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 836, Fri May 25 19:28:51 2001 UTC revision 1486, Tue May 11 17:04:01 2004 UTC
# Line 8  Line 8 
8  The form of an entry should be:  The form of an entry should be:
9    
10  Name:  Name:
11  Date:  Date: yyyy/mm/dd
12  Tag: <post-commit CVS tag>  Tag: <post-commit CVS tag>
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume (blume (at) tti - c (dot) org)
17    Date: 2004/05/11 12:05:00 CDT
18    Tag: blume-20040511-exports
19    Description:
20    
21    As per request by Adam Chlipala <adam@hcoop.net>, extended various
22    export lists in compiler-related .cm-files.
23    
24    ----------------------------------------------------------------------
25    Name: Matthias Blume (blume (at) tti - c (dot) org)
26    Date: 2004/05/11 11:35:00 CDT
27    Tag: blume-20040511-allsource
28    Description:
29    
30    The installer now honors the "src-smlnj" target again, although its meaning
31    has changed from "all sources required for the compiler" to "all sources
32    the installer knows about".  In other words, if you enable "src-smlnj"
33    in the "targets" file, then the installer will pull in sources for
34    everything.  (Notice that this refers to source code only.  Compiled
35    code is still only installed for modules that were requested explicitly
36    or which are required for other modules that were requested explicitly.)
37    
38    ----------------------------------------------------------------------
39    Name: Matthias Blume (blume (at) tti - c (dot) org)
40    Date: 2004/04/23 17:40:00 CDT
41    Tag: blume-20040423-ieee-scan
42    Description:
43    
44    Fixed IEEEReal.scan (and .fromString) so that if there is an overflow
45    in the exponent calculation we get INF or ZERO (depending on the mantissa
46    and the sign of the exponent).
47    
48    ----------------------------------------------------------------------
49    Name: Matthias Blume (blume (at) tti - c (dot) org)
50    Date: 2004/04/23 10:40:00 CDT
51    Tag: blume-20040423-ml-build
52    Description:
53    
54    The ml-build script now terminates with a non-0 status when something
55    goes wrong.
56    
57    ----------------------------------------------------------------------
58    Name: Matthias Blume (blume (at) tti - c (dot) org)
59    Date: 2004/04/22 16:35:00 CDT
60    Tag: blume-20040422-Option
61    Description:
62    
63    Made exception Option to be the same as exception Option.Option
64    (as it should be).
65    
66    ----------------------------------------------------------------------
67    Name: Allen Leung (leunga (at) reservoir (dot) com)
68    Date: 2004/03/19 14:40:00 EST
69    Tag: leunga-20040319-cygwin-nlffi
70    Description:
71    
72    Fixed the runtime so that ml-nlffi-lib runs on the cygwin version
73    of SML/NJ.  The problem is that
74    
75       lib = dlopen(NULL, ...)
76       f   = dlsym(lib, "malloc");
77    
78    does not work on Windows unless we explicitly export symbols
79    such as 'malloc' during linking.  We fixed this by explicitly
80    exporting the required symbols with the magic gcc incantation:
81    
82        -Wl,--export-all cygwin.def
83    
84    where cygwin.def is a file containing all the symbols that
85    we wish to export.
86    
87    I suspect this is a Windows problem and we'll have to
88    do the same (somehow with windows compilers) when
89    we build the native win32 version with the system
90    calls LoadLibrary/GetProcAddress.
91    
92    ----------------------------------------------------------------------
93    Name: Matthias Blume (blume (at) tti - c (dot) org)
94    Date: 2004/03/04 16:35:00 CST
95    Tag: blume-20040304-intinf-fmt
96    Description:
97    
98    Fixed problem with IntInf.fmt (sign would show up on the right instead
99    of on the left for BIN, OCT, and HEX).
100    
101    ----------------------------------------------------------------------
102    Name: Matthias Blume (blume (at) tti - c (dot) org)
103    Date: 2004/03/04 11:25:00 CST
104    Tag: blume-20040304-symlinks
105    Description:
106    
107    Fixed problem with installer script (unix only) where bin/ml-yacc and
108    friends pointed (via symlinks) to absolute locations instead of just
109    .run-sml.  This was reported by Vesa A Norrman.
110    
111    ----------------------------------------------------------------------
112    Name: Matthias Blume (blume (at) tti - c (dot) org)
113    Date: 2004/02/13 14:50:00 CST
114    Tag: Release_110_45
115    Description:
116    
117    New working version (110.45). New bootfiles.
118    
119    ----------------------------------------------------------------------
120    Name: Matthias Blume (blume (at) tti - c (dot) org)
121    Date: 2004/01/26 15:15:15 CST
122    Tag: blume-20040126-toplevel
123    Description:
124    
125    Improved handling of exceptions at the interactive toplevel.
126    
127    ----------------------------------------------------------------------
128    Name: Matthias Blume (blume (at) tti - c (dot) org)
129    Date: 2004/01/26 11:25:00
130    Tag: blume-20040126-app
131    Description:
132    
133    Type of top-level "app" corrected.
134    Added code for setting vp_limitPtrMask to Win32-specific runtime.
135    
136    ----------------------------------------------------------------------
137    Name: Matthias Blume (blume (at) tti - c (dot) org)
138    Date: 2003/11/18 17:10 CST
139    Tag: blume-20031118-basis-fiddle
140    Description:
141    
142    - changed Timer interface to what might become the spec
143    - POSIX_FLAGS -> BIT_FLAGS according to spec
144    - some other minor discrepancies wrt. spec eliminated
145    
146    ----------------------------------------------------------------------
147    Name: Matthias Blume (blume (at) tti - c (dot) org)
148    Date: 2003/11/06 12:00:00 CST
149    Tag: Release_110_44
150    Description:
151    
152    New working version (110.44). New bootfiles.
153    
154    ----------------------------------------------------------------------
155    Name: Matthias Blume (blume (at) tti - c (dot) org)
156    Date: 2003/11/04 11:50:00 CST
157    Tag: blume-20031104-move-libraries
158    Description:
159    
160    Eliminated the "dont_move_libraries" directive in config/targets.
161    (The mechanism was broken and could not be fixed easily.  Moreover,
162    there does not seem to be any reason not to move all libraries into
163    lib during installation.  I originally implemented this directive as a
164    backward-compatibility feature when I first introduced the new CM.
