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 1040, Mon Jan 28 16:55:24 2002 UTC revision 1487, Tue May 11 19:01:04 2004 UTC
# Line 13  Line 13 
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume (blume (at) tti - c (dot) org)
17    Date: 2004/05/11 14:05:00 CDT
18    Tag: blume-20040511-nlffi-netbsd
19    Description:
20    
21    Added support for nlffi under netbsd.  (Thanks to Vesa A. Norrman for
22    the patch.)
23    
24    ----------------------------------------------------------------------
25    Name: Matthias Blume (blume (at) tti - c (dot) org)
26    Date: 2004/05/11 12:05:00 CDT
27    Tag: blume-20040511-exports
28    Description:
29    
30    As per request by Adam Chlipala <adam@hcoop.net>, extended various
31    export lists in compiler-related .cm-files.
32    
33    ----------------------------------------------------------------------
34    Name: Matthias Blume (blume (at) tti - c (dot) org)
35    Date: 2004/05/11 11:35:00 CDT
36    Tag: blume-20040511-allsource
37    Description:
38    
39    The installer now honors the "src-smlnj" target again, although its meaning
40    has changed from "all sources required for the compiler" to "all sources
41    the installer knows about".  In other words, if you enable "src-smlnj"
42    in the "targets" file, then the installer will pull in sources for
43    everything.  (Notice that this refers to source code only.  Compiled
44    code is still only installed for modules that were requested explicitly
45    or which are required for other modules that were requested explicitly.)
46    
47    ----------------------------------------------------------------------
48    Name: Matthias Blume (blume (at) tti - c (dot) org)
49    Date: 2004/04/23 17:40:00 CDT
50    Tag: blume-20040423-ieee-scan
51    Description:
52    
53    Fixed IEEEReal.scan (and .fromString) so that if there is an overflow
54    in the exponent calculation we get INF or ZERO (depending on the mantissa
55    and the sign of the exponent).
56    
57    ----------------------------------------------------------------------
58    Name: Matthias Blume (blume (at) tti - c (dot) org)
59    Date: 2004/04/23 10:40:00 CDT
60    Tag: blume-20040423-ml-build
61    Description:
62    
63    The ml-build script now terminates with a non-0 status when something
64    goes wrong.
65    
66    ----------------------------------------------------------------------
67    Name: Matthias Blume (blume (at) tti - c (dot) org)
68    Date: 2004/04/22 16:35:00 CDT
69    Tag: blume-20040422-Option
70    Description:
71    
72    Made exception Option to be the same as exception Option.Option
73    (as it should be).
74    
75    ----------------------------------------------------------------------
76    Name: Allen Leung (leunga (at) reservoir (dot) com)
77    Date: 2004/03/19 14:40:00 EST
78    Tag: leunga-20040319-cygwin-nlffi
79    Description:
80    
81    Fixed the runtime so that ml-nlffi-lib runs on the cygwin version
82    of SML/NJ.  The problem is that
83    
84       lib = dlopen(NULL, ...)
85       f   = dlsym(lib, "malloc");
86    
87    does not work on Windows unless we explicitly export symbols
88    such as 'malloc' during linking.  We fixed this by explicitly
89    exporting the required symbols with the magic gcc incantation:
90    
91        -Wl,--export-all cygwin.def
92    
93    where cygwin.def is a file containing all the symbols that
94    we wish to export.
95    
96    I suspect this is a Windows problem and we'll have to
97    do the same (somehow with windows compilers) when
98    we build the native win32 version with the system
99    calls LoadLibrary/GetProcAddress.
100    
101    ----------------------------------------------------------------------
102    Name: Matthias Blume (blume (at) tti - c (dot) org)
103    Date: 2004/03/04 16:35:00 CST
104    Tag: blume-20040304-intinf-fmt
105    Description:
106    
107    Fixed problem with IntInf.fmt (sign would show up on the right instead
108    of on the left for BIN, OCT, and HEX).
109    
110    ----------------------------------------------------------------------
111    Name: Matthias Blume (blume (at) tti - c (dot) org)
112    Date: 2004/03/04 11:25:00 CST
113    Tag: blume-20040304-symlinks
114    Description:
115    
116    Fixed problem with installer script (unix only) where bin/ml-yacc and
117    friends pointed (via symlinks) to absolute locations instead of just
118    .run-sml.  This was reported by Vesa A Norrman.
119    
120    ----------------------------------------------------------------------
121    Name: Matthias Blume (blume (at) tti - c (dot) org)
122    Date: 2004/02/13 14:50:00 CST
123    Tag: Release_110_45
124    Description:
125    
126    New working version (110.45). New bootfiles.
127    
128    ----------------------------------------------------------------------
129    Name: Matthias Blume (blume (at) tti - c (dot) org)
130    Date: 2004/01/26 15:15:15 CST
131    Tag: blume-20040126-toplevel
132    Description:
133    
134    Improved handling of exceptions at the interactive toplevel.
135    
136    ----------------------------------------------------------------------
137    Name: Matthias Blume (blume (at) tti - c (dot) org)
138    Date: 2004/01/26 11:25:00
139    Tag: blume-20040126-app
140    Description:
141    
142    Type of top-level "app" corrected.
143    Added code for setting vp_limitPtrMask to Win32-specific runtime.
144    
145    ----------------------------------------------------------------------
146    Name: Matthias Blume (blume (at) tti - c (dot) org)
147    Date: 2003/11/18 17:10 CST
148    Tag: blume-20031118-basis-fiddle
149    Description:
150    
151    - changed Timer interface to what might become the spec
152    - POSIX_FLAGS -> BIT_FLAGS according to spec
153    - some other minor discrepancies wrt. spec eliminated
154    
155    ----------------------------------------------------------------------
156    Name: Matthias Blume (blume (at) tti - c (dot) org)
157    Date: 2003/11/06 12:00:00 CST
158    Tag: Release_110_44
159    Description:
160    
161    New working version (110.44). New bootfiles.
162    
163    ----------------------------------------------------------------------
164    Name: Matthias Blume (blume (at) tti - c (dot) org)
165    Date: 2003/11/04 11:50:00 CST
166    Tag: blume-20031104-move-libraries
167    Description:
168    
169    Eliminated the "dont_move_libraries" directive in config/targets.
170    (The mechanism was broken and could not be fixed easily.  Moreover,
171    there does not seem to be any reason not to move all libraries into
172    lib during installation.  I originally implemented this directive as a
173    backward-compatibility feature when I first introduced the new CM.
174    Now that things have been stable for a long time and going back to the
175    old CM is not an option, there is no reason to keep it around.)
176    
177    ----------------------------------------------------------------------
178    Name: Matthias Blume (blume (at) tti - c (dot) org)
179    Date: 2003/11/03 16:00:00 CST
180    Tag: blume-20031103-installdir
181    Description:
182    
183    Made installer honor INSTALLDIR variable again.  (Thanks to Chris
184    Richards for pointing out the problem and providing the solution.)
185    
186    ----------------------------------------------------------------------
187    Name: Matthias Blume (blume (at) tti - c (dot) org)
188    Date: 2003/10/01 17:05:00 CDT
189    Tag: blume-20031001-lal-mlrisc
190    Description:
191    
192    MLRISC bug fix from Lal.
193    
194    ----------------------------------------------------------------------
195    Name: Matthias Blume (blume (at) tti - c (dot) org)
196    Date: 2003/09/30 16:10:00 CDT
197    Tag: blume-20030930-primio-bat
198    Description:
199    
200    1. Added openVector, nullRd, and nullWr to PRIM_IO.
201    2. Improved .bat files (for Win32 port) to make things work under Win95.
202       (thanks to Aaron S. Hawley for this one)
203    
204    ----------------------------------------------------------------------
205    Name: Matthias Blume (blume (at) tti - c (dot) org)
206    Date: 2003/09/26 16:05:00 CDT
207    Tag: blume-20030926-wrappriv
208    Description:
209    
210    Added missing wrapper for privilege "primitive" in $smlnj/viscomp/core.cm.
211    
212    ----------------------------------------------------------------------
213    Name: Matthias Blume (blume (at) tti - c (dot) org)
214    Date: 2003/09/26 15:00:00 CDT
215    Tag: blume-20030926-110_43_3
216    Description:
217    
218    - additional cleanup
219    - version number bump, NEW BOOTFILES
220    
221    ----------------------------------------------------------------------
222    Name: Matthias Blume (blume (at) tti - c (dot) org)
223    Date: 2003/09/26 12:00:00 CDT
224    Tag: blume-20030926-ppautoload
225    Description:
226    
227    I modified the read-eval-print loop so that the autoloader gets
228    invoked whenever the prettyprinter tries to look up a symbol that
229    is not currently defined in the toplevel environment but which
230    appears in CM's autoload registry.  As a result, we see far fewer of
231    those ?.Foo.Bar.xxx names in the prettyprinter's output.
232    
233    In addition to this I tried to clean up some pieces of the Basis
234    implementation (e.g., Socket, Word8Array) in order to prevent other
235    instances of these ?.Foo.Bar.xxx names from being printed.
236    
237    The mechanism that picks names for types still needs some work, though.
238    (Right now it seems that if there is a type A.t which is defined to
239    be B.u, but B is unavailable at toplevel, then A.t gets printed as
240    "?.B.u" although the perhaps more sensible solution would be to use
241    "A.t" in this case.  In other words, the prettyprinter should follow
242    a chain of DEFtycs not farther than there are corresponding toplevel
243    names in the current environment.)
244    
245    ----------------------------------------------------------------------
246    Name: Matthias Blume (blume (at) tti - c (dot) org)
247    Date: 2003/09/24 16:31:00 CDT
248    Tag: blume-20030924-installer
249    Description:
250    
251    Another installer tweak:  All the ML code for the installer is now
252    compiled during CMB.make and put into a little library called
253    $smlnj/installer.cm.  The installation then simply invokes
254    
255       sml -m $smlnj/installer.cm
256    
257    and everything happens automagically.
258    
259    Win32:  ML code senses value of environment variable SMLNJ_HOME.
260    Unix:   ML code senses values of environment variables ROOT, CONFIGDIR,
261            and BINDIR.
262    
263    The new scheme guarantees that the ML code responsible for the installation
264    is in sync with the APIs of the main system.  Also, the installer is
265    somewhat faster because the installer script is precompiled.
266    
267    ----------------------------------------------------------------------
268    Name: Matthias Blume (blume (at) tti - c (dot) org)
269    Date: 2003/09/24 15:35:00 CDT
270    Tag: blume-20030924-synsock
271    Description:
272    
273    Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
274    but excludes all non-blocking operations.  Defined SOCKET (in Basis)
275    and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
276    implementations of non-blocking operations from CML's Socket
277    structure.
278    
279    ----------------------------------------------------------------------
280    Name: Matthias Blume (blume (at) tti - c (dot) org)
281    Date: 2003/09/24 15:10:05 CDT
282    Tag: blume-20030924-sockets
283    Description:
284    
285    1. Fixed SOCKET API and implementation to match Basis spec.
286       This required changing the internal representation of sockets to one
287       that remembers (for each socket file descriptor) whether it is currently
288       blocking or non-blocking.  This state is maintained lazily (i.e., a system
289       call is made only if the state actually needs to change).
