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 749, Fri Dec 8 15:16:36 2000 UTC revision 1427, Tue Nov 18 23:08:15 2003 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: 2003/11/18 17:10 CST
18    Tag: blume-20031118-basis-fiddle
19    Description:
20    
21    - changed Timer interface to what might become the spec
22    - POSIX_FLAGS -> BIT_FLAGS according to spec
23    - some other minor discrepancies wrt. spec eliminated
24    
25    ----------------------------------------------------------------------
26    Name: Matthias Blume (blume (at) tti - c (dot) org)
27    Date: 2003/11/06 12:00:00 CST
28    Tag: Release_110_44
29    Description:
30    
31    New working version (110.44). New bootfiles.
32    
33    ----------------------------------------------------------------------
34    Name: Matthias Blume (blume (at) tti - c (dot) org)
35    Date: 2003/11/04 11:50:00 CST
36    Tag: blume-20031104-move-libraries
37    Description:
38    
39    Eliminated the "dont_move_libraries" directive in config/targets.
40    (The mechanism was broken and could not be fixed easily.  Moreover,
41    there does not seem to be any reason not to move all libraries into
42    lib during installation.  I originally implemented this directive as a
43    backward-compatibility feature when I first introduced the new CM.
44    Now that things have been stable for a long time and going back to the
45    old CM is not an option, there is no reason to keep it around.)
46    
47    ----------------------------------------------------------------------
48    Name: Matthias Blume (blume (at) tti - c (dot) org)
49    Date: 2003/11/03 16:00:00 CST
50    Tag: blume-20031103-installdir
51    Description:
52    
53    Made installer honor INSTALLDIR variable again.  (Thanks to Chris
54    Richards for pointing out the problem and providing the solution.)
55    
56    ----------------------------------------------------------------------
57    Name: Matthias Blume (blume (at) tti - c (dot) org)
58    Date: 2003/10/01 17:05:00 CDT
59    Tag: blume-20031001-lal-mlrisc
60    Description:
61    
62    MLRISC bug fix from Lal.
63    
64    ----------------------------------------------------------------------
65    Name: Matthias Blume (blume (at) tti - c (dot) org)
66    Date: 2003/09/30 16:10:00 CDT
67    Tag: blume-20030930-primio-bat
68    Description:
69    
70    1. Added openVector, nullRd, and nullWr to PRIM_IO.
71    2. Improved .bat files (for Win32 port) to make things work under Win95.
72       (thanks to Aaron S. Hawley for this one)
73    
74    ----------------------------------------------------------------------
75    Name: Matthias Blume (blume (at) tti - c (dot) org)
76    Date: 2003/09/26 16:05:00 CDT
77    Tag: blume-20030926-wrappriv
78    Description:
79    
80    Added missing wrapper for privilege "primitive" in $smlnj/viscomp/core.cm.
81    
82    ----------------------------------------------------------------------
83    Name: Matthias Blume (blume (at) tti - c (dot) org)
84    Date: 2003/09/26 15:00:00 CDT
85    Tag: blume-20030926-110_43_3
86    Description:
87    
88    - additional cleanup
89    - version number bump, NEW BOOTFILES
90    
91    ----------------------------------------------------------------------
92    Name: Matthias Blume (blume (at) tti - c (dot) org)
93    Date: 2003/09/26 12:00:00 CDT
94    Tag: blume-20030926-ppautoload
95    Description:
96    
97    I modified the read-eval-print loop so that the autoloader gets
98    invoked whenever the prettyprinter tries to look up a symbol that
99    is not currently defined in the toplevel environment but which
100    appears in CM's autoload registry.  As a result, we see far fewer of
101    those ?.Foo.Bar.xxx names in the prettyprinter's output.
102    
103    In addition to this I tried to clean up some pieces of the Basis
104    implementation (e.g., Socket, Word8Array) in order to prevent other
105    instances of these ?.Foo.Bar.xxx names from being printed.
106    
107    The mechanism that picks names for types still needs some work, though.
108    (Right now it seems that if there is a type A.t which is defined to
109    be B.u, but B is unavailable at toplevel, then A.t gets printed as
110    "?.B.u" although the perhaps more sensible solution would be to use
111    "A.t" in this case.  In other words, the prettyprinter should follow
112    a chain of DEFtycs not farther than there are corresponding toplevel
113    names in the current environment.)
114    
115    ----------------------------------------------------------------------
116    Name: Matthias Blume (blume (at) tti - c (dot) org)
117    Date: 2003/09/24 16:31:00 CDT
118    Tag: blume-20030924-installer
119    Description:
120    
121    Another installer tweak:  All the ML code for the installer is now
122    compiled during CMB.make and put into a little library called
123    $smlnj/installer.cm.  The installation then simply invokes
124    
125       sml -m $smlnj/installer.cm
126    
127    and everything happens automagically.
128    
129    Win32:  ML code senses value of environment variable SMLNJ_HOME.
130    Unix:   ML code senses values of environment variables ROOT, CONFIGDIR,
131            and BINDIR.
132    
133    The new scheme guarantees that the ML code responsible for the installation
134    is in sync with the APIs of the main system.  Also, the installer is
135    somewhat faster because the installer script is precompiled.
136    
137    ----------------------------------------------------------------------
138    Name: Matthias Blume (blume (at) tti - c (dot) org)
139    Date: 2003/09/24 15:35:00 CDT
140    Tag: blume-20030924-synsock
141    Description:
142    
143    Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
144    but excludes all non-blocking operations.  Defined SOCKET (in Basis)
145    and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
146    implementations of non-blocking operations from CML's Socket
147    structure.
148    
149    ----------------------------------------------------------------------
150    Name: Matthias Blume (blume (at) tti - c (dot) org)
151    Date: 2003/09/24 15:10:05 CDT
152    Tag: blume-20030924-sockets
153    Description:
154    
155    1. Fixed SOCKET API and implementation to match Basis spec.
156       This required changing the internal representation of sockets to one
157       that remembers (for each socket file descriptor) whether it is currently
158       blocking or non-blocking.  This state is maintained lazily (i.e., a system
159       call is made only if the state actually needs to change).
160    
161    2. OS-specific details of sockets were moved into separate files, thus
162       making it possible to unify the bulk of the socket implementations
163       between Unix and Win32.
164    
165    3. CML's socket API changed accordingly.
166       (Note that we need to remove non-blocking functions from this API
167       since they are redundant in the case of CML!)
168    
169    4. CML's socket implementation now makes use of non-blocking functions
170       provided by Basis, thus removing all OS-dependent code from this part
171       of CML.
172    
173    5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
174    
175    ----------------------------------------------------------------------
176    Name: Matthias Blume (blume (at) tti - c (dot) org)
177    Date: 2003/09/22 12:10:00 CDT
178    Tag: blume-20030922-110_43_2
179    Description:
180    
181    Made a new interim version and bootfiles for developer's bootstrapping
182    convenience.
183    
184       110.43.2 -- NEW BOOTFILES
185    
186    ----------------------------------------------------------------------
187    Name: Matthias Blume (blume (at) tti - c (dot) org)
188    Date: 2003/09/19 15:55:00 CDT
189    Tag: blume-20030919-cmdir
190    Description:
191    
192    1. new-install.sh -> install.sh
193    2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
194    3. tweaked installer so that another name instead of .cm can be chosen
195       at install time (by setting the CM_DIR_ARC environment variable
196       during installation); once installation is complete, the name is
197       fixed
198    
199    ----------------------------------------------------------------------
200    Name: Matthias Blume (blume (at) tti - c (dot) org)
201    Date: 2003/09/18 16:00:00 CDT
202    Tag: blume-20030918-110_43_1
203    Description:
204    
205    Made a new interim version and bootfiles for developer's bootstrapping
206    convenience.
207    
208       110.43.1 -- NEW BOOTFILES
209    
210    ----------------------------------------------------------------------
211    Name: Matthias Blume (blume (at) tti - c (dot) org)
212    Date: 2003/09/18 15:20:00 CDT
213    Tag: blume-20030918-misc
214    Description:
215    
216    1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
217       this seems to be controversial at the moment) and used those in
218       Posix.ProcEnv.times.
219    
220    2. Added Time.{from,to}Nanoseconds to Time.
221    
222    3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
223       For example, fromLargeInt never needs to look at more than 3 "big
224       digits" to get its 53 bits of precision.
225    
226    ----------------------------------------------------------------------
227    Name: Matthias Blume (blume (at) tti - c (dot) org)
228    Date: 2003/09/17 16:30:00 CDT
229    Tag: blume-20030917-real32-slices
230    Description:
231    
232    Added an entry to the primitive environment
233    (compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
234    added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
235    
236    Removed some of the "magic" constants in real64.sml and replaced them
237    with code that generates these values from their corresponding
238    integer counterparts.
239    
240    Made all(?) the slice-related changes to the Basis and made everything
241    compile again...
242    
243    ----------------------------------------------------------------------
244    Name: Matthias Blume (blume (at) tti - c (dot) org)
245    Date: 2003/09/15 17:45:00 CDT
246    Tag: blume-20030915-rbase
247    Description:
248    
249    Fixed bug in Real.fromLargeInt.
250    
251    ----------------------------------------------------------------------
252    Name: Matthias Blume (blume (at) tti - c (dot) org)
253    Date: 2003/09/13 18:11:00 CDT
254    Tag: blume-20030913-libinstall
255    Description:
256    
257    Minor bugfix in config/libinstall (set anchor with path to
258    standalone tool after installing it, otherwise libraries that
259    need ml-lex or ml-yacc won't compile the first time the installer
260    runs).
261    
262    ----------------------------------------------------------------------
263    Name: Matthias Blume (blume (at) tti - c (dot) org)
264    Date: 2003/09/12 11:45:00 CDT
265    Tag: blume-20030912-various
266    Description:
267    
268     - fixed bug in Real.toLargeInt
269     - fixed bug in Posix.ProcEnv.times
270     - changed inputLine functions to return an option
271     - minor installer improvements / bugfixes
272     - changed default @SMLalloc parameter for x86/celeron to 64k
273    
274    ----------------------------------------------------------------------
275    Name: Matthias Blume (blume (at) tti - c (dot) org)
276    Date: 2003/09/09 22:00:00 CDT
277    Tag: Release_110_43
278    Description:
279    
280        New working release 110.43.  New bootfiles.
281    
282    ----------------------------------------------------------------------
283    Name: Matthias Blume (blume (at) tti - c (dot) org)
284    Date: 2003/09/09 19:20:00 CDT
285    Tag: blume-20030909-installer
286    Description:
287    
288    Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
289    Modified config/install.bat to take advantage of it.  Also modified
290    config/install.sh (and called it config/new-install.sh) to take advantage
291    of it on Unix systems.  (The SML code is (supposed to be) platform-
292    independent.)
293    
294    The installer can now install everything under Win32
295    as well as under *nix as long as it compiles.
296    
297    Other changes:
298    
299      - made CML compile again under Win32
300      - made eXene compile under Win32 (by providing a fake structure UnixSock
301        and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
302      - fixed a bug in nowhere: it assumed that type OS.Process.status is the
303        same as type int;  under Win32 it isn't
304      - fixed some slice-related problems in the win32-specific parts of CML
305      - added a functor argument "sameVol" to os-path-fn.sml in the Basis
306        (under Win32, the volume name is case-insensitive, and the
307        OS.Path code compares volume names for equality)
308    
309    ----------------------------------------------------------------------
310    Name: Matthias Blume (blume (at) tti - c (dot) org)
311    Date: 2003/09/08 11:55:00 CDT
312    Tag: blume-20030908-fullpath
313    Description:
314    
315    Made Win32 version of OS.FileSys.fullPath return current directory
316    when given an empty string.  This is what the spec says, and incidentally,
317    CM depends on it.  (CM otherwise goes into an infinite loop in certain
318    cases when presented with the name of a non-existing .cm file.)
319    
320    ----------------------------------------------------------------------
321    Name: Matthias Blume (blume (at) tti - c (dot) org)
322    Date: 2003/09/04 16:30:00 CDT
323    Tag: blume-20030905-slices-etc
324    Description:
325    
326      1. Changed interface to vectors and arrays in Basis to match
327         (draft) Basis spec.
328      2. Added signatures and implementations of slices according to
329         Basis spec.
330      3. Edited source code throughout the system to make it compile again
331         under 1. and 2.  (In some cases code had to be added to have it
332         match the new signatures.)
333      4. MLRISC should be backward-compatible: the copies of the originals
334         of files that needed to change under 3. were retained, the .cm files
335         check the compiler version number and use old versions when
336         appropriate.
337      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
338         dirstream -> string option (in accordance with Basis spec).
339      6. When generating code that counts lines, ml-lex used function
340         CharVector.foldli, taking advantage of its old interface.
341         This has been replaced with the corresponding code from
342         CharVectorSlice.  (html-lex must be re-lexed!)
343      7. BitArray in smlnj-lib/Util has been extended/modified to match the
344         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
345      8. Removed temporary additions (fromInternal, toInternal) from the
346         (now obsolete) IntInf in smlnj-lib/Util.
347      9. Cleaned up structure Byte.
348     10. Added localOffset, scan, and fromString to Date (according to spec).
349         Cleaned/corrected implementation of Date.
350         (Still need to check for correctness; implement better canonicalizeDate.)
351     11. Added "scan" to signature IEEE_REAL.
352     12. Some improvements to IntInf [in particular: efficiency-hack for
353         mod and rem when second operand is 2 (for parity checks).]
354     13. Changed representation of type Time.time, using a single IntInf.int
355         value counting microseconds.  This considerably simplified the
356         implementation of structure Time.  We now support negative time
357         values; scan and fromString handle signs.
358     14. Functor PrimIO now takes two additional arguments (VectorSlice and
359         ArraySlice).
360    
361    ----------------------------------------------------------------------
362    Name: Matthias Blume (blume (at) tti - c (dot) org)
363    Date: 2003/08/28 17:00:00 CDT
364    Tag: blume-20030828-intinf
365    Description:
366    
367    This is a major update which comes with a version number bump
368    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
369    and an implementation of IntInf in the Basis.
370    
371    There are a fairly large number of related changes and updates throughout
372    the system:
373    
374    Basis:
375            - Implemented IntInf.
376            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
377            - Added some missing Real64 operations, most notably Real.toLargeInt.
378            - Added FixedInt as a synonym for Int32.
379    
380    compiler:
381            * Added support for a built-in intinf type.
382              - literals
383              - pattern matching
384              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
385              - overloading on literals and operations
386    
387            This required adding a primitive type intinf, some additional
388            primops, and implementations for several non-trivial intinf
389            operations in Core.  (The intinf type is completely abstract
390            to the compiler; all operations get delegated back to the Core.)
391    
392            * Intinf equality is handled by polyequal.  However, the compiler
393            does not print its usual warning in this case (since polyequal
394            is the right thing to do there).
395    
396            * Improved the organization of structure InlineT.
397    
398            * A word about conversion primops:
399            If conversions involving intinf do not cancel out during
400            CPS contract, then the compiler must insert calls to Core functions.
401            Since all core access must be resolved already during the FLINT
402            translate phase, it would be too late a the time of CPS contract
403            to add new Core calls.  For this reason, conversion primops
404            for intinf carry two arguments:  1. the numeric argument that
405            they are supposed to convert, and 2. the Core function that
406            can help with this conversion if necessary.  If CPS contract
407            eliminates a primop, then the associated Core function becomes
408            dead and goes away.  Intinf conversion primops that do not get
409            eliminated by CPS contract get rewritten into calls of their
410            core functions by a separate, new phase.
411    
412    interactive system:
413            - Control.Print.intinfDepth controls max length of intinf constants
414            being printed. (Analogous to Control.Print.stringDepth.)
415            - Cleanup in printutil and pputil: got rid of unused stuff and
416            duplicates; replaced some of the code with code that makes better
417            use of library functionality.
418    
419    CM:
420            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
421            are no longer silent.
422    
423    CKIT:
424            Fixed mismatched uses of Int32 and LargeInt.  I always decided
425            in favor of LargeInt -- which is now the same as IntInf.
426            CKIT-knowledgable people should check whether this is what's
427            intended and otherwise change things back to using Int32 or
428            FixedInt.
429    
430    Throughout the code:
431            Started using IntInf.int literals and built-in operations
432            (e.g., comparison with 0) where this seems appropriate.
