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 1065, Thu Feb 14 15:05:16 2002 UTC revision 1482, Fri Apr 23 22:39:17 2004 UTC
# Line 13  Line 13 
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume (blume (at) tti - c (dot) org)
17    Date: 2004/04/23 17:40:00 CDT
18    Tag: blume-20040423-ieee-scan
19    Description:
20    
21    Fixed IEEEReal.scan (and .fromString) so that if there is an overflow
22    in the exponent calculation we get INF or ZERO (depending on the mantissa
23    and the sign of the exponent).
24    
25    ----------------------------------------------------------------------
26    Name: Matthias Blume (blume (at) tti - c (dot) org)
27    Date: 2004/04/23 10:40:00 CDT
28    Tag: blume-20040423-ml-build
29    Description:
30    
31    The ml-build script now terminates with a non-0 status when something
32    goes wrong.
33    
34    ----------------------------------------------------------------------
35    Name: Matthias Blume (blume (at) tti - c (dot) org)
36    Date: 2004/04/22 16:35:00 CDT
37    Tag: blume-20040422-Option
38    Description:
39    
40    Made exception Option to be the same as exception Option.Option
41    (as it should be).
42    
43    ----------------------------------------------------------------------
44    Name: Allen Leung (leunga (at) reservoir (dot) com)
45    Date: 2004/03/19 14:40:00 EST
46    Tag: leunga-20040319-cygwin-nlffi
47    Description:
48    
49    Fixed the runtime so that ml-nlffi-lib runs on the cygwin version
50    of SML/NJ.  The problem is that
51    
52       lib = dlopen(NULL, ...)
53       f   = dlsym(lib, "malloc");
54    
55    does not work on Windows unless we explicitly export symbols
56    such as 'malloc' during linking.  We fixed this by explicitly
57    exporting the required symbols with the magic gcc incantation:
58    
59        -Wl,--export-all cygwin.def
60    
61    where cygwin.def is a file containing all the symbols that
62    we wish to export.
63    
64    I suspect this is a Windows problem and we'll have to
65    do the same (somehow with windows compilers) when
66    we build the native win32 version with the system
67    calls LoadLibrary/GetProcAddress.
68    
69    ----------------------------------------------------------------------
70    Name: Matthias Blume (blume (at) tti - c (dot) org)
71    Date: 2004/03/04 16:35:00 CST
72    Tag: blume-20040304-intinf-fmt
73    Description:
74    
75    Fixed problem with IntInf.fmt (sign would show up on the right instead
76    of on the left for BIN, OCT, and HEX).
77    
78    ----------------------------------------------------------------------
79    Name: Matthias Blume (blume (at) tti - c (dot) org)
80    Date: 2004/03/04 11:25:00 CST
81    Tag: blume-20040304-symlinks
82    Description:
83    
84    Fixed problem with installer script (unix only) where bin/ml-yacc and
85    friends pointed (via symlinks) to absolute locations instead of just
86    .run-sml.  This was reported by Vesa A Norrman.
87    
88    ----------------------------------------------------------------------
89    Name: Matthias Blume (blume (at) tti - c (dot) org)
90    Date: 2004/02/13 14:50:00 CST
91    Tag: Release_110_45
92    Description:
93    
94    New working version (110.45). New bootfiles.
95    
96    ----------------------------------------------------------------------
97    Name: Matthias Blume (blume (at) tti - c (dot) org)
98    Date: 2004/01/26 15:15:15 CST
99    Tag: blume-20040126-toplevel
100    Description:
101    
102    Improved handling of exceptions at the interactive toplevel.
103    
104    ----------------------------------------------------------------------
105    Name: Matthias Blume (blume (at) tti - c (dot) org)
106    Date: 2004/01/26 11:25:00
107    Tag: blume-20040126-app
108    Description:
109    
110    Type of top-level "app" corrected.
111    Added code for setting vp_limitPtrMask to Win32-specific runtime.
112    
113    ----------------------------------------------------------------------
114    Name: Matthias Blume (blume (at) tti - c (dot) org)
115    Date: 2003/11/18 17:10 CST
116    Tag: blume-20031118-basis-fiddle
117    Description:
118    
119    - changed Timer interface to what might become the spec
120    - POSIX_FLAGS -> BIT_FLAGS according to spec
121    - some other minor discrepancies wrt. spec eliminated
122    
123    ----------------------------------------------------------------------
124    Name: Matthias Blume (blume (at) tti - c (dot) org)
125    Date: 2003/11/06 12:00:00 CST
126    Tag: Release_110_44
127    Description:
128    
129    New working version (110.44). New bootfiles.
130    
131    ----------------------------------------------------------------------
132    Name: Matthias Blume (blume (at) tti - c (dot) org)
133    Date: 2003/11/04 11:50:00 CST
134    Tag: blume-20031104-move-libraries
135    Description:
136    
137    Eliminated the "dont_move_libraries" directive in config/targets.
138    (The mechanism was broken and could not be fixed easily.  Moreover,
139    there does not seem to be any reason not to move all libraries into
140    lib during installation.  I originally implemented this directive as a
141    backward-compatibility feature when I first introduced the new CM.
142    Now that things have been stable for a long time and going back to the
143    old CM is not an option, there is no reason to keep it around.)
144    
145    ----------------------------------------------------------------------
146    Name: Matthias Blume (blume (at) tti - c (dot) org)
147    Date: 2003/11/03 16:00:00 CST
148    Tag: blume-20031103-installdir
149    Description:
150    
151    Made installer honor INSTALLDIR variable again.  (Thanks to Chris
152    Richards for pointing out the problem and providing the solution.)
153    
154    ----------------------------------------------------------------------
155    Name: Matthias Blume (blume (at) tti - c (dot) org)
156    Date: 2003/10/01 17:05:00 CDT
157    Tag: blume-20031001-lal-mlrisc
158    Description:
159    
160    MLRISC bug fix from Lal.
161    
162    ----------------------------------------------------------------------
163    Name: Matthias Blume (blume (at) tti - c (dot) org)
164    Date: 2003/09/30 16:10:00 CDT
165    Tag: blume-20030930-primio-bat
166    Description:
167    
168    1. Added openVector, nullRd, and nullWr to PRIM_IO.
169    2. Improved .bat files (for Win32 port) to make things work under Win95.
170       (thanks to Aaron S. Hawley for this one)
171    
172    ----------------------------------------------------------------------
173    Name: Matthias Blume (blume (at) tti - c (dot) org)
174    Date: 2003/09/26 16:05:00 CDT
175    Tag: blume-20030926-wrappriv
176    Description:
177    
178    Added missing wrapper for privilege "primitive" in $smlnj/viscomp/core.cm.
179    
180    ----------------------------------------------------------------------
181    Name: Matthias Blume (blume (at) tti - c (dot) org)
182    Date: 2003/09/26 15:00:00 CDT
183    Tag: blume-20030926-110_43_3
184    Description:
185    
186    - additional cleanup
187    - version number bump, NEW BOOTFILES
188    
189    ----------------------------------------------------------------------
190    Name: Matthias Blume (blume (at) tti - c (dot) org)
191    Date: 2003/09/26 12:00:00 CDT
192    Tag: blume-20030926-ppautoload
193    Description:
194    
195    I modified the read-eval-print loop so that the autoloader gets
196    invoked whenever the prettyprinter tries to look up a symbol that
197    is not currently defined in the toplevel environment but which
198    appears in CM's autoload registry.  As a result, we see far fewer of
199    those ?.Foo.Bar.xxx names in the prettyprinter's output.
200    
201    In addition to this I tried to clean up some pieces of the Basis
202    implementation (e.g., Socket, Word8Array) in order to prevent other
203    instances of these ?.Foo.Bar.xxx names from being printed.
204    
205    The mechanism that picks names for types still needs some work, though.
206    (Right now it seems that if there is a type A.t which is defined to
207    be B.u, but B is unavailable at toplevel, then A.t gets printed as
208    "?.B.u" although the perhaps more sensible solution would be to use
209    "A.t" in this case.  In other words, the prettyprinter should follow
210    a chain of DEFtycs not farther than there are corresponding toplevel
211    names in the current environment.)
212    
213    ----------------------------------------------------------------------
214    Name: Matthias Blume (blume (at) tti - c (dot) org)
215    Date: 2003/09/24 16:31:00 CDT
216    Tag: blume-20030924-installer
217    Description:
218    
219    Another installer tweak:  All the ML code for the installer is now
220    compiled during CMB.make and put into a little library called
221    $smlnj/installer.cm.  The installation then simply invokes
222    
223       sml -m $smlnj/installer.cm
224    
225    and everything happens automagically.
226    
227    Win32:  ML code senses value of environment variable SMLNJ_HOME.
228    Unix:   ML code senses values of environment variables ROOT, CONFIGDIR,
229            and BINDIR.
230    
231    The new scheme guarantees that the ML code responsible for the installation
232    is in sync with the APIs of the main system.  Also, the installer is
233    somewhat faster because the installer script is precompiled.
234    
235    ----------------------------------------------------------------------
236    Name: Matthias Blume (blume (at) tti - c (dot) org)
237    Date: 2003/09/24 15:35:00 CDT
238    Tag: blume-20030924-synsock
239    Description:
240    
241    Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
242    but excludes all non-blocking operations.  Defined SOCKET (in Basis)
243    and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
244    implementations of non-blocking operations from CML's Socket
245    structure.
246    
247    ----------------------------------------------------------------------
248    Name: Matthias Blume (blume (at) tti - c (dot) org)
249    Date: 2003/09/24 15:10:05 CDT
250    Tag: blume-20030924-sockets
251    Description:
252    
253    1. Fixed SOCKET API and implementation to match Basis spec.
254       This required changing the internal representation of sockets to one
255       that remembers (for each socket file descriptor) whether it is currently
256       blocking or non-blocking.  This state is maintained lazily (i.e., a system
257       call is made only if the state actually needs to change).