290    
291    2. OS-specific details of sockets were moved into separate files, thus
292       making it possible to unify the bulk of the socket implementations
293       between Unix and Win32.
294    
295    3. CML's socket API changed accordingly.
296       (Note that we need to remove non-blocking functions from this API
297       since they are redundant in the case of CML!)
298    
299    4. CML's socket implementation now makes use of non-blocking functions
300       provided by Basis, thus removing all OS-dependent code from this part
301       of CML.
302    
303    5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
304    
305    ----------------------------------------------------------------------
306    Name: Matthias Blume (blume (at) tti - c (dot) org)
307    Date: 2003/09/22 12:10:00 CDT
308    Tag: blume-20030922-110_43_2
309    Description:
310    
311    Made a new interim version and bootfiles for developer's bootstrapping
312    convenience.
313    
314       110.43.2 -- NEW BOOTFILES
315    
316    ----------------------------------------------------------------------
317    Name: Matthias Blume (blume (at) tti - c (dot) org)
318    Date: 2003/09/19 15:55:00 CDT
319    Tag: blume-20030919-cmdir
320    Description:
321    
322    1. new-install.sh -> install.sh
323    2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
324    3. tweaked installer so that another name instead of .cm can be chosen
325       at install time (by setting the CM_DIR_ARC environment variable
326       during installation); once installation is complete, the name is
327       fixed
328    
329    ----------------------------------------------------------------------
330    Name: Matthias Blume (blume (at) tti - c (dot) org)
331    Date: 2003/09/18 16:00:00 CDT
332    Tag: blume-20030918-110_43_1
333    Description:
334    
335    Made a new interim version and bootfiles for developer's bootstrapping
336    convenience.
337    
338       110.43.1 -- NEW BOOTFILES
339    
340    ----------------------------------------------------------------------
341    Name: Matthias Blume (blume (at) tti - c (dot) org)
342    Date: 2003/09/18 15:20:00 CDT
343    Tag: blume-20030918-misc
344    Description:
345    
346    1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
347       this seems to be controversial at the moment) and used those in
348       Posix.ProcEnv.times.
349    
350    2. Added Time.{from,to}Nanoseconds to Time.
351    
352    3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
353       For example, fromLargeInt never needs to look at more than 3 "big
354       digits" to get its 53 bits of precision.
355    
356    ----------------------------------------------------------------------
357    Name: Matthias Blume (blume (at) tti - c (dot) org)
358    Date: 2003/09/17 16:30:00 CDT
359    Tag: blume-20030917-real32-slices
360    Description:
361    
362    Added an entry to the primitive environment
363    (compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
364    added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
365    
366    Removed some of the "magic" constants in real64.sml and replaced them
367    with code that generates these values from their corresponding
368    integer counterparts.
369    
370    Made all(?) the slice-related changes to the Basis and made everything
371    compile again...
372    
373    ----------------------------------------------------------------------
374    Name: Matthias Blume (blume (at) tti - c (dot) org)
375    Date: 2003/09/15 17:45:00 CDT
376    Tag: blume-20030915-rbase
377    Description:
378    
379    Fixed bug in Real.fromLargeInt.
380    
381    ----------------------------------------------------------------------
382    Name: Matthias Blume (blume (at) tti - c (dot) org)
383    Date: 2003/09/13 18:11:00 CDT
384    Tag: blume-20030913-libinstall
385    Description:
386    
387    Minor bugfix in config/libinstall (set anchor with path to
388    standalone tool after installing it, otherwise libraries that
389    need ml-lex or ml-yacc won't compile the first time the installer
390    runs).
391    
392    ----------------------------------------------------------------------
393    Name: Matthias Blume (blume (at) tti - c (dot) org)
394    Date: 2003/09/12 11:45:00 CDT
395    Tag: blume-20030912-various
396    Description:
397    
398     - fixed bug in Real.toLargeInt
399     - fixed bug in Posix.ProcEnv.times
400     - changed inputLine functions to return an option
401     - minor installer improvements / bugfixes
402     - changed default @SMLalloc parameter for x86/celeron to 64k
403    
404    ----------------------------------------------------------------------
405    Name: Matthias Blume (blume (at) tti - c (dot) org)
406    Date: 2003/09/09 22:00:00 CDT
407    Tag: Release_110_43
408    Description:
409    
410        New working release 110.43.  New bootfiles.
411    
412    ----------------------------------------------------------------------
413    Name: Matthias Blume (blume (at) tti - c (dot) org)
414    Date: 2003/09/09 19:20:00 CDT
415    Tag: blume-20030909-installer
416    Description:
417    
418    Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
419    Modified config/install.bat to take advantage of it.  Also modified
420    config/install.sh (and called it config/new-install.sh) to take advantage
421    of it on Unix systems.  (The SML code is (supposed to be) platform-
422    independent.)
423    
424    The installer can now install everything under Win32
425    as well as under *nix as long as it compiles.
426    
427    Other changes:
428    
429      - made CML compile again under Win32
430      - made eXene compile under Win32 (by providing a fake structure UnixSock
431        and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
432      - fixed a bug in nowhere: it assumed that type OS.Process.status is the
433        same as type int;  under Win32 it isn't
434      - fixed some slice-related problems in the win32-specific parts of CML
435      - added a functor argument "sameVol" to os-path-fn.sml in the Basis
436        (under Win32, the volume name is case-insensitive, and the
437        OS.Path code compares volume names for equality)
438    
439    ----------------------------------------------------------------------
440    Name: Matthias Blume (blume (at) tti - c (dot) org)
441    Date: 2003/09/08 11:55:00 CDT
442    Tag: blume-20030908-fullpath
443    Description:
444    
445    Made Win32 version of OS.FileSys.fullPath return current directory
446    when given an empty string.  This is what the spec says, and incidentally,
447    CM depends on it.  (CM otherwise goes into an infinite loop in certain
448    cases when presented with the name of a non-existing .cm file.)
449    
450    ----------------------------------------------------------------------
451    Name: Matthias Blume (blume (at) tti - c (dot) org)
452    Date: 2003/09/04 16:30:00 CDT
453    Tag: blume-20030905-slices-etc
454    Description:
455    
456      1. Changed interface to vectors and arrays in Basis to match
457         (draft) Basis spec.
458      2. Added signatures and implementations of slices according to
459         Basis spec.
460      3. Edited source code throughout the system to make it compile again
461         under 1. and 2.  (In some cases code had to be added to have it
462         match the new signatures.)
463      4. MLRISC should be backward-compatible: the copies of the originals
464         of files that needed to change under 3. were retained, the .cm files
465         check the compiler version number and use old versions when
466         appropriate.
467      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
468         dirstream -> string option (in accordance with Basis spec).
469      6. When generating code that counts lines, ml-lex used function
470         CharVector.foldli, taking advantage of its old interface.
471         This has been replaced with the corresponding code from
472         CharVectorSlice.  (html-lex must be re-lexed!)
473      7. BitArray in smlnj-lib/Util has been extended/modified to match the
474         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
475      8. Removed temporary additions (fromInternal, toInternal) from the
476         (now obsolete) IntInf in smlnj-lib/Util.
477      9. Cleaned up structure Byte.
478     10. Added localOffset, scan, and fromString to Date (according to spec).
479         Cleaned/corrected implementation of Date.
480         (Still need to check for correctness; implement better canonicalizeDate.)
481     11. Added "scan" to signature IEEE_REAL.
482     12. Some improvements to IntInf [in particular: efficiency-hack for
483         mod and rem when second operand is 2 (for parity checks).]
484     13. Changed representation of type Time.time, using a single IntInf.int
485         value counting microseconds.  This considerably simplified the
486         implementation of structure Time.  We now support negative time
487         values; scan and fromString handle signs.
488     14. Functor PrimIO now takes two additional arguments (VectorSlice and
489         ArraySlice).
490    
491    ----------------------------------------------------------------------
492    Name: Matthias Blume (blume (at) tti - c (dot) org)
493    Date: 2003/08/28 17:00:00 CDT
494    Tag: blume-20030828-intinf
495    Description:
496    
497    This is a major update which comes with a version number bump
498    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
499    and an implementation of IntInf in the Basis.
500    
501    There are a fairly large number of related changes and updates throughout
502    the system:
503    
504    Basis:
505            - Implemented IntInf.
506            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
507            - Added some missing Real64 operations, most notably Real.toLargeInt.
508            - Added FixedInt as a synonym for Int32.
509    
510    compiler:
511            * Added support for a built-in intinf type.
512              - literals
513              - pattern matching
514              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
515              - overloading on literals and operations
516    
517            This required adding a primitive type intinf, some additional
518            primops, and implementations for several non-trivial intinf
519            operations in Core.  (The intinf type is completely abstract
520            to the compiler; all operations get delegated back to the Core.)
521    
522            * Intinf equality is handled by polyequal.  However, the compiler
523            does not print its usual warning in this case (since polyequal
524            is the right thing to do there).
525    
526            * Improved the organization of structure InlineT.
527    
528            * A word about conversion primops:
529            If conversions involving intinf do not cancel out during
530            CPS contract, then the compiler must insert calls to Core functions.
531            Since all core access must be resolved already during the FLINT
532            translate phase, it would be too late a the time of CPS contract
533            to add new Core calls.  For this reason, conversion primops
534            for intinf carry two arguments:  1. the numeric argument that
535            they are supposed to convert, and 2. the Core function that
536            can help with this conversion if necessary.  If CPS contract
537            eliminates a primop, then the associated Core function becomes
538            dead and goes away.  Intinf conversion primops that do not get
539            eliminated by CPS contract get rewritten into calls of their
540            core functions by a separate, new phase.
541    
542    interactive system:
543            - Control.Print.intinfDepth controls max length of intinf constants
544            being printed. (Analogous to Control.Print.stringDepth.)
545            - Cleanup in printutil and pputil: got rid of unused stuff and
546            duplicates; replaced some of the code with code that makes better
547            use of library functionality.
548    
549    CM:
550            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
551            are no longer silent.
552    
553    CKIT:
554            Fixed mismatched uses of Int32 and LargeInt.  I always decided
555            in favor of LargeInt -- which is now the same as IntInf.
556            CKIT-knowledgable people should check whether this is what's
557            intended and otherwise change things back to using Int32 or
558            FixedInt.
559    
560    Throughout the code:
561            Started using IntInf.int literals and built-in operations
562            (e.g., comparison with 0) where this seems appropriate.
563    
564    
565    ----------------------------------------------------------------------
566    Name: Dave MacQueen (dbm@cs.uchicago.edu)
567    Date: 2003/08/13 11:36:00 CDT
568    Tag: dbm-20030813-mcz-merge1
569    Description:
570    
571    Merging changes from the mcz-branch development branch into trunk.
572    These changes involve replacement of the emulated old prettyprinter
573    interface with direct use of the SML/NJ Lib PP library, and fixing
574    of a couple of bugs (895, 1186) relating to error messages.  A new
575    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
576    has been added.
577    
578    ----------------------------------------------------------------------
579    Name: Matthias Blume (blume (at) tti - c (dot) org)
580    Date: 2003/08/11 15:45:00 CDT
581    Tag: blume-20030811-windows
582    Description:
583    
584    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
585    
586        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
587    
588    This patch restores SML/NJ's ability to run under win32.  There are a
589    number of changes, including fixes for several bugs that had gone
590    unnoticed until now:
591    
592      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
593        trying to run the win32 version from within cygwin.)