165    Now that things have been stable for a long time and going back to the
166    old CM is not an option, there is no reason to keep it around.)
167    
168    ----------------------------------------------------------------------
169    Name: Matthias Blume (blume (at) tti - c (dot) org)
170    Date: 2003/11/03 16:00:00 CST
171    Tag: blume-20031103-installdir
172    Description:
173    
174    Made installer honor INSTALLDIR variable again.  (Thanks to Chris
175    Richards for pointing out the problem and providing the solution.)
176    
177    ----------------------------------------------------------------------
178    Name: Matthias Blume (blume (at) tti - c (dot) org)
179    Date: 2003/10/01 17:05:00 CDT
180    Tag: blume-20031001-lal-mlrisc
181    Description:
182    
183    MLRISC bug fix from Lal.
184    
185    ----------------------------------------------------------------------
186    Name: Matthias Blume (blume (at) tti - c (dot) org)
187    Date: 2003/09/30 16:10:00 CDT
188    Tag: blume-20030930-primio-bat
189    Description:
190    
191    1. Added openVector, nullRd, and nullWr to PRIM_IO.
192    2. Improved .bat files (for Win32 port) to make things work under Win95.
193       (thanks to Aaron S. Hawley for this one)
194    
195    ----------------------------------------------------------------------
196    Name: Matthias Blume (blume (at) tti - c (dot) org)
197    Date: 2003/09/26 16:05:00 CDT
198    Tag: blume-20030926-wrappriv
199    Description:
200    
201    Added missing wrapper for privilege "primitive" in $smlnj/viscomp/core.cm.
202    
203    ----------------------------------------------------------------------
204    Name: Matthias Blume (blume (at) tti - c (dot) org)
205    Date: 2003/09/26 15:00:00 CDT
206    Tag: blume-20030926-110_43_3
207    Description:
208    
209    - additional cleanup
210    - version number bump, NEW BOOTFILES
211    
212    ----------------------------------------------------------------------
213    Name: Matthias Blume (blume (at) tti - c (dot) org)
214    Date: 2003/09/26 12:00:00 CDT
215    Tag: blume-20030926-ppautoload
216    Description:
217    
218    I modified the read-eval-print loop so that the autoloader gets
219    invoked whenever the prettyprinter tries to look up a symbol that
220    is not currently defined in the toplevel environment but which
221    appears in CM's autoload registry.  As a result, we see far fewer of
222    those ?.Foo.Bar.xxx names in the prettyprinter's output.
223    
224    In addition to this I tried to clean up some pieces of the Basis
225    implementation (e.g., Socket, Word8Array) in order to prevent other
226    instances of these ?.Foo.Bar.xxx names from being printed.
227    
228    The mechanism that picks names for types still needs some work, though.
229    (Right now it seems that if there is a type A.t which is defined to
230    be B.u, but B is unavailable at toplevel, then A.t gets printed as
231    "?.B.u" although the perhaps more sensible solution would be to use
232    "A.t" in this case.  In other words, the prettyprinter should follow
233    a chain of DEFtycs not farther than there are corresponding toplevel
234    names in the current environment.)
235    
236    ----------------------------------------------------------------------
237    Name: Matthias Blume (blume (at) tti - c (dot) org)
238    Date: 2003/09/24 16:31:00 CDT
239    Tag: blume-20030924-installer
240    Description:
241    
242    Another installer tweak:  All the ML code for the installer is now
243    compiled during CMB.make and put into a little library called
244    $smlnj/installer.cm.  The installation then simply invokes
245    
246       sml -m $smlnj/installer.cm
247    
248    and everything happens automagically.
249    
250    Win32:  ML code senses value of environment variable SMLNJ_HOME.
251    Unix:   ML code senses values of environment variables ROOT, CONFIGDIR,
252            and BINDIR.
253    
254    The new scheme guarantees that the ML code responsible for the installation
255    is in sync with the APIs of the main system.  Also, the installer is
256    somewhat faster because the installer script is precompiled.
257    
258    ----------------------------------------------------------------------
259    Name: Matthias Blume (blume (at) tti - c (dot) org)
260    Date: 2003/09/24 15:35:00 CDT
261    Tag: blume-20030924-synsock
262    Description:
263    
264    Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
265    but excludes all non-blocking operations.  Defined SOCKET (in Basis)
266    and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
267    implementations of non-blocking operations from CML's Socket
268    structure.
269    
270    ----------------------------------------------------------------------
271    Name: Matthias Blume (blume (at) tti - c (dot) org)
272    Date: 2003/09/24 15:10:05 CDT
273    Tag: blume-20030924-sockets
274    Description:
275    
276    1. Fixed SOCKET API and implementation to match Basis spec.
277       This required changing the internal representation of sockets to one
278       that remembers (for each socket file descriptor) whether it is currently
279       blocking or non-blocking.  This state is maintained lazily (i.e., a system
280       call is made only if the state actually needs to change).
281    
282    2. OS-specific details of sockets were moved into separate files, thus
283       making it possible to unify the bulk of the socket implementations
284       between Unix and Win32.
285    
286    3. CML's socket API changed accordingly.
287       (Note that we need to remove non-blocking functions from this API
288       since they are redundant in the case of CML!)
289    
290    4. CML's socket implementation now makes use of non-blocking functions
291       provided by Basis, thus removing all OS-dependent code from this part
292       of CML.
293    
294    5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
295    
296    ----------------------------------------------------------------------
297    Name: Matthias Blume (blume (at) tti - c (dot) org)
298    Date: 2003/09/22 12:10:00 CDT
299    Tag: blume-20030922-110_43_2
300    Description:
301    
302    Made a new interim version and bootfiles for developer's bootstrapping
303    convenience.