258    
259    2. OS-specific details of sockets were moved into separate files, thus
260       making it possible to unify the bulk of the socket implementations
261       between Unix and Win32.
262    
263    3. CML's socket API changed accordingly.
264       (Note that we need to remove non-blocking functions from this API
265       since they are redundant in the case of CML!)
266    
267    4. CML's socket implementation now makes use of non-blocking functions
268       provided by Basis, thus removing all OS-dependent code from this part
269       of CML.
270    
271    5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
272    
273    ----------------------------------------------------------------------
274    Name: Matthias Blume (blume (at) tti - c (dot) org)
275    Date: 2003/09/22 12:10:00 CDT
276    Tag: blume-20030922-110_43_2
277    Description:
278    
279    Made a new interim version and bootfiles for developer's bootstrapping
280    convenience.
281    
282       110.43.2 -- NEW BOOTFILES
283    
284    ----------------------------------------------------------------------
285    Name: Matthias Blume (blume (at) tti - c (dot) org)
286    Date: 2003/09/19 15:55:00 CDT
287    Tag: blume-20030919-cmdir
288    Description:
289    
290    1. new-install.sh -> install.sh
291    2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
292    3. tweaked installer so that another name instead of .cm can be chosen
293       at install time (by setting the CM_DIR_ARC environment variable
294       during installation); once installation is complete, the name is
295       fixed
296    
297    ----------------------------------------------------------------------
298    Name: Matthias Blume (blume (at) tti - c (dot) org)
299    Date: 2003/09/18 16:00:00 CDT
300    Tag: blume-20030918-110_43_1
301    Description:
302    
303    Made a new interim version and bootfiles for developer's bootstrapping
304    convenience.
305    
306       110.43.1 -- NEW BOOTFILES
307    
308    ----------------------------------------------------------------------
309    Name: Matthias Blume (blume (at) tti - c (dot) org)
310    Date: 2003/09/18 15:20:00 CDT
311    Tag: blume-20030918-misc
312    Description:
313    
314    1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
315       this seems to be controversial at the moment) and used those in
316       Posix.ProcEnv.times.
317    
318    2. Added Time.{from,to}Nanoseconds to Time.
319    
320    3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
321       For example, fromLargeInt never needs to look at more than 3 "big
322       digits" to get its 53 bits of precision.
323    
324    ----------------------------------------------------------------------
325    Name: Matthias Blume (blume (at) tti - c (dot) org)
326    Date: 2003/09/17 16:30:00 CDT
327    Tag: blume-20030917-real32-slices
328    Description:
329    
330    Added an entry to the primitive environment
331    (compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
332    added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
333    
334    Removed some of the "magic" constants in real64.sml and replaced them
335    with code that generates these values from their corresponding
336    integer counterparts.
337    
338    Made all(?) the slice-related changes to the Basis and made everything
339    compile again...
340    
341    ----------------------------------------------------------------------
342    Name: Matthias Blume (blume (at) tti - c (dot) org)
343    Date: 2003/09/15 17:45:00 CDT
344    Tag: blume-20030915-rbase
345    Description:
346    
347    Fixed bug in Real.fromLargeInt.
348    
349    ----------------------------------------------------------------------
350    Name: Matthias Blume (blume (at) tti - c (dot) org)
351    Date: 2003/09/13 18:11:00 CDT
352    Tag: blume-20030913-libinstall
353    Description:
354    
355    Minor bugfix in config/libinstall (set anchor with path to
356    standalone tool after installing it, otherwise libraries that
357    need ml-lex or ml-yacc won't compile the first time the installer
358    runs).
359    
360    ----------------------------------------------------------------------
361    Name: Matthias Blume (blume (at) tti - c (dot) org)
362    Date: 2003/09/12 11:45:00 CDT
363    Tag: blume-20030912-various
364    Description:
365    
366     - fixed bug in Real.toLargeInt
367     - fixed bug in Posix.ProcEnv.times
368     - changed inputLine functions to return an option
369     - minor installer improvements / bugfixes
370     - changed default @SMLalloc parameter for x86/celeron to 64k
371    
372    ----------------------------------------------------------------------
373    Name: Matthias Blume (blume (at) tti - c (dot) org)
374    Date: 2003/09/09 22:00:00 CDT
375    Tag: Release_110_43
376    Description:
377    
378        New working release 110.43.  New bootfiles.
379    
380    ----------------------------------------------------------------------
381    Name: Matthias Blume (blume (at) tti - c (dot) org)
382    Date: 2003/09/09 19:20:00 CDT
383    Tag: blume-20030909-installer
384    Description:
385    
386    Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
387    Modified config/install.bat to take advantage of it.  Also modified
388    config/install.sh (and called it config/new-install.sh) to take advantage
389    of it on Unix systems.  (The SML code is (supposed to be) platform-
390    independent.)
391    
392    The installer can now install everything under Win32
393    as well as under *nix as long as it compiles.
394    
395    Other changes:
396    
397      - made CML compile again under Win32
398      - made eXene compile under Win32 (by providing a fake structure UnixSock
399        and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
400      - fixed a bug in nowhere: it assumed that type OS.Process.status is the
401        same as type int;  under Win32 it isn't
402      - fixed some slice-related problems in the win32-specific parts of CML
403      - added a functor argument "sameVol" to os-path-fn.sml in the Basis
404        (under Win32, the volume name is case-insensitive, and the
405        OS.Path code compares volume names for equality)
406    
407    ----------------------------------------------------------------------
408    Name: Matthias Blume (blume (at) tti - c (dot) org)
409    Date: 2003/09/08 11:55:00 CDT
410    Tag: blume-20030908-fullpath
411    Description:
412    
413    Made Win32 version of OS.FileSys.fullPath return current directory
414    when given an empty string.  This is what the spec says, and incidentally,
415    CM depends on it.  (CM otherwise goes into an infinite loop in certain
416    cases when presented with the name of a non-existing .cm file.)
417    
418    ----------------------------------------------------------------------
419    Name: Matthias Blume (blume (at) tti - c (dot) org)
420    Date: 2003/09/04 16:30:00 CDT
421    Tag: blume-20030905-slices-etc
422    Description:
423    
424      1. Changed interface to vectors and arrays in Basis to match
425         (draft) Basis spec.
426      2. Added signatures and implementations of slices according to
427         Basis spec.
428      3. Edited source code throughout the system to make it compile again
429         under 1. and 2.  (In some cases code had to be added to have it
430         match the new signatures.)
431      4. MLRISC should be backward-compatible: the copies of the originals
432         of files that needed to change under 3. were retained, the .cm files
433         check the compiler version number and use old versions when
434         appropriate.
435      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
436         dirstream -> string option (in accordance with Basis spec).
437      6. When generating code that counts lines, ml-lex used function
438         CharVector.foldli, taking advantage of its old interface.
439         This has been replaced with the corresponding code from
440         CharVectorSlice.  (html-lex must be re-lexed!)
441      7. BitArray in smlnj-lib/Util has been extended/modified to match the
442         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
443      8. Removed temporary additions (fromInternal, toInternal) from the
444         (now obsolete) IntInf in smlnj-lib/Util.
445      9. Cleaned up structure Byte.
446     10. Added localOffset, scan, and fromString to Date (according to spec).
447         Cleaned/corrected implementation of Date.
448         (Still need to check for correctness; implement better canonicalizeDate.)
449     11. Added "scan" to signature IEEE_REAL.
450     12. Some improvements to IntInf [in particular: efficiency-hack for
451         mod and rem when second operand is 2 (for parity checks).]
452     13. Changed representation of type Time.time, using a single IntInf.int
453         value counting microseconds.  This considerably simplified the
454         implementation of structure Time.  We now support negative time
455         values; scan and fromString handle signs.
456     14. Functor PrimIO now takes two additional arguments (VectorSlice and
457         ArraySlice).
458    
459    ----------------------------------------------------------------------
460    Name: Matthias Blume (blume (at) tti - c (dot) org)
461    Date: 2003/08/28 17:00:00 CDT
462    Tag: blume-20030828-intinf
463    Description:
464    
465    This is a major update which comes with a version number bump
466    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
467    and an implementation of IntInf in the Basis.
468    
469    There are a fairly large number of related changes and updates throughout
470    the system:
471    
472    Basis:
473            - Implemented IntInf.
474            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
475            - Added some missing Real64 operations, most notably Real.toLargeInt.
476            - Added FixedInt as a synonym for Int32.
477    
478    compiler:
479            * Added support for a built-in intinf type.
480              - literals
481              - pattern matching
482              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
483              - overloading on literals and operations
484    
485            This required adding a primitive type intinf, some additional
486            primops, and implementations for several non-trivial intinf
487            operations in Core.  (The intinf type is completely abstract
488            to the compiler; all operations get delegated back to the Core.)
489    
490            * Intinf equality is handled by polyequal.  However, the compiler
491            does not print its usual warning in this case (since polyequal
492            is the right thing to do there).
493    
494            * Improved the organization of structure InlineT.
495    
496            * A word about conversion primops:
497            If conversions involving intinf do not cancel out during
498            CPS contract, then the compiler must insert calls to Core functions.
499            Since all core access must be resolved already during the FLINT
500            translate phase, it would be too late a the time of CPS contract
501            to add new Core calls.  For this reason, conversion primops
502            for intinf carry two arguments:  1. the numeric argument that
503            they are supposed to convert, and 2. the Core function that
504            can help with this conversion if necessary.  If CPS contract
505            eliminates a primop, then the associated Core function becomes
506            dead and goes away.  Intinf conversion primops that do not get
507            eliminated by CPS contract get rewritten into calls of their
508            core functions by a separate, new phase.
509    
510    interactive system:
511            - Control.Print.intinfDepth controls max length of intinf constants
512            being printed. (Analogous to Control.Print.stringDepth.)
513            - Cleanup in printutil and pputil: got rid of unused stuff and
514            duplicates; replaced some of the code with code that makes better
515            use of library functionality.