594    
595      - There are a number of simple .bat scripts that substitute for their
596        corresponding Unix shell-scripts.  (See below.)
597    
598      - The internals of ml-build have been modified slightly.  The main
599        difference is that instead of calling ".link-sml" (or link-sml.bat)
600        using OS.Process.system, the ML process delegates this task back
601        to the script.  Otherwise problems arise in mixed environments such
602        as Cygwin where scripts look and work like Unix scripts, but
603        where OS.Process.system cannot run them.
604    
605      - In CM, the srcpath pickler used native pathname syntax -- which
606        is incorrect in the case of cross-compilation.  The new pickle format
607        is independent of platform-specific naming conventions.
608    
609      - Path configuration files (such as lib/pathconfig) can now choose
610        between native and standard syntax.  Placing a line of the form
611    
612            standard!
613    
614        into the file causes all subsequent paths to be interpreted using
615        CM standard pathname syntax (= Unix conventions); a line
616    
617            native!
618    
619        switches back to native style.  This was needed so that
620        path config files can be written portably, see src/system/pathconfig.
621    
622      - Runtime system:
623    
624          - win32-filesys.c:  get_file_time and set_file_time now
625            access modification time, not creation time.
626    
627          - I/O code made aware of new array representation.
628    
629          - Bug fixes in X86.prim.masm.
630    
631       - src/system/makeml made aware of win32. (For use under cygwin
632            and other Unix-environments for windows.)
633    
634       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
635         which caused BinIO.inputAll to fail consistently.
636    
637    .bat scripts:
638    
639       Windows .bat scripts assume that SMLNJ_HOME is defined.
640    
641       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
642         applications (sml, ml-yacc, ml-lex).
643       - ml-build.bat: analogous to ml-build.
644       - config\install.bat:  Analogous to config/install.sh.  This requires
645         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
646         (nmake etc. must be on the path, and vcvars32 must have been run.)
647         Moreover, sources for ml-lex and ml-yacc need to exist under src,
648         and the bootfile hierarchy must have been unpacked under
649         sml.boot.x86-win32.
650         The script is very primitive and does a poor job at error checking.
651         It only installs the base system, ml-lex, and ml-yacc.  No other
652         libraries are being installed (i.e., you get only those that
653         are part of the compiler.)
654       - link-sml.bat:  analogous to .link-sml, but not currently used
655    
656    Unrelated bug fixes:
657    
658      - ml-nlffigen now exports structures ST_* corresponding to incomplete
659        types.
660      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
661        otherwise.)
662    
663    ----------------------------------------------------------------------
664    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
665    Date: 2003/06/17
666    Tag: macqueen-20030617-bug895
667    Description:
668    
669    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
670    Tag will be used for new development branch (mcz-branch) for
671    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
672    
673    ----------------------------------------------------------------------
674    Name: Matthias Blume (blume (at) tti - c (dot) org)
675    Date: 2003/05/27 16:55:00 CDT
676    Tag: blume-20030527-polyeq
677    Description:
678    
679    Tried to eliminated most cases of polymorphic equality.
680    
681    ----------------------------------------------------------------------
682    Name: Matthias Blume (blume (at) tti - c (dot) org)
683    Date: 2003/05/21 17:45:00 CDT
684    Tag: blume-20030517-complete
685    Description:
686    
687    Two changes:
688    
689      1. Added a flag for controlling whether non-exhaustive bindings will
690         be treated as errors (default is false).
691      2. Cleaned up the *entire* source tree so that CMB.make goes through
692         without a single non-exhaustive match- or bind warning.
693    
694    ----------------------------------------------------------------------
695    Name: Matthias Blume (blume (at) tti - c (dot) org)
696    Date: 2003/05/17 10:20:00 CDT
697    Tag: blume-20030517-absyn
698    Description:
699    
700    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
701    
702       This mainly affects the quality of error messages.  However, some
703       of the code is now more straightforward than before. (Treatment of
704       the above four constructs in translate.sml is much simpler than
705       the "macro-expansion" that was going on before.  Plus, the mach-
706       compiler no longer gets invoked just to be able to compile an
707       if-expression.)
708    
709    2. The ErrorMsg.Error exception is now caught and absorbed by the
710       interactive loop.
711    
712    ----------------------------------------------------------------------
713    Name: Allen Leung
714    Date: 2003/05/16 13:05:00 CDT
715    Tag: leunga-20030516-cygwin-runtime
716    Description:
717    
718        Ported the runtime system to cygwin, which uses the unix
719    x86-unix bin files.  Missing/buggy features:
720    
721       o getnetbyname, getnetbyaddr: these functions seem to be missing in
722         the Cygwin library.
723       o Ctrl-C handling may be flaky.
724       o Windows system calls and Windows I/O are not supported.
725    
726       A new set of binfiles is located at:
727    
728         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
729    
730       This is only needed for bootstrapping the cygwin version of smlnj.
731       Other x86 versions can use the existing binfiles.
732    
733    ----------------------------------------------------------------------
734    Name: Matthias Blume
735    Date: 2003/04/08 15:42:00 CDT
736    Tag: blume-20030408-listpair
737    Description:
738    
739    1. Added a target 'mlrisc' to installer.
740    
741    2. Added missing elements to structure ListPair.
742    
743    ----------------------------------------------------------------------
744    Name: Allen Leung
745    Date: 2003/01/07 10:40:00 EST
746    Tag: leunga-20030107-int-rem
747    Description:
748    
749    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
750    The arguments to the SUBL instruction were swapped.
751    
752    ----------------------------------------------------------------------
753    Name: Matthias Blume
754    Date: 2002/12/12 16:25:00 EST
755    Tag: blume-20021212-risc-ra
756    Description:
757    
758    Fixed a serious bug in the rewrite code for FP spilling/reloading that
759    sent the RA into an infinite loop when floating point registers get
760    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
761    110.37 and 110.38.)
762    There was another set of potential problems related to the handling of
763    MLRISC annotations (but those did not yet cause real problems, apparently).
764    
765    ----------------------------------------------------------------------
766    Name: Matthias Blume
767    Date: 2002/12/06 22:40:00 EST
768    Tag: blume-20021206-cm-fileid
769    Description:
770    
771    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
772    This fixes the problem of CM getting confused by files that suddenly
773    change their identity (e.g., by getting unlinked and recreated by some
774    text editor such as vi).  There might be a better/cheaper/cleaner way
775    of doing this, but for now this will have to do.
776    
777    ----------------------------------------------------------------------
778    Name: Matthias Blume
779    Date: 2002/10/28 09:50:00 EST
780    Tag: blume-20021028-typecheck
781    Description:
782    
783    Exported structure Typecheck from $smlnj/viscomp/core.cm.
784    
785    ----------------------------------------------------------------------
786    Name: Matthias Blume
787    Date: 2002/10/17 09:10:00 EDT
788    Tag: Release_110_42
789    Description:
790    
791    In good old tradition, there has been a slight hiccup so that we have
792    to patch 110.42 after the fact.   The old release tag has been replaced
793    (see below).
794    
795    The change solves a problem with two competing approaches the
796    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
797    each other's way.
798    
799    This change only affects the runtime system code and the installer script.
800    (No new bootfiles.)
801    
802    ----------------------------------------------------------------------
803    Name: Matthias Blume
804    Date: 2002/10/16 12:00:00 EDT
805    Tag: Release_110_42_removed
806    Description:
807    
808    New working release.  New bootfiles.
809    
810    ----------------------------------------------------------------------
811    Name: Matthias Blume
812    Date: 2002/10/10 13:10:00 EDT
813    Tag: blume-20021010-ppc-divs
814    Description:
815    
816    The mltree operator DIVS must be implemented with an overflow check on
817    the PPC because the hardware indicates divide-by-zero using "overflow" as
818    well.
819    
820    ----------------------------------------------------------------------
821    Name: Matthias Blume
822    Date: 2002/07/23 11:45:00
823    Tag: blume-20020723-smlnj-home
824    Description:
825    
826    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
827    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
828    CM_PATHCONFIG is also set), the path configuration file is assumed
829    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
830    the entire tree to some other place and everything will "just work".
831    
832    (Companion commands such as ml-build and ml-makedepend also sense this
833    variable.)
834    
835    ----------------------------------------------------------------------
836    Name: Matthias Blume
837    Date: 2002/07/12 21:19:00 EDT
838    Tag: blume-20020712-liveness
839    Description:
840    
841    Exported two useful "step" functions from liveness module (MLRISC).
842    
843    ----------------------------------------------------------------------
844    Name: Matthias Blume
845    Date: 2002/07/05 16:00 EDT
846    Tag: Release_110_41
847    Description:
848    
849    New working release.  New bootfiles.
850    
851    ----------------------------------------------------------------------
852    Name: Matthias Blume
853    Date: 2002/07/05 10:25:00 EDT
854    Tag: blume-20020705-btimp
855    Description:
856    
857    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
858    other clients can set up backtracing support.
859    
860    ----------------------------------------------------------------------
861    Name: Matthias Blume
862    Date: 2002/06/25 17:23:00 EDT
863    Tag: blume-20020625-fpmax
864    Description:
865    
866    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
867    case.  (The sense of the isNaN test was reversed -- which made min and
868    max always return their first argument.)
869    
870    ----------------------------------------------------------------------
871    Name: Matthias Blume
872    Date: 2002/06/11
873    Tag: blume-20020611-unixpath
874    Description:
875    
876    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
877    
878    ----------------------------------------------------------------------
879    Name: Matthias Blume
880    Date: 2002/06/10 16:35:00 EDT
881    Tag: blume-20020610-ieeereal
882    Description:
883    
884    I back-ported my implementation of IEEEReal.fromString from the
885    idlbasis-devel branch so that we can test it.
886    
887    Another small change is that ppDec tries to give more information
888    than just "<sig>" in the case of functors.  However, this code is
889    broken in some mysterious way if the functor's body's signature
890    has not been declared by ascription but gets inferred from the
891    implementation.  This needs fixing...
892    
893    ----------------------------------------------------------------------
894    Name: Matthias Blume
895    Date: 2002/05/31
896    Tag: blume-20020531-btrace-mode
897    Description:
898    
899    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
900    the wayside when I switched over to using Controls everywhere.)
901    
902    ----------------------------------------------------------------------
903    Name: Lal George
904    Date: 2002/05/23 12:21:40 EDT
905    Tag: george-20020523-visual-labels
906    Description:
907    
908    Labels are now displayed in the graphical output to make
909    the fall-through and target blocks obvious.
910    
911    ----------------------------------------------------------------------
912    Name: Matthias Blume
913    Date: 2002/05/22 11:03:00 EDT
914    Tag: blume-20020522-shrink
915    Description:
916    
917    John tweaked yesterday's fix for 1131 to handle an out-of-memory
918    situation that comes up when allocating huge arrays.
919    
920    ----------------------------------------------------------------------
921    Name: Matthias Blume
922    Date: 2002/05/21 16:00:00 EDT
923    Tag: Release_110_40
924    Description:
925    
926    New working release (110.40).  New bootfiles.
927    
928    [Also: John Reppy fixed GC bug 1131.]