304    
305       110.43.2 -- NEW BOOTFILES
306    
307    ----------------------------------------------------------------------
308    Name: Matthias Blume (blume (at) tti - c (dot) org)
309    Date: 2003/09/19 15:55:00 CDT
310    Tag: blume-20030919-cmdir
311    Description:
312    
313    1. new-install.sh -> install.sh
314    2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
315    3. tweaked installer so that another name instead of .cm can be chosen
316       at install time (by setting the CM_DIR_ARC environment variable
317       during installation); once installation is complete, the name is
318       fixed
319    
320    ----------------------------------------------------------------------
321    Name: Matthias Blume (blume (at) tti - c (dot) org)
322    Date: 2003/09/18 16:00:00 CDT
323    Tag: blume-20030918-110_43_1
324    Description:
325    
326    Made a new interim version and bootfiles for developer's bootstrapping
327    convenience.
328    
329       110.43.1 -- NEW BOOTFILES
330    
331    ----------------------------------------------------------------------
332    Name: Matthias Blume (blume (at) tti - c (dot) org)
333    Date: 2003/09/18 15:20:00 CDT
334    Tag: blume-20030918-misc
335    Description:
336    
337    1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
338       this seems to be controversial at the moment) and used those in
339       Posix.ProcEnv.times.
340    
341    2. Added Time.{from,to}Nanoseconds to Time.
342    
343    3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
344       For example, fromLargeInt never needs to look at more than 3 "big
345       digits" to get its 53 bits of precision.
346    
347    ----------------------------------------------------------------------
348    Name: Matthias Blume (blume (at) tti - c (dot) org)
349    Date: 2003/09/17 16:30:00 CDT
350    Tag: blume-20030917-real32-slices
351    Description:
352    
353    Added an entry to the primitive environment
354    (compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
355    added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
356    
357    Removed some of the "magic" constants in real64.sml and replaced them
358    with code that generates these values from their corresponding
359    integer counterparts.
360    
361    Made all(?) the slice-related changes to the Basis and made everything
362    compile again...
363    
364    ----------------------------------------------------------------------
365    Name: Matthias Blume (blume (at) tti - c (dot) org)
366    Date: 2003/09/15 17:45:00 CDT
367    Tag: blume-20030915-rbase
368    Description:
369    
370    Fixed bug in Real.fromLargeInt.
371    
372    ----------------------------------------------------------------------
373    Name: Matthias Blume (blume (at) tti - c (dot) org)
374    Date: 2003/09/13 18:11:00 CDT
375    Tag: blume-20030913-libinstall
376    Description:
377    
378    Minor bugfix in config/libinstall (set anchor with path to
379    standalone tool after installing it, otherwise libraries that
380    need ml-lex or ml-yacc won't compile the first time the installer
381    runs).
382    
383    ----------------------------------------------------------------------
384    Name: Matthias Blume (blume (at) tti - c (dot) org)
385    Date: 2003/09/12 11:45:00 CDT
386    Tag: blume-20030912-various
387    Description:
388    
389     - fixed bug in Real.toLargeInt
390     - fixed bug in Posix.ProcEnv.times
391     - changed inputLine functions to return an option
392     - minor installer improvements / bugfixes
393     - changed default @SMLalloc parameter for x86/celeron to 64k
394    
395    ----------------------------------------------------------------------
396    Name: Matthias Blume (blume (at) tti - c (dot) org)
397    Date: 2003/09/09 22:00:00 CDT
398    Tag: Release_110_43
399    Description:
400    
401        New working release 110.43.  New bootfiles.
402    
403    ----------------------------------------------------------------------
404    Name: Matthias Blume (blume (at) tti - c (dot) org)
405    Date: 2003/09/09 19:20:00 CDT
406    Tag: blume-20030909-installer
407    Description:
408    
409    Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
410    Modified config/install.bat to take advantage of it.  Also modified
411    config/install.sh (and called it config/new-install.sh) to take advantage
412    of it on Unix systems.  (The SML code is (supposed to be) platform-
413    independent.)
414    
415    The installer can now install everything under Win32
416    as well as under *nix as long as it compiles.
417    
418    Other changes:
419    
420      - made CML compile again under Win32
421      - made eXene compile under Win32 (by providing a fake structure UnixSock
422        and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
423      - fixed a bug in nowhere: it assumed that type OS.Process.status is the
424        same as type int;  under Win32 it isn't
425      - fixed some slice-related problems in the win32-specific parts of CML
426      - added a functor argument "sameVol" to os-path-fn.sml in the Basis
427        (under Win32, the volume name is case-insensitive, and the
428        OS.Path code compares volume names for equality)
429    
430    ----------------------------------------------------------------------
431    Name: Matthias Blume (blume (at) tti - c (dot) org)
432    Date: 2003/09/08 11:55:00 CDT
433    Tag: blume-20030908-fullpath
434    Description:
435    
436    Made Win32 version of OS.FileSys.fullPath return current directory
437    when given an empty string.  This is what the spec says, and incidentally,
438    CM depends on it.  (CM otherwise goes into an infinite loop in certain
439    cases when presented with the name of a non-existing .cm file.)
440    
441    ----------------------------------------------------------------------
442    Name: Matthias Blume (blume (at) tti - c (dot) org)
443    Date: 2003/09/04 16:30:00 CDT
444    Tag: blume-20030905-slices-etc
445    Description:
446    
447      1. Changed interface to vectors and arrays in Basis to match
448         (draft) Basis spec.
449      2. Added signatures and implementations of slices according to
450         Basis spec.
451      3. Edited source code throughout the system to make it compile again
452         under 1. and 2.  (In some cases code had to be added to have it
453         match the new signatures.)
454      4. MLRISC should be backward-compatible: the copies of the originals
455         of files that needed to change under 3. were retained, the .cm files
456         check the compiler version number and use old versions when
457         appropriate.
458      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
459         dirstream -> string option (in accordance with Basis spec).
460      6. When generating code that counts lines, ml-lex used function
461         CharVector.foldli, taking advantage of its old interface.
462         This has been replaced with the corresponding code from
463         CharVectorSlice.  (html-lex must be re-lexed!)
464      7. BitArray in smlnj-lib/Util has been extended/modified to match the
465         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
466      8. Removed temporary additions (fromInternal, toInternal) from the
467         (now obsolete) IntInf in smlnj-lib/Util.
468      9. Cleaned up structure Byte.
469     10. Added localOffset, scan, and fromString to Date (according to spec).
470         Cleaned/corrected implementation of Date.
471         (Still need to check for correctness; implement better canonicalizeDate.)
472     11. Added "scan" to signature IEEE_REAL.