516    
517    CM:
518            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
519            are no longer silent.
520    
521    CKIT:
522            Fixed mismatched uses of Int32 and LargeInt.  I always decided
523            in favor of LargeInt -- which is now the same as IntInf.
524            CKIT-knowledgable people should check whether this is what's
525            intended and otherwise change things back to using Int32 or
526            FixedInt.
527    
528    Throughout the code:
529            Started using IntInf.int literals and built-in operations
530            (e.g., comparison with 0) where this seems appropriate.
531    
532    
533    ----------------------------------------------------------------------
534    Name: Dave MacQueen (dbm@cs.uchicago.edu)
535    Date: 2003/08/13 11:36:00 CDT
536    Tag: dbm-20030813-mcz-merge1
537    Description:
538    
539    Merging changes from the mcz-branch development branch into trunk.
540    These changes involve replacement of the emulated old prettyprinter
541    interface with direct use of the SML/NJ Lib PP library, and fixing
542    of a couple of bugs (895, 1186) relating to error messages.  A new
543    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
544    has been added.
545    
546    ----------------------------------------------------------------------
547    Name: Matthias Blume (blume (at) tti - c (dot) org)
548    Date: 2003/08/11 15:45:00 CDT
549    Tag: blume-20030811-windows
550    Description:
551    
552    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
553    
554        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
555    
556    This patch restores SML/NJ's ability to run under win32.  There are a
557    number of changes, including fixes for several bugs that had gone
558    unnoticed until now:
559    
560      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
561        trying to run the win32 version from within cygwin.)
562    
563      - There are a number of simple .bat scripts that substitute for their
564        corresponding Unix shell-scripts.  (See below.)
565    
566      - The internals of ml-build have been modified slightly.  The main
567        difference is that instead of calling ".link-sml" (or link-sml.bat)
568        using OS.Process.system, the ML process delegates this task back
569        to the script.  Otherwise problems arise in mixed environments such
570        as Cygwin where scripts look and work like Unix scripts, but
571        where OS.Process.system cannot run them.
572    
573      - In CM, the srcpath pickler used native pathname syntax -- which
574        is incorrect in the case of cross-compilation.  The new pickle format
575        is independent of platform-specific naming conventions.
576    
577      - Path configuration files (such as lib/pathconfig) can now choose
578        between native and standard syntax.  Placing a line of the form
579    
580            standard!
581    
582        into the file causes all subsequent paths to be interpreted using
583        CM standard pathname syntax (= Unix conventions); a line
584    
585            native!
586    
587        switches back to native style.  This was needed so that
588        path config files can be written portably, see src/system/pathconfig.
589    
590      - Runtime system:
591    
592          - win32-filesys.c:  get_file_time and set_file_time now
593            access modification time, not creation time.
594    
595          - I/O code made aware of new array representation.
596    
597          - Bug fixes in X86.prim.masm.
598    
599       - src/system/makeml made aware of win32. (For use under cygwin
600            and other Unix-environments for windows.)
601    
602       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
603         which caused BinIO.inputAll to fail consistently.
604    
605    .bat scripts:
606    
607       Windows .bat scripts assume that SMLNJ_HOME is defined.
608    
609       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
610         applications (sml, ml-yacc, ml-lex).
611       - ml-build.bat: analogous to ml-build.
612       - config\install.bat:  Analogous to config/install.sh.  This requires
613         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
614         (nmake etc. must be on the path, and vcvars32 must have been run.)
615         Moreover, sources for ml-lex and ml-yacc need to exist under src,
616         and the bootfile hierarchy must have been unpacked under
617         sml.boot.x86-win32.
618         The script is very primitive and does a poor job at error checking.
619         It only installs the base system, ml-lex, and ml-yacc.  No other
620         libraries are being installed (i.e., you get only those that
621         are part of the compiler.)
622       - link-sml.bat:  analogous to .link-sml, but not currently used
623    
624    Unrelated bug fixes:
625    
626      - ml-nlffigen now exports structures ST_* corresponding to incomplete
627        types.
628      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
629        otherwise.)
630    
631    ----------------------------------------------------------------------
632    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
633    Date: 2003/06/17
634    Tag: macqueen-20030617-bug895
635    Description:
636    
637    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
638    Tag will be used for new development branch (mcz-branch) for
639    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
640    
641    ----------------------------------------------------------------------
642    Name: Matthias Blume (blume (at) tti - c (dot) org)
643    Date: 2003/05/27 16:55:00 CDT
644    Tag: blume-20030527-polyeq
645    Description:
646    
647    Tried to eliminated most cases of polymorphic equality.
648    
649    ----------------------------------------------------------------------
650    Name: Matthias Blume (blume (at) tti - c (dot) org)
651    Date: 2003/05/21 17:45:00 CDT
652    Tag: blume-20030517-complete
653    Description:
654    
655    Two changes:
656    
657      1. Added a flag for controlling whether non-exhaustive bindings will
658         be treated as errors (default is false).
659      2. Cleaned up the *entire* source tree so that CMB.make goes through
660         without a single non-exhaustive match- or bind warning.
661    
662    ----------------------------------------------------------------------
663    Name: Matthias Blume (blume (at) tti - c (dot) org)
664    Date: 2003/05/17 10:20:00 CDT
665    Tag: blume-20030517-absyn
666    Description:
667    
668    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
669    
670       This mainly affects the quality of error messages.  However, some
671       of the code is now more straightforward than before. (Treatment of
672       the above four constructs in translate.sml is much simpler than
673       the "macro-expansion" that was going on before.  Plus, the mach-
674       compiler no longer gets invoked just to be able to compile an
675       if-expression.)
676    
677    2. The ErrorMsg.Error exception is now caught and absorbed by the
678       interactive loop.
679    
680    ----------------------------------------------------------------------
681    Name: Allen Leung
682    Date: 2003/05/16 13:05:00 CDT
683    Tag: leunga-20030516-cygwin-runtime
684    Description:
685    
686        Ported the runtime system to cygwin, which uses the unix
687    x86-unix bin files.  Missing/buggy features:
688    
689       o getnetbyname, getnetbyaddr: these functions seem to be missing in
690         the Cygwin library.
691       o Ctrl-C handling may be flaky.
692       o Windows system calls and Windows I/O are not supported.
693    
694       A new set of binfiles is located at:
695    
696         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
697    
698       This is only needed for bootstrapping the cygwin version of smlnj.
699       Other x86 versions can use the existing binfiles.
700    
701    ----------------------------------------------------------------------
702    Name: Matthias Blume
703    Date: 2003/04/08 15:42:00 CDT
704    Tag: blume-20030408-listpair
705    Description:
706    
707    1. Added a target 'mlrisc' to installer.
708    
709    2. Added missing elements to structure ListPair.
710    
711    ----------------------------------------------------------------------
712    Name: Allen Leung
713    Date: 2003/01/07 10:40:00 EST
714    Tag: leunga-20030107-int-rem
715    Description:
716    
717    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
718    The arguments to the SUBL instruction were swapped.
719    
720    ----------------------------------------------------------------------
721    Name: Matthias Blume
722    Date: 2002/12/12 16:25:00 EST
723    Tag: blume-20021212-risc-ra
724    Description:
725    
726    Fixed a serious bug in the rewrite code for FP spilling/reloading that
727    sent the RA into an infinite loop when floating point registers get
728    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
729    110.37 and 110.38.)
730    There was another set of potential problems related to the handling of
731    MLRISC annotations (but those did not yet cause real problems, apparently).
732    
733    ----------------------------------------------------------------------
734    Name: Matthias Blume
735    Date: 2002/12/06 22:40:00 EST
736    Tag: blume-20021206-cm-fileid
737    Description:
738    
739    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
740    This fixes the problem of CM getting confused by files that suddenly
741    change their identity (e.g., by getting unlinked and recreated by some
742    text editor such as vi).  There might be a better/cheaper/cleaner way
743    of doing this, but for now this will have to do.
744    
745    ----------------------------------------------------------------------
746    Name: Matthias Blume
747    Date: 2002/10/28 09:50:00 EST
748    Tag: blume-20021028-typecheck
749    Description:
750    
751    Exported structure Typecheck from $smlnj/viscomp/core.cm.
752    
753    ----------------------------------------------------------------------
754    Name: Matthias Blume
755    Date: 2002/10/17 09:10:00 EDT
756    Tag: Release_110_42
757    Description:
758    
759    In good old tradition, there has been a slight hiccup so that we have
760    to patch 110.42 after the fact.   The old release tag has been replaced
761    (see below).
762    
763    The change solves a problem with two competing approaches the
764    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
765    each other's way.
766    
767    This change only affects the runtime system code and the installer script.
768    (No new bootfiles.)
769    
770    ----------------------------------------------------------------------
771    Name: Matthias Blume
772    Date: 2002/10/16 12:00:00 EDT
773    Tag: Release_110_42_removed
774    Description:
775    
776    New working release.  New bootfiles.
777    
778    ----------------------------------------------------------------------
779    Name: Matthias Blume
780    Date: 2002/10/10 13:10:00 EDT
781    Tag: blume-20021010-ppc-divs
782    Description:
783    
784    The mltree operator DIVS must be implemented with an overflow check on
785    the PPC because the hardware indicates divide-by-zero using "overflow" as
786    well.
787    
788    ----------------------------------------------------------------------
789    Name: Matthias Blume
790    Date: 2002/07/23 11:45:00
791    Tag: blume-20020723-smlnj-home
792    Description:
793    
794    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
795    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
796    CM_PATHCONFIG is also set), the path configuration file is assumed
797    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
798    the entire tree to some other place and everything will "just work".
799    
800    (Companion commands such as ml-build and ml-makedepend also sense this
801    variable.)