929    
930    ----------------------------------------------------------------------
931    Name: Matthias Blume
932    Date: 2002/05/21 12:35:00 EDT
933    Tag: blume-20020521-cmdoc
934    Description:
935    
936    CM documentation update.
937    
938    ----------------------------------------------------------------------
939    Name: Matthias Blume
940    Date: 2002/05/21 10:55:00 EDT
941    Tag: blume-20020521-misc
942    Description:
943    
944    - John tweaked runtime to be silent on heap export (except when
945      GC messages are on).
946    - I added a few more things (cross-compiling versions of CMB) to
947      config/preloads (as suggestions).
948    
949    ----------------------------------------------------------------------
950    Name: Matthias Blume
951    Date: 2002/05/20 22:25:00 EDT
952    Tag: blume-20020520-controls
953    Description:
954    
955    - Added ControlUtil structure to control-lib.cm.
956    - Use it throughout.
957    - Used Controls facility to define MLRISC controls (as opposed to
958      registering MLRISC control ref cells with Controls after the fact)
959    - Fixed messed-up controls priorities.
960    
961    * Removed again all the stuff from config/preloads that one wouldn't
962      be able to preload at the time the initial heap image is built.
963      (Many libraries, e.g., CML, do not exist yet at this time.  The
964       only libraries that can be preloaded via config/preloads are those
965       that come bundled with the bootfiles.)
966    
967    ----------------------------------------------------------------------
968    Name: Matthias Blume
969    Date: 2002/05/20 10:59:00 EDT
970    Tag: blume-20020520-preloads
971    Description:
972    
973    Added a lot of commented-out suggestions for things to be included
974    in config/preloads.
975    
976    ----------------------------------------------------------------------
977    Name: Allen Leung
978    Date: 2002/05/18 14:20:00 EDT
979    Tag: leunga-20020518-mdl
980    Description:
981    
982    o Made the mdl tool stuff compile and run again.
983    o I've disabled all the stuff that depends on RTL specifications; they
984      are all badly broken anyway.
985    
986    ----------------------------------------------------------------------
987    Name: Matthias Blume
988    Date: 2002/05/17 16:49:00 EDT
989    Tag: blume-20020517-controls
990    Description:
991    
992    0. John Reppy made several modifications to the SML/NJ library.
993       In particular, there is a shiny new controls-lib.cm.
994    
995    1. Pushed new controls interface through compiler so that everything
996       compiles again.
997    
998    2. Added FormatComb and FORMAT_COMB to the CML version of the
999       SML/NJ library (so that CML compiles again).
1000    
1001    3. Modified init scripts because XXX_DEFAULT environment variables
1002       are no longer with us.  (Boot-time initialization is now done
1003       using the same environment variables that are also used for
1004       startup-time initialization of controls.)
1005    
1006    ----------------------------------------------------------------------
1007    Name: Lal George
1008    Date: 2002/05/15 09:20:10 EDT
1009    Tag: george-20020515-pseudo-op-decls
1010    Description:
1011    
1012            All pseudo-ops emitted before the first segment declaration
1013    such as TEXT, DATA, and BSS directives are assumed to be global
1014    declarations and are emitted first in the assembly file. This is
1015    useful in a number of situations where one has pseudo-ops that are not
1016    specific to any segment, and also works around the constraint that one
1017    cannot have client pseudo-ops in the TEXT segment.
1018    
1019            Because no segment is associated with these declarations it is
1020    an error to allocate any space or objects before the first segment
1021    directive and an exception will be raised. However, we cannot make
1022    this check for client pseudo-ops.
1023    
1024            These top level declarations are a field in the CFG graph_info.
1025    In theory you can continue to add to this field after the CFG has been
1026    built -- provided you know what you are doing;-)
1027    
1028    ----------------------------------------------------------------------
1029    Name: Matthias Blume
1030    Date: 2002/05/13 16:40:00 EDT
1031    Tag: blume-20020513-pp-etc
1032    Description:
1033    
1034    A few minor bugfixes:
1035    
1036      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
1037        (Bogus printouts for redefined bindings still occur.  Compiler
1038        bug should no longer occur now.  We need to redo the prettyprinter
1039        from scratch.)
1040    
1041      - CM pathname printer now also adds escape sequences for ( and )
1042    
1043      - commend and docu fixes for ml-nlffi
1044    
1045    ----------------------------------------------------------------------
1046    Name: Matthias Blume
1047    Date: 2002/05/10 16:40:00 EDT
1048    Tag: blume-20020510-erg-textio
1049    Description:
1050    
1051    Applied the following bugfix provided by Emden Gansner:
1052    
1053        Output is corrupted when outputSubstr is used rather than output.
1054    
1055        The problem occurs when a substring
1056    
1057            ss = (s, dataStart, dataLen)
1058    
1059        where dataStart > 0, fills a stream buffer with avail bytes left.
1060        avail bytes of s, starting at index dataStart, are copied into the
1061        buffer, the buffer is flushed, and then the remaining dataLen-avail
1062        bytes of ss are copied into the beginning of the buffer. Instead of
1063        starting this copy at index dataStart+avail in s, the current code
1064        starts the copy at index avail.
1065    
1066        Fix:
1067        In text-io-fn.sml, change line 695 from
1068             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
1069        to
1070             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
1071    
1072    ----------------------------------------------------------------------
1073    Name: Matthias Blume
1074    Date: 2002/04/12 13:55:00 EDT
1075    Tag: blume-20020412-assyntax
1076    Description:
1077    
1078    1. Grabbed newer assyntax.h from the XFree86 project.
1079    2. Fiddled with how to compile X86.prim.asm without warnings.
1080    3. (Very) Minor cleanup in CM.
1081    
1082    ----------------------------------------------------------------------
1083    Name: Matthias Blume
1084    Date: 2002/04/01 (no joke!) 17:07:00 EST
1085    Tag: blume-20020401-x86div
1086    Description:
1087    
1088    Added full support for div/mod/rem/quot on the x86, using the machine
1089    instruction's two results (without clumsily recomputing the remainder)
1090    directly where appropriate.
1091    
1092    Some more extensive power-of-two support was added to the x86 instruction
1093    selector (avoiding expensive divs, mods, and muls where they can be
1094    replaced with cheaper shifts and masks).  However, this sort of thing
1095    ought to be done earlier, e.g., within the CPS optimizer so that
1096    all architectures benefit from it.
1097    
1098    The compiler compiles to a fixed point, but changes might be somewhat
1099    fragile nevertheless.  Please, report any strange things that you might
1100    see wrt. div/mod/quot/rem...
1101    
1102    ----------------------------------------------------------------------
1103    Name: Matthias Blume
1104    Date: 2002/03/29 17:22:00
1105    Tag: blume-20020329-div
1106    Description:
1107    
1108    Fixed my broken div/mod logic.  Unfortunately, this means that the
1109    inline code for div/mod now has one more comparison than before.
1110    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
1111    The problem was with quotient = 0, because that alone does not tell
1112    us which way the rounding went.  One then has to look at whether
1113    remainder and divisor have the same sign...  :(
1114    
1115    Anyway, I replaced the bootfiles with fresh ones...
1116    
1117    ----------------------------------------------------------------------
1118    Name: Matthias Blume
1119    Date: 2002/03/29 14:10:00 EST
1120    Tag: blume-20020329-inlprims
1121    Description:
1122    
1123    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
1124    
1125    Primops have changed. This means that the bin/boot-file formats have
1126    changed as well.
1127    
1128    To make sure that there is no confusion, I made a new version.
1129    
1130    
1131    CHANGES:
1132    
1133    * removed REMT from mltree (remainder should never overflow).
1134    
1135    * added primops to deal with divisions of all flavors to the frontend
1136    
1137    * handled these primops all the way through so they map to their respective
1138      MLRISC support
1139    
1140    * used these primops in the implementation of Int, Int32, Word, Word32
1141    
1142    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
1143    
1144    * parameterized INLMIN, INLMAX, and INLABS by a numkind
1145    
1146    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
1147      floating point
1148    
1149    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
1150      and Real (but Real.abs maps to a separate floating-point-only primop)
1151    
1152    
1153    TODO items:
1154    
1155    * Hacked Alpha32 instruction selection, disabling the selection of REMx
1156      instructions because the machine instruction encoder cannot handle
1157      them.  (Hppa, PPC, and Sparc instruction selection did not handle
1158      REM in the first place, and REM is supported by the x86 machine coder.)
1159    
1160    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
1161      selection phase.  (The two can be streamlined because the hardware
1162      delivers both quotient and remainder at the same time anyway.)
1163    
1164    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
1165      (Currently the behavior is inconsistent both across architectures and
1166      wrt. the draft Basis spec.)
1167    
1168    * Word8 should eventually be handled natively, too.
1169    
1170    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
1171      as if there currently is no execution path that could trigger it in
1172      SML/NJ.  (The assumptions underlying functions arith and promotable do not
1173      hold for things like multiplication and division.)
1174    
1175    ----------------------------------------------------------------------
1176    Name: Matthias Blume
1177    Date: 2002/03/27 16:27:00 EST
1178    Tag: blume-20020327-mlrisc-divisions
1179    Description:
1180    
1181    Added support for all four division operations (ML's div, mod, quot,
1182    and rem) to MLRISC.  In the course of doing so, I also rationalized
1183    the naming (no more annoying switch-around of DIV and QUOT), by
1184    parameterizing the operation by div_rounding_mode (which can be either
1185    DIV_TO_ZERO or DIV_TO_NEGINF).
1186    
1187    The generic MLTreeGen functor takes care of compiling all four
1188    operations down to only round-to-zero div.
1189    
1190    Missing pieces:
1191    
1192      * Doing something smarter than relying on MLTreeGen on architectures
1193        like, e.g., the x86 where hardware division delivers both quotient and
1194        remainder at the same time.  With this, the implementation of the
1195        round-to-neginf operations could be further streamlined.
1196    
1197      * Remove inlining support for div/mod/rem from the frontend and replace it
1198        with primops that get carried through to the backend.  Do this for all
1199        int and word types.
1200    
1201    ----------------------------------------------------------------------
1202    Name: Matthias Blume
1203    Date: 2002/03/25 17:25:00 EST
1204    Tag: blume-20020325-divmod
1205    Description:
1206    
1207    I improved (hopefully without breaking them) the implementation of Int.div,
1208    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
1209    advantage of the following observations:
1210    
1211      Let  q = x quot y      r = x rem y
1212           d = x div  y      m = x mod y
1213    
1214    where "quot" is the round-to-zero version of integer division that
1215    hardware usually provides.  Then we have:
1216    
1217         r = x - q * y        where neither the * nor the - will overflow
1218         d = if q >= 0 orelse x = q * y then q else q - 1
1219                              where neither the * nor the - will overflow
1220         m = if q >= 0 orelse r = 0 then r else r + y
1221                              where the + will not overflow
1222    
1223    This results in substantial simplification of the generated code.
1224    The following table shows the number of CFG nodes and edges generated
1225    for
1226            fun f (x, y) = x OPER y
1227            (* with OPER \in div, mod, quot, rem *)
1228    
1229    
1230        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
1231        --------------------------------------------------------
1232         div |         24 |         39 |         12 |         16
1233         mod |         41 |         71 |         12 |         16
1234        quot |          8 |         10 |          8 |         10
1235         rem |         10 |         14 |          8 |         10
1236    
1237    
1238    ----------------------------------------------------------------------
1239    Name: Matthias Blume
1240    Date: 2002/03/25 22:06:00 EST
1241    Tag: blume-20020325-cprotobug
1242    Description:
1243    
1244    Fixed a bug in cproto (c prototype decoder).
1245    
1246    ----------------------------------------------------------------------
1247    Name: Matthias Blume
1248    Date: 2002/03/25 16:00:00 EST
1249    Tag: blume-20020325-raw-primops
1250    Description:
1251    
1252    I did some cleanup to Allen's new primop code and
1253    replaced yesterday's bootfiles with new ones.