473     12. Some improvements to IntInf [in particular: efficiency-hack for
474         mod and rem when second operand is 2 (for parity checks).]
475     13. Changed representation of type Time.time, using a single IntInf.int
476         value counting microseconds.  This considerably simplified the
477         implementation of structure Time.  We now support negative time
478         values; scan and fromString handle signs.
479     14. Functor PrimIO now takes two additional arguments (VectorSlice and
480         ArraySlice).
481    
482    ----------------------------------------------------------------------
483    Name: Matthias Blume (blume (at) tti - c (dot) org)
484    Date: 2003/08/28 17:00:00 CDT
485    Tag: blume-20030828-intinf
486    Description:
487    
488    This is a major update which comes with a version number bump
489    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
490    and an implementation of IntInf in the Basis.
491    
492    There are a fairly large number of related changes and updates throughout
493    the system:
494    
495    Basis:
496            - Implemented IntInf.
497            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
498            - Added some missing Real64 operations, most notably Real.toLargeInt.
499            - Added FixedInt as a synonym for Int32.
500    
501    compiler:
502            * Added support for a built-in intinf type.
503              - literals
504              - pattern matching
505              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
506              - overloading on literals and operations
507    
508            This required adding a primitive type intinf, some additional
509            primops, and implementations for several non-trivial intinf
510            operations in Core.  (The intinf type is completely abstract
511            to the compiler; all operations get delegated back to the Core.)
512    
513            * Intinf equality is handled by polyequal.  However, the compiler
514            does not print its usual warning in this case (since polyequal
515            is the right thing to do there).
516    
517            * Improved the organization of structure InlineT.
518    
519            * A word about conversion primops:
520            If conversions involving intinf do not cancel out during
521            CPS contract, then the compiler must insert calls to Core functions.
522            Since all core access must be resolved already during the FLINT
523            translate phase, it would be too late a the time of CPS contract
524            to add new Core calls.  For this reason, conversion primops
525            for intinf carry two arguments:  1. the numeric argument that
526            they are supposed to convert, and 2. the Core function that
527            can help with this conversion if necessary.  If CPS contract
528            eliminates a primop, then the associated Core function becomes
529            dead and goes away.  Intinf conversion primops that do not get
530            eliminated by CPS contract get rewritten into calls of their
531            core functions by a separate, new phase.
532    
533    interactive system:
534            - Control.Print.intinfDepth controls max length of intinf constants
535            being printed. (Analogous to Control.Print.stringDepth.)
536            - Cleanup in printutil and pputil: got rid of unused stuff and
537            duplicates; replaced some of the code with code that makes better
538            use of library functionality.
539    
540    CM:
541            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
542            are no longer silent.
543    
544    CKIT:
545            Fixed mismatched uses of Int32 and LargeInt.  I always decided
546            in favor of LargeInt -- which is now the same as IntInf.
547            CKIT-knowledgable people should check whether this is what's
548            intended and otherwise change things back to using Int32 or
549            FixedInt.
550    
551    Throughout the code:
552            Started using IntInf.int literals and built-in operations
553            (e.g., comparison with 0) where this seems appropriate.
554    
555    
556    ----------------------------------------------------------------------
557    Name: Dave MacQueen (dbm@cs.uchicago.edu)
558    Date: 2003/08/13 11:36:00 CDT
559    Tag: dbm-20030813-mcz-merge1
560    Description:
561    
562    Merging changes from the mcz-branch development branch into trunk.
563    These changes involve replacement of the emulated old prettyprinter
564    interface with direct use of the SML/NJ Lib PP library, and fixing
565    of a couple of bugs (895, 1186) relating to error messages.  A new
566    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
567    has been added.
568    
569    ----------------------------------------------------------------------
570    Name: Matthias Blume (blume (at) tti - c (dot) org)
571    Date: 2003/08/11 15:45:00 CDT
572    Tag: blume-20030811-windows
573    Description:
574    
575    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
576    
577        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
578    
579    This patch restores SML/NJ's ability to run under win32.  There are a
580    number of changes, including fixes for several bugs that had gone
581    unnoticed until now:
582    
583      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
584        trying to run the win32 version from within cygwin.)
585    
586      - There are a number of simple .bat scripts that substitute for their
587        corresponding Unix shell-scripts.  (See below.)
588    
589      - The internals of ml-build have been modified slightly.  The main
590        difference is that instead of calling ".link-sml" (or link-sml.bat)
591        using OS.Process.system, the ML process delegates this task back
592        to the script.  Otherwise problems arise in mixed environments such
593        as Cygwin where scripts look and work like Unix scripts, but
594        where OS.Process.system cannot run them.
595    
596      - In CM, the srcpath pickler used native pathname syntax -- which
597        is incorrect in the case of cross-compilation.  The new pickle format
598        is independent of platform-specific naming conventions.
599    
600      - Path configuration files (such as lib/pathconfig) can now choose
601        between native and standard syntax.  Placing a line of the form
602    
603            standard!
604    
605        into the file causes all subsequent paths to be interpreted using
606        CM standard pathname syntax (= Unix conventions); a line
607    
608            native!
609    
610        switches back to native style.  This was needed so that
611        path config files can be written portably, see src/system/pathconfig.
612    
613      - Runtime system:
614    
615          - win32-filesys.c:  get_file_time and set_file_time now
616            access modification time, not creation time.
617    
618          - I/O code made aware of new array representation.
619    
620          - Bug fixes in X86.prim.masm.
621    
622       - src/system/makeml made aware of win32. (For use under cygwin
623            and other Unix-environments for windows.)
624    
625       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
626         which caused BinIO.inputAll to fail consistently.
627    
628    .bat scripts:
629    
630       Windows .bat scripts assume that SMLNJ_HOME is defined.
631    
632       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
633         applications (sml, ml-yacc, ml-lex).
634       - ml-build.bat: analogous to ml-build.
635       - config\install.bat:  Analogous to config/install.sh.  This requires
636         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
637         (nmake etc. must be on the path, and vcvars32 must have been run.)
638         Moreover, sources for ml-lex and ml-yacc need to exist under src,
639         and the bootfile hierarchy must have been unpacked under
640         sml.boot.x86-win32.
