
3360    The result was that all of a sudden the previously interrupted
3361    compilation would continue on its own.  This was because of
3362    an over-optimization: CM did not bother to clean out certain queues
3363    when no servers were attached "anyway", resulting in the contents
3364    of these queues to grab control when new servers did get attached.)
3365    
3366    There is also another minor update to the CM manual.
3367    
3368    ----------------------------------------------------------------------
3369    Name: Matthias Blume
3370    Date: 2001/06/26 16:15:00 EDT
3371    Tag: blume-20010626-cmdoc
3372    Description:
3373    
3374    Minor typo fixed in CM manual (syntax diagram for libraries).
3375    
3376    ----------------------------------------------------------------------
3377    Name: Matthias Blume
3378    Date: 2001/06/25 22:55:00 EDT
3379    Tag: blume-20010625-x86pc
3380    Description:
3381    
3382    Fixed a nasty bug in the X86 assembly code that caused signal
3383    handlers to fail (crash) randomly.
3384    
3385    ----------------------------------------------------------------------
3386    Name: Matthias Blume
3387    Date: 2001/06/25 12:05:00 EDT
3388    Tag: blume-20010625-nlffigen
3389    Description:
3390    
3391    This update fixes a number of minor bugs in ml-nlffigen as reported by
3392    Nick Carter <nbc@andrew.cmu.edu>.
3393    
3394      1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
3395      2. Default names for generated files are now derived from the name of
3396         the C file *without its directory*.  In particular, this causes generated
3397         files to be placed locally even if the C file is in some system directory.
3398      3. Default names for generated signatures and structures are also derived
3399         from the C file name without its directory.  This avoids silly things
3400         like "structure GL/GL".
3401         (Other silly names are still possible because ml-nlffigen does not do
3402          a thorough check of whether generated names are legal ML identifiers.
3403          When in doubt, use command line arguments to force particular names.)
3404    
3405    ----------------------------------------------------------------------
3406    Name: Matthias Blume
3407    Date: 2001/06/21 12:25:00 EDT
3408    Tag: blume-20010621-eXene
3409    Description:
3410    
3411    eXene now compiles and (sort of) works again.
3412    
3413    The library name (for version > 110.33) is $/eXene.cm.
3414    
3415    I also added an new example in src/eXene/examples/nbody.  See the
3416    README file there for details.
3417    
3418    ----------------------------------------------------------------------
3419    Name: Matthias Blume
3420    Date: 2001/06/20 16:40:00 EDT
3421    Tag: blume-20010620-cml
3422    Description:
3423    
3424    CML now compiles and works again.
3425    
3426    Libraries (for version > 110.33):
3427    
3428      $cml/cml.cm            Main CML library.
3429      $cml/basis.cm          CML's version of $/basis.cm.
3430      $cml/cml-internal.cm   Internal helper library.
3431      $cml/core-cml.cm       Internal helper library.
3432      $cml-lib/trace-cml.cm  Tracing facility.
3433      $cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
3434    
3435    The installer (config/install.sh) has been taught how to properly
3436    install this stuff.
3437    
3438    ----------------------------------------------------------------------
3439    Name: Matthias Blume
3440    Date: 2001/06/19 17:55:00 EDT
3441    Tag: blume-20010619-instantiate
3442    Description:
3443    
3444    This un-breaks the fix for bug 1432.
3445    (The bug was originally fixed in 110.9 but I broke it again some
3446    time after that.)
3447    
3448    ----------------------------------------------------------------------
3449    Name: Matthias Blume
3450    Date: 2001/06/19 17:25:00 EDT
3451    Tag: blume-20010619-signals
3452    Description:
3453    
3454    This should (hopefully) fix the long-standing signal handling bug.
3455    (The runtime system was constructing a continuation record with an
3456    incorrect descriptor which would cause the GC to drop data on the floor...)
3457    
3458    ----------------------------------------------------------------------
3459    Name: Matthias Blume
3460    Date: 2001/06/15 15:05:00 EDT
3461    Tag: blume-20010615-moresparc
3462    Description:
3463    
3464    Here is a short late-hour update related to Sparc c-calls:
3465    
3466     -- made handling of double-word arguments a bit smarter
3467    
3468     -- instruction selection phase tries to collapse certain clumsily
3469        constructed ML-Trees; typical example:
3470    
3471            ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
3472    
3473        This currently has no further impact on SML/NJ since mlriscGen does
3474        not seem to generate such patterns in the first place, and c-calls
3475        (which did generate them in the beginning) has meanwhile been fixed
3476        so as to avoid them as well.
3477    
3478    ----------------------------------------------------------------------
3479    Name: Matthias Blume
3480    Date: 2001/06/15 15:05:00 EDT
3481    Tag: blume-20010615-sparc
3482    Description:
3483    
3484    The purpose of this update is to provide an implementation of NLFFI
3485    on Sparc machines.
3486    
3487    Here are the changes in detail:
3488    
3489       * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
3490       the Sparc implementation of the c-calls API.