433    
434    
435    ----------------------------------------------------------------------
436    Name: Dave MacQueen (dbm@cs.uchicago.edu)
437    Date: 2003/08/13 11:36:00 CDT
438    Tag: dbm-20030813-mcz-merge1
439    Description:
440    
441    Merging changes from the mcz-branch development branch into trunk.
442    These changes involve replacement of the emulated old prettyprinter
443    interface with direct use of the SML/NJ Lib PP library, and fixing
444    of a couple of bugs (895, 1186) relating to error messages.  A new
445    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
446    has been added.
447    
448    ----------------------------------------------------------------------
449    Name: Matthias Blume (blume (at) tti - c (dot) org)
450    Date: 2003/08/11 15:45:00 CDT
451    Tag: blume-20030811-windows
452    Description:
453    
454    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
455    
456        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
457    
458    This patch restores SML/NJ's ability to run under win32.  There are a
459    number of changes, including fixes for several bugs that had gone
460    unnoticed until now:
461    
462      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
463        trying to run the win32 version from within cygwin.)
464    
465      - There are a number of simple .bat scripts that substitute for their
466        corresponding Unix shell-scripts.  (See below.)
467    
468      - The internals of ml-build have been modified slightly.  The main
469        difference is that instead of calling ".link-sml" (or link-sml.bat)
470        using OS.Process.system, the ML process delegates this task back
471        to the script.  Otherwise problems arise in mixed environments such
472        as Cygwin where scripts look and work like Unix scripts, but
473        where OS.Process.system cannot run them.
474    
475      - In CM, the srcpath pickler used native pathname syntax -- which
476        is incorrect in the case of cross-compilation.  The new pickle format
477        is independent of platform-specific naming conventions.
478    
479      - Path configuration files (such as lib/pathconfig) can now choose
480        between native and standard syntax.  Placing a line of the form
481    
482            standard!
483    
484        into the file causes all subsequent paths to be interpreted using
485        CM standard pathname syntax (= Unix conventions); a line
486    
487            native!
488    
489        switches back to native style.  This was needed so that
490        path config files can be written portably, see src/system/pathconfig.
491    
492      - Runtime system:
493    
494          - win32-filesys.c:  get_file_time and set_file_time now
495            access modification time, not creation time.
496    
497          - I/O code made aware of new array representation.
498    
499          - Bug fixes in X86.prim.masm.
500    
501       - src/system/makeml made aware of win32. (For use under cygwin
502            and other Unix-environments for windows.)
503    
504       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
505         which caused BinIO.inputAll to fail consistently.
506    
507    .bat scripts:
508    
509       Windows .bat scripts assume that SMLNJ_HOME is defined.
510    
511       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
512         applications (sml, ml-yacc, ml-lex).
513       - ml-build.bat: analogous to ml-build.
514       - config\install.bat:  Analogous to config/install.sh.  This requires
515         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
516         (nmake etc. must be on the path, and vcvars32 must have been run.)
517         Moreover, sources for ml-lex and ml-yacc need to exist under src,
518         and the bootfile hierarchy must have been unpacked under
519         sml.boot.x86-win32.
520         The script is very primitive and does a poor job at error checking.
521         It only installs the base system, ml-lex, and ml-yacc.  No other
522         libraries are being installed (i.e., you get only those that
523         are part of the compiler.)
524       - link-sml.bat:  analogous to .link-sml, but not currently used
525    
526    Unrelated bug fixes:
527    
528      - ml-nlffigen now exports structures ST_* corresponding to incomplete
529        types.
530      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
531        otherwise.)
532    
533    ----------------------------------------------------------------------
534    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
535    Date: 2003/06/17
536    Tag: macqueen-20030617-bug895
537    Description:
538    
539    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
540    Tag will be used for new development branch (mcz-branch) for
541    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
542    
543    ----------------------------------------------------------------------
544    Name: Matthias Blume (blume (at) tti - c (dot) org)
545    Date: 2003/05/27 16:55:00 CDT
546    Tag: blume-20030527-polyeq
547    Description:
548    
549    Tried to eliminated most cases of polymorphic equality.
550    
551    ----------------------------------------------------------------------
552    Name: Matthias Blume (blume (at) tti - c (dot) org)
553    Date: 2003/05/21 17:45:00 CDT
554    Tag: blume-20030517-complete
555    Description:
556    
557    Two changes:
558    
559      1. Added a flag for controlling whether non-exhaustive bindings will
560         be treated as errors (default is false).
561      2. Cleaned up the *entire* source tree so that CMB.make goes through
562         without a single non-exhaustive match- or bind warning.
563    
564    ----------------------------------------------------------------------
565    Name: Matthias Blume (blume (at) tti - c (dot) org)
566    Date: 2003/05/17 10:20:00 CDT
567    Tag: blume-20030517-absyn
568    Description:
569    
570    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
571    
572       This mainly affects the quality of error messages.  However, some
573       of the code is now more straightforward than before. (Treatment of
574       the above four constructs in translate.sml is much simpler than
575       the "macro-expansion" that was going on before.  Plus, the mach-
576       compiler no longer gets invoked just to be able to compile an
577       if-expression.)
578    
579    2. The ErrorMsg.Error exception is now caught and absorbed by the
580       interactive loop.
581    
582    ----------------------------------------------------------------------
583    Name: Allen Leung
584    Date: 2003/05/16 13:05:00 CDT
585    Tag: leunga-20030516-cygwin-runtime
586    Description:
587    
588        Ported the runtime system to cygwin, which uses the unix
589    x86-unix bin files.  Missing/buggy features:
590    
591       o getnetbyname, getnetbyaddr: these functions seem to be missing in
592         the Cygwin library.
593       o Ctrl-C handling may be flaky.
594       o Windows system calls and Windows I/O are not supported.
595    
596       A new set of binfiles is located at:
597    
598         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
599    
600       This is only needed for bootstrapping the cygwin version of smlnj.
601       Other x86 versions can use the existing binfiles.
602    
603    ----------------------------------------------------------------------
604    Name: Matthias Blume
605    Date: 2003/04/08 15:42:00 CDT
606    Tag: blume-20030408-listpair
607    Description:
608    
609    1. Added a target 'mlrisc' to installer.
610    
611    2. Added missing elements to structure ListPair.
612    
613    ----------------------------------------------------------------------
614    Name: Allen Leung
615    Date: 2003/01/07 10:40:00 EST
616    Tag: leunga-20030107-int-rem
617    Description:
618    
619    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
620    The arguments to the SUBL instruction were swapped.
621    
622    ----------------------------------------------------------------------
623    Name: Matthias Blume
624    Date: 2002/12/12 16:25:00 EST
625    Tag: blume-20021212-risc-ra
626    Description:
627    
628    Fixed a serious bug in the rewrite code for FP spilling/reloading that
629    sent the RA into an infinite loop when floating point registers get
630    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
631    110.37 and 110.38.)
632    There was another set of potential problems related to the handling of
633    MLRISC annotations (but those did not yet cause real problems, apparently).
634    
635    ----------------------------------------------------------------------
636    Name: Matthias Blume
637    Date: 2002/12/06 22:40:00 EST
638    Tag: blume-20021206-cm-fileid
639    Description:
640    
641    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
642    This fixes the problem of CM getting confused by files that suddenly
643    change their identity (e.g., by getting unlinked and recreated by some
644    text editor such as vi).  There might be a better/cheaper/cleaner way
645    of doing this, but for now this will have to do.
646    
647    ----------------------------------------------------------------------
648    Name: Matthias Blume
649    Date: 2002/10/28 09:50:00 EST
650    Tag: blume-20021028-typecheck
651    Description:
652    
653    Exported structure Typecheck from $smlnj/viscomp/core.cm.
654    
655    ----------------------------------------------------------------------
656    Name: Matthias Blume
657    Date: 2002/10/17 09:10:00 EDT
658    Tag: Release_110_42
659    Description:
660    
661    In good old tradition, there has been a slight hiccup so that we have
662    to patch 110.42 after the fact.   The old release tag has been replaced
663    (see below).
664    
665    The change solves a problem with two competing approaches the
666    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
667    each other's way.
668    
669    This change only affects the runtime system code and the installer script.
670    (No new bootfiles.)
671    
672    ----------------------------------------------------------------------
673    Name: Matthias Blume
674    Date: 2002/10/16 12:00:00 EDT
675    Tag: Release_110_42_removed
676    Description:
677    
678    New working release.  New bootfiles.
679    
680    ----------------------------------------------------------------------
681    Name: Matthias Blume
682    Date: 2002/10/10 13:10:00 EDT
683    Tag: blume-20021010-ppc-divs
684    Description:
685    
686    The mltree operator DIVS must be implemented with an overflow check on
687    the PPC because the hardware indicates divide-by-zero using "overflow" as
688    well.
689    
690    ----------------------------------------------------------------------
691    Name: Matthias Blume
692    Date: 2002/07/23 11:45:00
693    Tag: blume-20020723-smlnj-home
694    Description:
695    
696    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
697    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
698    CM_PATHCONFIG is also set), the path configuration file is assumed
699    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
700    the entire tree to some other place and everything will "just work".
701    
702    (Companion commands such as ml-build and ml-makedepend also sense this
703    variable.)
704    
705    ----------------------------------------------------------------------
706    Name: Matthias Blume
707    Date: 2002/07/12 21:19:00 EDT
708    Tag: blume-20020712-liveness
709    Description:
710    
711    Exported two useful "step" functions from liveness module (MLRISC).
712    
713    ----------------------------------------------------------------------
714    Name: Matthias Blume
715    Date: 2002/07/05 16:00 EDT
716    Tag: Release_110_41
717    Description:
718    
719    New working release.  New bootfiles.
720    
721    ----------------------------------------------------------------------
722    Name: Matthias Blume
723    Date: 2002/07/05 10:25:00 EDT
724    Tag: blume-20020705-btimp
725    Description:
726    
727    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
728    other clients can set up backtracing support.
729    
730    ----------------------------------------------------------------------
731    Name: Matthias Blume
732    Date: 2002/06/25 17:23:00 EDT
733    Tag: blume-20020625-fpmax
734    Description:
735    
736    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
737    case.  (The sense of the isNaN test was reversed -- which made min and
738    max always return their first argument.)
739    
740    ----------------------------------------------------------------------
741    Name: Matthias Blume
742    Date: 2002/06/11
743    Tag: blume-20020611-unixpath
744    Description:
745    
746    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
747    
748    ----------------------------------------------------------------------
749    Name: Matthias Blume
750    Date: 2002/06/10 16:35:00 EDT
751    Tag: blume-20020610-ieeereal
752    Description:
753    
754    I back-ported my implementation of IEEEReal.fromString from the
755    idlbasis-devel branch so that we can test it.
756    
757    Another small change is that ppDec tries to give more information
758    than just "<sig>" in the case of functors.  However, this code is
759    broken in some mysterious way if the functor's body's signature
760    has not been declared by ascription but gets inferred from the
761    implementation.  This needs fixing...
762    
763    ----------------------------------------------------------------------
764    Name: Matthias Blume
765    Date: 2002/05/31
766    Tag: blume-20020531-btrace-mode
767    Description:
768    
769    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
770    the wayside when I switched over to using Controls everywhere.)
771    
772    ----------------------------------------------------------------------
773    Name: Lal George
774    Date: 2002/05/23 12:21:40 EDT
775    Tag: george-20020523-visual-labels
776    Description:
777    
778    Labels are now displayed in the graphical output to make
779    the fall-through and target blocks obvious.
780    
781    ----------------------------------------------------------------------
782    Name: Matthias Blume
783    Date: 2002/05/22 11:03:00 EDT
784    Tag: blume-20020522-shrink
785    Description:
786    
787    John tweaked yesterday's fix for 1131 to handle an out-of-memory
788    situation that comes up when allocating huge arrays.
789    
790    ----------------------------------------------------------------------
791    Name: Matthias Blume
792    Date: 2002/05/21 16:00:00 EDT
793    Tag: Release_110_40
794    Description:
795    
796    New working release (110.40).  New bootfiles.
797    
798    [Also: John Reppy fixed GC bug 1131.]
799    
800    ----------------------------------------------------------------------
801    Name: Matthias Blume
802    Date: 2002/05/21 12:35:00 EDT
803    Tag: blume-20020521-cmdoc
804    Description:
805    
806    CM documentation update.
807    
808    ----------------------------------------------------------------------
809    Name: Matthias Blume
810    Date: 2002/05/21 10:55:00 EDT
811    Tag: blume-20020521-misc
812    Description:
813    
814    - John tweaked runtime to be silent on heap export (except when
815      GC messages are on).
816    - I added a few more things (cross-compiling versions of CMB) to
817      config/preloads (as suggestions).
818    
819    ----------------------------------------------------------------------
820    Name: Matthias Blume
821    Date: 2002/05/20 22:25:00 EDT
822    Tag: blume-20020520-controls
823    Description:
824    
825    - Added ControlUtil structure to control-lib.cm.
826    - Use it throughout.
827    - Used Controls facility to define MLRISC controls (as opposed to
828      registering MLRISC control ref cells with Controls after the fact)
829    - Fixed messed-up controls priorities.
830    
831    * Removed again all the stuff from config/preloads that one wouldn't
832      be able to preload at the time the initial heap image is built.
833      (Many libraries, e.g., CML, do not exist yet at this time.  The
834       only libraries that can be preloaded via config/preloads are those
835       that come bundled with the bootfiles.)
836    
837    ----------------------------------------------------------------------
838    Name: Matthias Blume
839    Date: 2002/05/20 10:59:00 EDT
840    Tag: blume-20020520-preloads
841    Description:
842    
843    Added a lot of commented-out suggestions for things to be included
844    in config/preloads.
845    
846    ----------------------------------------------------------------------
847    Name: Allen Leung
848    Date: 2002/05/18 14:20:00 EDT
849    Tag: leunga-20020518-mdl
850    Description:
851    
852    o Made the mdl tool stuff compile and run again.
853    o I've disabled all the stuff that depends on RTL specifications; they
854      are all badly broken anyway.
855    
856    ----------------------------------------------------------------------
857    Name: Matthias Blume
858    Date: 2002/05/17 16:49:00 EDT
859    Tag: blume-20020517-controls
860    Description:
861    
862    0. John Reppy made several modifications to the SML/NJ library.
863       In particular, there is a shiny new controls-lib.cm.
864    
865    1. Pushed new controls interface through compiler so that everything
866       compiles again.
867    
868    2. Added FormatComb and FORMAT_COMB to the CML version of the
869       SML/NJ library (so that CML compiles again).
870    
871    3. Modified init scripts because XXX_DEFAULT environment variables
872       are no longer with us.  (Boot-time initialization is now done
873       using the same environment variables that are also used for
874       startup-time initialization of controls.)
875    
876    ----------------------------------------------------------------------
877    Name: Lal George
878    Date: 2002/05/15 09:20:10 EDT
879    Tag: george-20020515-pseudo-op-decls
880    Description:
881    
882            All pseudo-ops emitted before the first segment declaration
883    such as TEXT, DATA, and BSS directives are assumed to be global
884    declarations and are emitted first in the assembly file. This is
885    useful in a number of situations where one has pseudo-ops that are not
886    specific to any segment, and also works around the constraint that one
887    cannot have client pseudo-ops in the TEXT segment.
888    
889            Because no segment is associated with these declarations it is
890    an error to allocate any space or objects before the first segment
891    directive and an exception will be raised. However, we cannot make
892    this check for client pseudo-ops.
893    
894            These top level declarations are a field in the CFG graph_info.
895    In theory you can continue to add to this field after the CFG has been
896    built -- provided you know what you are doing;-)
897    
898    ----------------------------------------------------------------------
899    Name: Matthias Blume
900    Date: 2002/05/13 16:40:00 EDT
901    Tag: blume-20020513-pp-etc
902    Description:
903    
904    A few minor bugfixes:
905    
906      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
907        (Bogus printouts for redefined bindings still occur.  Compiler
908        bug should no longer occur now.  We need to redo the prettyprinter
909        from scratch.)
910    
911      - CM pathname printer now also adds escape sequences for ( and )
912    
913      - commend and docu fixes for ml-nlffi
914    
915    ----------------------------------------------------------------------
916    Name: Matthias Blume
917    Date: 2002/05/10 16:40:00 EDT
918    Tag: blume-20020510-erg-textio
919    Description:
920    
921    Applied the following bugfix provided by Emden Gansner:
922    
923        Output is corrupted when outputSubstr is used rather than output.
924    
925        The problem occurs when a substring
926    
927            ss = (s, dataStart, dataLen)
928    
929        where dataStart > 0, fills a stream buffer with avail bytes left.