641         The script is very primitive and does a poor job at error checking.
642         It only installs the base system, ml-lex, and ml-yacc.  No other
643         libraries are being installed (i.e., you get only those that
644         are part of the compiler.)
645       - link-sml.bat:  analogous to .link-sml, but not currently used
646    
647    Unrelated bug fixes:
648    
649      - ml-nlffigen now exports structures ST_* corresponding to incomplete
650        types.
651      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
652        otherwise.)
653    
654    ----------------------------------------------------------------------
655    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
656    Date: 2003/06/17
657    Tag: macqueen-20030617-bug895
658    Description:
659    
660    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
661    Tag will be used for new development branch (mcz-branch) for
662    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
663    
664    ----------------------------------------------------------------------
665    Name: Matthias Blume (blume (at) tti - c (dot) org)
666    Date: 2003/05/27 16:55:00 CDT
667    Tag: blume-20030527-polyeq
668    Description:
669    
670    Tried to eliminated most cases of polymorphic equality.
671    
672    ----------------------------------------------------------------------
673    Name: Matthias Blume (blume (at) tti - c (dot) org)
674    Date: 2003/05/21 17:45:00 CDT
675    Tag: blume-20030517-complete
676    Description:
677    
678    Two changes:
679    
680      1. Added a flag for controlling whether non-exhaustive bindings will
681         be treated as errors (default is false).
682      2. Cleaned up the *entire* source tree so that CMB.make goes through
683         without a single non-exhaustive match- or bind warning.
684    
685    ----------------------------------------------------------------------
686    Name: Matthias Blume (blume (at) tti - c (dot) org)
687    Date: 2003/05/17 10:20:00 CDT
688    Tag: blume-20030517-absyn
689    Description:
690    
691    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
692    
693       This mainly affects the quality of error messages.  However, some
694       of the code is now more straightforward than before. (Treatment of
695       the above four constructs in translate.sml is much simpler than
696       the "macro-expansion" that was going on before.  Plus, the mach-
697       compiler no longer gets invoked just to be able to compile an
698       if-expression.)
699    
700    2. The ErrorMsg.Error exception is now caught and absorbed by the
701       interactive loop.
702    
703    ----------------------------------------------------------------------
704    Name: Allen Leung
705    Date: 2003/05/16 13:05:00 CDT
706    Tag: leunga-20030516-cygwin-runtime
707    Description:
708    
709        Ported the runtime system to cygwin, which uses the unix
710    x86-unix bin files.  Missing/buggy features:
711    
712       o getnetbyname, getnetbyaddr: these functions seem to be missing in
713         the Cygwin library.
714       o Ctrl-C handling may be flaky.
715       o Windows system calls and Windows I/O are not supported.
716    
717       A new set of binfiles is located at:
718    
719         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
720    
721       This is only needed for bootstrapping the cygwin version of smlnj.
722       Other x86 versions can use the existing binfiles.
723    
724    ----------------------------------------------------------------------
725    Name: Matthias Blume
726    Date: 2003/04/08 15:42:00 CDT
727    Tag: blume-20030408-listpair
728    Description:
729    
730    1. Added a target 'mlrisc' to installer.
731    
732    2. Added missing elements to structure ListPair.
733    
734    ----------------------------------------------------------------------
735    Name: Allen Leung
736    Date: 2003/01/07 10:40:00 EST
737    Tag: leunga-20030107-int-rem
738    Description:
739    
740    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
741    The arguments to the SUBL instruction were swapped.
742    
743    ----------------------------------------------------------------------
744    Name: Matthias Blume
745    Date: 2002/12/12 16:25:00 EST
746    Tag: blume-20021212-risc-ra
747    Description:
748    
749    Fixed a serious bug in the rewrite code for FP spilling/reloading that
750    sent the RA into an infinite loop when floating point registers get
751    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
752    110.37 and 110.38.)
753    There was another set of potential problems related to the handling of
754    MLRISC annotations (but those did not yet cause real problems, apparently).
755    
756    ----------------------------------------------------------------------
757    Name: Matthias Blume
758    Date: 2002/12/06 22:40:00 EST
759    Tag: blume-20021206-cm-fileid
760    Description:
761    
762    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
763    This fixes the problem of CM getting confused by files that suddenly
764    change their identity (e.g., by getting unlinked and recreated by some
765    text editor such as vi).  There might be a better/cheaper/cleaner way
766    of doing this, but for now this will have to do.
767    
768    ----------------------------------------------------------------------
769    Name: Matthias Blume
770    Date: 2002/10/28 09:50:00 EST
771    Tag: blume-20021028-typecheck
772    Description:
773    
774    Exported structure Typecheck from $smlnj/viscomp/core.cm.
775    
776    ----------------------------------------------------------------------
777    Name: Matthias Blume
778    Date: 2002/10/17 09:10:00 EDT
779    Tag: Release_110_42
780    Description:
781    
782    In good old tradition, there has been a slight hiccup so that we have
783    to patch 110.42 after the fact.   The old release tag has been replaced
784    (see below).
785    
786    The change solves a problem with two competing approaches the
787    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
788    each other's way.
789    
790    This change only affects the runtime system code and the installer script.
791    (No new bootfiles.)
792    
793    ----------------------------------------------------------------------
794    Name: Matthias Blume
795    Date: 2002/10/16 12:00:00 EDT
796    Tag: Release_110_42_removed
797    Description:
798    
799    New working release.  New bootfiles.
800    
801    ----------------------------------------------------------------------
802    Name: Matthias Blume
803    Date: 2002/10/10 13:10:00 EDT
804    Tag: blume-20021010-ppc-divs
805    Description:
806    
807    The mltree operator DIVS must be implemented with an overflow check on
808    the PPC because the hardware indicates divide-by-zero using "overflow" as
809    well.
810    
811    ----------------------------------------------------------------------
812    Name: Matthias Blume
813    Date: 2002/07/23 11:45:00
814    Tag: blume-20020723-smlnj-home
815    Description:
816    
817    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
818    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
819    CM_PATHCONFIG is also set), the path configuration file is assumed
820    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
821    the entire tree to some other place and everything will "just work".
822    
823    (Companion commands such as ml-build and ml-makedepend also sense this
824    variable.)