3491       * The Sparc backend of SML/NJ has been modified to uniformely use %fp
3492       for accessing the ML frame.  Thus, we have a real frame pointer and
3493       can freely modify %sp without need for an omit-frame-ptr phase.
3494       The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
3495       this case.
3496       * ml-nlffigen has been taught to produce code for different architectures
3497       and calling conventions.
3498       * In a way similar to what was done in the x86 case, the Sparc
3499       backend uses its own specific extension to mltree.  (For example,
3500       it needs to be able to generate UNIMP instructions which are part
3501       of the calling convention.)
3502       * ml-nlffi-lib was reorganized to make it more modular (in particular,
3503       to make it easier to plug in new machine- and os-dependent parts).
3504    
3505    There are some other fairly unrelated bug fixes and cleanups as well:
3506    
3507       * I further hacked the .cm files for MLRISC tools (like MDLGen) so
3508       that they properly share their libraries with existing SML/NJ libraries.
3509       * I fixed a minor cosmetic bug in CM, supressing certain spurious
3510       follow-up error messages.
3511       * Updates to CM/CMB documentation.
3512    
3513    TODO items:
3514    
3515       * MLRISC should use a different register as its asmTemp on the Sparc.
3516         (The current %o2 is a really bad choice because it is part of the
3517         calling conventions, so things might interfere in unexpected ways.)
3518    
3519    ----------------------------------------------------------------------
3520    Name: Matthias Blume
3521    Date: 2001/06/07
3522    Tag: blume-20010607-calls
3523    Description:
3524    
3525    A number of internal changes related to C calls and calling conventions:
3526    
3527    1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
3528       number of bytes popped implicitly (by the callee).  In most cases
3529       this field is 0 but on x86/win32 it is some non-zero value.  This
3530       is information provided for the benefit of the "omit-frameptr" pass.
3531    2. The CALL instruction on the x86 carries a similar "pops" field.
3532       The instruction selection phase copies its value from the ML-Tree
3533       CALL statement.
3534    3. On all other architectures, the instruction selection phase checks
3535       whether "pops=0" and complains if not.
3536    4. The c-calls implementation for x86 now accepts two calling conventions:
3537       "ccall" and "stdcall".  When "ccall" is selected, the caller cleans
3538       up after the call and pops is set to 0.  For "stdcall", the caller
3539       does nothing, leaving the cleanup to the callee; pops is set to
3540       the number of bytes that were pushed onto the stack.
3541    5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
3542       distinguish between "ccall" and "stdcall".
3543    6. The UNIMP instruction has been added to the supported Sparc instruction
3544       set. (This is needed for implementing the official C calling convention
3545       on this architecture.)
3546    7. I fixed some of the .cm files under src/MLRISC/Tools to make them
3547       work with the latest CM.
3548    
3549    ----------------------------------------------------------------------
3550    Name: Matthias Blume
3551    Date: 2001/06/05 15:10:00 EDT
3552    Tag: blume-20010605-cm-index
3553    Description:
3554    
3555    0. The "lambdasplit" parameter for class "sml" in CM has been documented.
3556    
3557    1. CM can now generate "index files".  These are human-readable files
3558       that list on a per-.cm-file basis each toplevel symbol defined or
3559       imported.  The location of the index file for
3560       <p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
3561       To enable index-file generation, set CM.Control.generate_index to true
3562       or export an environment-symbol: export CM_GENERATE_INDEX=true.
3563    
3564       The CM manual has been updated accordingly.
3565    
3566    2. I made some slight modifications to the c-calls API in MLRISC.
3567    
3568         a) There is now a callback to support saving/restoring of
3569            dedicated but caller-save registers around the actual call
3570            instruction.
3571         b) One can optionally specify a comment-annotation for the
3572            call instruction.
3573    
3574    3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
3575       (For example, the comment annotation shows the C prototype of
3576        the function being called.)
3577    
3578    ----------------------------------------------------------------------
3579    Name: Matthias Blume
3580    Date: 2001/06/01 13:30:00 EDT
3581    Tag: blume-20010601-nlffi-cleanup
3582    Description:
3583    
3584    This is mostly a cleanup of MLFFI stuff:
3585    
3586       - some signature files have been put into a more exposed place
3587       - the ugly 'f type parameter is gone (simplifies types tremendously!)
3588       - ml-nlffigen changed accordingly
3589       - tutorial updated
3590    
3591    Other changes:
3592    
3593       - author's affiliation in CM manual(s) updated
3594       - some more recognized keywords added to Allen's sml.sty
3595    
3596    ----------------------------------------------------------------------
3597    Name: Matthias Blume
3598    Date: 2001/05/25 15:30:00 EDT
3599    Tag: blume-20010525-iptr
3600    Description:
3601    
3602      - put the official 110.33-README (as it appears on the ftp server) under
3603        CVS
3604      - fixed a small bug related to incomplete pointer types in
3605        ml-nlffigen
3606      - small cosmetic change to the ml-nlffi-lib's "arr" type constructor
3607        (it does not need the 'f type parameter)
3608    
3609    ----------------------------------------------------------------------
3610    Name: Matthias Blume