802    
803    ----------------------------------------------------------------------
804    Name: Matthias Blume
805    Date: 2002/07/12 21:19:00 EDT
806    Tag: blume-20020712-liveness
807    Description:
808    
809    Exported two useful "step" functions from liveness module (MLRISC).
810    
811    ----------------------------------------------------------------------
812    Name: Matthias Blume
813    Date: 2002/07/05 16:00 EDT
814    Tag: Release_110_41
815    Description:
816    
817    New working release.  New bootfiles.
818    
819    ----------------------------------------------------------------------
820    Name: Matthias Blume
821    Date: 2002/07/05 10:25:00 EDT
822    Tag: blume-20020705-btimp
823    Description:
824    
825    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
826    other clients can set up backtracing support.
827    
828    ----------------------------------------------------------------------
829    Name: Matthias Blume
830    Date: 2002/06/25 17:23:00 EDT
831    Tag: blume-20020625-fpmax
832    Description:
833    
834    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
835    case.  (The sense of the isNaN test was reversed -- which made min and
836    max always return their first argument.)
837    
838    ----------------------------------------------------------------------
839    Name: Matthias Blume
840    Date: 2002/06/11
841    Tag: blume-20020611-unixpath
842    Description:
843    
844    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
845    
846    ----------------------------------------------------------------------
847    Name: Matthias Blume
848    Date: 2002/06/10 16:35:00 EDT
849    Tag: blume-20020610-ieeereal
850    Description:
851    
852    I back-ported my implementation of IEEEReal.fromString from the
853    idlbasis-devel branch so that we can test it.
854    
855    Another small change is that ppDec tries to give more information
856    than just "<sig>" in the case of functors.  However, this code is
857    broken in some mysterious way if the functor's body's signature
858    has not been declared by ascription but gets inferred from the
859    implementation.  This needs fixing...
860    
861    ----------------------------------------------------------------------
862    Name: Matthias Blume
863    Date: 2002/05/31
864    Tag: blume-20020531-btrace-mode
865    Description:
866    
867    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
868    the wayside when I switched over to using Controls everywhere.)
869    
870    ----------------------------------------------------------------------
871    Name: Lal George
872    Date: 2002/05/23 12:21:40 EDT
873    Tag: george-20020523-visual-labels
874    Description:
875    
876    Labels are now displayed in the graphical output to make
877    the fall-through and target blocks obvious.
878    
879    ----------------------------------------------------------------------
880    Name: Matthias Blume
881    Date: 2002/05/22 11:03:00 EDT
882    Tag: blume-20020522-shrink
883    Description:
884    
885    John tweaked yesterday's fix for 1131 to handle an out-of-memory
886    situation that comes up when allocating huge arrays.
887    
888    ----------------------------------------------------------------------
889    Name: Matthias Blume
890    Date: 2002/05/21 16:00:00 EDT
891    Tag: Release_110_40
892    Description:
893    
894    New working release (110.40).  New bootfiles.
895    
896    [Also: John Reppy fixed GC bug 1131.]
897    
898    ----------------------------------------------------------------------
899    Name: Matthias Blume
900    Date: 2002/05/21 12:35:00 EDT
901    Tag: blume-20020521-cmdoc
902    Description:
903    
904    CM documentation update.
905    
906    ----------------------------------------------------------------------
907    Name: Matthias Blume
908    Date: 2002/05/21 10:55:00 EDT
909    Tag: blume-20020521-misc
910    Description:
911    
912    - John tweaked runtime to be silent on heap export (except when
913      GC messages are on).
914    - I added a few more things (cross-compiling versions of CMB) to
915      config/preloads (as suggestions).
916    
917    ----------------------------------------------------------------------
918    Name: Matthias Blume
919    Date: 2002/05/20 22:25:00 EDT
920    Tag: blume-20020520-controls
921    Description:
922    
923    - Added ControlUtil structure to control-lib.cm.
924    - Use it throughout.
925    - Used Controls facility to define MLRISC controls (as opposed to
926      registering MLRISC control ref cells with Controls after the fact)
927    - Fixed messed-up controls priorities.
928    
929    * Removed again all the stuff from config/preloads that one wouldn't
930      be able to preload at the time the initial heap image is built.
931      (Many libraries, e.g., CML, do not exist yet at this time.  The
932       only libraries that can be preloaded via config/preloads are those
933       that come bundled with the bootfiles.)
934    
935    ----------------------------------------------------------------------
936    Name: Matthias Blume
937    Date: 2002/05/20 10:59:00 EDT
938    Tag: blume-20020520-preloads
939    Description:
940    
941    Added a lot of commented-out suggestions for things to be included
942    in config/preloads.
943    
944    ----------------------------------------------------------------------
945    Name: Allen Leung
946    Date: 2002/05/18 14:20:00 EDT
947    Tag: leunga-20020518-mdl
948    Description:
949    
950    o Made the mdl tool stuff compile and run again.
951    o I've disabled all the stuff that depends on RTL specifications; they
952      are all badly broken anyway.
953    
954    ----------------------------------------------------------------------
955    Name: Matthias Blume
956    Date: 2002/05/17 16:49:00 EDT
957    Tag: blume-20020517-controls
958    Description:
959    
960    0. John Reppy made several modifications to the SML/NJ library.
961       In particular, there is a shiny new controls-lib.cm.
962    
963    1. Pushed new controls interface through compiler so that everything
964       compiles again.
965    
966    2. Added FormatComb and FORMAT_COMB to the CML version of the
967       SML/NJ library (so that CML compiles again).
968    
969    3. Modified init scripts because XXX_DEFAULT environment variables
970       are no longer with us.  (Boot-time initialization is now done
971       using the same environment variables that are also used for
972       startup-time initialization of controls.)
973    
974    ----------------------------------------------------------------------
975    Name: Lal George
976    Date: 2002/05/15 09:20:10 EDT
977    Tag: george-20020515-pseudo-op-decls
978    Description:
979    
980            All pseudo-ops emitted before the first segment declaration
981    such as TEXT, DATA, and BSS directives are assumed to be global
982    declarations and are emitted first in the assembly file. This is
983    useful in a number of situations where one has pseudo-ops that are not
984    specific to any segment, and also works around the constraint that one
985    cannot have client pseudo-ops in the TEXT segment.
986    
987            Because no segment is associated with these declarations it is
988    an error to allocate any space or objects before the first segment
989    directive and an exception will be raised. However, we cannot make
990    this check for client pseudo-ops.
991    
992            These top level declarations are a field in the CFG graph_info.
993    In theory you can continue to add to this field after the CFG has been
994    built -- provided you know what you are doing;-)
995    
996    ----------------------------------------------------------------------
997    Name: Matthias Blume
998    Date: 2002/05/13 16:40:00 EDT
999    Tag: blume-20020513-pp-etc
1000    Description:
1001    
1002    A few minor bugfixes:
1003    
1004      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
1005        (Bogus printouts for redefined bindings still occur.  Compiler
1006        bug should no longer occur now.  We need to redo the prettyprinter
1007        from scratch.)
1008    
1009      - CM pathname printer now also adds escape sequences for ( and )
1010    
1011      - commend and docu fixes for ml-nlffi
1012    
1013    ----------------------------------------------------------------------
1014    Name: Matthias Blume
1015    Date: 2002/05/10 16:40:00 EDT
1016    Tag: blume-20020510-erg-textio
1017    Description:
1018    
1019    Applied the following bugfix provided by Emden Gansner:
1020    
1021        Output is corrupted when outputSubstr is used rather than output.
1022    
1023        The problem occurs when a substring
1024    
1025            ss = (s, dataStart, dataLen)
1026    
1027        where dataStart > 0, fills a stream buffer with avail bytes left.
1028        avail bytes of s, starting at index dataStart, are copied into the
1029        buffer, the buffer is flushed, and then the remaining dataLen-avail
1030        bytes of ss are copied into the beginning of the buffer. Instead of
1031        starting this copy at index dataStart+avail in s, the current code
1032        starts the copy at index avail.
1033    
1034        Fix:
1035        In text-io-fn.sml, change line 695 from
1036             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
1037        to
1038             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
1039    
1040    ----------------------------------------------------------------------
1041    Name: Matthias Blume
1042    Date: 2002/04/12 13:55:00 EDT
1043    Tag: blume-20020412-assyntax
1044    Description:
1045    
1046    1. Grabbed newer assyntax.h from the XFree86 project.
1047    2. Fiddled with how to compile X86.prim.asm without warnings.
1048    3. (Very) Minor cleanup in CM.
1049    
1050    ----------------------------------------------------------------------
1051    Name: Matthias Blume
1052    Date: 2002/04/01 (no joke!) 17:07:00 EST
1053    Tag: blume-20020401-x86div
1054    Description:
1055    
1056    Added full support for div/mod/rem/quot on the x86, using the machine
1057    instruction's two results (without clumsily recomputing the remainder)
1058    directly where appropriate.
1059    
1060    Some more extensive power-of-two support was added to the x86 instruction
1061    selector (avoiding expensive divs, mods, and muls where they can be
1062    replaced with cheaper shifts and masks).  However, this sort of thing
1063    ought to be done earlier, e.g., within the CPS optimizer so that
1064    all architectures benefit from it.
1065    
1066    The compiler compiles to a fixed point, but changes might be somewhat
1067    fragile nevertheless.  Please, report any strange things that you might
1068    see wrt. div/mod/quot/rem...
1069    
1070    ----------------------------------------------------------------------
1071    Name: Matthias Blume
1072    Date: 2002/03/29 17:22:00
1073    Tag: blume-20020329-div
1074    Description:
1075    
1076    Fixed my broken div/mod logic.  Unfortunately, this means that the
1077    inline code for div/mod now has one more comparison than before.
1078    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
1079    The problem was with quotient = 0, because that alone does not tell
1080    us which way the rounding went.  One then has to look at whether
1081    remainder and divisor have the same sign...  :(
1082    
1083    Anyway, I replaced the bootfiles with fresh ones...
1084    
1085    ----------------------------------------------------------------------
1086    Name: Matthias Blume
1087    Date: 2002/03/29 14:10:00 EST
1088    Tag: blume-20020329-inlprims
1089    Description:
1090    
1091    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
1092    
1093    Primops have changed. This means that the bin/boot-file formats have
1094    changed as well.