825    
826    ----------------------------------------------------------------------
827    Name: Matthias Blume
828    Date: 2002/07/12 21:19:00 EDT
829    Tag: blume-20020712-liveness
830    Description:
831    
832    Exported two useful "step" functions from liveness module (MLRISC).
833    
834    ----------------------------------------------------------------------
835    Name: Matthias Blume
836    Date: 2002/07/05 16:00 EDT
837    Tag: Release_110_41
838    Description:
839    
840    New working release.  New bootfiles.
841    
842    ----------------------------------------------------------------------
843    Name: Matthias Blume
844    Date: 2002/07/05 10:25:00 EDT
845    Tag: blume-20020705-btimp
846    Description:
847    
848    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
849    other clients can set up backtracing support.
850    
851    ----------------------------------------------------------------------
852    Name: Matthias Blume
853    Date: 2002/06/25 17:23:00 EDT
854    Tag: blume-20020625-fpmax
855    Description:
856    
857    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
858    case.  (The sense of the isNaN test was reversed -- which made min and
859    max always return their first argument.)
860    
861    ----------------------------------------------------------------------
862    Name: Matthias Blume
863    Date: 2002/06/11
864    Tag: blume-20020611-unixpath
865    Description:
866    
867    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
868    
869    ----------------------------------------------------------------------
870    Name: Matthias Blume
871    Date: 2002/06/10 16:35:00 EDT
872    Tag: blume-20020610-ieeereal
873    Description:
874    
875    I back-ported my implementation of IEEEReal.fromString from the
876    idlbasis-devel branch so that we can test it.
877    
878    Another small change is that ppDec tries to give more information
879    than just "<sig>" in the case of functors.  However, this code is
880    broken in some mysterious way if the functor's body's signature
881    has not been declared by ascription but gets inferred from the
882    implementation.  This needs fixing...
883    
884    ----------------------------------------------------------------------
885    Name: Matthias Blume
886    Date: 2002/05/31
887    Tag: blume-20020531-btrace-mode
888    Description:
889    
890    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
891    the wayside when I switched over to using Controls everywhere.)
892    
893    ----------------------------------------------------------------------
894    Name: Lal George
895    Date: 2002/05/23 12:21:40 EDT
896    Tag: george-20020523-visual-labels
897    Description:
898    
899    Labels are now displayed in the graphical output to make
900    the fall-through and target blocks obvious.
901    
902    ----------------------------------------------------------------------
903    Name: Matthias Blume
904    Date: 2002/05/22 11:03:00 EDT
905    Tag: blume-20020522-shrink
906    Description:
907    
908    John tweaked yesterday's fix for 1131 to handle an out-of-memory
909    situation that comes up when allocating huge arrays.
910    
911    ----------------------------------------------------------------------
912    Name: Matthias Blume
913    Date: 2002/05/21 16:00:00 EDT
914    Tag: Release_110_40
915    Description:
916    
917    New working release (110.40).  New bootfiles.
918    
919    [Also: John Reppy fixed GC bug 1131.]
920    
921    ----------------------------------------------------------------------
922    Name: Matthias Blume
923    Date: 2002/05/21 12:35:00 EDT
924    Tag: blume-20020521-cmdoc
925    Description:
926    
927    CM documentation update.
928    
929    ----------------------------------------------------------------------
930    Name: Matthias Blume
931    Date: 2002/05/21 10:55:00 EDT
932    Tag: blume-20020521-misc
933    Description:
934    
935    - John tweaked runtime to be silent on heap export (except when
936      GC messages are on).
937    - I added a few more things (cross-compiling versions of CMB) to
938      config/preloads (as suggestions).
939    
940    ----------------------------------------------------------------------
941    Name: Matthias Blume
942    Date: 2002/05/20 22:25:00 EDT
943    Tag: blume-20020520-controls
944    Description:
945    
946    - Added ControlUtil structure to control-lib.cm.
947    - Use it throughout.
948    - Used Controls facility to define MLRISC controls (as opposed to
949      registering MLRISC control ref cells with Controls after the fact)
950    - Fixed messed-up controls priorities.
951    
952    * Removed again all the stuff from config/preloads that one wouldn't
953      be able to preload at the time the initial heap image is built.
954      (Many libraries, e.g., CML, do not exist yet at this time.  The
955       only libraries that can be preloaded via config/preloads are those
956       that come bundled with the bootfiles.)
957    
958    ----------------------------------------------------------------------
959    Name: Matthias Blume
960    Date: 2002/05/20 10:59:00 EDT
961    Tag: blume-20020520-preloads
962    Description:
963    
964    Added a lot of commented-out suggestions for things to be included
965    in config/preloads.
966    
967    ----------------------------------------------------------------------
968    Name: Allen Leung
969    Date: 2002/05/18 14:20:00 EDT
970    Tag: leunga-20020518-mdl
971    Description:
972    
973    o Made the mdl tool stuff compile and run again.
974    o I've disabled all the stuff that depends on RTL specifications; they
975      are all badly broken anyway.
976    
977    ----------------------------------------------------------------------
978    Name: Matthias Blume
979    Date: 2002/05/17 16:49:00 EDT
980    Tag: blume-20020517-controls
981    Description:
982    
983    0. John Reppy made several modifications to the SML/NJ library.
984       In particular, there is a shiny new controls-lib.cm.
985    
986    1. Pushed new controls interface through compiler so that everything
987       compiles again.
988    
989    2. Added FormatComb and FORMAT_COMB to the CML version of the
990       SML/NJ library (so that CML compiles again).
991    
992    3. Modified init scripts because XXX_DEFAULT environment variables
993       are no longer with us.  (Boot-time initialization is now done
994       using the same environment variables that are also used for
995       startup-time initialization of controls.)
996    
997    ----------------------------------------------------------------------
998    Name: Lal George
999    Date: 2002/05/15 09:20:10 EDT
1000    Tag: george-20020515-pseudo-op-decls
1001    Description:
1002    
1003            All pseudo-ops emitted before the first segment declaration
1004    such as TEXT, DATA, and BSS directives are assumed to be global
1005    declarations and are emitted first in the assembly file. This is
1006    useful in a number of situations where one has pseudo-ops that are not
1007    specific to any segment, and also works around the constraint that one
1008    cannot have client pseudo-ops in the TEXT segment.