1254    (But they are stored in the same place.)
1255    
1256    ----------------------------------------------------------------------
1257    Name: Matthias Blume
1258    Date: 2002/03/24 22:40:00 EST
1259    Tag: blume-20020324-bootfiles
1260    Description:
1261    
1262    Made the bootfiles that Allen asked for.
1263    
1264    ----------------------------------------------------------------------
1265    Name: Allen Leung
1266    Date: 2002/03/23 15:50:00 EST
1267    Tag: leunga-20020323-flint-cps-rcc-primops
1268    Description:
1269    
1270      1. Changes to FLINT primops:
1271    
1272        (* make a call to a C-function;
1273         * The primop carries C function prototype information and specifies
1274         * which of its (ML-) arguments are floating point. C prototype
1275         * information is for use by the backend, ML information is for
1276         * use by the CPS converter. *)
1277      | RAW_CCALL of { c_proto: CTypes.c_proto,
1278                       ml_args: ccall_type list,
1279                       ml_res_opt: ccall_type option,
1280                       reentrant : bool
1281                     } option
1282       (* Allocate uninitialized storage on the heap.
1283        * The record is meant to hold short-lived C objects, i.e., they
1284        * are not ML pointers.  With the tag, the representation is
1285        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
1286        *)
1287      | RAW_RECORD of {tag:bool,sz:int}
1288      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
1289    
1290      2.  These CPS primops are now overloaded:
1291    
1292           rawload of {kind:numkind}
1293           rawstore of {kind:numkind}
1294    
1295          The one argument form is:
1296    
1297             rawload {kind} address
1298    
1299          The two argument form is:
1300    
1301             rawload {kind} [ml object, byte-offset]
1302    
1303      3. RAW_CCALL/RCC now takes two extra arguments:
1304    
1305         a. The first is whether the C call is reentrant, i.e., whether
1306            ML state should be saved and restored.
1307         b. The second argument is a string argument specifying the name of
1308            library and the C function.
1309    
1310         These things are currently not handled in the code generator, yet.
1311    
1312      4. In CProto,
1313    
1314         An encoding type of "bool" means "ml object" and is mapped into
1315         C prototype of PTR.  Note that "bool" is different than "string",
1316         even though "string" is also mapped into PTR, because "bool"
1317         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
1318    
1319      5. Pickler/unpicker
1320    
1321         Changed to handle RAW_RECORD and newest RAW_CCALL
1322    
1323      6. MLRiscGen,
1324    
1325         1. Changed to handle the new rawload/rawstore/rawrecord operators.
1326         2. Code for handling C Calls has been moved to a new module CPSCCalls,
1327            in the file CodeGen/cpscompile/cps-c-calls.sml
1328    
1329      7. Added the conditional move operator
1330    
1331             condmove of branch
1332    
1333         to cps.  Generation of this is still buggy so it is currently
1334         disabled.
1335    
1336    ----------------------------------------------------------------------
1337    Name: Lal George
1338    Date: 2002/03/22 14:18:25 EST
1339    Tag: george-20020322-cps-branch-prob
1340    Description:
1341    
1342    Implemented the Ball-Larus branch prediction-heuristics, and
1343    incorporated graphical viewers for control flow graphs.
1344    
1345    Ball-Larus Heuristics:
1346    ---------------------
1347    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1348    
1349    By design it uses the Dempster-Shafer theory for combining
1350    probabilities.  For example, in the function:
1351    
1352        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1353    
1354    the ball-larus heuristics predicts that the n=0 is unlikely
1355    (OH-heuristic), and the 'then' branch is unlikely because of the
1356    RH-heuristic -- giving the 'then' branch an even lower combined
1357    probability using the Dempster-Shafer theory.
1358    
1359    Finally, John Reppy's loop analysis in MLRISC, further lowers the
1360    probability of the 'then' branch because of the loop in the else
1361    branch.
1362    
1363    
1364    Graphical Viewing:
1365    ------------------
1366    I merely plugged in Allen's graphical viewers into the compiler. The
1367    additional code is not much. At the top level, saying:
1368    
1369            Control.MLRISC.getFlag "cfg-graphical-view" := true;
1370    
1371    will display the graphical view of the control flow graph just before
1372    back-patching.  daVinci must be in your path for this to work. If
1373    daVinci is not available, then the default viewer can be changed
1374    using:
1375    
1376            Control.MLRISC.getString "viewer"
1377    
1378    which can be set to "dot" or "vcg" for the corresponding viewers. Of
1379    course, these viewers must be in your path.
1380    
1381    The above will display the compilation unit at the level of clusters,
1382    many of which are small, boring, and un-interesting. Also setting:
1383    
1384            Control.MLRISC.getInt "cfg-graphical-view_size"
1385    
1386    will display clusters that are larger than the value set by the above.
1387    
1388    
1389    ----------------------------------------------------------------------
1390    Name: Matthias Blume
1391    Date: 2002/03/21 22:20:00 EST
1392    Tag: blume-20020321-kmp-bugfix
1393    Description:
1394    
1395    Changed the interface to the KMP routine in PreString and fixed
1396    a minor bug in one place where it was used.
1397    
1398    ----------------------------------------------------------------------
1399    Name: Allen Leung
1400    Date: 2002/03/21 20:30:00 EST
1401    Tag: leunga-20020321-cfg
1402    Description:
1403    
1404      Fixed a potential problem in cfg edge splitting.
1405    
1406    ----------------------------------------------------------------------
1407    Name: Allen Leung
1408    Date: 2002/03/21 17:15:00 EST
1409    Tag: leunga-20020321-x86-fp-cfg
1410    Description:
1411    
1412      1. Recoded the buggy parts of x86-fp.
1413    
1414         a. All the block reordering code has been removed.
1415            We now depend on the block placement phases to do this work.
1416    
1417         b. Critical edge splitting code has been simplified and moved into the
1418            CFG modules, as where they belong.
1419    
1420         Both of these were quite buggy and complex.  The code is now much, much
1421         simpler.
1422    
1423      2. X86 backend.
1424    
1425         a. Added instructions for 64-bit support.  Instruction selection for
1426            64-bit has not been committed, however, since that
1427            requires changes to MLTREE which haven't been approved by
1428            Lal and John.
1429    
1430         b. Added support for FUCOMI and FUCOMIP when generating code for
1431            PentiumPro and above.  We only generate these instructions in
1432            the fast-fp mode.
1433    
1434         c. Added cases for JP and JNP in X86FreqProps.
1435    
1436      3. CFG
1437    
1438         CFG now has a bunch of methods for edge splitting and merging.
1439    
1440      4. Machine description.
1441    
1442         John's simplification of MLTREE_BASIS.fcond broke a few machine
1443         description things:
1444    
1445         rtl-build.{sig,sml} and hppa.mdl fixed.
1446    
1447         NOTE: the machine description stuff in the repository is still broken.
1448               Again, I can't put my fixes in because that involves
1449               changes to MLTREE.
1450    
1451    ----------------------------------------------------------------------
1452    Name: Matthias Blume
1453    Date: 2002/03/20 15:55:00 EST
1454    Tag: blume-20020320-kmp
1455    Description:
1456    
1457    Implemented Knuth-Morris-Pratt string matching in PreString and used
1458    it for String.isSubstring, Substring.isSubstring, and
1459    Substring.position.
1460    
1461    (Might need some stress-testing.  Simple examples worked fine.)
1462    
1463    ----------------------------------------------------------------------
1464    Name: Matthias Blume
1465    Date: 2002/03/19 16:37:00 EST
1466    Tag: blume-20020319-witnesses
1467    Description:
1468    
1469    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1470    
1471    This implements a generic mechanism for changing constness qualifiers
1472    anywhere within big C types without resorting to outright "casts".
1473    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1474    modify the constness at the outermost level.)
1475    The implementation of "convert" is based on the idea of "witness"
1476    values -- values that are not used by the operation but whose types
1477    "testify" to their applicability.  On the implementation side, "convert"
1478    is simply a projection (returning its second curried argument).  With
1479    cross-module inlining, it should not result in any machine code being
1480    generated.
1481    
1482    ----------------------------------------------------------------------
1483    Name: Matthias Blume
1484    Date: 2002/03/15 16:40:00 EST
1485    Tag: blume-20020315-basis
1486    Description:
1487    
1488    Provided (preliminary?) implementations for
1489    
1490      {String,Substring}.{concatWith,isSuffix,isSubstring}
1491    
1492    and
1493    
1494      Substring.full
1495    
1496    Those are in the Basis spec but they were missing in SML/NJ.
1497    
1498    ----------------------------------------------------------------------
1499    Name: Matthias Blume
1500    Date: 2002/03/14 21:30:00 EST
1501    Tag: blume-20020314-controls
1502    Description:
1503    
1504    Controls:
1505    ---------
1506    
1507    1. Factored out the recently-added Controls : CONTROLS stuff and put
1508       it into its own library $/controls-lib.cm.  The source tree for
1509       this is under src/smlnj-lib/Controls.
1510    
1511    2. Changed the names of types and functions in this interface, so they
1512       make a bit more "sense":
1513    
1514          module -> registry
1515          'a registry -> 'a group
1516    
1517    3. The interface now deals in ref cells only.  The getter/setter interface
1518       is (mostly) gone.
1519    
1520    4. Added a function that lets one register an already-existing ref cell.
1521    
1522    5. Made the corresponding modifications to the rest of the code so that
1523       everything compiles again.
1524    
1525    6. Changed the implementation of Controls.MLRISC back to something closer
1526       to the original.  In particular, this module (and therefore MLRISC)
1527       does not depend on Controls.  There now is some link-time code in
1528       int-sys.sml that registers the MLRISC controls with the Controls
1529       module.
1530    
1531    CM:
1532    ---
1533    
1534      * One can now specify the lambda-split aggressiveness in init.cmi.
1535    
1536    ----------------------------------------------------------------------
1537    Name: Allen Leung
1538    Date: 2002/03/13 17:30:00 EST
1539    Tag: leunga-20020313-x86-fp-unary
1540    Description:
1541    
1542    Bug fix for:
1543    
1544    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1545    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1546    > - fun f(x,(y,z)) = Real.~ y;
1547    > [autoloading]
1548    > [autoloading done]
1549    >       fchsl   (%eax), 184(%esp)
1550    > Error: MLRisc bug: X86MCEmitter.emitInstr
1551    >
1552    > uncaught exception Error
1553    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1554    
1555    The problem was that the code generator did not generate any fp registers
1556    in this case, and the ra didn't know that it needed to run the X86FP phase to
1557    translate the pseudo fp instruction.   This only happened with unary fp
1558    operators in certain situations.