930        avail bytes of s, starting at index dataStart, are copied into the
931        buffer, the buffer is flushed, and then the remaining dataLen-avail
932        bytes of ss are copied into the beginning of the buffer. Instead of
933        starting this copy at index dataStart+avail in s, the current code
934        starts the copy at index avail.
935    
936        Fix:
937        In text-io-fn.sml, change line 695 from
938             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
939        to
940             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
941    
942    ----------------------------------------------------------------------
943    Name: Matthias Blume
944    Date: 2002/04/12 13:55:00 EDT
945    Tag: blume-20020412-assyntax
946    Description:
947    
948    1. Grabbed newer assyntax.h from the XFree86 project.
949    2. Fiddled with how to compile X86.prim.asm without warnings.
950    3. (Very) Minor cleanup in CM.
951    
952    ----------------------------------------------------------------------
953    Name: Matthias Blume
954    Date: 2002/04/01 (no joke!) 17:07:00 EST
955    Tag: blume-20020401-x86div
956    Description:
957    
958    Added full support for div/mod/rem/quot on the x86, using the machine
959    instruction's two results (without clumsily recomputing the remainder)
960    directly where appropriate.
961    
962    Some more extensive power-of-two support was added to the x86 instruction
963    selector (avoiding expensive divs, mods, and muls where they can be
964    replaced with cheaper shifts and masks).  However, this sort of thing
965    ought to be done earlier, e.g., within the CPS optimizer so that
966    all architectures benefit from it.
967    
968    The compiler compiles to a fixed point, but changes might be somewhat
969    fragile nevertheless.  Please, report any strange things that you might
970    see wrt. div/mod/quot/rem...
971    
972    ----------------------------------------------------------------------
973    Name: Matthias Blume
974    Date: 2002/03/29 17:22:00
975    Tag: blume-20020329-div
976    Description:
977    
978    Fixed my broken div/mod logic.  Unfortunately, this means that the
979    inline code for div/mod now has one more comparison than before.
980    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
981    The problem was with quotient = 0, because that alone does not tell
982    us which way the rounding went.  One then has to look at whether
983    remainder and divisor have the same sign...  :(
984    
985    Anyway, I replaced the bootfiles with fresh ones...
986    
987    ----------------------------------------------------------------------
988    Name: Matthias Blume
989    Date: 2002/03/29 14:10:00 EST
990    Tag: blume-20020329-inlprims
991    Description:
992    
993    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
994    
995    Primops have changed. This means that the bin/boot-file formats have
996    changed as well.
997    
998    To make sure that there is no confusion, I made a new version.
999    
1000    
1001    CHANGES:
1002    
1003    * removed REMT from mltree (remainder should never overflow).
1004    
1005    * added primops to deal with divisions of all flavors to the frontend
1006    
1007    * handled these primops all the way through so they map to their respective
1008      MLRISC support
1009    
1010    * used these primops in the implementation of Int, Int32, Word, Word32
1011    
1012    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
1013    
1014    * parameterized INLMIN, INLMAX, and INLABS by a numkind
1015    
1016    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
1017      floating point
1018    
1019    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
1020      and Real (but Real.abs maps to a separate floating-point-only primop)
1021    
1022    
1023    TODO items:
1024    
1025    * Hacked Alpha32 instruction selection, disabling the selection of REMx
1026      instructions because the machine instruction encoder cannot handle
1027      them.  (Hppa, PPC, and Sparc instruction selection did not handle
1028      REM in the first place, and REM is supported by the x86 machine coder.)
1029    
1030    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
1031      selection phase.  (The two can be streamlined because the hardware
1032      delivers both quotient and remainder at the same time anyway.)
1033    
1034    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
1035      (Currently the behavior is inconsistent both across architectures and
1036      wrt. the draft Basis spec.)
1037    
1038    * Word8 should eventually be handled natively, too.
1039    
1040    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
1041      as if there currently is no execution path that could trigger it in
1042      SML/NJ.  (The assumptions underlying functions arith and promotable do not
1043      hold for things like multiplication and division.)
1044    
1045    ----------------------------------------------------------------------
1046    Name: Matthias Blume
1047    Date: 2002/03/27 16:27:00 EST
1048    Tag: blume-20020327-mlrisc-divisions
1049    Description:
1050    
1051    Added support for all four division operations (ML's div, mod, quot,
1052    and rem) to MLRISC.  In the course of doing so, I also rationalized
1053    the naming (no more annoying switch-around of DIV and QUOT), by
1054    parameterizing the operation by div_rounding_mode (which can be either
1055    DIV_TO_ZERO or DIV_TO_NEGINF).
1056    
1057    The generic MLTreeGen functor takes care of compiling all four
1058    operations down to only round-to-zero div.
1059    
1060    Missing pieces:
1061    
1062      * Doing something smarter than relying on MLTreeGen on architectures
1063        like, e.g., the x86 where hardware division delivers both quotient and
1064        remainder at the same time.  With this, the implementation of the
1065        round-to-neginf operations could be further streamlined.
1066    
1067      * Remove inlining support for div/mod/rem from the frontend and replace it
1068        with primops that get carried through to the backend.  Do this for all
1069        int and word types.
1070    
1071    ----------------------------------------------------------------------
1072    Name: Matthias Blume
1073    Date: 2002/03/25 17:25:00 EST
1074    Tag: blume-20020325-divmod
1075    Description:
1076    
1077    I improved (hopefully without breaking them) the implementation of Int.div,
1078    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
1079    advantage of the following observations:
1080    
1081      Let  q = x quot y      r = x rem y
1082           d = x div  y      m = x mod y
1083    
1084    where "quot" is the round-to-zero version of integer division that
1085    hardware usually provides.  Then we have:
1086    
1087         r = x - q * y        where neither the * nor the - will overflow
1088         d = if q >= 0 orelse x = q * y then q else q - 1
1089                              where neither the * nor the - will overflow
1090         m = if q >= 0 orelse r = 0 then r else r + y
1091                              where the + will not overflow
1092    
1093    This results in substantial simplification of the generated code.
1094    The following table shows the number of CFG nodes and edges generated
1095    for
1096            fun f (x, y) = x OPER y
1097            (* with OPER \in div, mod, quot, rem *)
1098    
1099    
1100        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
1101        --------------------------------------------------------
1102         div |         24 |         39 |         12 |         16
1103         mod |         41 |         71 |         12 |         16
1104        quot |          8 |         10 |          8 |         10
1105         rem |         10 |         14 |          8 |         10
1106    
1107    
1108    ----------------------------------------------------------------------
1109    Name: Matthias Blume
1110    Date: 2002/03/25 22:06:00 EST
1111    Tag: blume-20020325-cprotobug
1112    Description:
1113    
1114    Fixed a bug in cproto (c prototype decoder).
1115    
1116    ----------------------------------------------------------------------
1117    Name: Matthias Blume
1118    Date: 2002/03/25 16:00:00 EST
1119    Tag: blume-20020325-raw-primops
1120    Description:
1121    
1122    I did some cleanup to Allen's new primop code and
1123    replaced yesterday's bootfiles with new ones.
1124    (But they are stored in the same place.)
1125    
1126    ----------------------------------------------------------------------
1127    Name: Matthias Blume
1128    Date: 2002/03/24 22:40:00 EST
1129    Tag: blume-20020324-bootfiles
1130    Description:
1131    
1132    Made the bootfiles that Allen asked for.
1133    
1134    ----------------------------------------------------------------------
1135    Name: Allen Leung
1136    Date: 2002/03/23 15:50:00 EST
1137    Tag: leunga-20020323-flint-cps-rcc-primops
1138    Description:
1139    
1140      1. Changes to FLINT primops:
1141    
1142        (* make a call to a C-function;
1143         * The primop carries C function prototype information and specifies
1144         * which of its (ML-) arguments are floating point. C prototype
1145         * information is for use by the backend, ML information is for
1146         * use by the CPS converter. *)
1147      | RAW_CCALL of { c_proto: CTypes.c_proto,
1148                       ml_args: ccall_type list,
1149                       ml_res_opt: ccall_type option,
1150                       reentrant : bool
1151                     } option
1152       (* Allocate uninitialized storage on the heap.
1153        * The record is meant to hold short-lived C objects, i.e., they
1154        * are not ML pointers.  With the tag, the representation is
1155        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
1156        *)
1157      | RAW_RECORD of {tag:bool,sz:int}
1158      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
1159    
1160      2.  These CPS primops are now overloaded:
1161    
1162           rawload of {kind:numkind}
1163           rawstore of {kind:numkind}
1164    
1165          The one argument form is:
1166    
1167             rawload {kind} address
1168    
1169          The two argument form is:
1170    
1171             rawload {kind} [ml object, byte-offset]
1172    
1173      3. RAW_CCALL/RCC now takes two extra arguments:
1174    
1175         a. The first is whether the C call is reentrant, i.e., whether
1176            ML state should be saved and restored.
1177         b. The second argument is a string argument specifying the name of
1178            library and the C function.
1179    
1180         These things are currently not handled in the code generator, yet.
1181    
1182      4. In CProto,
1183    
1184         An encoding type of "bool" means "ml object" and is mapped into
1185         C prototype of PTR.  Note that "bool" is different than "string",
1186         even though "string" is also mapped into PTR, because "bool"
1187         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
1188    
1189      5. Pickler/unpicker
1190    
1191         Changed to handle RAW_RECORD and newest RAW_CCALL
1192    
1193      6. MLRiscGen,
1194    
1195         1. Changed to handle the new rawload/rawstore/rawrecord operators.
1196         2. Code for handling C Calls has been moved to a new module CPSCCalls,
1197            in the file CodeGen/cpscompile/cps-c-calls.sml
1198    
1199      7. Added the conditional move operator
1200    
1201             condmove of branch
1202    
1203         to cps.  Generation of this is still buggy so it is currently
1204         disabled.
1205    
1206    ----------------------------------------------------------------------
1207    Name: Lal George
1208    Date: 2002/03/22 14:18:25 EST
1209    Tag: george-20020322-cps-branch-prob
1210    Description:
1211    
1212    Implemented the Ball-Larus branch prediction-heuristics, and
1213    incorporated graphical viewers for control flow graphs.
1214    
1215    Ball-Larus Heuristics:
1216    ---------------------
1217    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1218    
1219    By design it uses the Dempster-Shafer theory for combining
1220    probabilities.  For example, in the function:
1221    
1222        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1223    
1224    the ball-larus heuristics predicts that the n=0 is unlikely
1225    (OH-heuristic), and the 'then' branch is unlikely because of the
1226    RH-heuristic -- giving the 'then' branch an even lower combined
1227    probability using the Dempster-Shafer theory.
1228    
1229    Finally, John Reppy's loop analysis in MLRISC, further lowers the
1230    probability of the 'then' branch because of the loop in the else
1231    branch.
1232    
1233    
1234    Graphical Viewing:
1235    ------------------
1236    I merely plugged in Allen's graphical viewers into the compiler. The
1237    additional code is not much. At the top level, saying:
1238    
1239            Control.MLRISC.getFlag "cfg-graphical-view" := true;
1240    
1241    will display the graphical view of the control flow graph just before
1242    back-patching.  daVinci must be in your path for this to work. If
1243    daVinci is not available, then the default viewer can be changed
1244    using:
1245    
1246            Control.MLRISC.getString "viewer"
1247    
1248    which can be set to "dot" or "vcg" for the corresponding viewers. Of
1249    course, these viewers must be in your path.
1250    
1251    The above will display the compilation unit at the level of clusters,
1252    many of which are small, boring, and un-interesting. Also setting:
1253    
1254            Control.MLRISC.getInt "cfg-graphical-view_size"
1255    
1256    will display clusters that are larger than the value set by the above.
1257    
1258    
1259    ----------------------------------------------------------------------
1260    Name: Matthias Blume
1261    Date: 2002/03/21 22:20:00 EST
1262    Tag: blume-20020321-kmp-bugfix
1263    Description:
1264    
1265    Changed the interface to the KMP routine in PreString and fixed
1266    a minor bug in one place where it was used.
1267    
1268    ----------------------------------------------------------------------
1269    Name: Allen Leung
1270    Date: 2002/03/21 20:30:00 EST
1271    Tag: leunga-20020321-cfg
1272    Description:
1273    
1274      Fixed a potential problem in cfg edge splitting.
1275    
1276    ----------------------------------------------------------------------
1277    Name: Allen Leung
1278    Date: 2002/03/21 17:15:00 EST
1279    Tag: leunga-20020321-x86-fp-cfg
1280    Description:
1281    
1282      1. Recoded the buggy parts of x86-fp.
1283    
1284         a. All the block reordering code has been removed.
1285            We now depend on the block placement phases to do this work.
1286    
1287         b. Critical edge splitting code has been simplified and moved into the
1288            CFG modules, as where they belong.
1289    
1290         Both of these were quite buggy and complex.  The code is now much, much
1291         simpler.
1292    
1293      2. X86 backend.
1294    
1295         a. Added instructions for 64-bit support.  Instruction selection for
1296            64-bit has not been committed, however, since that
1297            requires changes to MLTREE which haven't been approved by
1298            Lal and John.
1299    
1300         b. Added support for FUCOMI and FUCOMIP when generating code for
1301            PentiumPro and above.  We only generate these instructions in
1302            the fast-fp mode.
1303    
1304         c. Added cases for JP and JNP in X86FreqProps.
1305    
1306      3. CFG
1307    
1308         CFG now has a bunch of methods for edge splitting and merging.
1309    
1310      4. Machine description.
1311    
1312         John's simplification of MLTREE_BASIS.fcond broke a few machine
1313         description things:
1314    
1315         rtl-build.{sig,sml} and hppa.mdl fixed.
1316    
1317         NOTE: the machine description stuff in the repository is still broken.
1318               Again, I can't put my fixes in because that involves
1319               changes to MLTREE.
1320    
1321    ----------------------------------------------------------------------
1322    Name: Matthias Blume
1323    Date: 2002/03/20 15:55:00 EST
1324    Tag: blume-20020320-kmp
1325    Description:
1326    
1327    Implemented Knuth-Morris-Pratt string matching in PreString and used
1328    it for String.isSubstring, Substring.isSubstring, and
1329    Substring.position.
1330    
1331    (Might need some stress-testing.  Simple examples worked fine.)
1332    
1333    ----------------------------------------------------------------------
1334    Name: Matthias Blume
1335    Date: 2002/03/19 16:37:00 EST
1336    Tag: blume-20020319-witnesses
1337    Description:
1338    
1339    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1340    
1341    This implements a generic mechanism for changing constness qualifiers
1342    anywhere within big C types without resorting to outright "casts".
1343    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1344    modify the constness at the outermost level.)
1345    The implementation of "convert" is based on the idea of "witness"
1346    values -- values that are not used by the operation but whose types
1347    "testify" to their applicability.  On the implementation side, "convert"
1348    is simply a projection (returning its second curried argument).  With
1349    cross-module inlining, it should not result in any machine code being
1350    generated.
1351    
1352    ----------------------------------------------------------------------
1353    Name: Matthias Blume
1354    Date: 2002/03/15 16:40:00 EST
1355    Tag: blume-20020315-basis
1356    Description:
1357    
1358    Provided (preliminary?) implementations for
1359    
1360      {String,Substring}.{concatWith,isSuffix,isSubstring}
1361    
1362    and
1363    
1364      Substring.full
1365    
1366    Those are in the Basis spec but they were missing in SML/NJ.
1367    
1368    ----------------------------------------------------------------------
1369    Name: Matthias Blume
1370    Date: 2002/03/14 21:30:00 EST
1371    Tag: blume-20020314-controls
1372    Description:
1373    
1374    Controls:
1375    ---------
1376    
1377    1. Factored out the recently-added Controls : CONTROLS stuff and put
1378       it into its own library $/controls-lib.cm.  The source tree for
1379       this is under src/smlnj-lib/Controls.
1380    
1381    2. Changed the names of types and functions in this interface, so they
1382       make a bit more "sense":
1383    
1384          module -> registry
1385          'a registry -> 'a group
1386    
1387    3. The interface now deals in ref cells only.  The getter/setter interface
1388       is (mostly) gone.