1009    
1010            Because no segment is associated with these declarations it is
1011    an error to allocate any space or objects before the first segment
1012    directive and an exception will be raised. However, we cannot make
1013    this check for client pseudo-ops.
1014    
1015            These top level declarations are a field in the CFG graph_info.
1016    In theory you can continue to add to this field after the CFG has been
1017    built -- provided you know what you are doing;-)
1018    
1019    ----------------------------------------------------------------------
1020    Name: Matthias Blume
1021    Date: 2002/05/13 16:40:00 EDT
1022    Tag: blume-20020513-pp-etc
1023    Description:
1024    
1025    A few minor bugfixes:
1026    
1027      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
1028        (Bogus printouts for redefined bindings still occur.  Compiler
1029        bug should no longer occur now.  We need to redo the prettyprinter
1030        from scratch.)
1031    
1032      - CM pathname printer now also adds escape sequences for ( and )
1033    
1034      - commend and docu fixes for ml-nlffi
1035    
1036    ----------------------------------------------------------------------
1037    Name: Matthias Blume
1038    Date: 2002/05/10 16:40:00 EDT
1039    Tag: blume-20020510-erg-textio
1040    Description:
1041    
1042    Applied the following bugfix provided by Emden Gansner:
1043    
1044        Output is corrupted when outputSubstr is used rather than output.
1045    
1046        The problem occurs when a substring
1047    
1048            ss = (s, dataStart, dataLen)
1049    
1050        where dataStart > 0, fills a stream buffer with avail bytes left.
1051        avail bytes of s, starting at index dataStart, are copied into the
1052        buffer, the buffer is flushed, and then the remaining dataLen-avail
1053        bytes of ss are copied into the beginning of the buffer. Instead of
1054        starting this copy at index dataStart+avail in s, the current code
1055        starts the copy at index avail.
1056    
1057        Fix:
1058        In text-io-fn.sml, change line 695 from
1059             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
1060        to
1061             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
1062    
1063    ----------------------------------------------------------------------
1064    Name: Matthias Blume
1065    Date: 2002/04/12 13:55:00 EDT
1066    Tag: blume-20020412-assyntax
1067    Description:
1068    
1069    1. Grabbed newer assyntax.h from the XFree86 project.
1070    2. Fiddled with how to compile X86.prim.asm without warnings.
1071    3. (Very) Minor cleanup in CM.
1072    
1073    ----------------------------------------------------------------------
1074    Name: Matthias Blume
1075    Date: 2002/04/01 (no joke!) 17:07:00 EST
1076    Tag: blume-20020401-x86div
1077    Description:
1078    
1079    Added full support for div/mod/rem/quot on the x86, using the machine
1080    instruction's two results (without clumsily recomputing the remainder)
1081    directly where appropriate.
1082    
1083    Some more extensive power-of-two support was added to the x86 instruction
1084    selector (avoiding expensive divs, mods, and muls where they can be
1085    replaced with cheaper shifts and masks).  However, this sort of thing
1086    ought to be done earlier, e.g., within the CPS optimizer so that
1087    all architectures benefit from it.
1088    
1089    The compiler compiles to a fixed point, but changes might be somewhat
1090    fragile nevertheless.  Please, report any strange things that you might
1091    see wrt. div/mod/quot/rem...
1092    
1093    ----------------------------------------------------------------------
1094    Name: Matthias Blume
1095    Date: 2002/03/29 17:22:00
1096    Tag: blume-20020329-div
1097    Description:
1098    
1099    Fixed my broken div/mod logic.  Unfortunately, this means that the
1100    inline code for div/mod now has one more comparison than before.
1101    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
1102    The problem was with quotient = 0, because that alone does not tell
1103    us which way the rounding went.  One then has to look at whether
1104    remainder and divisor have the same sign...  :(
1105    
1106    Anyway, I replaced the bootfiles with fresh ones...
1107    
1108    ----------------------------------------------------------------------
1109    Name: Matthias Blume
1110    Date: 2002/03/29 14:10:00 EST
1111    Tag: blume-20020329-inlprims
1112    Description:
1113    
1114    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
1115    
1116    Primops have changed. This means that the bin/boot-file formats have
1117    changed as well.
1118    
1119    To make sure that there is no confusion, I made a new version.
1120    
1121    
1122    CHANGES:
1123    
1124    * removed REMT from mltree (remainder should never overflow).
1125    
1126    * added primops to deal with divisions of all flavors to the frontend
1127    
1128    * handled these primops all the way through so they map to their respective
1129      MLRISC support
1130    
1131    * used these primops in the implementation of Int, Int32, Word, Word32
1132    
1133    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
1134    
1135    * parameterized INLMIN, INLMAX, and INLABS by a numkind
1136    
1137    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
1138      floating point
1139    
1140    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
1141      and Real (but Real.abs maps to a separate floating-point-only primop)
1142    
1143    
1144    TODO items:
1145    
1146    * Hacked Alpha32 instruction selection, disabling the selection of REMx
1147      instructions because the machine instruction encoder cannot handle
1148      them.  (Hppa, PPC, and Sparc instruction selection did not handle
1149      REM in the first place, and REM is supported by the x86 machine coder.)
1150    
1151    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
1152      selection phase.  (The two can be streamlined because the hardware
1153      delivers both quotient and remainder at the same time anyway.)
1154    
1155    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
1156      (Currently the behavior is inconsistent both across architectures and
1157      wrt. the draft Basis spec.)
1158    
1159    * Word8 should eventually be handled natively, too.
1160    
1161    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
1162      as if there currently is no execution path that could trigger it in
1163      SML/NJ.  (The assumptions underlying functions arith and promotable do not
1164      hold for things like multiplication and division.)
1165    
1166    ----------------------------------------------------------------------
1167    Name: Matthias Blume
1168    Date: 2002/03/27 16:27:00 EST
1169    Tag: blume-20020327-mlrisc-divisions
1170    Description:
1171    
1172    Added support for all four division operations (ML's div, mod, quot,
1173    and rem) to MLRISC.  In the course of doing so, I also rationalized
1174    the naming (no more annoying switch-around of DIV and QUOT), by
1175    parameterizing the operation by div_rounding_mode (which can be either
1176    DIV_TO_ZERO or DIV_TO_NEGINF).
1177    
1178    The generic MLTreeGen functor takes care of compiling all four
1179    operations down to only round-to-zero div.