1559    
1560    ----------------------------------------------------------------------
1561    Name: Matthias Blume
1562    Date: 2002/03/13 14:00:00 EST
1563    Tag: blume-20020313-overload-etc
1564    Description:
1565    
1566    1. Added _overload as a synonym for overload for backward compatibility.
1567       (Control.overloadKW must be true for either version to be accepted.)
1568    
1569    2. Fixed bug in install script that caused more things to be installed
1570       than what was requested in config/targets.
1571    
1572    3. Made CM aware of the (_)overload construct so that autoloading
1573       works.
1574    
1575    ----------------------------------------------------------------------
1576    Name: Matthias Blume
1577    Date: 2002/03/12 22:03:00 EST
1578    Tag: blume-20020312-url
1579    Description:
1580    
1581    Forgot to update BOOT and srcarchiveurl.
1582    
1583    ----------------------------------------------------------------------
1584    Name: Matthias Blume
1585    Date: 2002/03/12 17:30:00 EST
1586    Tag: blume-20020312-version110392
1587    Description:
1588    
1589    Yet another version number bump (because of small changes to the
1590    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1591    
1592    Changes:
1593    
1594      The new pid generation scheme described a few weeks ago was overly
1595      complicated.  I implemented a new mechanism that is simpler and
1596      provides a bit more "stability":  Once CM has seen a compilation
1597      unit, it keeps its identity constant (as long as you do not delete
1598      those crucial CM/GUID/* files).  This means that when you change
1599      an interface, compile, then go back to the old interface, and
1600      compile again, you arrive at the original pid.
1601    
1602      There now also is a mechanism that instructs CM to use the plain
1603      environment hash as a module's pid (effectively making its GUID
1604      the empty string).  For this, "noguid" must be specified as an
1605      option to the .sml file in question within its .cm file.
1606      This is most useful for code that is being generated by tools such
1607      as ml-nlffigen (because during development programmers tend to
1608      erase the tool's entire output directory tree including CM's cached
1609      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1610      revert to the old, broken behavior of SML/NJ, but in specific cases
1611      where there is no danger of interface confusion, its use is ok
1612      (I think).
1613    
1614      ml-nlffigen by default generates "noguid" annotations.  They can be
1615      turned off by specifying -guid in its command line.
1616    
1617    ----------------------------------------------------------------------
1618    Name: Lal George
1619    Date: 2002/03/12 12 14:42:36 EST
1620    Tag: george-20020312-frequency-computation
1621    Description:
1622    
1623    Integrated jump chaining and static block frequency into the
1624    compiler. More details and numbers later.
1625    
1626    ----------------------------------------------------------------------
1627    Name: Lal George
1628    Date: 2002/03/11 11 22:38:53 EST
1629    Tag: george-20020311-jump-chain-elim
1630    Description:
1631    
1632    Tested the jump chain elimination on all architectures (except the
1633    hppa).  This is on by default right now and is profitable for the
1634    alpha and x86, however, it may not be profitable for the sparc and ppc
1635    when compiling the compiler.
1636    
1637    The gc test will typically jump to a label at the end of the cluster,
1638    where there is another jump to an external cluster containing the actual
1639    code to invoke gc. This is to allow factoring of common gc invocation
1640    sequences. That is to say, we generate:
1641    
1642            f:
1643               testgc
1644               ja   L1      % jump if above to L1
1645    
1646            L1:
1647               jmp L2
1648    
1649    
1650    After jump chain elimination the 'ja L1' instructions is converted to
1651    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1652    up being implemented in their long form (if L2 is far away) using:
1653    
1654            jbe     L3      % jump if below or equal to L3
1655            jmp     L2
1656         L3:
1657            ...
1658    
1659    
1660    For large compilation units L2  may be far away.
1661    
1662    
1663    ----------------------------------------------------------------------
1664    Name: Matthias Blume
1665    Date: 2002/03/11 13:30:00 EST
1666    Tag: blume-20020311-mltreeeval
1667    Description:
1668    
1669    A functor parameter was missing.
1670    
1671    ----------------------------------------------------------------------
1672    Name: Allen Leung
1673    Date: 2002/03/11 10:30:00 EST
1674    Tag: leunga-20020311-runtime-string0
1675    Description:
1676    
1677       The representation of the empty string now points to a
1678    legal null terminated C string instead of unit.  It is now possible
1679    to convert an ML string into C string with InlineT.CharVector.getData.
1680    This compiles into one single machine instruction.
1681    
1682    ----------------------------------------------------------------------
1683    Name: Allen Leung
1684    Date: 2002/03/10 23:55:00 EST
1685    Tag: leunga-20020310-x86-call
1686    Description:
1687    
1688       Added machine generation for CALL instruction (relative displacement mode)
1689    
1690    ----------------------------------------------------------------------
1691    Name: Matthias Blume
1692    Date: 2002/03/08 16:05:00
1693    Tag: blume-20020308-entrypoints
1694    Description:
1695    
1696    Version number bumped to 110.39.1.  NEW BOOTFILES!
1697    
1698    Entrypoints: non-zero offset into a code object where execution should begin.
1699    
1700    - Added the notion of an entrypoint to CodeObj.
1701    - Added reading/writing of entrypoint info to Binfile.
1702    - Made runtime system bootloader aware of entrypoints.
1703    - Use the address of the label of the first function given to mlriscGen
1704      as the entrypoint.  This address is currently always 0, but it will
1705      not be 0 once we turn on block placement.
1706    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1707      with entry points) from mlriscGen.
1708    
1709    ----------------------------------------------------------------------
1710    Name: Allen Leung
1711    Date: 2002/03/07 20:45:00 EST
1712    Tag: leunga-20020307-x86-cmov
1713    Description:
1714    
1715       Bug fixes for CMOVcc on x86.
1716    
1717       1. Added machine code generation for CMOVcc
1718       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1719       3. CMOVcc cannot have an immediate operand as argument.
1720    
1721    ----------------------------------------------------------------------
1722    Name: Matthias Blume
1723    Date: 2002/03/07 16:15:00 EST
1724    Tag: blume-20020307-controls
1725    Description:
1726    
1727    This is a very large but mostly boring patch which makes (almost)
1728    every tuneable compiler knob (i.e., pretty much everything under
1729    Control.* plus a few other things) configurable via both the command
1730    line and environment variables in the style CM did its configuration
1731    until now.
1732    
1733    Try starting sml with '-h' (or, if you are brave, '-H')
1734    
1735    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1736    implements the underlying generic mechanism.
1737    
1738    The interface to some of the existing such facilities has changed somewhat.
1739    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1740    (The getFoo interface is still there for backward-compatibility, but its
1741    use is deprecated.)
1742    
1743    The ml-build script passes -Cxxx=yyy command-line arguments through so
1744    that one can now twiddle the compiler settings when using this "batch"
1745    compiler.
1746    
1747    TODO items:
1748    
1749    We should go through and throw out all controls that are no longer
1750    connected to anything.  Moreover, we should go through and provide
1751    meaningful (and correct!) documentation strings for those controls
1752    that still are connected.
1753    
1754    Currently, multiple calls to Controls.new are accepted (only the first
1755    has any effect).  Eventually we should make sure that every control
1756    is being made (via Controls.new) exactly once.  Future access can then
1757    be done using Controls.acc.
1758    
1759    Finally, it would probably be a good idea to use the getter-setter
1760    interface to controls rather than ref cells.  For the time being, both
1761    styles are provided by the Controls module, but getter-setter pairs are
1762    better if thread-safety is of any concern because they can be wrapped.
1763    
1764    *****************************************
1765    
1766    One bug fix: The function blockPlacement in three of the MLRISC
1767    backpatch files used to be hard-wired to one of two possibilities at
1768    link time (according to the value of the placementFlag).  But (I
1769    think) it should rather sense the flag every time.
1770    
1771    *****************************************
1772    
1773    Other assorted changes (by other people who did not supply a HISTORY entry):
1774    
1775    1. the cross-module inliner now works much better (Monnier)
1776    2. representation of weights, frequencies, and probabilities in MLRISC
1777       changed in preparation of using those for weighted block placement
1778       (Reppy, George)
1779    
1780    ----------------------------------------------------------------------
1781    Name: Lal George
1782    Date: 2002/03/07 14:44:24 EST 2002
1783    Tag: george-20020307-weighted-block-placement
1784    
1785    Tested the weighted block placement optimization on all architectures
1786    (except the hppa) using AMPL to generate the block and edge frequencies.
1787    Changes were required in the machine properties to correctly
1788    categorize trap instructions. There is an MLRISC flag
1789    "weighted-block-placement" that can be used to enable weighted block
1790    placement, but this will be ineffective without block/edge
1791    frequencies (coming soon).
1792    
1793    
1794    ----------------------------------------------------------------------
1795    Name: Lal George
1796    Date: 2002/03/05 17:24:48 EST
1797    Tag: george-20020305-linkage-cluster
1798    
1799    In order to support the block placement optimization, a new cluster
1800    is generated as the very first cluster (called the linkage cluster).
1801    It contains a single jump to the 'real' entry point for the compilation
1802    unit. Block placement has no effect on the linkage cluster itself, but
1803    all the other clusters  have full freedom in the manner in which they
1804    reorder blocks or functions.
1805    
1806    On the x86 the typical linkage code that is generated is:
1807       ----------------------
1808            .align 2
1809       L0:
1810            addl    $L1-L0, 72(%esp)
1811            jmp     L1
1812    
1813    
1814            .align  2
1815       L1:
1816       ----------------------
1817    
1818    72(%esp) is the memory location for the stdlink register. This
1819    must contain the address of the CPS function being called. In the
1820    above example, it contains the address of  L0; before
1821    calling L1 (the real entry point for the compilation unit), it
1822    must contain the address for L1, and hence
1823    
1824            addl $L1-L0, 72(%esp)
1825    
1826    I have tested this on all architectures except the hppa.The increase
1827    in code size is of course negligible
1828    
1829    ----------------------------------------------------------------------
1830    Name: Allen Leung
1831    Date: 2002/03/03 13:20:00 EST
1832    Tag: leunga-20020303-mlrisc-tools
1833    
1834      Added #[ ... ] expressions to mlrisc tools
1835    
1836    ----------------------------------------------------------------------
1837    Name: Matthias Blume
1838    Date: 2002/02/27 12:29:00 EST
1839    Tag: blume-20020227-cdebug
1840    Description:
1841    
1842    - made types in structure C and C_Debug to be equal
1843    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1844    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1845    
1846    ----------------------------------------------------------------------
1847    Name: Matthias Blume
1848    Date: 2002/02/26 12:00:00 EST
1849    Tag: blume-20020226-ffi
1850    Description:
1851    
1852    1. Fixed a minor bug in CM's "noweb" tool:
1853       If numbering is turned off, then truly don't number (i.e., do not
1854       supply the -L option to noweb).  The previous behavior was to supply
1855       -L'' -- which caused noweb to use the "default" line numbering scheme.