1389    
1390    4. Added a function that lets one register an already-existing ref cell.
1391    
1392    5. Made the corresponding modifications to the rest of the code so that
1393       everything compiles again.
1394    
1395    6. Changed the implementation of Controls.MLRISC back to something closer
1396       to the original.  In particular, this module (and therefore MLRISC)
1397       does not depend on Controls.  There now is some link-time code in
1398       int-sys.sml that registers the MLRISC controls with the Controls
1399       module.
1400    
1401    CM:
1402    ---
1403    
1404      * One can now specify the lambda-split aggressiveness in init.cmi.
1405    
1406    ----------------------------------------------------------------------
1407    Name: Allen Leung
1408    Date: 2002/03/13 17:30:00 EST
1409    Tag: leunga-20020313-x86-fp-unary
1410    Description:
1411    
1412    Bug fix for:
1413    
1414    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1415    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1416    > - fun f(x,(y,z)) = Real.~ y;
1417    > [autoloading]
1418    > [autoloading done]
1419    >       fchsl   (%eax), 184(%esp)
1420    > Error: MLRisc bug: X86MCEmitter.emitInstr
1421    >
1422    > uncaught exception Error
1423    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1424    
1425    The problem was that the code generator did not generate any fp registers
1426    in this case, and the ra didn't know that it needed to run the X86FP phase to
1427    translate the pseudo fp instruction.   This only happened with unary fp
1428    operators in certain situations.
1429    
1430    ----------------------------------------------------------------------
1431    Name: Matthias Blume
1432    Date: 2002/03/13 14:00:00 EST
1433    Tag: blume-20020313-overload-etc
1434    Description:
1435    
1436    1. Added _overload as a synonym for overload for backward compatibility.
1437       (Control.overloadKW must be true for either version to be accepted.)
1438    
1439    2. Fixed bug in install script that caused more things to be installed
1440       than what was requested in config/targets.
1441    
1442    3. Made CM aware of the (_)overload construct so that autoloading
1443       works.
1444    
1445    ----------------------------------------------------------------------
1446    Name: Matthias Blume
1447    Date: 2002/03/12 22:03:00 EST
1448    Tag: blume-20020312-url
1449    Description:
1450    
1451    Forgot to update BOOT and srcarchiveurl.
1452    
1453    ----------------------------------------------------------------------
1454    Name: Matthias Blume
1455    Date: 2002/03/12 17:30:00 EST
1456    Tag: blume-20020312-version110392
1457    Description:
1458    
1459    Yet another version number bump (because of small changes to the
1460    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1461    
1462    Changes:
1463    
1464      The new pid generation scheme described a few weeks ago was overly
1465      complicated.  I implemented a new mechanism that is simpler and
1466      provides a bit more "stability":  Once CM has seen a compilation
1467      unit, it keeps its identity constant (as long as you do not delete
1468      those crucial CM/GUID/* files).  This means that when you change
1469      an interface, compile, then go back to the old interface, and
1470      compile again, you arrive at the original pid.
1471    
1472      There now also is a mechanism that instructs CM to use the plain
1473      environment hash as a module's pid (effectively making its GUID
1474      the empty string).  For this, "noguid" must be specified as an
1475      option to the .sml file in question within its .cm file.
1476      This is most useful for code that is being generated by tools such
1477      as ml-nlffigen (because during development programmers tend to
1478      erase the tool's entire output directory tree including CM's cached
1479      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1480      revert to the old, broken behavior of SML/NJ, but in specific cases
1481      where there is no danger of interface confusion, its use is ok
1482      (I think).
1483    
1484      ml-nlffigen by default generates "noguid" annotations.  They can be
1485      turned off by specifying -guid in its command line.
1486    
1487    ----------------------------------------------------------------------
1488    Name: Lal George
1489    Date: 2002/03/12 12 14:42:36 EST
1490    Tag: george-20020312-frequency-computation
1491    Description:
1492    
1493    Integrated jump chaining and static block frequency into the
1494    compiler. More details and numbers later.
1495    
1496    ----------------------------------------------------------------------
1497    Name: Lal George
1498    Date: 2002/03/11 11 22:38:53 EST
1499    Tag: george-20020311-jump-chain-elim
1500    Description:
1501    
1502    Tested the jump chain elimination on all architectures (except the
1503    hppa).  This is on by default right now and is profitable for the
1504    alpha and x86, however, it may not be profitable for the sparc and ppc
1505    when compiling the compiler.
1506    
1507    The gc test will typically jump to a label at the end of the cluster,
1508    where there is another jump to an external cluster containing the actual
1509    code to invoke gc. This is to allow factoring of common gc invocation
1510    sequences. That is to say, we generate:
1511    
1512            f:
1513               testgc
1514               ja   L1      % jump if above to L1
1515    
1516            L1:
1517               jmp L2
1518    
1519    
1520    After jump chain elimination the 'ja L1' instructions is converted to
1521    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1522    up being implemented in their long form (if L2 is far away) using:
1523    
1524            jbe     L3      % jump if below or equal to L3
1525            jmp     L2
1526         L3:
1527            ...
1528    
1529    
1530    For large compilation units L2  may be far away.
1531    
1532    
1533    ----------------------------------------------------------------------
1534    Name: Matthias Blume
1535    Date: 2002/03/11 13:30:00 EST
1536    Tag: blume-20020311-mltreeeval
1537    Description:
1538    
1539    A functor parameter was missing.
1540    
1541    ----------------------------------------------------------------------
1542    Name: Allen Leung
1543    Date: 2002/03/11 10:30:00 EST
1544    Tag: leunga-20020311-runtime-string0
1545    Description:
1546    
1547       The representation of the empty string now points to a
1548    legal null terminated C string instead of unit.  It is now possible
1549    to convert an ML string into C string with InlineT.CharVector.getData.
1550    This compiles into one single machine instruction.
1551    
1552    ----------------------------------------------------------------------
1553    Name: Allen Leung
1554    Date: 2002/03/10 23:55:00 EST
1555    Tag: leunga-20020310-x86-call
1556    Description:
1557    
1558       Added machine generation for CALL instruction (relative displacement mode)
1559    
1560    ----------------------------------------------------------------------
1561    Name: Matthias Blume
1562    Date: 2002/03/08 16:05:00
1563    Tag: blume-20020308-entrypoints
1564    Description:
1565    
1566    Version number bumped to 110.39.1.  NEW BOOTFILES!
1567    
1568    Entrypoints: non-zero offset into a code object where execution should begin.
1569    
1570    - Added the notion of an entrypoint to CodeObj.
1571    - Added reading/writing of entrypoint info to Binfile.
1572    - Made runtime system bootloader aware of entrypoints.
1573    - Use the address of the label of the first function given to mlriscGen
1574      as the entrypoint.  This address is currently always 0, but it will
1575      not be 0 once we turn on block placement.
1576    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1577      with entry points) from mlriscGen.
1578    
1579    ----------------------------------------------------------------------
1580    Name: Allen Leung
1581    Date: 2002/03/07 20:45:00 EST
1582    Tag: leunga-20020307-x86-cmov
1583    Description:
1584    
1585       Bug fixes for CMOVcc on x86.
1586    
1587       1. Added machine code generation for CMOVcc
1588       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1589       3. CMOVcc cannot have an immediate operand as argument.
1590    
1591    ----------------------------------------------------------------------
1592    Name: Matthias Blume
1593    Date: 2002/03/07 16:15:00 EST
1594    Tag: blume-20020307-controls
1595    Description:
1596    
1597    This is a very large but mostly boring patch which makes (almost)
1598    every tuneable compiler knob (i.e., pretty much everything under
1599    Control.* plus a few other things) configurable via both the command
1600    line and environment variables in the style CM did its configuration
1601    until now.
1602    
1603    Try starting sml with '-h' (or, if you are brave, '-H')
1604    
1605    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1606    implements the underlying generic mechanism.
1607    
1608    The interface to some of the existing such facilities has changed somewhat.
1609    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1610    (The getFoo interface is still there for backward-compatibility, but its
1611    use is deprecated.)
1612    
1613    The ml-build script passes -Cxxx=yyy command-line arguments through so
1614    that one can now twiddle the compiler settings when using this "batch"
1615    compiler.
1616    
1617    TODO items:
1618    
1619    We should go through and throw out all controls that are no longer
1620    connected to anything.  Moreover, we should go through and provide
1621    meaningful (and correct!) documentation strings for those controls
1622    that still are connected.
1623    
1624    Currently, multiple calls to Controls.new are accepted (only the first
1625    has any effect).  Eventually we should make sure that every control
1626    is being made (via Controls.new) exactly once.  Future access can then
1627    be done using Controls.acc.
1628    
1629    Finally, it would probably be a good idea to use the getter-setter
1630    interface to controls rather than ref cells.  For the time being, both
1631    styles are provided by the Controls module, but getter-setter pairs are
1632    better if thread-safety is of any concern because they can be wrapped.
1633    
1634    *****************************************
1635    
1636    One bug fix: The function blockPlacement in three of the MLRISC
1637    backpatch files used to be hard-wired to one of two possibilities at
1638    link time (according to the value of the placementFlag).  But (I
1639    think) it should rather sense the flag every time.
1640    
1641    *****************************************
1642    
1643    Other assorted changes (by other people who did not supply a HISTORY entry):
1644    
1645    1. the cross-module inliner now works much better (Monnier)
1646    2. representation of weights, frequencies, and probabilities in MLRISC
1647       changed in preparation of using those for weighted block placement
1648       (Reppy, George)
1649    
1650    ----------------------------------------------------------------------
1651    Name: Lal George
1652    Date: 2002/03/07 14:44:24 EST 2002
1653    Tag: george-20020307-weighted-block-placement
1654    
1655    Tested the weighted block placement optimization on all architectures
1656    (except the hppa) using AMPL to generate the block and edge frequencies.
1657    Changes were required in the machine properties to correctly
1658    categorize trap instructions. There is an MLRISC flag
1659    "weighted-block-placement" that can be used to enable weighted block
1660    placement, but this will be ineffective without block/edge
1661    frequencies (coming soon).
1662    
1663    
1664    ----------------------------------------------------------------------
1665    Name: Lal George
1666    Date: 2002/03/05 17:24:48 EST
1667    Tag: george-20020305-linkage-cluster
1668    
1669    In order to support the block placement optimization, a new cluster
1670    is generated as the very first cluster (called the linkage cluster).
1671    It contains a single jump to the 'real' entry point for the compilation
1672    unit. Block placement has no effect on the linkage cluster itself, but
1673    all the other clusters  have full freedom in the manner in which they
1674    reorder blocks or functions.
1675    
1676    On the x86 the typical linkage code that is generated is:
1677       ----------------------
1678            .align 2
1679       L0:
1680            addl    $L1-L0, 72(%esp)
1681            jmp     L1
1682    
1683    
1684            .align  2
1685       L1:
1686       ----------------------
1687    
1688    72(%esp) is the memory location for the stdlink register. This
1689    must contain the address of the CPS function being called. In the
1690    above example, it contains the address of  L0; before
1691    calling L1 (the real entry point for the compilation unit), it
1692    must contain the address for L1, and hence
1693    
1694            addl $L1-L0, 72(%esp)
1695    
1696    I have tested this on all architectures except the hppa.The increase
1697    in code size is of course negligible
1698    
1699    ----------------------------------------------------------------------
1700    Name: Allen Leung
1701    Date: 2002/03/03 13:20:00 EST
1702    Tag: leunga-20020303-mlrisc-tools
1703    
1704      Added #[ ... ] expressions to mlrisc tools
1705    
1706    ----------------------------------------------------------------------
1707    Name: Matthias Blume
1708    Date: 2002/02/27 12:29:00 EST
1709    Tag: blume-20020227-cdebug
1710    Description:
1711    
1712    - made types in structure C and C_Debug to be equal
1713    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1714    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1715    
1716    ----------------------------------------------------------------------
1717    Name: Matthias Blume
1718    Date: 2002/02/26 12:00:00 EST
1719    Tag: blume-20020226-ffi
1720    Description:
1721    
1722    1. Fixed a minor bug in CM's "noweb" tool:
1723       If numbering is turned off, then truly don't number (i.e., do not
1724       supply the -L option to noweb).  The previous behavior was to supply
1725       -L'' -- which caused noweb to use the "default" line numbering scheme.
1726       Thanks to Chris Richards for pointing this out (and supplying the fix).
1727    
1728    2. Once again, I reworked some aspects of the FFI:
1729    
1730       A. The incomplete/complete type business:
1731    
1732       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1733         gone!
1734       - ML types representing an incomplete type are now *equal* to
1735         ML types representing their corresponding complete types (just like
1736         in C).  This is still safe because ml-nlffigen will not generate
1737         RTTI for incomplete types, nor will it generate functions that
1738         require access to such RTTI.   But when ML code generated from both
1739         incomplete and complete versions of the C type meet, the ML types
1740         are trivially interoperable.
1741    
1742         NOTE:  These changes restore the full generality of the translation
1743         (which was previously lost when I eliminated functorization)!
1744    
1745       B. Enum types:
1746    
1747       - Structure C now has a type constructor "enum" that is similar to
1748         how the "su" constructor works.  However, "enum" is not a phantom
1749         type because each "T enum" has values (and is isomorphic to
1750         MLRep.Signed.int).
1751       - There are generic access operations for enum objects (using
1752         MLRep.Signed.int).
1753       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1754         * The structure contains the definition of type "mlrep" (the ML-side
1755         representation type of the enum).  Normally, mlrep is the same
1756         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1757         then mlrep will be defined as a datatype -- thus facilitating
1758         pattern matching on mlrep values.
1759         ("-ec" will be suppressed if there are duplicate values in an
1760          enumeration.)
1761         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1762         will be generated for each C enum constant xxx.
1763         * Conversion functions m2i and i2m convert between mlrep and
1764         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1765         * Coversion functions c and ml convert between mlrep and "tag enum".
1766         * Access functions (get/set) fetch and store mlrep values.
1767       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1768         enumerations are merged into one single enumeration represented by
1769         structure E_'.
1770    
1771    ----------------------------------------------------------------------
1772    Name: Allen Leung
1773    Date: 2002/02/25 04:45:00 EST
1774    Tag: leunga-20020225-cps-spill
1775    
1776    This is a new implementation of the CPS spill phase.
1777    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1778    In case of problems, replace it with the old file spill.sml
1779    
1780    The current compiler runs into some serious performance problems when
1781    constructing a large record.  This can happen when we try to compile a
1782    structure with many items.  Even a very simple structure like the following
1783    makes the compiler slow down.
1784    
1785        structure Foo = struct
1786           val x_1 = 0w1 : Word32.int
1787           val x_2 = 0w2 : Word32.int
1788           val x_3 = 0w3 : Word32.int
1789           ...
1790           val x_N = 0wN : Word32.int
1791        end
1792    
1793    The following table shows the compile time, from N=1000 to N=4000,
1794    with the old compiler:
1795    
1796    N
1797    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1798           MLRISC ra                               0.06u  0.00s  0.05g
1799              (spills = 0 reloads = 0)
1800           TOTAL                                   0.63u  0.07s  0.21g
1801    
1802    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1803           MLRISC ra                               5.68u  0.59s  3.93g
1804              (spills = 0 reloads = 0)
1805           TOTAL                                   14.71u  0.99s  4.81g
1806    
1807    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1808           MLRISC ra                               5.54u  0.65s  3.91g
1809              (spills = 543 reloads = 1082)
1810           TOTAL                                   65.40u  3.13s  6.00g
1811    
1812    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1813           MLRISC ra                               0.80u  0.10s  0.55g
1814              (spills = 42 reloads = 84)
1815           TOTAL                                   129.42u  5.10s  4.13g
1816    
1817    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1818           MLRISC ra                               2.69u  0.27s  1.38g
1819              (spills = 62 reloads = 124)
1820           TOTAL                                   682.48u  19.61s  13.99g
1821    
1822    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1823           MLRISC ra                               4.96u  0.27s  2.72g
1824              (spills = 85 reloads = 170)
1825           TOTAL                                   2375.26u  57.21s  48.00g
1826    
1827    As you can see the old cps spill module suffers from some serious
1828    performance problem.  But since I cannot decipher the old code fully,
1829    instead of patching the problems up, I'm reimplementing it
1830    with a different algorithm.  The new code is more modular,
1831    smaller when compiled, and substantially faster
1832    (O(n log n) time and O(n) space).  Timing of the new spill module:
1833    
1834    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1835          MLRISC ra                               0.25u  0.02s  0.15g
1836             (spills=1 reloads=3)
1837          TOTAL                                   7.74u  0.34s  1.62g
1838    
1839    Implementation details:
1840    
1841    As far as I can tell, the purpose of the CPS spill module is to make sure the
1842    number of live variables at any program point (the bandwidth)
1843    does not exceed a certain limit, which is determined by the
1844    size of the spill area.