1095    
1096    To make sure that there is no confusion, I made a new version.
1097    
1098    
1099    CHANGES:
1100    
1101    * removed REMT from mltree (remainder should never overflow).
1102    
1103    * added primops to deal with divisions of all flavors to the frontend
1104    
1105    * handled these primops all the way through so they map to their respective
1106      MLRISC support
1107    
1108    * used these primops in the implementation of Int, Int32, Word, Word32
1109    
1110    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
1111    
1112    * parameterized INLMIN, INLMAX, and INLABS by a numkind
1113    
1114    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
1115      floating point
1116    
1117    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
1118      and Real (but Real.abs maps to a separate floating-point-only primop)
1119    
1120    
1121    TODO items:
1122    
1123    * Hacked Alpha32 instruction selection, disabling the selection of REMx
1124      instructions because the machine instruction encoder cannot handle
1125      them.  (Hppa, PPC, and Sparc instruction selection did not handle
1126      REM in the first place, and REM is supported by the x86 machine coder.)
1127    
1128    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
1129      selection phase.  (The two can be streamlined because the hardware
1130      delivers both quotient and remainder at the same time anyway.)
1131    
1132    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
1133      (Currently the behavior is inconsistent both across architectures and
1134      wrt. the draft Basis spec.)
1135    
1136    * Word8 should eventually be handled natively, too.
1137    
1138    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
1139      as if there currently is no execution path that could trigger it in
1140      SML/NJ.  (The assumptions underlying functions arith and promotable do not
1141      hold for things like multiplication and division.)
1142    
1143    ----------------------------------------------------------------------
1144    Name: Matthias Blume
1145    Date: 2002/03/27 16:27:00 EST
1146    Tag: blume-20020327-mlrisc-divisions
1147    Description:
1148    
1149    Added support for all four division operations (ML's div, mod, quot,
1150    and rem) to MLRISC.  In the course of doing so, I also rationalized
1151    the naming (no more annoying switch-around of DIV and QUOT), by
1152    parameterizing the operation by div_rounding_mode (which can be either
1153    DIV_TO_ZERO or DIV_TO_NEGINF).
1154    
1155    The generic MLTreeGen functor takes care of compiling all four
1156    operations down to only round-to-zero div.
1157    
1158    Missing pieces:
1159    
1160      * Doing something smarter than relying on MLTreeGen on architectures
1161        like, e.g., the x86 where hardware division delivers both quotient and
1162        remainder at the same time.  With this, the implementation of the
1163        round-to-neginf operations could be further streamlined.
1164    
1165      * Remove inlining support for div/mod/rem from the frontend and replace it
1166        with primops that get carried through to the backend.  Do this for all
1167        int and word types.
1168    
1169    ----------------------------------------------------------------------
1170    Name: Matthias Blume
1171    Date: 2002/03/25 17:25:00 EST
1172    Tag: blume-20020325-divmod
1173    Description:
1174    
1175    I improved (hopefully without breaking them) the implementation of Int.div,
1176    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
1177    advantage of the following observations:
1178    
1179      Let  q = x quot y      r = x rem y
1180           d = x div  y      m = x mod y
1181    
1182    where "quot" is the round-to-zero version of integer division that
1183    hardware usually provides.  Then we have:
1184    
1185         r = x - q * y        where neither the * nor the - will overflow
1186         d = if q >= 0 orelse x = q * y then q else q - 1
1187                              where neither the * nor the - will overflow
1188         m = if q >= 0 orelse r = 0 then r else r + y
1189                              where the + will not overflow
1190    
1191    This results in substantial simplification of the generated code.
1192    The following table shows the number of CFG nodes and edges generated
1193    for
1194            fun f (x, y) = x OPER y
1195            (* with OPER \in div, mod, quot, rem *)
1196    
1197    
1198        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
1199        --------------------------------------------------------
1200         div |         24 |         39 |         12 |         16
1201         mod |         41 |         71 |         12 |         16
1202        quot |          8 |         10 |          8 |         10
1203         rem |         10 |         14 |          8 |         10
1204    
1205    
1206    ----------------------------------------------------------------------
1207    Name: Matthias Blume
1208    Date: 2002/03/25 22:06:00 EST
1209    Tag: blume-20020325-cprotobug
1210    Description:
1211    
1212    Fixed a bug in cproto (c prototype decoder).
1213    
1214    ----------------------------------------------------------------------
1215    Name: Matthias Blume
1216    Date: 2002/03/25 16:00:00 EST
1217    Tag: blume-20020325-raw-primops
1218    Description:
1219    
1220    I did some cleanup to Allen's new primop code and
1221    replaced yesterday's bootfiles with new ones.
1222    (But they are stored in the same place.)
1223    
1224    ----------------------------------------------------------------------
1225    Name: Matthias Blume
1226    Date: 2002/03/24 22:40:00 EST
1227    Tag: blume-20020324-bootfiles
1228    Description:
1229    
1230    Made the bootfiles that Allen asked for.
1231    
1232    ----------------------------------------------------------------------
1233    Name: Allen Leung
1234    Date: 2002/03/23 15:50:00 EST
1235    Tag: leunga-20020323-flint-cps-rcc-primops
1236    Description:
1237    
1238      1. Changes to FLINT primops:
1239    
1240        (* make a call to a C-function;
1241         * The primop carries C function prototype information and specifies
1242         * which of its (ML-) arguments are floating point. C prototype
1243         * information is for use by the backend, ML information is for
1244         * use by the CPS converter. *)
1245      | RAW_CCALL of { c_proto: CTypes.c_proto,
1246                       ml_args: ccall_type list,
1247                       ml_res_opt: ccall_type option,
1248                       reentrant : bool
1249                     } option
1250       (* Allocate uninitialized storage on the heap.
1251        * The record is meant to hold short-lived C objects, i.e., they
1252        * are not ML pointers.  With the tag, the representation is
1253        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
1254        *)
1255      | RAW_RECORD of {tag:bool,sz:int}
1256      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
1257    
1258      2.  These CPS primops are now overloaded:
1259    
1260           rawload of {kind:numkind}
1261           rawstore of {kind:numkind}
1262    
1263          The one argument form is:
1264    
1265             rawload {kind} address
1266    
1267          The two argument form is:
1268    
1269             rawload {kind} [ml object, byte-offset]
1270    
1271      3. RAW_CCALL/RCC now takes two extra arguments:
1272    
1273         a. The first is whether the C call is reentrant, i.e., whether
1274            ML state should be saved and restored.
1275         b. The second argument is a string argument specifying the name of
1276            library and the C function.
1277    
1278         These things are currently not handled in the code generator, yet.
1279    
1280      4. In CProto,
1281    
1282         An encoding type of "bool" means "ml object" and is mapped into
1283         C prototype of PTR.  Note that "bool" is different than "string",
1284         even though "string" is also mapped into PTR, because "bool"
1285         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
1286    
1287      5. Pickler/unpicker
1288    
1289         Changed to handle RAW_RECORD and newest RAW_CCALL
1290    
1291      6. MLRiscGen,
1292    
1293         1. Changed to handle the new rawload/rawstore/rawrecord operators.
1294         2. Code for handling C Calls has been moved to a new module CPSCCalls,
1295            in the file CodeGen/cpscompile/cps-c-calls.sml
1296    
1297      7. Added the conditional move operator
1298    
1299             condmove of branch
1300    
1301         to cps.  Generation of this is still buggy so it is currently
1302         disabled.
1303    
1304    ----------------------------------------------------------------------
1305    Name: Lal George
1306    Date: 2002/03/22 14:18:25 EST
1307    Tag: george-20020322-cps-branch-prob
1308    Description:
1309    
1310    Implemented the Ball-Larus branch prediction-heuristics, and
1311    incorporated graphical viewers for control flow graphs.
1312    
1313    Ball-Larus Heuristics:
1314    ---------------------
1315    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1316    
1317    By design it uses the Dempster-Shafer theory for combining
1318    probabilities.  For example, in the function:
1319    
1320        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1321    
1322    the ball-larus heuristics predicts that the n=0 is unlikely
1323    (OH-heuristic), and the 'then' branch is unlikely because of the
1324    RH-heuristic -- giving the 'then' branch an even lower combined
1325    probability using the Dempster-Shafer theory.
1326    
1327    Finally, John Reppy's loop analysis in MLRISC, further lowers the
1328    probability of the 'then' branch because of the loop in the else
1329    branch.
1330    
1331    
1332    Graphical Viewing:
1333    ------------------
1334    I merely plugged in Allen's graphical viewers into the compiler. The
1335    additional code is not much. At the top level, saying:
1336    
1337            Control.MLRISC.getFlag "cfg-graphical-view" := true;
1338    
1339    will display the graphical view of the control flow graph just before
1340    back-patching.  daVinci must be in your path for this to work. If
1341    daVinci is not available, then the default viewer can be changed
1342    using:
1343    
1344            Control.MLRISC.getString "viewer"
1345    
1346    which can be set to "dot" or "vcg" for the corresponding viewers. Of
1347    course, these viewers must be in your path.
1348    
1349    The above will display the compilation unit at the level of clusters,
1350    many of which are small, boring, and un-interesting. Also setting:
1351    
1352            Control.MLRISC.getInt "cfg-graphical-view_size"
1353    
1354    will display clusters that are larger than the value set by the above.
1355    
1356    
1357    ----------------------------------------------------------------------
1358    Name: Matthias Blume
1359    Date: 2002/03/21 22:20:00 EST
1360    Tag: blume-20020321-kmp-bugfix
1361    Description:
1362    
1363    Changed the interface to the KMP routine in PreString and fixed
1364    a minor bug in one place where it was used.