1856       Thanks to Chris Richards for pointing this out (and supplying the fix).
1857    
1858    2. Once again, I reworked some aspects of the FFI:
1859    
1860       A. The incomplete/complete type business:
1861    
1862       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1863         gone!
1864       - ML types representing an incomplete type are now *equal* to
1865         ML types representing their corresponding complete types (just like
1866         in C).  This is still safe because ml-nlffigen will not generate
1867         RTTI for incomplete types, nor will it generate functions that
1868         require access to such RTTI.   But when ML code generated from both
1869         incomplete and complete versions of the C type meet, the ML types
1870         are trivially interoperable.
1871    
1872         NOTE:  These changes restore the full generality of the translation
1873         (which was previously lost when I eliminated functorization)!
1874    
1875       B. Enum types:
1876    
1877       - Structure C now has a type constructor "enum" that is similar to
1878         how the "su" constructor works.  However, "enum" is not a phantom
1879         type because each "T enum" has values (and is isomorphic to
1880         MLRep.Signed.int).
1881       - There are generic access operations for enum objects (using
1882         MLRep.Signed.int).
1883       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1884         * The structure contains the definition of type "mlrep" (the ML-side
1885         representation type of the enum).  Normally, mlrep is the same
1886         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1887         then mlrep will be defined as a datatype -- thus facilitating
1888         pattern matching on mlrep values.
1889         ("-ec" will be suppressed if there are duplicate values in an
1890          enumeration.)
1891         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1892         will be generated for each C enum constant xxx.
1893         * Conversion functions m2i and i2m convert between mlrep and
1894         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1895         * Coversion functions c and ml convert between mlrep and "tag enum".
1896         * Access functions (get/set) fetch and store mlrep values.
1897       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1898         enumerations are merged into one single enumeration represented by
1899         structure E_'.
1900    
1901    ----------------------------------------------------------------------
1902    Name: Allen Leung
1903    Date: 2002/02/25 04:45:00 EST
1904    Tag: leunga-20020225-cps-spill
1905    
1906    This is a new implementation of the CPS spill phase.
1907    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1908    In case of problems, replace it with the old file spill.sml
1909    
1910    The current compiler runs into some serious performance problems when
1911    constructing a large record.  This can happen when we try to compile a
1912    structure with many items.  Even a very simple structure like the following
1913    makes the compiler slow down.
1914    
1915        structure Foo = struct
1916           val x_1 = 0w1 : Word32.int
1917           val x_2 = 0w2 : Word32.int
1918           val x_3 = 0w3 : Word32.int
1919           ...
1920           val x_N = 0wN : Word32.int
1921        end
1922    
1923    The following table shows the compile time, from N=1000 to N=4000,
1924    with the old compiler:
1925    
1926    N
1927    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1928           MLRISC ra                               0.06u  0.00s  0.05g
1929              (spills = 0 reloads = 0)
1930           TOTAL                                   0.63u  0.07s  0.21g
1931    
1932    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1933           MLRISC ra                               5.68u  0.59s  3.93g
1934              (spills = 0 reloads = 0)
1935           TOTAL                                   14.71u  0.99s  4.81g
1936    
1937    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1938           MLRISC ra                               5.54u  0.65s  3.91g
1939              (spills = 543 reloads = 1082)
1940           TOTAL                                   65.40u  3.13s  6.00g
1941    
1942    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1943           MLRISC ra                               0.80u  0.10s  0.55g
1944              (spills = 42 reloads = 84)
1945           TOTAL                                   129.42u  5.10s  4.13g
1946    
1947    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1948           MLRISC ra                               2.69u  0.27s  1.38g
1949              (spills = 62 reloads = 124)
1950           TOTAL                                   682.48u  19.61s  13.99g
1951    
1952    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1953           MLRISC ra                               4.96u  0.27s  2.72g
1954              (spills = 85 reloads = 170)
1955           TOTAL                                   2375.26u  57.21s  48.00g
1956    
1957    As you can see the old cps spill module suffers from some serious
1958    performance problem.  But since I cannot decipher the old code fully,
1959    instead of patching the problems up, I'm reimplementing it
1960    with a different algorithm.  The new code is more modular,
1961    smaller when compiled, and substantially faster
1962    (O(n log n) time and O(n) space).  Timing of the new spill module:
1963    
1964    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1965          MLRISC ra                               0.25u  0.02s  0.15g
1966             (spills=1 reloads=3)
1967          TOTAL                                   7.74u  0.34s  1.62g
1968    
1969    Implementation details:
1970    
1971    As far as I can tell, the purpose of the CPS spill module is to make sure the
1972    number of live variables at any program point (the bandwidth)
1973    does not exceed a certain limit, which is determined by the
1974    size of the spill area.
1975    
1976    When the bandwidth is too large, we decrease the register pressure by
1977    packing live variables into spill records.  How we achieve this is
1978    completely different than what we did in the old code.
1979    
1980    First, there is something about the MLRiscGen code generator
1981    that we should be aware of:
1982    
1983    o MLRiscGen performs code motion!
1984    
1985       In particular, it will move floating point computations and
1986       address computations involving only the heap pointer to
1987       their use sites (if there is only a single use).
1988       What this means is that if we have a CPS record construction
1989       statement
1990    
1991           RECORD(k,vl,w,e)
1992    
1993       we should never count the new record address w as live if w
1994       has only one use (which is often the case).
1995    
1996       We should do something similar to floating point, but the transformation
1997       there is much more complex, so I won't deal with that.
1998    
1999    Secondly, there are now two new cps primops at our disposal:
2000    
2001     1. rawrecord of record_kind option
2002        This pure operator allocates some uninitialized storage from the heap.
2003        There are two forms:
2004    
2005         rawrecord NONE [INT n]  allocates a tagless record of length n
2006         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
2007                                     and initializes the tag.
2008    
2009     2. rawupdate of cty
2010          rawupdate cty (v,i,x)
2011          Assigns to x to the ith component of record v.
2012          The storelist is not updated.
2013    
2014    We use these new primops for both spilling and increment record construction.
2015    
2016     1. Spilling.
2017    
2018        This is implemented with a linear scan algorithm (but generalized
2019        to trees).  The algorithm will create a single spill record at the
2020        beginning of the cps function and use rawupdate to spill to it,
2021        and SELECT or SELp to reload from it.  So both spills and reloads
2022        are fine-grain operations.  In contrast, in the old algorithm
2023        "spills" have to be bundled together in records.
2024    
2025        Ideally, we should sink the spill record construction to where
2026        it is needed.  We can even split the spill record into multiple ones
2027        at the places where they are needed.  But CPS is not a good
2028        representation for global code motion, so I'll keep it simple and
2029        am not attempting this.
2030    
2031     2. Incremental record construction (aka record splitting).
2032    
2033        Long records with many component values which are simulatenously live
2034        (recall that single use record addresses are not considered to
2035         be live) are constructed with rawrecord and rawupdate.
2036        We allocate space on the heap with rawrecord first, then gradually
2037        fill it in with rawupdate.  This is the technique suggested to me
2038        by Matthias.
2039    
2040        Some restrictions on when this is applicable:
2041        1. It is not a VECTOR record.  The code generator currently does not handle
2042           this case. VECTOR record uses double indirection like arrays.
2043        2. All the record component values are defined in the same "basic block"
2044           as the record constructor.  This is to prevent speculative
2045           record construction.
2046    
2047    ----------------------------------------------------------------------
2048    Name: Allen Leung
2049    Date: 2002/02/22 01:02:00 EST
2050    Tag: leunga-20020222-mlrisc-tools
2051    
2052    Minor bug fixes in the parser and rewriter
2053    
2054    ----------------------------------------------------------------------
2055    Name: Allen Leung
2056    Date: 2002/02/21 20:20:00 EST
2057    Tag: leunga-20020221-peephole
2058    
2059    Regenerated the peephole files.  Some contained typos in the specification
2060    and some didn't compile because of pretty printing bugs in the old version
2061    of 'nowhere'.
2062    
2063    ----------------------------------------------------------------------
2064    Name: Allen Leung
2065    Date: 2002/02/19 20:20:00 EST
2066    Tag: leunga-20020219-mlrisc-tools
2067    Description:
2068    
2069       Minor bug fixes to the mlrisc-tools library:
2070    
2071       1.  Fixed up parsing colon suffixed keywords
2072       2.  Added the ability to shut the error messages up
2073       3.  Reimplemented the pretty printer and fixed up/improved
2074           the pretty printing of handle and -> types.
2075       4.  Fixed up generation of literal symbols in the nowhere tool.
2076       5.  Added some SML keywords to to sml.sty
2077    
2078    ----------------------------------------------------------------------
2079    Name: Matthias Blume
2080    Date: 2002/02/19 16:20:00 EST
2081    Tag: blume-20020219-cmffi
2082    Description:
2083    
2084    A wild mix of changes, some minor, some major:
2085    
2086    * All C FFI-related libraries are now anchored under $c:
2087        $/c.cm      --> $c/c.cm
2088        $/c-int.cm  --> $c/internals/c-int.cm
2089        $/memory.cm --> $c/memory/memory.cm
2090    
2091    * "make" tool (in CM) now treats its argument pathname slightly
2092      differently:
2093        1. If the native expansion is an absolute name, then before invoking
2094           the "make" command on it, CM will apply OS.Path.mkRelative
2095           (with relativeTo = OS.FileSys.getDir()) to it.
2096        2. The argument will be passed through to subsequent phases of CM
2097           processing without "going native".  In particular, if the argument
2098           was an anchored path, then "make" will not lose track of that anchor.
2099    
2100    * Compiler backends now "know" their respective C calling conventions
2101      instead of having to be told about it by ml-nlffigen.  This relieves
2102      ml-nlffigen from one of its burdens.
2103    
2104    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
2105    
2106    * Export C_DEBUG and C_Debug from $c/c.cm.
2107    
2108    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
2109      subtyping relationship between incomplete pointers and their complete
2110      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
2111      with the convention of instantiating 'o with ('t, 'c) obj whenever
2112      the pointer target type is complete.  In the incomplete case, 'o
2113      will be instantiated with some "'c iobj" -- a type obtained by
2114      using one of the functors PointerToIncompleteType or PointerToCompleteType.
2115    
2116      Operations that work on both incomplete and complete pointer types are
2117      typed as taking an 'o ptr while operations that require the target to
2118      be known are typed as taking some ('t, 'c) obj ptr.
2119    
2120      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
2121      where void is an eqtype without any values.  This makes it possible
2122      to work on voidptr values using functions meant to operate on light
2123      incomplete pointers.
2124    
2125    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
2126      been vastly simplified.