1845    
1846    When the bandwidth is too large, we decrease the register pressure by
1847    packing live variables into spill records.  How we achieve this is
1848    completely different than what we did in the old code.
1849    
1850    First, there is something about the MLRiscGen code generator
1851    that we should be aware of:
1852    
1853    o MLRiscGen performs code motion!
1854    
1855       In particular, it will move floating point computations and
1856       address computations involving only the heap pointer to
1857       their use sites (if there is only a single use).
1858       What this means is that if we have a CPS record construction
1859       statement
1860    
1861           RECORD(k,vl,w,e)
1862    
1863       we should never count the new record address w as live if w
1864       has only one use (which is often the case).
1865    
1866       We should do something similar to floating point, but the transformation
1867       there is much more complex, so I won't deal with that.
1868    
1869    Secondly, there are now two new cps primops at our disposal:
1870    
1871     1. rawrecord of record_kind option
1872        This pure operator allocates some uninitialized storage from the heap.
1873        There are two forms:
1874    
1875         rawrecord NONE [INT n]  allocates a tagless record of length n
1876         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1877                                     and initializes the tag.
1878    
1879     2. rawupdate of cty
1880          rawupdate cty (v,i,x)
1881          Assigns to x to the ith component of record v.
1882          The storelist is not updated.
1883    
1884    We use these new primops for both spilling and increment record construction.
1885    
1886     1. Spilling.
1887    
1888        This is implemented with a linear scan algorithm (but generalized
1889        to trees).  The algorithm will create a single spill record at the
1890        beginning of the cps function and use rawupdate to spill to it,
1891        and SELECT or SELp to reload from it.  So both spills and reloads
1892        are fine-grain operations.  In contrast, in the old algorithm
1893        "spills" have to be bundled together in records.
1894    
1895        Ideally, we should sink the spill record construction to where
1896        it is needed.  We can even split the spill record into multiple ones
1897        at the places where they are needed.  But CPS is not a good
1898        representation for global code motion, so I'll keep it simple and
1899        am not attempting this.
1900    
1901     2. Incremental record construction (aka record splitting).
1902    
1903        Long records with many component values which are simulatenously live
1904        (recall that single use record addresses are not considered to
1905         be live) are constructed with rawrecord and rawupdate.
1906        We allocate space on the heap with rawrecord first, then gradually
1907        fill it in with rawupdate.  This is the technique suggested to me
1908        by Matthias.
1909    
1910        Some restrictions on when this is applicable:
1911        1. It is not a VECTOR record.  The code generator currently does not handle
1912           this case. VECTOR record uses double indirection like arrays.
1913        2. All the record component values are defined in the same "basic block"
1914           as the record constructor.  This is to prevent speculative
1915           record construction.
1916    
1917    ----------------------------------------------------------------------
1918    Name: Allen Leung
1919    Date: 2002/02/22 01:02:00 EST
1920    Tag: leunga-20020222-mlrisc-tools
1921    
1922    Minor bug fixes in the parser and rewriter
1923    
1924    ----------------------------------------------------------------------
1925    Name: Allen Leung
1926    Date: 2002/02/21 20:20:00 EST
1927    Tag: leunga-20020221-peephole
1928    
1929    Regenerated the peephole files.  Some contained typos in the specification
1930    and some didn't compile because of pretty printing bugs in the old version
1931    of 'nowhere'.
1932    
1933    ----------------------------------------------------------------------
1934    Name: Allen Leung
1935    Date: 2002/02/19 20:20:00 EST
1936    Tag: leunga-20020219-mlrisc-tools
1937    Description:
1938    
1939       Minor bug fixes to the mlrisc-tools library:
1940    
1941       1.  Fixed up parsing colon suffixed keywords
1942       2.  Added the ability to shut the error messages up
1943       3.  Reimplemented the pretty printer and fixed up/improved
1944           the pretty printing of handle and -> types.
1945       4.  Fixed up generation of literal symbols in the nowhere tool.
1946       5.  Added some SML keywords to to sml.sty
1947    
1948    ----------------------------------------------------------------------
1949    Name: Matthias Blume
1950    Date: 2002/02/19 16:20:00 EST
1951    Tag: blume-20020219-cmffi
1952    Description:
1953    
1954    A wild mix of changes, some minor, some major:
1955    
1956    * All C FFI-related libraries are now anchored under $c:
1957        $/c.cm      --> $c/c.cm
1958        $/c-int.cm  --> $c/internals/c-int.cm
1959        $/memory.cm --> $c/memory/memory.cm
1960    
1961    * "make" tool (in CM) now treats its argument pathname slightly
1962      differently:
1963        1. If the native expansion is an absolute name, then before invoking
1964           the "make" command on it, CM will apply OS.Path.mkRelative
1965           (with relativeTo = OS.FileSys.getDir()) to it.
1966        2. The argument will be passed through to subsequent phases of CM
1967           processing without "going native".  In particular, if the argument
1968           was an anchored path, then "make" will not lose track of that anchor.
1969    
1970    * Compiler backends now "know" their respective C calling conventions
1971      instead of having to be told about it by ml-nlffigen.  This relieves
1972      ml-nlffigen from one of its burdens.
1973    
1974    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1975    
1976    * Export C_DEBUG and C_Debug from $c/c.cm.
1977    
1978    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1979      subtyping relationship between incomplete pointers and their complete
1980      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1981      with the convention of instantiating 'o with ('t, 'c) obj whenever
1982      the pointer target type is complete.  In the incomplete case, 'o
1983      will be instantiated with some "'c iobj" -- a type obtained by
1984      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1985    
1986      Operations that work on both incomplete and complete pointer types are
1987      typed as taking an 'o ptr while operations that require the target to
1988      be known are typed as taking some ('t, 'c) obj ptr.
1989    
1990      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1991      where void is an eqtype without any values.  This makes it possible
1992      to work on voidptr values using functions meant to operate on light
1993      incomplete pointers.
1994    
1995    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1996      been vastly simplified.
1997    
1998    ----------------------------------------------------------------------
1999    Name: Matthias Blume
2000    Date: 2002/02/19 10:48:00 EST
2001    Tag: blume-20020219-pqfix
2002    Description:
2003    
2004    Applied Chris Okasaki's bug fix for priority queues.
2005    
2006    ----------------------------------------------------------------------
2007    Name: Matthias Blume
2008    Date: 2002/02/15 17:05:00
2009    Tag: Release_110_39
2010    Description:
2011    
2012    Last-minute retagging is becoming a tradition... :-(
2013    
2014    This is the working release 110.39.
2015    
2016    ----------------------------------------------------------------------
2017    Name: Matthias Blume
2018    Date: 2002/02/15 16:00:00 EST
2019    Tag: Release_110_39-orig
2020    Description:
2021    
2022    Working release 110.39.  New bootfiles.
2023    
2024    (Update: There was a small bug in the installer so it wouldn't work
2025    with all shells.  So I retagged. -Matthias)
2026    
2027    ----------------------------------------------------------------------
2028    Name: Matthias Blume
2029    Date: 2002/02/15 14:17:00 EST
2030    Tag: blume-20020215-showbindings
2031    Description:
2032    
2033    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
2034    the latter can be useful for exploring what bindings are available at
2035    the interactive prompt.  (The first function returns only the list
2036    of symbols that are really bound, the second prints those but also the
2037    ones that CM's autoloading mechanism knows about.)
2038    
2039    ----------------------------------------------------------------------
2040    Name: Matthias Blume
2041    Date: 2002/02/15 12:08:00 EST
2042    Tag: blume-20020215-iptrs
2043    Description:
2044    
2045    Two improvements to ml-nlffigen:
2046    
2047      1. Write files only if they do not exist or if their current contents
2048         do not coincide with what's being written.  (That is, avoid messing
2049         with the time stamps unless absolutely necessary.)
2050    
2051      2. Implement a "repository" mechanism for generated files related
2052         to "incomplete pointer types".   See the README file for details.
2053    
2054    ----------------------------------------------------------------------
2055    Name: Matthias Blume
2056    Date: 2002/02/14 11:50:00 EST
2057    Tag: blume-20020214-quote
2058    Description:
2059    
2060    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
2061    because of the new and improved tag generation scheme.  (Thanks to Allen
2062    Leung for pointing it out.)
2063    
2064    ----------------------------------------------------------------------
2065    Name: Lal George
2066    Date: 2002/02/14 09:55:27 EST 2002
2067    Tag: george-20020214-isabelle-bug
2068    Description:
2069    
2070    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
2071    of Isabelle on the x86.
2072    
2073    From Allen:
2074    -----------
2075     I've found the problem:
2076    
2077         in ra-core.sml, I use the counter "blocked" to keep track of the
2078         true number of elements in the freeze queue.  When the counter goes
2079         to zero, I skip examining the queue.  But I've messed up the
2080         bookkeeping in combine():
2081    
2082             else ();
2083             case !ucol of
2084               PSEUDO => (if !cntv > 0 then
2085                     (if !cntu > 0 then blocked := !blocked - 1 else ();
2086                                        ^^^^^^^^^^^^^^^^^^^^^^^
2087                      moveu := mergeMoveList(!movev, !moveu)
2088                     )
2089                  else ();
2090    
2091         combine() is called to coalesce two nodes u and v.
2092         I think I was thinking that if the move counts of u and v are both
2093         greater than zero then after they are coalesced then one node is
2094         removed from the freeze queue.  Apparently I was thinking that
2095         both u and v are of low degree, but that's clearly not necessarily true.
2096    
2097    
2098    02/12/2002:
2099        Here's the patch.  HOL now compiles.
2100    
2101        I don't know how this impact on performance (compile
2102        time or runtime).  This bug caused the RA (especially on the x86)
2103        to go thru the potential spill phase when there are still nodes on the
2104        freeze queue.
2105    
2106    
2107    
2108    
2109    ----------------------------------------------------------------------
2110    Name: Matthias Blume
2111    Date: 2002/02/13 22:40:00 EST
2112    Tag: blume-20020213-fptr-rtti
2113    Description:
2114    
2115    Fixed a bug in ml-nlffigen that was introduced with one of the previous
2116    updates.
2117    
2118    ----------------------------------------------------------------------
2119    Name: Matthias Blume
2120    Date: 2002/02/13 16:41:00 EST
2121    Tag: blume-20020213-cmlpq
2122    Description:
2123    
2124    Added new priority queue export symbols (which have just been added to
2125    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
2126    not compile and the installer would choke.)
2127    
2128    ----------------------------------------------------------------------
2129    Name: Matthias Blume
2130    Date: 2002/02/13 16:15:00 EST
2131    Tag: blume-20020213-various
2132    Description:
2133    
2134    1. More tweaks to ml-nlffigen:
2135    
2136       - better internal datastructures (resulting in slight speedup)
2137       - "-match" option requires exact match
2138       - "localized" gensym counters (untagged structs/unions nested within
2139         other structs/unions or within typedefs get a fresh counter; their
2140         tag will be prefixed by a concatenation of their parents' tags)
2141       - bug fixes (related to calculation of transitive closure of types
2142         to be included in the output)
2143    
2144    2. Minor Basis updates:
2145    
2146       - added implementations for List.collate and Option.app
2147    
2148    ----------------------------------------------------------------------
2149    Name: Matthias Blume
2150    Date: 2002/02/11 15:55:00 EST
2151    Tag: blume-20020211-gensym
2152    Description:
2153    
2154    Added a "-gensym" option to command line of ml-nlffigen.  This can be
2155    used to specify a "stem" -- a string that is inserted in all "gensym'd"
2156    names (ML structure names that correspond to unnamed C structs, unions,
2157    and enums), so that separate runs of ml-nlffigen do not clash.
2158    
2159    ----------------------------------------------------------------------
2160    Name: Matthias Blume
2161    Date: 2002/02/11 12:05:00 EST
2162    Tag: blume-20020211-gensml
2163    Description:
2164    
2165    A quick fix for a problem with GenSML (in the pgraph-util library):
2166    Make generation of toplevel "local" optional.  (Strictly speaking,
2167    signature definitions within "local" are not legal SML.)
2168    
2169    Other than that: updates to INSTALL and cm/TODO.
2170    
2171    ----------------------------------------------------------------------
2172    Name: Matthias Blume
2173    Date: 2002/02/08 15:00:00 EST
2174    Tag: blume-20020208-uniquepid
2175    Description:
2176    
2177    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
2178    
2179    1. The installer (config/install.sh) has gotten smarter:
2180    
2181         - Configuration options are a bit easier to specify now
2182           (in config/targets).
2183         - Bug in recognizing .tar.bz2 files fixed.
2184         - Installer automatically resolves dependencies between
2185           configuration options (e.g., if you ask for eXene, you will
2186           also get cml -- regardless whether you asked for it or not).
2187         - Installer can run in "quieter mode" by setting the environment
2188           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
2189           "completely silent", though.
2190         - Build HashCons library as part of smlnj-lib.
2191    
2192    2. A new scheme for assigning persistent identifiers to compilation
2193       units (and, by extension, to types etc.) has been put into place.
2194       This fixes a long-standing bug where types and even dynamic values
2195       can get internally confused, thereby compromising type safety
2196       (abstraction) and dynamic correctness.  See
2197    
2198         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
2199    
2200       for an example of how things could go wrong until now.
2201    
2202       The downside of the new scheme is that pids are not quite as
2203       persistent as they used to be: CM will generate a fresh pid
2204       for every compilation unit that it thinks it sees for the first
2205       time.  That means that if you compile starting from a clean, fresh
2206       source tree at two different times, you end up with different
2207       binaries.
2208    
2209       Cutoff recompilation, however, has not been compromised because
2210       CM keeps pid information in special caches between runs.
2211    
2212    ----------------------------------------------------------------------
2213    Name: Lal George
2214    Date: 2002/02/07 15:34:13 EST 2002
2215    Tag: <none>
2216    Description:
2217    
2218    Compilers that generate assembly code may produce  global labels
2219    whose value is resolved at link time. The various peephole optimization
2220    modules did not take this in account.
2221    
2222    TODO. The Labels.addrOf function should really return an option
2223    type so that clients are forced to deal with this issue, rather
2224    than an exception being raised.
2225    
2226    ----------------------------------------------------------------------
2227    Name: Lal George
2228    Date: 2002/02/06 13:55:02 EST
2229    Tag: george-20020206-ra-breakup
2230    Description:
2231    
2232    1. A bug fix from Allen.
2233    
2234        A typo causes extra fstp %st(0)'s to be generated at compensation
2235        edges, which might cause stack underflow traps at runtime.  This
2236        occurs in fft where there are extraneous fstps right before the 'into'
2237        trap instruction (in this case they are harmless since none of the
2238        integers overflow.)
2239    
2240    2. Pulled out various utility modules that were embedded in the modules
2241       of the register allocator. I need these modules for other purposes, but
2242       they are not complete enough to put into a library (just yet).
2243    ----------------------------------------------------------------------
2244    Name: Matthias Blume
2245    Date: 2002/01/31 16:05:00 EST
2246    Tag: blume-20020131-sparc-ccalls
2247    Description:
2248    
2249    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
2250       of extra stack space by mistake.  Fixed.
2251    
2252    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
2253    
2254    ----------------------------------------------------------------------
2255    Name: Allen Leung
2256    Date: 2002/01/30
2257    Tag: leunga-20020130-nowhere-bug-fix
2258    Description:
2259    
2260       MLRISC bug fixes:
2261       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
2262       2. MachineInt.fromString was negating its value.
2263    
2264    ----------------------------------------------------------------------
2265    Name: Matthias Blume
2266    Date: 2002/01/29
2267    Tag: blume-20020129-INSTALL
2268    Description:
2269    
2270    - Added somewhat detailed installation instructions (file INSTALL).
2271    - Fixed curl-detection bug in config/install.sh.