1365    
1366    ----------------------------------------------------------------------
1367    Name: Allen Leung
1368    Date: 2002/03/21 20:30:00 EST
1369    Tag: leunga-20020321-cfg
1370    Description:
1371    
1372      Fixed a potential problem in cfg edge splitting.
1373    
1374    ----------------------------------------------------------------------
1375    Name: Allen Leung
1376    Date: 2002/03/21 17:15:00 EST
1377    Tag: leunga-20020321-x86-fp-cfg
1378    Description:
1379    
1380      1. Recoded the buggy parts of x86-fp.
1381    
1382         a. All the block reordering code has been removed.
1383            We now depend on the block placement phases to do this work.
1384    
1385         b. Critical edge splitting code has been simplified and moved into the
1386            CFG modules, as where they belong.
1387    
1388         Both of these were quite buggy and complex.  The code is now much, much
1389         simpler.
1390    
1391      2. X86 backend.
1392    
1393         a. Added instructions for 64-bit support.  Instruction selection for
1394            64-bit has not been committed, however, since that
1395            requires changes to MLTREE which haven't been approved by
1396            Lal and John.
1397    
1398         b. Added support for FUCOMI and FUCOMIP when generating code for
1399            PentiumPro and above.  We only generate these instructions in
1400            the fast-fp mode.
1401    
1402         c. Added cases for JP and JNP in X86FreqProps.
1403    
1404      3. CFG
1405    
1406         CFG now has a bunch of methods for edge splitting and merging.
1407    
1408      4. Machine description.
1409    
1410         John's simplification of MLTREE_BASIS.fcond broke a few machine
1411         description things:
1412    
1413         rtl-build.{sig,sml} and hppa.mdl fixed.
1414    
1415         NOTE: the machine description stuff in the repository is still broken.
1416               Again, I can't put my fixes in because that involves
1417               changes to MLTREE.
1418    
1419    ----------------------------------------------------------------------
1420    Name: Matthias Blume
1421    Date: 2002/03/20 15:55:00 EST
1422    Tag: blume-20020320-kmp
1423    Description:
1424    
1425    Implemented Knuth-Morris-Pratt string matching in PreString and used
1426    it for String.isSubstring, Substring.isSubstring, and
1427    Substring.position.
1428    
1429    (Might need some stress-testing.  Simple examples worked fine.)
1430    
1431    ----------------------------------------------------------------------
1432    Name: Matthias Blume
1433    Date: 2002/03/19 16:37:00 EST
1434    Tag: blume-20020319-witnesses
1435    Description:
1436    
1437    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1438    
1439    This implements a generic mechanism for changing constness qualifiers
1440    anywhere within big C types without resorting to outright "casts".
1441    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1442    modify the constness at the outermost level.)
1443    The implementation of "convert" is based on the idea of "witness"
1444    values -- values that are not used by the operation but whose types
1445    "testify" to their applicability.  On the implementation side, "convert"
1446    is simply a projection (returning its second curried argument).  With
1447    cross-module inlining, it should not result in any machine code being
1448    generated.
1449    
1450    ----------------------------------------------------------------------
1451    Name: Matthias Blume
1452    Date: 2002/03/15 16:40:00 EST
1453    Tag: blume-20020315-basis
1454    Description:
1455    
1456    Provided (preliminary?) implementations for
1457    
1458      {String,Substring}.{concatWith,isSuffix,isSubstring}
1459    
1460    and
1461    
1462      Substring.full
1463    
1464    Those are in the Basis spec but they were missing in SML/NJ.
1465    
1466    ----------------------------------------------------------------------
1467    Name: Matthias Blume
1468    Date: 2002/03/14 21:30:00 EST
1469    Tag: blume-20020314-controls
1470    Description:
1471    
1472    Controls:
1473    ---------
1474    
1475    1. Factored out the recently-added Controls : CONTROLS stuff and put
1476       it into its own library $/controls-lib.cm.  The source tree for
1477       this is under src/smlnj-lib/Controls.
1478    
1479    2. Changed the names of types and functions in this interface, so they
1480       make a bit more "sense":
1481    
1482          module -> registry
1483          'a registry -> 'a group
1484    
1485    3. The interface now deals in ref cells only.  The getter/setter interface
1486       is (mostly) gone.
1487    
1488    4. Added a function that lets one register an already-existing ref cell.
1489    
1490    5. Made the corresponding modifications to the rest of the code so that
1491       everything compiles again.
1492    
1493    6. Changed the implementation of Controls.MLRISC back to something closer
1494       to the original.  In particular, this module (and therefore MLRISC)
1495       does not depend on Controls.  There now is some link-time code in
1496       int-sys.sml that registers the MLRISC controls with the Controls
1497       module.
1498    
1499    CM:
1500    ---
1501    
1502      * One can now specify the lambda-split aggressiveness in init.cmi.
1503    
1504    ----------------------------------------------------------------------
1505    Name: Allen Leung
1506    Date: 2002/03/13 17:30:00 EST
1507    Tag: leunga-20020313-x86-fp-unary
1508    Description:
1509    
1510    Bug fix for:
1511    
1512    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1513    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1514    > - fun f(x,(y,z)) = Real.~ y;
1515    > [autoloading]
1516    > [autoloading done]
1517    >       fchsl   (%eax), 184(%esp)
1518    > Error: MLRisc bug: X86MCEmitter.emitInstr
1519    >
1520    > uncaught exception Error
1521    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1522    
1523    The problem was that the code generator did not generate any fp registers
1524    in this case, and the ra didn't know that it needed to run the X86FP phase to
1525    translate the pseudo fp instruction.   This only happened with unary fp
1526    operators in certain situations.
1527    
1528    ----------------------------------------------------------------------
1529    Name: Matthias Blume
1530    Date: 2002/03/13 14:00:00 EST
1531    Tag: blume-20020313-overload-etc
1532    Description:
1533    
1534    1. Added _overload as a synonym for overload for backward compatibility.
1535       (Control.overloadKW must be true for either version to be accepted.)
1536    
1537    2. Fixed bug in install script that caused more things to be installed
1538       than what was requested in config/targets.
1539    
1540    3. Made CM aware of the (_)overload construct so that autoloading
1541       works.
1542    
1543    ----------------------------------------------------------------------
1544    Name: Matthias Blume
1545    Date: 2002/03/12 22:03:00 EST
1546    Tag: blume-20020312-url
1547    Description:
1548    
1549    Forgot to update BOOT and srcarchiveurl.
1550    
1551    ----------------------------------------------------------------------
1552    Name: Matthias Blume
1553    Date: 2002/03/12 17:30:00 EST
1554    Tag: blume-20020312-version110392
1555    Description:
1556    
1557    Yet another version number bump (because of small changes to the
1558    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1559    
1560    Changes:
1561    
1562      The new pid generation scheme described a few weeks ago was overly
1563      complicated.  I implemented a new mechanism that is simpler and
1564      provides a bit more "stability":  Once CM has seen a compilation
1565      unit, it keeps its identity constant (as long as you do not delete
1566      those crucial CM/GUID/* files).  This means that when you change
1567      an interface, compile, then go back to the old interface, and
1568      compile again, you arrive at the original pid.
1569    
1570      There now also is a mechanism that instructs CM to use the plain
1571      environment hash as a module's pid (effectively making its GUID
1572      the empty string).  For this, "noguid" must be specified as an
1573      option to the .sml file in question within its .cm file.
1574      This is most useful for code that is being generated by tools such
1575      as ml-nlffigen (because during development programmers tend to
1576      erase the tool's entire output directory tree including CM's cached
1577      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1578      revert to the old, broken behavior of SML/NJ, but in specific cases
1579      where there is no danger of interface confusion, its use is ok
1580      (I think).
1581    
1582      ml-nlffigen by default generates "noguid" annotations.  They can be
1583      turned off by specifying -guid in its command line.
1584    
1585    ----------------------------------------------------------------------
1586    Name: Lal George
1587    Date: 2002/03/12 12 14:42:36 EST
1588    Tag: george-20020312-frequency-computation
1589    Description:
1590    
1591    Integrated jump chaining and static block frequency into the
1592    compiler. More details and numbers later.
1593    
1594    ----------------------------------------------------------------------
1595    Name: Lal George
1596    Date: 2002/03/11 11 22:38:53 EST
1597    Tag: george-20020311-jump-chain-elim
1598    Description:
1599    
1600    Tested the jump chain elimination on all architectures (except the
1601    hppa).  This is on by default right now and is profitable for the
1602    alpha and x86, however, it may not be profitable for the sparc and ppc
1603    when compiling the compiler.
1604    
1605    The gc test will typically jump to a label at the end of the cluster,
1606    where there is another jump to an external cluster containing the actual
1607    code to invoke gc. This is to allow factoring of common gc invocation
1608    sequences. That is to say, we generate:
1609    
1610            f:
1611               testgc
1612               ja   L1      % jump if above to L1
1613    
1614            L1:
1615               jmp L2
1616    
1617    
1618    After jump chain elimination the 'ja L1' instructions is converted to
1619    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1620    up being implemented in their long form (if L2 is far away) using:
1621    
1622            jbe     L3      % jump if below or equal to L3
1623            jmp     L2
1624         L3:
1625            ...
1626    
1627    
1628    For large compilation units L2  may be far away.
1629    
1630    
1631    ----------------------------------------------------------------------
1632    Name: Matthias Blume
1633    Date: 2002/03/11 13:30:00 EST
1634    Tag: blume-20020311-mltreeeval
1635    Description:
1636    
1637    A functor parameter was missing.
1638    
1639    ----------------------------------------------------------------------
1640    Name: Allen Leung
1641    Date: 2002/03/11 10:30:00 EST
1642    Tag: leunga-20020311-runtime-string0
1643    Description:
1644    
1645       The representation of the empty string now points to a
1646    legal null terminated C string instead of unit.  It is now possible
1647    to convert an ML string into C string with InlineT.CharVector.getData.