2127    
2128    ----------------------------------------------------------------------
2129    Name: Matthias Blume
2130    Date: 2002/02/19 10:48:00 EST
2131    Tag: blume-20020219-pqfix
2132    Description:
2133    
2134    Applied Chris Okasaki's bug fix for priority queues.
2135    
2136    ----------------------------------------------------------------------
2137    Name: Matthias Blume
2138    Date: 2002/02/15 17:05:00
2139    Tag: Release_110_39
2140    Description:
2141    
2142    Last-minute retagging is becoming a tradition... :-(
2143    
2144    This is the working release 110.39.
2145    
2146    ----------------------------------------------------------------------
2147    Name: Matthias Blume
2148    Date: 2002/02/15 16:00:00 EST
2149    Tag: Release_110_39-orig
2150    Description:
2151    
2152    Working release 110.39.  New bootfiles.
2153    
2154    (Update: There was a small bug in the installer so it wouldn't work
2155    with all shells.  So I retagged. -Matthias)
2156    
2157    ----------------------------------------------------------------------
2158    Name: Matthias Blume
2159    Date: 2002/02/15 14:17:00 EST
2160    Tag: blume-20020215-showbindings
2161    Description:
2162    
2163    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
2164    the latter can be useful for exploring what bindings are available at
2165    the interactive prompt.  (The first function returns only the list
2166    of symbols that are really bound, the second prints those but also the
2167    ones that CM's autoloading mechanism knows about.)
2168    
2169    ----------------------------------------------------------------------
2170    Name: Matthias Blume
2171    Date: 2002/02/15 12:08:00 EST
2172    Tag: blume-20020215-iptrs
2173    Description:
2174    
2175    Two improvements to ml-nlffigen:
2176    
2177      1. Write files only if they do not exist or if their current contents
2178         do not coincide with what's being written.  (That is, avoid messing
2179         with the time stamps unless absolutely necessary.)
2180    
2181      2. Implement a "repository" mechanism for generated files related
2182         to "incomplete pointer types".   See the README file for details.
2183    
2184    ----------------------------------------------------------------------
2185    Name: Matthias Blume
2186    Date: 2002/02/14 11:50:00 EST
2187    Tag: blume-20020214-quote
2188    Description:
2189    
2190    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
2191    because of the new and improved tag generation scheme.  (Thanks to Allen
2192    Leung for pointing it out.)
2193    
2194    ----------------------------------------------------------------------
2195    Name: Lal George
2196    Date: 2002/02/14 09:55:27 EST 2002
2197    Tag: george-20020214-isabelle-bug
2198    Description:
2199    
2200    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
2201    of Isabelle on the x86.
2202    
2203    From Allen:
2204    -----------
2205     I've found the problem:
2206    
2207         in ra-core.sml, I use the counter "blocked" to keep track of the
2208         true number of elements in the freeze queue.  When the counter goes
2209         to zero, I skip examining the queue.  But I've messed up the
2210         bookkeeping in combine():
2211    
2212             else ();
2213             case !ucol of
2214               PSEUDO => (if !cntv > 0 then
2215                     (if !cntu > 0 then blocked := !blocked - 1 else ();
2216                                        ^^^^^^^^^^^^^^^^^^^^^^^
2217                      moveu := mergeMoveList(!movev, !moveu)
2218                     )
2219                  else ();
2220    
2221         combine() is called to coalesce two nodes u and v.
2222         I think I was thinking that if the move counts of u and v are both
2223         greater than zero then after they are coalesced then one node is
2224         removed from the freeze queue.  Apparently I was thinking that
2225         both u and v are of low degree, but that's clearly not necessarily true.
2226    
2227    
2228    02/12/2002:
2229        Here's the patch.  HOL now compiles.
2230    
2231        I don't know how this impact on performance (compile
2232        time or runtime).  This bug caused the RA (especially on the x86)
2233        to go thru the potential spill phase when there are still nodes on the
2234        freeze queue.
2235    
2236    
2237    
2238    
2239    ----------------------------------------------------------------------
2240    Name: Matthias Blume
2241    Date: 2002/02/13 22:40:00 EST
2242    Tag: blume-20020213-fptr-rtti
2243    Description:
2244    
2245    Fixed a bug in ml-nlffigen that was introduced with one of the previous
2246    updates.
2247    
2248    ----------------------------------------------------------------------
2249    Name: Matthias Blume
2250    Date: 2002/02/13 16:41:00 EST
2251    Tag: blume-20020213-cmlpq
2252    Description:
2253    
2254    Added new priority queue export symbols (which have just been added to
2255    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
2256    not compile and the installer would choke.)
2257    
2258    ----------------------------------------------------------------------
2259    Name: Matthias Blume
2260    Date: 2002/02/13 16:15:00 EST
2261    Tag: blume-20020213-various
2262    Description:
2263    
2264    1. More tweaks to ml-nlffigen:
2265    
2266       - better internal datastructures (resulting in slight speedup)
2267       - "-match" option requires exact match
2268       - "localized" gensym counters (untagged structs/unions nested within
2269         other structs/unions or within typedefs get a fresh counter; their
2270         tag will be prefixed by a concatenation of their parents' tags)
2271       - bug fixes (related to calculation of transitive closure of types
2272         to be included in the output)
2273    
2274    2. Minor Basis updates:
2275    
2276       - added implementations for List.collate and Option.app
2277    
2278    ----------------------------------------------------------------------
2279    Name: Matthias Blume
2280    Date: 2002/02/11 15:55:00 EST
2281    Tag: blume-20020211-gensym
2282    Description:
2283    
2284    Added a "-gensym" option to command line of ml-nlffigen.  This can be
2285    used to specify a "stem" -- a string that is inserted in all "gensym'd"
2286    names (ML structure names that correspond to unnamed C structs, unions,
2287    and enums), so that separate runs of ml-nlffigen do not clash.
2288    
2289    ----------------------------------------------------------------------
2290    Name: Matthias Blume
2291    Date: 2002/02/11 12:05:00 EST
2292    Tag: blume-20020211-gensml
2293    Description:
2294    
2295    A quick fix for a problem with GenSML (in the pgraph-util library):
2296    Make generation of toplevel "local" optional.  (Strictly speaking,
2297    signature definitions within "local" are not legal SML.)
2298    
2299    Other than that: updates to INSTALL and cm/TODO.
2300    
2301    ----------------------------------------------------------------------
2302    Name: Matthias Blume
2303    Date: 2002/02/08 15:00:00 EST
2304    Tag: blume-20020208-uniquepid
2305    Description:
2306    
2307    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
2308    
2309    1. The installer (config/install.sh) has gotten smarter:
2310    
2311         - Configuration options are a bit easier to specify now
2312           (in config/targets).
2313         - Bug in recognizing .tar.bz2 files fixed.
2314         - Installer automatically resolves dependencies between
2315           configuration options (e.g., if you ask for eXene, you will
2316           also get cml -- regardless whether you asked for it or not).
2317         - Installer can run in "quieter mode" by setting the environment
2318           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
2319           "completely silent", though.
2320         - Build HashCons library as part of smlnj-lib.
2321    
2322    2. A new scheme for assigning persistent identifiers to compilation
2323       units (and, by extension, to types etc.) has been put into place.
2324       This fixes a long-standing bug where types and even dynamic values
2325       can get internally confused, thereby compromising type safety
2326       (abstraction) and dynamic correctness.  See
2327    
2328         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
2329    
2330       for an example of how things could go wrong until now.
2331    
2332       The downside of the new scheme is that pids are not quite as
2333       persistent as they used to be: CM will generate a fresh pid
2334       for every compilation unit that it thinks it sees for the first
2335       time.  That means that if you compile starting from a clean, fresh
2336       source tree at two different times, you end up with different
2337       binaries.
2338    
2339       Cutoff recompilation, however, has not been compromised because
2340       CM keeps pid information in special caches between runs.
2341    
2342    ----------------------------------------------------------------------
2343    Name: Lal George
2344    Date: 2002/02/07 15:34:13 EST 2002
2345    Tag: <none>
2346    Description:
2347    
2348    Compilers that generate assembly code may produce  global labels
2349    whose value is resolved at link time. The various peephole optimization
2350    modules did not take this in account.
2351    
2352    TODO. The Labels.addrOf function should really return an option
2353    type so that clients are forced to deal with this issue, rather
2354    than an exception being raised.
2355    
2356    ----------------------------------------------------------------------
2357    Name: Lal George
2358    Date: 2002/02/06 13:55:02 EST
2359    Tag: george-20020206-ra-breakup
2360    Description:
2361    
2362    1. A bug fix from Allen.
2363    
2364        A typo causes extra fstp %st(0)'s to be generated at compensation
2365        edges, which might cause stack underflow traps at runtime.  This
2366        occurs in fft where there are extraneous fstps right before the 'into'
2367        trap instruction (in this case they are harmless since none of the
2368        integers overflow.)
2369    
2370    2. Pulled out various utility modules that were embedded in the modules
2371       of the register allocator. I need these modules for other purposes, but
2372       they are not complete enough to put into a library (just yet).
2373    ----------------------------------------------------------------------
2374    Name: Matthias Blume
2375    Date: 2002/01/31 16:05:00 EST
2376    Tag: blume-20020131-sparc-ccalls
2377    Description:
2378    
2379    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
2380       of extra stack space by mistake.  Fixed.
2381    
2382    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
2383    
2384    ----------------------------------------------------------------------
2385    Name: Allen Leung
2386    Date: 2002/01/30
2387    Tag: leunga-20020130-nowhere-bug-fix
2388    Description:
2389    
2390       MLRISC bug fixes:
2391       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
2392       2. MachineInt.fromString was negating its value.
2393    
2394    ----------------------------------------------------------------------
2395    Name: Matthias Blume
2396    Date: 2002/01/29
2397    Tag: blume-20020129-INSTALL
2398    Description:
2399    
2400    - Added somewhat detailed installation instructions (file INSTALL).
2401    - Fixed curl-detection bug in config/install.sh.
2402    - It is now possible to select the URL getter using the URLGETTER
2403      environment variable:
2404    
2405          not set / "unknown"      --> automatic detection (script tries wget,
2406                                       curl, and lynx)
2407          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
2408                                       how to properly invoke them)
2409          other                    --> use $URLGETTER directly, it must take
2410                                       precisely two command-line arguments
2411                                       (source URL and destination file name)
2412    
2413    ----------------------------------------------------------------------
2414    Name: Matthias Blume
2415    Date: 2002/01/28
2416    Tag: blume-20020128-sparc-ccalls
2417    Description:
2418    
2419    - Fixed problem with calculation of "used" registers in sparc-c-calls.
2420    - Make use of the allocParam argument in sparc-c-calls.
2421    
2422    ----------------------------------------------------------------------
2423    Name: Matthias Blume
2424    Date: 2002/01/28
2425    Tag: blume-20020128-allocParam
2426    Description:
2427    
2428    John Reppy:  Changes c-calls API to accept client-callback for
2429    allocating extra stack space.
2430    me: Corresponding changes to mlriscGen (using a dummy argument that
2431        does not change the current behavior).
2432    
2433    ----------------------------------------------------------------------
2434  Name: Matthias Blume  Name: Matthias Blume
2435  Date: 2002/01/28 12:00:00  Date: 2002/01/28 12:00:00
2436  Tag: Release_110_38  Tag: Release_110_38
# Line 573  Line 2991 
2991    
2992  ----------------------------------------------------------------------  ----------------------------------------------------------------------
2993  Name: Matthias Blume  Name: Matthias Blume
 >>>>>>> 1.169  
2994  Date: 2001/09/18 15:35:00 EDT  Date: 2001/09/18 15:35:00 EDT
2995  Tag: blume-20010918-readme11036  Tag: blume-20010918-readme11036
2996  Description:  Description:

Legend:
Removed from v.1040  
changed lines
  Added in v.1487

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