2272    - It is now possible to select the URL getter using the URLGETTER
2273      environment variable:
2274    
2275          not set / "unknown"      --> automatic detection (script tries wget,
2276                                       curl, and lynx)
2277          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
2278                                       how to properly invoke them)
2279          other                    --> use $URLGETTER directly, it must take
2280                                       precisely two command-line arguments
2281                                       (source URL and destination file name)
2282    
2283    ----------------------------------------------------------------------
2284    Name: Matthias Blume
2285    Date: 2002/01/28
2286    Tag: blume-20020128-sparc-ccalls
2287    Description:
2288    
2289    - Fixed problem with calculation of "used" registers in sparc-c-calls.
2290    - Make use of the allocParam argument in sparc-c-calls.
2291    
2292    ----------------------------------------------------------------------
2293    Name: Matthias Blume
2294    Date: 2002/01/28
2295    Tag: blume-20020128-allocParam
2296    Description:
2297    
2298    John Reppy:  Changes c-calls API to accept client-callback for
2299    allocating extra stack space.
2300    me: Corresponding changes to mlriscGen (using a dummy argument that
2301        does not change the current behavior).
2302    
2303    ----------------------------------------------------------------------
2304    Name: Matthias Blume
2305    Date: 2002/01/28 12:00:00
2306    Tag: Release_110_38
2307    Description:
2308    
2309    This time for real!!!
2310    
2311    ----------------------------------------------------------------------
2312    Name: Matthias Blume
2313    Date: 2002/01/28 10:56:00 EST
2314    Tag: blume-20020128-retraction
2315    Description:
2316    
2317    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
2318       with blume-Release_110_38-retracted.)
2319    
2320    1. Fixed a problem with incorrect rounding modes in real64.sml.
2321       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
2322    
2323    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
2324       and enums fixed.  The naming of corresponding ML identifiers should
2325       now be consistent again.
2326    
2327    ----------------------------------------------------------------------
2328    Name: Allen Leung
2329    Date: 2002/01/27
2330    Tag: leunga-20020127-nowhere
2331    Description:
2332    
2333       Added a target called nowhere in the configuration scripts.
2334       Enabling this will build the MLRISC 'nowhere' tool (for translating
2335       programs with where-clauses into legal SML code) during installation.
2336    
2337    ----------------------------------------------------------------------
2338    Name: Matthias Blume
2339    Date: 2002/01/25 21:27:00 EST
2340    Tag: blume-Release_110_38-retracted
2341    Description:
2342    
2343    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
2344    
2345    README will be added later.
2346    
2347    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
2348    (Release_110_38) removed.  Reason: Last-minute bug fixes.
2349    
2350    ----------------------------------------------------------------------
2351    Name: Matthias Blume
2352    Date: 2002/01/25
2353    Tag: blume-20020125-ffi
2354    Description:
2355    
2356    A large number of tweaks and improvements to ml-nlffi-lib and
2357    ml-nlffigen:
2358    
2359       - ML represenation types have been streamlined
2360       - getter and setter functions work with concrete values, not abstract
2361         ones where possible
2362       - ml-nlffigen command line more flexible (see README file there)
2363       - some bugs have been fixed (hopefully)
2364    
2365    ----------------------------------------------------------------------
2366    Name: Lal George
2367    Date: 2002/01/24
2368    Tag: george-20020124-risc-ra-interface
2369    Description:
2370    
2371       There is a dramatic simplification in the interface to the
2372       register allocator for RISC architectures as a result of making
2373       parallel copy instructions explicit.
2374    
2375    ----------------------------------------------------------------------
2376    Name: Matthias Blume
2377    Date: 2002/01/22
2378    Tag: blume-20020122-x86-ccalls
2379    Description:
2380    
2381    Bug fix for c-calls on x86 (having to do with how char- and
2382    short-arguments are being handled).
2383    
2384    ----------------------------------------------------------------------
2385    Name: Matthias Blume
2386    Date: 2002/01/21
2387    Tag: blume-20020121-ff
2388    Description:
2389    
2390    Another day of fiddling with the FFI...
2391    
2392    1. Bug fix/workaround:  CKIT does not complain about negative array
2393       dimensions, so ml-nlffigen has to guard itself against this possibility.
2394       (Otherwise a negative dimension would send it into an infinite loop.)
2395    
2396    2. Some of the abstract types (light objects, light pointers, most "base"
2397       types) in structure C are now eqtypes.
2398    
2399    3. Added constructors and test functions for NULL function pointers.
2400    
2401    ----------------------------------------------------------------------
2402    Name: Matthias Blume
2403    Date: 2002/01/18
2404    Tag: blume-20020118-ready-for-new-release
2405    Description:
2406    
2407    Made config/srcarchiveurl point to a new place.  (Will provide boot
2408    files shortly.)
2409    
2410    Maybe we christen this to be 110.38?
2411    
2412    ----------------------------------------------------------------------
2413    Name: Matthias Blume
2414    Date: 2002/01/18
2415    Tag: blume-20020118-more-ffifiddle
2416    Description:
2417    
2418    Today's FFI fiddling:
2419    
2420      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
2421        C.Set.* that use concrete (MLRep.*) arguments and results instead
2422        of abstract ones.
2423    
2424      - Provided word-style bit operations etc. for "int" representation
2425        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
2426        and Long.
2427    
2428    ----------------------------------------------------------------------
2429    Name: Matthias Blume
2430    Date: 2002/01/18
2431    Tag: blume-20020118-use-x86-fp
2432    Description:
2433    
2434    Now that x86-fast-fp seems to be working, I turned it back on again
2435    by default.  (Seems to work fine now, even with the FFI.)
2436    
2437    Other than that, I added some documentation about the FFI to
2438    src/ml-nlffigen/README and updated the FFI test examples in
2439    src/ml-nlffi-lib/Tests/*.
2440    
2441    ----------------------------------------------------------------------
2442    Name: Allen Leung
2443    Date: 2002/01/17
2444    Tag: leunga-20020117-x86-fast-fp-call
2445    Description:
2446    
2447       1. Fixed a problem with handling return fp values when x86's fast fp
2448          mode is turned on.
2449    
2450       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
2451          of %f32.
2452    
2453       3. Added a constructor INT32lit to the ast of MLRISC tools.
2454    
2455    ----------------------------------------------------------------------
2456    Name: Matthias Blume
2457    Date: 2002/01/16
2458    Tag: blume-20020116-ffifiddle
2459    Description:
2460    
2461    More fiddling with the FFI interface:
2462    
2463     - Make constness 'c instead of rw wherever possible.  This eliminates
2464       the need for certain explicit coercions.  (However, due to ML's
2465       value polymorphism, there will still be many cases where explicit
2466       coercions are necessary.  Phantom types are not the whole answer
2467       to modeling a subtyping relationship in ML.)
2468    
2469     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
2470    
2471     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
2472       clumsy workaround for varargs problem.)
2473    
2474    ----------------------------------------------------------------------
2475    Name: Lal George
2476    Date: 2002/01/15
2477    Tag: <none>
2478    Description:
2479    
2480    1. Since COPY instructions are no longer native to the architecture,
2481       a generic functor can be used to implement the expandCopies function.
2482    
2483    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
2484       TEXT segment.
2485    
2486    ----------------------------------------------------------------------
2487    Name: Matthias Blume
2488    Date: 2002/01/15
2489    Tag: blume-20020115-ffiupdates
2490    Description:
2491    
2492    1. Fix for bug resulting in single-precision float values being returned
2493       incorrectly from FFI calls.
2494    
2495    2. Small modifications to C FFI API:
2496    
2497        - memory-allocation routines return straight objects (no options)
2498          and raise an exception in out-of-memory situations
2499        - unsafe extensions to cast between function pointers and pointers
2500          from/to ints
2501        - added structure C_Debug as an alternative to structure C where
2502          pointer-dereferencing (|*| and |*!) always check for null-pointers
2503        - added open_lib' to DynLinkage;  open_lib' works like open_lib
2504          but also takes a (possibly empty) list of existing library handles
2505          that the current library depends on
2506    
2507    ----------------------------------------------------------------------
2508    Name: Matthias Blume
2509    Date: 2002/01/10
2510    Tag: blume-20020110-newffigen
2511    Description:
2512    
2513    1. Updates to portable graph code.
2514    
2515    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
2516       more scalable now so that even huge interfaces such as the one
2517       for GTK compile in finite time and space. :-)
2518       See src/ml-nlffigen/README for details on what's new.
2519    
2520    ----------------------------------------------------------------------
2521    Name: Lal George
2522    Date: 2001/01/09 14:31:35 EST 2002
2523    Tag: george-20011206-rm-native-copy
2524    Description:
2525    
2526            Removed the native COPY and FCOPY instructions
2527            from all the architectures and replaced it with the
2528            explicit COPY instruction from the previous commit.
2529    
2530            It is now possible to simplify many of the optimizations
2531            modules that manipulate copies. This has not been
2532            done in this change.
2533    
2534    ----------------------------------------------------------------------
2535    Name: Lal George
2536    Date: 2001/12/06 16:50:13 EST 2001
2537    Tag: george-20011206-mlrisc-instruction
2538    Description:
2539    
2540    Changed the representation of instructions from being fully abstract
2541    to being partially concrete. That is to say:
2542    
2543      from
2544            type instruction
2545    
2546      to
2547            type instr                              (* machine instruction *)
2548    
2549            datatype instruction =
2550                LIVE of {regs: C.cellset, spilled: C.cellset}
2551              | KILL of {regs: C.cellset, spilled: C.cellset}
2552              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
2553              | ANNOTATION of {i: instruction, a: Annotations.annotation}
2554              | INSTR of instr
2555    
2556    This makes the handling of certain special instructions that appear on
2557    all architectures easier and uniform.
2558    
2559    LIVE and KILL say that a list of registers are live or killed at the
2560    program point where they appear. No spill code is generated when an
2561    element of the 'regs' field is spilled, but the register is moved to
2562    the 'spilled' (which is present, more for debugging than anything else).
2563    
2564    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
2565    We used to generate:
2566    
2567            DEFFREG f1
2568            f1 := f2 + f3
2569            trapb
2570    
2571    but now generate:
2572    
2573            f1 := f2 + f3
2574            trapb
2575            LIVE {regs=[f1,f2,f3], spilled=[]}
2576    
2577    Furthermore, the DEFFREG (hack) required that all floating point instruction
2578    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
2579    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
2580    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
2581    this architecture).
2582    
2583    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
2584    all the architectures. This will result in further simplification of the
2585    register allocator that must be aware of them for coalescing purposes, and
2586    will also simplify certain aspects of the machine description that provides
2587    callbacks related to parallel copies.
2588    
2589    ANNOTATION should be obvious, and now INSTR represents the honest to God
2590    machine instruction set!
2591    
2592    The <arch>/instructions/<arch>Instr.sml files define certain utility
2593    functions for making porting easier -- essentially converting upper case
2594    to lower case. All machine instructions (of type instr) are in upper case,
2595    and the lower case form generates an MLRISC instruction. For example on
2596    the alpha we have:
2597    
2598      datatype instr =
2599         LDA of {r:cell, b:cell, d:operand}
2600       | ...
2601    
2602      val lda : {r:cell, b:cell, d:operand} -> instruction
2603        ...
2604    
2605    where lda is just (INSTR o LDA), etc.
2606    
2607    ----------------------------------------------------------------------
2608    Name: Matthias Blume
2609    Date: 2001/11/22 21:40:00 EST
2610    Tag: Release_110_37
2611    Description:
2612    
2613    Release 110.37.  This time for real.
2614    
2615    ----------------------------------------------------------------------
2616    Name: Matthias Blume
2617    Date: 2001/11/21 16:35:00 EST
2618    Tag: blume-20011121-foot-in-mouth
2619    Description:
2620    
2621    Removed the "Release_110_37" tag because of a serious bug.
2622    This will be re-tagged once the bug is fixed.
2623    
2624    ----------------------------------------------------------------------
2625    Name: Matthias Blume
2626    Date: 2001/11/21 16:14:00 EST
2627    Tag: blume-20011121-forgottenfile
2628    Description:
2629    
2630    Forgot to add a file.  (Just a .tex-file -- part of
2631    the CM manual source.)
2632    
2633    ----------------------------------------------------------------------
2634    Name: Matthias Blume
2635    Date: 2001/11/21 16:10:00 EST
2636    Tag: blume-20011121-invalid_110_37
2637    Description:
2638    
2639    Note: I removed the original tag "Release_110_37" from this commit
2640          because we found a serious bug in all non-x86 backends.
2641          - Matthias
2642    
2643    1. Modifications to the SML/NJ code generator and to the runtime system
2644       so that code object name strings are directly inserted into code
2645       objects at code generation time.  The only business the runtime system
2646       has with this is now to read the name strings on occasions.
2647       (The encoding of the name string has also changed somewhat.)
2648    
2649    2. CM now implements a simple "set calculus" for specifying export lists.
2650       In particular, it is now possible to refer to the export lists of
2651       other libraries/groups/sources and form unions as well as differences.
2652       See the latest CM manual for details.
2653    
2654    3. An separate notion of "proxy" libraries has again be eliminated from
2655       CM's model.  (Proxy libraries are now simply a special case of using
2656       the export list calculus.)
2657    
2658    4. Some of the existing libraries now take advantage of the new set
2659       calculus.
2660       (Notice that not all libraries have been converted because some
2661       of the existing .cm-files are supposed to be backward compatible
2662       with 110.0.x.)
2663    
2664    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2665       "General.exnMessage"!  The former relies on a certain hook to be
2666       initialized, and that often does not happen in the stand-alone case.)
2667    
2668    ----------------------------------------------------------------------
2669    Name: Lal George
2670    Date: 2001/11/21  13:56:18 EST
2671    Tag: george-2001121-pseudo-ops
2672    Description:
2673    
2674      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2675      ought to never be any question of incompatabilities with
2676      pseudo-op syntax expected by host assemblers.
2677    
2678      For now, only modules supporting GAS syntax are implemented
2679      but more should follow, such as MASM, and vendor assembler
2680      syntax, e.g. IBM as, Sun as, etc.
2681    
2682    ----------------------------------------------------------------------
2683    Name: Matthias Blume
2684    Date: 2001/11/14 11:52:00 EST
2685    Tag: blume-20011114-srcname
2686    Description:
2687    
2688    1. Routed the name of the current source file to mlriscgen where it
2689       should be directly emitted into the code object.  (This last part
2690       is yet to be done.)
2691    
2692    2. Some cleanup of the pgraph code to make it match the proposal that
2693       I put out the other day.  (The proposal notwithstanding, things are
2694       still in flux here.)
2695    
2696    ----------------------------------------------------------------------
2697    Name: Lal George
2698    Date: 2001/11/14 09:44:04 EST
2699    Tag:
2700    Description:
2701    
2702      Fix for a backpatching bug reported by Allen.
2703    
2704      Because the boundary between short and long span-dependent
2705      instructions is +/- 128, there are an astounding number of
2706      span-dependent instructions whose size is over estimated.
2707    
2708      Allen came up with the idea of letting the size of span
2709      dependent instructions be non-monotonic, for a maxIter
2710      number of times, after which the size must be monotonically
2711      increasing.
2712    
2713      This table shows the number of span-dependent instructions
2714      whose size was over-estimated as a function of maxIter, for the
2715      file Parse/parse/ml.grm.sml:
2716    
2717         maxIter            # of instructions:
2718            10                      687
2719            20                      438
2720            30                      198
2721            40                        0
2722    
2723      In compiling the compiler, there is no significant difference in
2724      compilation speed between maxIter=10 and maxIter=40. Actually,
2725      my measurements showed that maxIter=40 was a tad faster than
2726      maxIter=10! Also 96% of the  files in the compiler reach a fix
2727      point within 13 iterations, so fixing maxIter at 40, while high,
2728      is okay.
2729    
2730    ----------------------------------------------------------------------
2731    Name: Matthias Blume
2732    Date: 2001/10/31 15:25:00 EST
2733    Tag: blume-20011031-pgraph
2734    Description:
2735    
2736    CKIT:
2737    * Changed the "Function" constructor of type Ast.ctype to carry optional
2738      argument identifiers.
2739    * Changed the return type of TypeUtil.getFunction accordingly.
2740    * Type equality ignores the argument names.
2741    * TypeUtil.composite tries to preserve argument names but gives up quickly
2742      if there is a mismatch.