1648    This compiles into one single machine instruction.
1649    
1650    ----------------------------------------------------------------------
1651    Name: Allen Leung
1652    Date: 2002/03/10 23:55:00 EST
1653    Tag: leunga-20020310-x86-call
1654    Description:
1655    
1656       Added machine generation for CALL instruction (relative displacement mode)
1657    
1658    ----------------------------------------------------------------------
1659    Name: Matthias Blume
1660    Date: 2002/03/08 16:05:00
1661    Tag: blume-20020308-entrypoints
1662    Description:
1663    
1664    Version number bumped to 110.39.1.  NEW BOOTFILES!
1665    
1666    Entrypoints: non-zero offset into a code object where execution should begin.
1667    
1668    - Added the notion of an entrypoint to CodeObj.
1669    - Added reading/writing of entrypoint info to Binfile.
1670    - Made runtime system bootloader aware of entrypoints.
1671    - Use the address of the label of the first function given to mlriscGen
1672      as the entrypoint.  This address is currently always 0, but it will
1673      not be 0 once we turn on block placement.
1674    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1675      with entry points) from mlriscGen.
1676    
1677    ----------------------------------------------------------------------
1678    Name: Allen Leung
1679    Date: 2002/03/07 20:45:00 EST
1680    Tag: leunga-20020307-x86-cmov
1681    Description:
1682    
1683       Bug fixes for CMOVcc on x86.
1684    
1685       1. Added machine code generation for CMOVcc
1686       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1687       3. CMOVcc cannot have an immediate operand as argument.
1688    
1689    ----------------------------------------------------------------------
1690    Name: Matthias Blume
1691    Date: 2002/03/07 16:15:00 EST
1692    Tag: blume-20020307-controls
1693    Description:
1694    
1695    This is a very large but mostly boring patch which makes (almost)
1696    every tuneable compiler knob (i.e., pretty much everything under
1697    Control.* plus a few other things) configurable via both the command
1698    line and environment variables in the style CM did its configuration
1699    until now.
1700    
1701    Try starting sml with '-h' (or, if you are brave, '-H')
1702    
1703    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1704    implements the underlying generic mechanism.
1705    
1706    The interface to some of the existing such facilities has changed somewhat.
1707    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1708    (The getFoo interface is still there for backward-compatibility, but its
1709    use is deprecated.)
1710    
1711    The ml-build script passes -Cxxx=yyy command-line arguments through so
1712    that one can now twiddle the compiler settings when using this "batch"
1713    compiler.
1714    
1715    TODO items:
1716    
1717    We should go through and throw out all controls that are no longer
1718    connected to anything.  Moreover, we should go through and provide
1719    meaningful (and correct!) documentation strings for those controls
1720    that still are connected.
1721    
1722    Currently, multiple calls to Controls.new are accepted (only the first
1723    has any effect).  Eventually we should make sure that every control
1724    is being made (via Controls.new) exactly once.  Future access can then
1725    be done using Controls.acc.
1726    
1727    Finally, it would probably be a good idea to use the getter-setter
1728    interface to controls rather than ref cells.  For the time being, both
1729    styles are provided by the Controls module, but getter-setter pairs are
1730    better if thread-safety is of any concern because they can be wrapped.
1731    
1732    *****************************************
1733    
1734    One bug fix: The function blockPlacement in three of the MLRISC
1735    backpatch files used to be hard-wired to one of two possibilities at
1736    link time (according to the value of the placementFlag).  But (I
1737    think) it should rather sense the flag every time.
1738    
1739    *****************************************
1740    
1741    Other assorted changes (by other people who did not supply a HISTORY entry):
1742    
1743    1. the cross-module inliner now works much better (Monnier)
1744    2. representation of weights, frequencies, and probabilities in MLRISC
1745       changed in preparation of using those for weighted block placement
1746       (Reppy, George)
1747    
1748    ----------------------------------------------------------------------
1749    Name: Lal George
1750    Date: 2002/03/07 14:44:24 EST 2002
1751    Tag: george-20020307-weighted-block-placement
1752    
1753    Tested the weighted block placement optimization on all architectures
1754    (except the hppa) using AMPL to generate the block and edge frequencies.
1755    Changes were required in the machine properties to correctly
1756    categorize trap instructions. There is an MLRISC flag
1757    "weighted-block-placement" that can be used to enable weighted block
1758    placement, but this will be ineffective without block/edge
1759    frequencies (coming soon).
1760    
1761    
1762    ----------------------------------------------------------------------
1763    Name: Lal George
1764    Date: 2002/03/05 17:24:48 EST
1765    Tag: george-20020305-linkage-cluster
1766    
1767    In order to support the block placement optimization, a new cluster
1768    is generated as the very first cluster (called the linkage cluster).
1769    It contains a single jump to the 'real' entry point for the compilation
1770    unit. Block placement has no effect on the linkage cluster itself, but
1771    all the other clusters  have full freedom in the manner in which they
1772    reorder blocks or functions.
1773    
1774    On the x86 the typical linkage code that is generated is:
1775       ----------------------
1776            .align 2
1777       L0:
1778            addl    $L1-L0, 72(%esp)
1779            jmp     L1
1780    
1781    
1782            .align  2
1783       L1:
1784       ----------------------
1785    
1786    72(%esp) is the memory location for the stdlink register. This
1787    must contain the address of the CPS function being called. In the
1788    above example, it contains the address of  L0; before
1789    calling L1 (the real entry point for the compilation unit), it
1790    must contain the address for L1, and hence
1791    
1792            addl $L1-L0, 72(%esp)
1793    
1794    I have tested this on all architectures except the hppa.The increase
1795    in code size is of course negligible
1796    
1797    ----------------------------------------------------------------------
1798    Name: Allen Leung
1799    Date: 2002/03/03 13:20:00 EST
1800    Tag: leunga-20020303-mlrisc-tools
1801    
1802      Added #[ ... ] expressions to mlrisc tools
1803    
1804    ----------------------------------------------------------------------
1805    Name: Matthias Blume
1806    Date: 2002/02/27 12:29:00 EST
1807    Tag: blume-20020227-cdebug
1808    Description:
1809    
1810    - made types in structure C and C_Debug to be equal
1811    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1812    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1813    
1814    ----------------------------------------------------------------------
1815    Name: Matthias Blume
1816    Date: 2002/02/26 12:00:00 EST
1817    Tag: blume-20020226-ffi
1818    Description:
1819    
1820    1. Fixed a minor bug in CM's "noweb" tool:
1821       If numbering is turned off, then truly don't number (i.e., do not
1822       supply the -L option to noweb).  The previous behavior was to supply
1823       -L'' -- which caused noweb to use the "default" line numbering scheme.
1824       Thanks to Chris Richards for pointing this out (and supplying the fix).
1825    
1826    2. Once again, I reworked some aspects of the FFI:
1827    
1828       A. The incomplete/complete type business:
1829    
1830       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1831         gone!
1832       - ML types representing an incomplete type are now *equal* to
1833         ML types representing their corresponding complete types (just like
1834         in C).  This is still safe because ml-nlffigen will not generate
1835         RTTI for incomplete types, nor will it generate functions that
1836         require access to such RTTI.   But when ML code generated from both
1837         incomplete and complete versions of the C type meet, the ML types
1838         are trivially interoperable.
1839    
1840         NOTE:  These changes restore the full generality of the translation
1841         (which was previously lost when I eliminated functorization)!
1842    
1843       B. Enum types:
1844    
1845       - Structure C now has a type constructor "enum" that is similar to
1846         how the "su" constructor works.  However, "enum" is not a phantom
1847         type because each "T enum" has values (and is isomorphic to
1848         MLRep.Signed.int).
1849       - There are generic access operations for enum objects (using
1850         MLRep.Signed.int).
1851       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1852         * The structure contains the definition of type "mlrep" (the ML-side
1853         representation type of the enum).  Normally, mlrep is the same
1854         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1855         then mlrep will be defined as a datatype -- thus facilitating
1856         pattern matching on mlrep values.
1857         ("-ec" will be suppressed if there are duplicate values in an
1858          enumeration.)
1859         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1860         will be generated for each C enum constant xxx.
1861         * Conversion functions m2i and i2m convert between mlrep and
1862         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1863         * Coversion functions c and ml convert between mlrep and "tag enum".
1864         * Access functions (get/set) fetch and store mlrep values.
1865       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1866         enumerations are merged into one single enumeration represented by
1867         structure E_'.
1868    
1869    ----------------------------------------------------------------------
1870    Name: Allen Leung
1871    Date: 2002/02/25 04:45:00 EST
1872    Tag: leunga-20020225-cps-spill
1873    
1874    This is a new implementation of the CPS spill phase.
1875    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1876    In case of problems, replace it with the old file spill.sml
1877    
1878    The current compiler runs into some serious performance problems when
1879    constructing a large record.  This can happen when we try to compile a
1880    structure with many items.  Even a very simple structure like the following
1881    makes the compiler slow down.
1882    
1883        structure Foo = struct
1884           val x_1 = 0w1 : Word32.int
1885           val x_2 = 0w2 : Word32.int
1886           val x_3 = 0w3 : Word32.int
1887           ...