2743    
2744    installation script:
2745    * attempts to use "curl" if available (unless "wget" is available as well)
2746    
2747    CM:
2748    * has an experimental implementation of "portable graphs" which I will
2749      soon propose as an implementation-independent library format
2750    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2751    
2752    NLFFI-LIB:
2753    * some cleanup (all cosmetic)
2754    
2755    NLFFIGEN:
2756    * temporarily disabled the mechanism that suppresses ML output for
2757      C definitions whose identifiers start with an underscore character
2758    * generate val bindings for enum constants
2759    * user can request that only one style (light or heavy) is being used;
2760      default is to use both (command-line arguments: -heavy and -light)
2761    * fixed bug in handling of function types involving incomplete pointers
2762    * generate ML entry points that take record arguments (i.e., using
2763      named arguments) for C functions that have a prototype with named
2764      arguments
2765      (see changes to CKIT)
2766    
2767    ----------------------------------------------------------------------
2768    Name: Allen Leung
2769    Date: 2001/10/27 20:34:00 EDT
2770    Tag: leunga-20011027-x86-fast-fp-call
2771    Description:
2772    
2773       Fixed the bug described in blume-20010920-slowfp.
2774    
2775       The fix involves
2776          1. generating FCOPYs in FSTP in ia32-svid
2777          2. marking a CALL with the appropriate annotation
2778    
2779    ----------------------------------------------------------------------
2780    Name: Matthias Blume
2781    Date: 2001/10/16 11:32:00 EDT
2782    Tag: blume-20011016-netbsd
2783    Description:
2784    
2785    Underscore patch from Chris Richards (fixing problem with compiling
2786    runtime system under recent NetBSD).
2787    
2788    ----------------------------------------------------------------------
2789    Name: Allen Leung
2790    Date: 2001/10/12 17:18:32 EDT 2001
2791    Tag: leung-20011012-x86-printflowgraph
2792    Description:
2793    
2794    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2795    
2796    ----------------------------------------------------------------------
2797    Name: Lal George
2798    Date: 2001/10/11 23:51:34 EDT
2799    Tag: george-20011011-too-many-instrs
2800    Description:
2801    
2802    The representation of a program point never expected to see more
2803    than 65536 instructions in a basic block!
2804    
2805    ----------------------------------------------------------------------
2806    Name: Lal George
2807    Date: 2001/10/09 09:41:37 EDT
2808    Tag: george-20011008-mlrisc-labels
2809    Description:
2810    
2811    Changed the machine description files to support printing of
2812    local and global labels in assembly code, based on host assembler
2813    conventions.
2814    
2815    ----------------------------------------------------------------------
2816    Name: Matthias Blume
2817    Date: 2001/09/25 15:25:00 EDT
2818    Tag: blume-20010925-exninfo
2819    Description:
2820    
2821    I provided a non-hook implementation of exnName (at the toplevel) and
2822    made the "dummy" implementation of exnMessage (at the toplevel) more
2823    useful: if nothing gets "hooked in", then at least you are going to
2824    see the exception name and a message indicating why you don't see more.
2825    
2826    [For the time being, programs that need exnMessage and want to use
2827    ml-build should either use General.exnMessage (strongly recommended) or
2828    refer to structure General at some other point so that CM sees a
2829    static dependency.]
2830    
2831    [Similar remarks go for "print" and "use":  If you want to use their
2832    functionality in stand-alone programs generated by ml-build, then use
2833    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2834    
2835    ----------------------------------------------------------------------
2836    Name: Matthias Blume
2837    Date: 2001/09/20 17:28:00 EDT
2838    Tag: blume-20010920-slowfp
2839    Description:
2840    
2841    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2842    
2843    ----------------------------------------------------------------------
2844    Name: Matthias Blume
2845    Date: 2001/09/20 17:20:00 EDT
2846    Tag: blume-20010920-canonicalpaths
2847    Description:
2848    
2849    0. Updated the BOOT file (something that I forgot to do earlier).
2850    
2851    1. Small internal change to CM so that it avoids "/../" in filenames
2852       as much as possible (but only where it is safe).
2853    
2854    2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2855       that arguments that contain delimiters are passed through correctly.
2856       This change also means that all "special" arguments of the form
2857       @SMLxxx... must come first.
2858    
2859    3. Changed install script to put relative anchor names for tool commands
2860       into pathconfig.
2861    
2862    ----------------------------------------------------------------------
2863    Name: Matthias Blume
2864    Date: 2001/09/18 15:35:00 EDT
2865    Tag: blume-20010918-readme11036
2866    Description:
2867    
2868    Added README files.
2869    
2870    ----------------------------------------------------------------------
2871    Name: Matthias Blume
2872    Date: 2001/09/18 11:45:00 EDT
2873    Tag: Release_110_36 (retag)
2874    Description:
2875    
2876    Fixed mistake in config/preloads. Retagged as 110.36.
2877    
2878    ----------------------------------------------------------------------
2879    Name: Matthias Blume
2880    Date: 2001/09/18 09:40:00 EDT
2881    Tag: Release_110_36_orig (tag changed)
2882    Description:
2883    
2884    New version (110.36).  New bootfiles.
2885    
2886    ----------------------------------------------------------------------
2887    Name: Matthias Blume
2888    Date: 2001/09/14 16:15:00 EDT
2889    Tag: blume-20010914-x86fastfp
2890    Description:
2891    
2892    John committed some changes that Allen made, in particular a (hopefully)
2893    correctly working version of the x86-fp module.
2894    
2895    I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
2896    flag to "true".  Everything seems to compile to a fixpoint ok, and
2897    "mandelbrot" speeds up by about 15%.
2898    
2899    ----------------------------------------------------------------------
2900    Name: Matthias Blume
2901    Date: 2001/09/13 11:20:00 EDT
2902    Tag: blume-20010913-minimal
2903    Description:
2904    
2905    1. Stefan Monnier's patch to fix a miscompilation problem that
2906       was brought to light by John Reppy's work on Moby.
2907    
2908    2. Implemented a minimal "structure Compiler" that contains just
2909       "version" and "architecture".  The minimal version will be
2910       available when the full version is not.  This is for backward-
2911       compatibility with code that wants to test Compiler.version.
2912    
2913    ----------------------------------------------------------------------
2914    Name: Matthias Blume
2915    Date: 2001/08/28 14:03:00 EDT
2916    Tag: blume-20010828-ml-lex
2917    Description:
2918    
2919    Fix for bug 1581, received from Neophytos Michael.
2920    
2921    ----------------------------------------------------------------------
2922    Name: Matthias Blume
2923    Date: 2001/08/27 11:20:00 EDT
2924    Tag: blume-20010827-readme11035
2925    Description:
2926    
2927    Fleshed out the README file for 110.35.
2928    
2929    ----------------------------------------------------------------------
2930    Name: Matthias Blume
2931    Date: 2001/08/24 17:10:00 EDT
2932    Tag: Release_110_35
2933    Description:
2934    
2935    New version number (110.35).  New bootfiles.
2936    
2937    ----------------------------------------------------------------------
2938    Name: Lal George
2939    Date: 2001/08/24 13:47:18 EDT 2001
2940    Tag: george-20010824-MLRISC-graphs
2941    Description:
2942    
2943     removed clusters from MLRISC completely and replaced with graphs.
2944    
2945    ----------------------------------------------------------------------
2946    Name: Matthias Blume
2947    Date: 2001/08/23 17:50:00 EDT
2948    Tag: blume-20010823-toplevel
2949    Description:
2950    
2951    - some reorganization of the code that implements various kinds of
2952      environments in the compiler (static, dynamic, symbolic, combined)
2953    - re-implemented the EnvRef module so that evalStream works properly
2954      (if the stream contains references to "use", "CM.make", etc.)
2955    - cleaned up evalloop.sml and interact.sml (but they need more cleaning)
2956    
2957    ----------------------------------------------------------------------
2958    Name: Matthias Blume
2959    Date: 2001/08/20 15:50 EDT
2960    Tag: blume20010820-slipup
2961    Description:
2962    
2963    I forgot to commit a few files.  Here they are...
2964    
2965    ----------------------------------------------------------------------
2966    Name: Matthias Blume
2967    Date: 2001/08/20 15:35:00 EDT
2968    Tag: blume-20010820-debugprof
2969    Description:
2970    
2971    !!!! NEW BOOTFILES !!!!
2972    
2973    This is another round of reorganizing the compiler sources.  This
2974    time the main goal was to factor out all the "instrumentation"
2975    passes (for profiling and backtracing) into their own library.
2976    The difficulty was to do it in such a way that it does not depend
2977    on elaborate.cm but only on elabdata.cm.
2978    
2979    Therefore there have been further changes to both elaborate.cm and
2980    elabdata.cm -- more "generic" things have been moved from the former
2981    to the latter.  As a result, I was forced to split the assignment
2982    of numbers indicating "primtyc"s into two portions: SML-generic and
2983    SML/NJ-specific.  Since it would have been awkward to maintain,
2984    I bit the bullet and actually _changed_ the mapping between these
2985    numbers and primtycs.  The bottom line of this is that you need
2986    a new set of bin- and bootfiles.
2987    
2988    I have built new bootfiles for all architectures, so doing a fresh
2989    checkout and config/install.sh should be all you need.
2990    
2991    The newly created library's name is
2992    
2993        $smlnj/viscomp/debugprof.cm
2994    
2995    and its sources live under
2996    
2997        src/compiler/DebugProf
2998    
2999    ----------------------------------------------------------------------
3000    Name: Matthias Blume
3001    Date: 2001/08/15 17:15:00 EDT
3002    Tag: blume-20010815-compreorg
3003    Description:
3004    
3005    This is a first cut at reorganizing the CM libraries that make up the
3006    core of the compiler.  The idea is to separate out pieces that could
3007    be used independently by tools, e.g., the parser, the typechecker, etc.
3008    
3009    The current status is a step in this direction, but it is not quite
3010    satisfactory yet.  Expect more changes in the future.
3011    
3012    Here is the current (new) organization...
3013    
3014        What used to be $smlnj/viscomp/core.cm is now divided into
3015        six CM libraries:
3016    
3017             $smlnj/viscomp/basics.cm
3018                           /parser.cm
3019                           /elabdata.cm
3020                           /elaborate.cm
3021                           /execute.cm
3022                           /core.cm
3023    
3024        The CM files for these libraries live under src/system/smlnj/viscomp.
3025        All these libraries are proxy libraries that contain precisely
3026        one CM library component.  Here are the locations of the components
3027        (all within the src/compiler tree):
3028    
3029             Basics/basics.cm
3030             Parse/parser.cm
3031             ElabData/elabdata.cm
3032             Elaborator/elaborate.cm
3033             Execution/execute.cm
3034             core.cm
3035    
3036         [This organization is the same that has been used already
3037         for a while for the architecture-specific parts of the visible
3038         compiler and for the old version of core.cm.]
3039    
3040         As you will notice, many source files have been moved from their
3041         respective original locations to a new home in one of the above
3042         subtrees.
3043    
3044         The division of labor between the new libraries is the following:
3045    
3046             basics.cm:
3047                - Simple, basic definitions that pertain to many (or all) of
3048                  the other libraries.
3049             parser.cm:
3050                - The SML parser, producing output of type Ast.dec.
3051                - The type family for Ast is also defined and exported here.
3052             elabdata.cm:
3053                - The datatypes that describe input and output of the elaborator.
3054                  This includes types, absyn, and static environments.
3055             elaborator.cm:
3056                - The SML/NJ type checker and elaborator.
3057                  This maps an Ast.dec (with a given static environment) to
3058                  an Absyn.dec (with a new static environment).
3059                - This libraries implements certain modules that used to be
3060                  structures as functors (to remove dependencies on FLINT).
3061             execute.cm:
3062                - Everything having to do with executing binary code objects.
3063                - Dynamic environments.
3064             core.cm:
3065                - SML/NJ-specific instantiations of the elaborator and MLRISC.
3066                - Top-level modules.
3067                - FLINT (this should eventually become its own library)
3068    
3069    Notes:
3070    
3071    I am not 100% happy with the way I separated the elaborator (and its
3072    data structures) from FLINT.  Two instances of the same problem:
3073    
3074        1. Data structures contain certain fields that carry FLINT-specific
3075           information.  I hacked around this using exn and the property list
3076           module from smlnj-lib.  But the fact that there are middle-end
3077           specific fields around at all is a bit annoying.
3078    
3079        2. The elaborator calculates certain FLINT-related information.  I tried
3080           to make this as abstract as I could using functorization, but, again,
3081           the fact that the elaborator has to perform calculations on behalf
3082           of the middle-end at all is not nice.
3083    
3084        3. Having to used exn and property lists is unfortunate because it
3085           weakens type checking.  The other alternative (parameterizing
3086           nearly *everything*) is not appealing, though.
3087    
3088    I removed the "rebinding =" warning hack because due to the new organization
3089    it was awkward to maintain it.  As a result, the compiler now issues some of
3090    these warnings when compiling init.cmi during bootstrap compilation. On
3091    the plus side, you also get a warning when you do, for example:
3092       val op = = Int32.+
3093    which was not the case up to now.
3094    
3095    I placed "assign" and "deref" into the _Core structure so that the
3096    code that deals with the "lazy" keyword can find them there.  This
3097    removes the need for having access to the primitive environment
3098    during elaboration.
3099    
3100    ----------------------------------------------------------------------
3101    Name: Matthias Blume
3102    Date: 2001/08/13
3103    Tag: blume-20010813-closures
3104    Description:
3105    
3106    This fix was sent to us by Zhong Shao.  It is supposed to improve the
3107    performance of certain loops by avoiding needless closure allocation.
3108    
3109    ----------------------------------------------------------------------
3110    Name: Lal George
3111    Date: 2001/07/31 10:03:23 EDT 2001
3112    Tag: george-20010731-x86-fmalloc
3113    Description: Fixed bug in x86 calls
3114    
3115        There was a bug where call instructions would mysteriously
3116        vanish. The call instruction had to be one that returned
3117        a floating point value.
3118    
3119    ----------------------------------------------------------------------
3120    Name: Lal George
3121    Date: 2001/07/19 16:36:29 EDT 2001
3122    Tag: george-20010719-simple-cells
3123    Description:
3124    
3125    I have dramatically simplified the interface for CELLS in MLRISC.
3126    
3127    In summary, the cells interface is broken up into three parts:
3128    
3129      1. CellsBasis : CELLS_BASIS
3130    
3131            CellsBasis is a top level structure and common for all
3132            architectures.  it contains the definitions of basic datatypes
3133            and utility  functions over these types.
3134    
3135      2. functor Cells() : CELLS
3136    
3137            Cells generates an interface for CELLS that incorporates the
3138            specific resources on the target architecture, such as the
3139            presence of special register classes, their number and size,
3140            and various useful substructures.
3141    
3142      3. <ARCH>CELLS
3143    
3144            e.g. SparcCells: SPARCCELLS
3145    
3146            <ARCH>CELLS usually contains additional bindings for special
3147            registers  on the architecture, such as:
3148    
3149                    val r0 : cell           (* register zero *)
3150                    val y : cell            (* Y register *)
3151                    val psr : cell          (* processor status register *)
3152                    ...
3153    
3154            The structure returned by applying the Cells functor is opened
3155            in this interface.
3156    
3157    The main implication of all this is that the datatypes for cells is
3158    split between CellsBasis and CELLS -- a fairly simple change for user
3159    code.
3160    
3161    In the old scheme the CELLS interface had a definitional binding of
3162    the form:
3163    
3164            signature CELLS = sig
3165    
3166               structure CellsBasis = CellsBasis
3167    
3168               ...
3169    
3170            end
3171    
3172    With all the sharing constraints that goes on in MLRISC, this old
3173    design  quickly leads to errors such as:
3174    
3175            "structure definition spec inside of sharing ... "
3176    
3177    
3178    and appears to require an unacceptable amount of sharing and where
3179    constraint hackery.
3180    
3181    I think this error message (the interaction of definitional specs and
3182    sharing) requires more explanation on our web page.
3183    
3184    ----------------------------------------------------------------------
3185    Name: Matthias Blume
3186    Date: 2001/07/19 15:00:00 EDT
3187    Tag: blume-20010719-libreorg
3188    Description:
3189    
3190    This update puts together a fairly extensive but straightforward change
3191    to the way the libraries that implement the interactive system are
3192    organized:
3193    
3194       The biggest change is the elimination of structure Compiler.  As a
3195       replacement for this structure, there is now a CM library
3196       (known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
3197       that exports all the substructures of the original structure Compiler
3198       directly.  So instead of saying Compiler.Foo.bar one now simply
3199       says Foo.bar.  (The CM libraries actually export a collection of
3200       structures that is richer than the collection of substructures of
3201       structure Compiler.)
3202    
3203       To make the transition smooth, there is a separate library called
3204       $smlnj/compiler/compiler.cm which puts together and exports the
3205       original structure Compiler (or at least something very close to it).
3206    
3207       There are five members of the original structure Compiler
3208       that are not exported directly but which instead became members
3209       of a new structure Backend (described by signature BACKEND).  These are:
3210       structure Profile (: PROFILE), structure Compile (: COMPILE), structure
3211       Interact (: INTERACT), structure Machine (: MACHINE), and val
3212       architecture (: string).
3213    
3214       Structure Compiler.Version has become structure CompilerVersion.
3215    
3216       Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
3217       by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
3218       or x86, respectively.
3219       Each of these exports the same frontend structures that
3220       $smlnj/compiler.cm exports.  But they do not have a structure Backend
3221       and instead export some structure <Arch>Backend where <Arch> is Alpha32,
3222       Hppa, PPC, Sparc, or X86, respectively.
3223    
3224       Library $smlnj/compiler/all.cm exports the union of the exports of
3225       $smlnj/compiler/<arch>.cm
3226    
3227       There are no structures <Arch>Compiler anymore, use
3228       $smlnj/compiler/<arch>.cm instead.
3229    
3230       Library host-compiler-0.cm is gone.  Instead, the internal library
3231       that instantiates CM is now called cm0.cm.  Selection of the host
3232       compiler (backend) is no longer done here but. (Responsibility for it
3233       now lies with $smlnj/compiler/current.cm.  This seems to be more
3234       logical.)
3235    
3236       Many individual files have been moved or renamed.  Some files have
3237       been split into multiple files, and some "dead" files have been deleted.
3238    
3239    Aside from these changes to library organization, there are also changes
3240    to the way the code itself is organized:
3241    
3242       Structure Binfile has been re-implemented in such a way that it no
3243       longer needs any knowledge of the compiler.  It exclusively deals
3244       with the details of binfile layout.  It no longer invokes the
3245       compiler (for the purpose of creating new prospective binfile
3246       content), and it no longer has any knowledge of how to interpret
3247       pickles.
3248    
3249       Structure Compile (: COMPILE) has been stripped down to the bare
3250       essentials of compilation.  It no longer deals with linking/execution.
3251       The interface has been cleaned up considerably.
3252    
3253       Utility routines for dealing with linking and execution have been
3254       moved into their own substructures.
3255    
3256       (The ultimate goal of these changes is to provide a light-weight
3257       binfile loader/linker (at least for, e.g., stable libraries) that
3258       does not require CM or the compiler to be present.)
3259    
3260    CM documentation has been updated to reflect the changes to library
3261    organization.
3262    
3263    ----------------------------------------------------------------------
3264    Name: Matthias Blume
3265    Date: 2001/07/10 17:30:00 EDT
3266    Tag: Release_110_34
3267    Description:
3268    
3269    Minor tweak to 110.34 (re-tagged):
3270    
3271      - README.html file added to CVS repository
3272      - runtime compiles properly under FreeBSD 3.X and 4.X
3273    
3274    ----------------------------------------------------------------------
3275    Name: Matthias Blume
3276    Date: 2001/07/10 17:30:00 EDT
3277    Tag: Release_110_34
3278    Description:
3279    
3280    New version number (110.34). New bootfiles.
3281    
3282    ----------------------------------------------------------------------
3283    Name: Matthias Blume
3284    Date: 2001/07/09 16:00:00 EDT
3285    Tag: blume-20010709-more-varargs
3286    Description:
3287    
3288    I changed the handling of varargs in ml-nlffigen again:
3289    The ellipsis ... will now simply be ignored (with an accompanying warning).
3290    
3291    The immediate effect is that you can actually call a varargs function
3292    from ML -- but you can't actually supply any arguments beyond the ones
3293    specified explicitly.  (For example, you can call printf with its format
3294    string, but you cannot pass additional arguments.)
3295    
3296    This behavior is only marginally more useful than the one before, but
3297    it has the advantage that a function or, more importantly, a function
3298    type never gets dropped on the floor, thus avoiding follow-up problems with
3299    other types that refer to the offending one.
3300    
3301    ----------------------------------------------------------------------
3302    Name: Matthias Blume
3303    Date: 2001/07/09 11:25:00 EDT
3304    Tag: blume-20010709-varargs
3305    Description:
3306    
3307    1. ckit-lib.cm now exports structure Error
3308    2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
3309       with a warning accompanied by a source location.  Moreover, it
3310       merely skips the offending function or type and proceeds with the
3311       rest of its work.u  As a result, one can safely feed C code containing
3312       "..." to ml-nlffigen.
3313    3. There are some internal improvements to CM, providing slightly
3314       more general string substitutions in the tools subsystem.
3315    
3316    ----------------------------------------------------------------------
3317    Name: Matthias Blume
3318    Date: 2001/06/27 15:10:00 EDT
3319    Tag: blume-20010627-concur
3320    Description:
3321    
3322    Fixed a small bug in CM's handling of parallel compilation.
3323    (You could observe the bug by Control-C-interrupting an ordinary
3324    CMB.make or CM.stabilize and then attaching some compile servers.
3325    The result was that all of a sudden the previously interrupted
3326    compilation would continue on its own.  This was because of
3327    an over-optimization: CM did not bother to clean out certain queues
3328    when no servers were attached "anyway", resulting in the contents
3329    of these queues to grab control when new servers did get attached.)
3330    
3331    There is also another minor update to the CM manual.
3332    
3333    ----------------------------------------------------------------------
3334    Name: Matthias Blume
3335    Date: 2001/06/26 16:15:00 EDT
3336    Tag: blume-20010626-cmdoc
3337    Description:
3338    
3339    Minor typo fixed in CM manual (syntax diagram for libraries).
3340    
3341    ----------------------------------------------------------------------
3342    Name: Matthias Blume
3343    Date: 2001/06/25 22:55:00 EDT
3344    Tag: blume-20010625-x86pc
3345    Description:
3346    
3347    Fixed a nasty bug in the X86 assembly code that caused signal
3348    handlers to fail (crash) randomly.
3349    
3350    ----------------------------------------------------------------------
3351    Name: Matthias Blume
3352    Date: 2001/06/25 12:05:00 EDT
3353    Tag: blume-20010625-nlffigen
3354    Description:
3355    
3356    This update fixes a number of minor bugs in ml-nlffigen as reported by
3357    Nick Carter <nbc@andrew.cmu.edu>.
3358    
3359      1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
3360      2. Default names for generated files are now derived from the name of
3361         the C file *without its directory*.  In particular, this causes generated
3362         files to be placed locally even if the C file is in some system directory.
3363      3. Default names for generated signatures and structures are also derived
3364         from the C file name without its directory.  This avoids silly things
3365         like "structure GL/GL".
3366         (Other silly names are still possible because ml-nlffigen does not do
3367          a thorough check of whether generated names are legal ML identifiers.
3368          When in doubt, use command line arguments to force particular names.)
3369    
3370    ----------------------------------------------------------------------
3371    Name: Matthias Blume
3372    Date: 2001/06/21 12:25:00 EDT
3373    Tag: blume-20010621-eXene
3374    Description:
3375    
3376    eXene now compiles and (sort of) works again.
3377    
3378    The library name (for version > 110.33) is $/eXene.cm.
3379    
3380    I also added an new example in src/eXene/examples/nbody.  See the
3381    README file there for details.
3382    
3383    ----------------------------------------------------------------------
3384    Name: Matthias Blume
3385    Date: 2001/06/20 16:40:00 EDT
3386    Tag: blume-20010620-cml
3387    Description:
3388    
3389    CML now compiles and works again.
3390    
3391    Libraries (for version > 110.33):
3392    
3393      $cml/cml.cm            Main CML library.
3394      $cml/basis.cm          CML's version of $/basis.cm.
3395      $cml/cml-internal.cm   Internal helper library.
3396      $cml/core-cml.cm       Internal helper library.
3397      $cml-lib/trace-cml.cm  Tracing facility.
3398      $cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
3399    
3400    The installer (config/install.sh) has been taught how to properly
3401    install this stuff.
3402    
3403    ----------------------------------------------------------------------
3404    Name: Matthias Blume
3405    Date: 2001/06/19 17:55:00 EDT
3406    Tag: blume-20010619-instantiate
3407    Description:
3408    
3409    This un-breaks the fix for bug 1432.
3410    (The bug was originally fixed in 110.9 but I broke it again some
3411    time after that.)
3412    
3413    ----------------------------------------------------------------------
3414    Name: Matthias Blume
3415    Date: 2001/06/19 17:25:00 EDT
3416    Tag: blume-20010619-signals
3417    Description:
3418    
3419    This should (hopefully) fix the long-standing signal handling bug.
3420    (The runtime system was constructing a continuation record with an
3421    incorrect descriptor which would cause the GC to drop data on the floor...)
3422    
3423    ----------------------------------------------------------------------
3424    Name: Matthias Blume
3425    Date: 2001/06/15 15:05:00 EDT
3426    Tag: blume-20010615-moresparc
3427    Description:
3428    
3429    Here is a short late-hour update related to Sparc c-calls:
3430    
3431     -- made handling of double-word arguments a bit smarter
3432    
3433     -- instruction selection phase tries to collapse certain clumsily
3434        constructed ML-Trees; typical example:
3435    
3436            ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
3437    
3438        This currently has no further impact on SML/NJ since mlriscGen does
3439        not seem to generate such patterns in the first place, and c-calls
3440        (which did generate them in the beginning) has meanwhile been fixed
3441        so as to avoid them as well.
3442    
3443    ----------------------------------------------------------------------
3444    Name: Matthias Blume
3445    Date: 2001/06/15 15:05:00 EDT
3446    Tag: blume-20010615-sparc
3447    Description:
3448    
3449    The purpose of this update is to provide an implementation of NLFFI
3450    on Sparc machines.
3451    
3452    Here are the changes in detail:
3453    
3454       * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
3455       the Sparc implementation of the c-calls API.
3456       * The Sparc backend of SML/NJ has been modified to uniformely use %fp
3457       for accessing the ML frame.  Thus, we have a real frame pointer and
3458       can freely modify %sp without need for an omit-frame-ptr phase.
3459       The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
3460       this case.
3461       * ml-nlffigen has been taught to produce code for different architectures
3462       and calling conventions.
3463       * In a way similar to what was done in the x86 case, the Sparc
3464       backend uses its own specific extension to mltree.  (For example,
3465       it needs to be able to generate UNIMP instructions which are part
3466       of the calling convention.)
3467       * ml-nlffi-lib was reorganized to make it more modular (in particular,
3468       to make it easier to plug in new machine- and os-dependent parts).
3469    
3470    There are some other fairly unrelated bug fixes and cleanups as well:
3471    
3472       * I further hacked the .cm files for MLRISC tools (like MDLGen) so
3473       that they properly share their libraries with existing SML/NJ libraries.
3474       * I fixed a minor cosmetic bug in CM, supressing certain spurious
3475       follow-up error messages.
3476       * Updates to CM/CMB documentation.
3477    
3478    TODO items:
3479    
3480       * MLRISC should use a different register as its asmTemp on the Sparc.
3481         (The current %o2 is a really bad choice because it is part of the
3482         calling conventions, so things might interfere in unexpected ways.)
3483    
3484    ----------------------------------------------------------------------
3485    Name: Matthias Blume
3486    Date: 2001/06/07
3487    Tag: blume-20010607-calls
3488    Description:
3489    
3490    A number of internal changes related to C calls and calling conventions:
3491    
3492    1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
3493       number of bytes popped implicitly (by the callee).  In most cases
3494       this field is 0 but on x86/win32 it is some non-zero value.  This
3495       is information provided for the benefit of the "omit-frameptr" pass.
3496    2. The CALL instruction on the x86 carries a similar "pops" field.
3497       The instruction selection phase copies its value from the ML-Tree
3498       CALL statement.
3499    3. On all other architectures, the instruction selection phase checks
3500       whether "pops=0" and complains if not.
3501    4. The c-calls implementation for x86 now accepts two calling conventions:
3502       "ccall" and "stdcall".  When "ccall" is selected, the caller cleans
3503       up after the call and pops is set to 0.  For "stdcall", the caller
3504       does nothing, leaving the cleanup to the callee; pops is set to
3505       the number of bytes that were pushed onto the stack.
3506    5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
3507       distinguish between "ccall" and "stdcall".
3508    6. The UNIMP instruction has been added to the supported Sparc instruction
3509       set. (This is needed for implementing the official C calling convention
3510       on this architecture.)
3511    7. I fixed some of the .cm files under src/MLRISC/Tools to make them
3512       work with the latest CM.
3513    
3514    ----------------------------------------------------------------------
3515    Name: Matthias Blume
3516    Date: 2001/06/05 15:10:00 EDT
3517    Tag: blume-20010605-cm-index
3518    Description:
3519    
3520    0. The "lambdasplit" parameter for class "sml" in CM has been documented.
3521    
3522    1. CM can now generate "index files".  These are human-readable files
3523       that list on a per-.cm-file basis each toplevel symbol defined or
3524       imported.  The location of the index file for
3525       <p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
3526       To enable index-file generation, set CM.Control.generate_index to true
3527       or export an environment-symbol: export CM_GENERATE_INDEX=true.
3528    
3529       The CM manual has been updated accordingly.
3530    
3531    2. I made some slight modifications to the c-calls API in MLRISC.
3532    
3533         a) There is now a callback to support saving/restoring of
3534            dedicated but caller-save registers around the actual call
3535            instruction.
3536         b) One can optionally specify a comment-annotation for the
3537            call instruction.
3538    
3539    3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
3540       (For example, the comment annotation shows the C prototype of
3541        the function being called.)
3542    
3543    ----------------------------------------------------------------------
3544    Name: Matthias Blume
3545    Date: 2001/06/01 13:30:00 EDT
3546    Tag: blume-20010601-nlffi-cleanup
3547    Description:
3548    
3549    This is mostly a cleanup of MLFFI stuff:
3550    
3551       - some signature files have been put into a more exposed place
3552       - the ugly 'f type parameter is gone (simplifies types tremendously!)
3553       - ml-nlffigen changed accordingly
3554       - tutorial updated
3555    
3556    Other changes:
3557    
3558       - author's affiliation in CM manual(s) updated
3559       - some more recognized keywords added to Allen's sml.sty
3560    
3561    ----------------------------------------------------------------------
3562    Name: Matthias Blume
3563    Date: 2001/05/25 15:30:00 EDT
3564    Tag: blume-20010525-iptr
3565    Description:
3566    
3567      - put the official 110.33-README (as it appears on the ftp server) under
3568        CVS
3569      - fixed a small bug related to incomplete pointer types in
3570        ml-nlffigen
3571      - small cosmetic change to the ml-nlffi-lib's "arr" type constructor
3572        (it does not need the 'f type parameter)
3573    
3574    ----------------------------------------------------------------------
3575    Name: Matthias Blume
3576    Date: 2001/05/23 14:30:00 EDT
3577    Tag: Release_110_33
3578    Description:
3579    
3580    New version number (110.33).  New bootfiles.
3581    
3582    ----------------------------------------------------------------------
3583    Name: Matthias Blume
3584    Date: 2001/05/22 18:06:00 EDT
3585    Tag: blume-20010522-targets
3586    Description:
3587    
3588    Made install.sh use file config/targets.customized if it exists, falling
3589    back to config/targets if it doesn't.  This way one can have a customized
3590    version of the targets file without touching the "real thing", thus
3591    eliminating the constant fear of accidentally checking something bogus
3592    back into the CVS repository...  (File config/targets.customized must
3593    not be added to the repository!)
3594    
3595    ----------------------------------------------------------------------
3596    Name: Matthias Blume
3597    Date: 2001/05/22 16:30:00 EDT
3598    Tag: blume-20010522-minitut
3599    Description:
3600    
3601    1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
3602       struct returns.
3603    2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
3604       incomplete, preliminary documentation for NLFFI.
3605    
3606    ----------------------------------------------------------------------
3607    Name: Matthias Blume
3608    Date: 2001/05/14 11:30:00 EDT
3609    Tag: blume-20010514-script
3610