1888           val x_N = 0wN : Word32.int
1889        end
1890    
1891    The following table shows the compile time, from N=1000 to N=4000,
1892    with the old compiler:
1893    
1894    N
1895    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1896           MLRISC ra                               0.06u  0.00s  0.05g
1897              (spills = 0 reloads = 0)
1898           TOTAL                                   0.63u  0.07s  0.21g
1899    
1900    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1901           MLRISC ra                               5.68u  0.59s  3.93g
1902              (spills = 0 reloads = 0)
1903           TOTAL                                   14.71u  0.99s  4.81g
1904    
1905    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1906           MLRISC ra                               5.54u  0.65s  3.91g
1907              (spills = 543 reloads = 1082)
1908           TOTAL                                   65.40u  3.13s  6.00g
1909    
1910    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1911           MLRISC ra                               0.80u  0.10s  0.55g
1912              (spills = 42 reloads = 84)
1913           TOTAL                                   129.42u  5.10s  4.13g
1914    
1915    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1916           MLRISC ra                               2.69u  0.27s  1.38g
1917              (spills = 62 reloads = 124)
1918           TOTAL                                   682.48u  19.61s  13.99g
1919    
1920    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1921           MLRISC ra                               4.96u  0.27s  2.72g
1922              (spills = 85 reloads = 170)
1923           TOTAL                                   2375.26u  57.21s  48.00g
1924    
1925    As you can see the old cps spill module suffers from some serious
1926    performance problem.  But since I cannot decipher the old code fully,
1927    instead of patching the problems up, I'm reimplementing it
1928    with a different algorithm.  The new code is more modular,
1929    smaller when compiled, and substantially faster
1930    (O(n log n) time and O(n) space).  Timing of the new spill module:
1931    
1932    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1933          MLRISC ra                               0.25u  0.02s  0.15g
1934             (spills=1 reloads=3)
1935          TOTAL                                   7.74u  0.34s  1.62g
1936    
1937    Implementation details:
1938    
1939    As far as I can tell, the purpose of the CPS spill module is to make sure the
1940    number of live variables at any program point (the bandwidth)
1941    does not exceed a certain limit, which is determined by the
1942    size of the spill area.
1943    
1944    When the bandwidth is too large, we decrease the register pressure by
1945    packing live variables into spill records.  How we achieve this is
1946    completely different than what we did in the old code.
1947    
1948    First, there is something about the MLRiscGen code generator
1949    that we should be aware of:
1950    
1951    o MLRiscGen performs code motion!
1952    
1953       In particular, it will move floating point computations and
1954       address computations involving only the heap pointer to
1955       their use sites (if there is only a single use).
1956       What this means is that if we have a CPS record construction
1957       statement
1958    
1959           RECORD(k,vl,w,e)
1960    
1961       we should never count the new record address w as live if w
1962       has only one use (which is often the case).
1963    
1964       We should do something similar to floating point, but the transformation
1965       there is much more complex, so I won't deal with that.
1966    
1967    Secondly, there are now two new cps primops at our disposal:
1968    
1969     1. rawrecord of record_kind option
1970        This pure operator allocates some uninitialized storage from the heap.
1971        There are two forms:
1972    
1973         rawrecord NONE [INT n]  allocates a tagless record of length n
1974         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1975                                     and initializes the tag.
1976    
1977     2. rawupdate of cty
1978          rawupdate cty (v,i,x)
1979          Assigns to x to the ith component of record v.
1980          The storelist is not updated.
1981    
1982    We use these new primops for both spilling and increment record construction.
1983    
1984     1. Spilling.
1985    
1986        This is implemented with a linear scan algorithm (but generalized
1987        to trees).  The algorithm will create a single spill record at the
1988        beginning of the cps function and use rawupdate to spill to it,
1989        and SELECT or SELp to reload from it.  So both spills and reloads
1990        are fine-grain operations.  In contrast, in the old algorithm
1991        "spills" have to be bundled together in records.
1992    
1993        Ideally, we should sink the spill record construction to where
1994        it is needed.  We can even split the spill record into multiple ones
1995        at the places where they are needed.  But CPS is not a good
1996        representation for global code motion, so I'll keep it simple and
1997        am not attempting this.
1998    
1999     2. Incremental record construction (aka record splitting).
2000    
2001        Long records with many component values which are simulatenously live
2002        (recall that single use record addresses are not considered to
2003         be live) are constructed with rawrecord and rawupdate.
2004        We allocate space on the heap with rawrecord first, then gradually
2005        fill it in with rawupdate.  This is the technique suggested to me
2006        by Matthias.
2007    
2008        Some restrictions on when this is applicable:
2009        1. It is not a VECTOR record.  The code generator currently does not handle
2010           this case. VECTOR record uses double indirection like arrays.
2011        2. All the record component values are defined in the same "basic block"
2012           as the record constructor.  This is to prevent speculative
2013           record construction.
2014    
2015    ----------------------------------------------------------------------
2016    Name: Allen Leung
2017    Date: 2002/02/22 01:02:00 EST
2018    Tag: leunga-20020222-mlrisc-tools
2019    
2020    Minor bug fixes in the parser and rewriter
2021    
2022    ----------------------------------------------------------------------
2023    Name: Allen Leung
2024    Date: 2002/02/21 20:20:00 EST
2025    Tag: leunga-20020221-peephole
2026    
2027    Regenerated the peephole files.  Some contained typos in the specification
2028    and some didn't compile because of pretty printing bugs in the old version
2029    of 'nowhere'.
2030    
2031    ----------------------------------------------------------------------
2032    Name: Allen Leung
2033    Date: 2002/02/19 20:20:00 EST
2034    Tag: leunga-20020219-mlrisc-tools
2035    Description:
2036    
2037       Minor bug fixes to the mlrisc-tools library:
2038    
2039       1.  Fixed up parsing colon suffixed keywords
2040       2.  Added the ability to shut the error messages up
2041       3.  Reimplemented the pretty printer and fixed up/improved
2042           the pretty printing of handle and -> types.
2043       4.  Fixed up generation of literal symbols in the nowhere tool.
2044       5.  Added some SML keywords to to sml.sty
2045    
2046    ----------------------------------------------------------------------
2047    Name: Matthias Blume
2048    Date: 2002/02/19 16:20:00 EST
2049    Tag: blume-20020219-cmffi
2050    Description:
2051    
2052    A wild mix of changes, some minor, some major:
2053    
2054    * All C FFI-related libraries are now anchored under $c:
2055        $/c.cm      --> $c/c.cm
2056        $/c-int.cm  --> $c/internals/c-int.cm
2057        $/memory.cm --> $c/memory/memory.cm
2058    
2059    * "make" tool (in CM) now treats its argument pathname slightly
2060      differently:
2061        1. If the native expansion is an absolute name, then before invoking
2062           the "make" command on it, CM will apply OS.Path.mkRelative
2063           (with relativeTo = OS.FileSys.getDir()) to it.
2064        2. The argument will be passed through to subsequent phases of CM
2065           processing without "going native".  In particular, if the argument
2066           was an anchored path, then "make" will not lose track of that anchor.
2067    
2068    * Compiler backends now "know" their respective C calling conventions
2069      instead of having to be told about it by ml-nlffigen.  This relieves
2070      ml-nlffigen from one of its burdens.
2071    
2072    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
2073    
2074    * Export C_DEBUG and C_Debug from $c/c.cm.
2075    
2076    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
2077      subtyping relationship between incomplete pointers and their complete
2078      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
2079      with the convention of instantiating 'o with ('t, 'c) obj whenever
2080      the pointer target type is complete.  In the incomplete case, 'o
2081      will be instantiated with some "'c iobj" -- a type obtained by
2082      using one of the functors PointerToIncompleteType or PointerToCompleteType.
2083    
2084      Operations that work on both incomplete and complete pointer types are
2085      typed as taking an 'o ptr while operations that require the target to
2086      be known are typed as taking some ('t, 'c) obj ptr.
2087    
2088      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
2089      where void is an eqtype without any values.  This makes it possible
2090      to work on voidptr values using functions meant to operate on light
2091      incomplete pointers.
2092    
2093    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
2094      been vastly simplified.
2095    
2096    ----------------------------------------------------------------------
2097    Name: Matthias Blume
2098    Date: 2002/02/19 10:48:00 EST
2099    Tag: blume-20020219-pqfix
2100    Description:
2101    
2102    Applied Chris Okasaki's bug fix for priority queues.
2103    
2104    ----------------------------------------------------------------------
2105    Name: Matthias Blume
2106    Date: 2002/02/15 17:05:00
2107    Tag: Release_110_39
2108    Description:
2109    
2110    Last-minute retagging is becoming a tradition... :-(
2111    
2112    This is the working release 110.39.
2113    
2114    ----------------------------------------------------------------------
2115    Name: Matthias Blume
2116    Date: 2002/02/15 16:00:00 EST
2117    Tag: Release_110_39-orig
2118    Description:
2119    
2120    Working release 110.39.  New bootfiles.
2121    
2122    (Update: There was a small bug in the installer so it wouldn't work
2123    with all shells.  So I retagged. -Matthias)
2124    
2125    ----------------------------------------------------------------------
2126    Name: Matthias Blume
2127    Date: 2002/02/15 14:17:00 EST
2128    Tag: blume-20020215-showbindings
2129    Description:
2130    
2131    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
2132    the latter can be useful for exploring what bindings are available at
2133    the interactive prompt.  (The first function returns only the list
2134    of symbols that are really bound, the second prints those but also the
2135    ones that CM's autoloading mechanism knows about.)
2136    
2137    ----------------------------------------------------------------------
2138    Name: Matthias Blume
2139    Date: 2002/02/15 12:08:00 EST
2140    Tag: blume-20020215-iptrs
2141    Description:
2142    
2143    Two improvements to ml-nlffigen:
2144    
2145      1. Write files only if they do not exist or if their current contents
2146         do not coincide with what's being written.  (That is, avoid messing
2147         with the time stamps unless absolutely necessary.)
2148    
2149      2. Implement a "repository" mechanism for generated files related
2150         to "incomplete pointer types".   See the README file for details.
2151    
2152    ----------------------------------------------------------------------
2153    Name: Matthias Blume
2154    Date: 2002/02/14 11:50:00 EST
2155    Tag: blume-20020214-quote
2156    Description:
2157    
2158    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
2159    because of the new and improved tag generation scheme.  (Thanks to Allen
2160    Leung for pointing it out.)
2161    
2162    ----------------------------------------------------------------------
2163  Name: Lal George  Name: Lal George
2164  Date: 2002/02/14 09:55:27 EST 2002  Date: 2002/02/14 09:55:27 EST 2002
2165  Tag: george-20020214-isabelle-bug  Tag: george-20020214-isabelle-bug

Legend:
Removed from v.1065  
changed lines
  Added in v.1482

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