1180    
1181    Missing pieces:
1182    
1183      * Doing something smarter than relying on MLTreeGen on architectures
1184        like, e.g., the x86 where hardware division delivers both quotient and
1185        remainder at the same time.  With this, the implementation of the
1186        round-to-neginf operations could be further streamlined.
1187    
1188      * Remove inlining support for div/mod/rem from the frontend and replace it
1189        with primops that get carried through to the backend.  Do this for all
1190        int and word types.
1191    
1192    ----------------------------------------------------------------------
1193    Name: Matthias Blume
1194    Date: 2002/03/25 17:25:00 EST
1195    Tag: blume-20020325-divmod
1196    Description:
1197    
1198    I improved (hopefully without breaking them) the implementation of Int.div,
1199    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
1200    advantage of the following observations:
1201    
1202      Let  q = x quot y      r = x rem y
1203           d = x div  y      m = x mod y
1204    
1205    where "quot" is the round-to-zero version of integer division that
1206    hardware usually provides.  Then we have:
1207    
1208         r = x - q * y        where neither the * nor the - will overflow
1209         d = if q >= 0 orelse x = q * y then q else q - 1
1210                              where neither the * nor the - will overflow
1211         m = if q >= 0 orelse r = 0 then r else r + y
1212                              where the + will not overflow
1213    
1214    This results in substantial simplification of the generated code.
1215    The following table shows the number of CFG nodes and edges generated
1216    for
1217            fun f (x, y) = x OPER y
1218            (* with OPER \in div, mod, quot, rem *)
1219    
1220    
1221        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
1222        --------------------------------------------------------
1223         div |         24 |         39 |         12 |         16
1224         mod |         41 |         71 |         12 |         16
1225        quot |          8 |         10 |          8 |         10
1226         rem |         10 |         14 |          8 |         10
1227    
1228    
1229    ----------------------------------------------------------------------
1230    Name: Matthias Blume
1231    Date: 2002/03/25 22:06:00 EST
1232    Tag: blume-20020325-cprotobug
1233    Description:
1234    
1235    Fixed a bug in cproto (c prototype decoder).
1236    
1237    ----------------------------------------------------------------------
1238    Name: Matthias Blume
1239    Date: 2002/03/25 16:00:00 EST
1240    Tag: blume-20020325-raw-primops
1241    Description:
1242    
1243    I did some cleanup to Allen's new primop code and
1244    replaced yesterday's bootfiles with new ones.
1245    (But they are stored in the same place.)
1246    
1247    ----------------------------------------------------------------------
1248    Name: Matthias Blume
1249    Date: 2002/03/24 22:40:00 EST
1250    Tag: blume-20020324-bootfiles
1251    Description:
1252    
1253    Made the bootfiles that Allen asked for.
1254    
1255    ----------------------------------------------------------------------
1256    Name: Allen Leung
1257    Date: 2002/03/23 15:50:00 EST
1258    Tag: leunga-20020323-flint-cps-rcc-primops
1259    Description:
1260    
1261      1. Changes to FLINT primops:
1262    
1263        (* make a call to a C-function;
1264         * The primop carries C function prototype information and specifies
1265         * which of its (ML-) arguments are floating point. C prototype
1266         * information is for use by the backend, ML information is for
1267         * use by the CPS converter. *)
1268      | RAW_CCALL of { c_proto: CTypes.c_proto,
1269                       ml_args: ccall_type list,
1270                       ml_res_opt: ccall_type option,
1271                       reentrant : bool
1272                     } option
1273       (* Allocate uninitialized storage on the heap.
1274        * The record is meant to hold short-lived C objects, i.e., they
1275        * are not ML pointers.  With the tag, the representation is
1276        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
1277        *)
1278      | RAW_RECORD of {tag:bool,sz:int}
1279      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
1280    
1281      2.  These CPS primops are now overloaded:
1282    
1283           rawload of {kind:numkind}
1284           rawstore of {kind:numkind}
1285    
1286          The one argument form is:
1287    
1288             rawload {kind} address
1289    
1290          The two argument form is:
1291    
1292             rawload {kind} [ml object, byte-offset]
1293    
1294      3. RAW_CCALL/RCC now takes two extra arguments:
1295    
1296         a. The first is whether the C call is reentrant, i.e., whether
1297            ML state should be saved and restored.
1298         b. The second argument is a string argument specifying the name of
1299            library and the C function.
1300    
1301         These things are currently not handled in the code generator, yet.
1302    
1303      4. In CProto,
1304    
1305         An encoding type of "bool" means "ml object" and is mapped into
1306         C prototype of PTR.  Note that "bool" is different than "string",
1307         even though "string" is also mapped into PTR, because "bool"
1308         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
1309    
1310      5. Pickler/unpicker
1311    
1312         Changed to handle RAW_RECORD and newest RAW_CCALL
1313    
1314      6. MLRiscGen,
1315    
1316         1. Changed to handle the new rawload/rawstore/rawrecord operators.
1317         2. Code for handling C Calls has been moved to a new module CPSCCalls,
1318            in the file CodeGen/cpscompile/cps-c-calls.sml
1319    
1320      7. Added the conditional move operator
1321    
1322             condmove of branch
1323    
1324         to cps.  Generation of this is still buggy so it is currently
1325         disabled.
1326    
1327    ----------------------------------------------------------------------
1328    Name: Lal George
1329    Date: 2002/03/22 14:18:25 EST
1330    Tag: george-20020322-cps-branch-prob
1331    Description:
1332    
1333    Implemented the Ball-Larus branch prediction-heuristics, and
1334    incorporated graphical viewers for control flow graphs.
1335    
1336    Ball-Larus Heuristics:
1337    ---------------------
1338    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1339    
1340    By design it uses the Dempster-Shafer theory for combining
1341    probabilities.  For example, in the function:
1342    
1343        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1344    
1345    the ball-larus heuristics predicts that the n=0 is unlikely
1346    (OH-heuristic), and the 'then' branch is unlikely because of the
1347    RH-heuristic -- giving the 'then' branch an even lower combined
1348    probability using the Dempster-Shafer theory.
1349    
1350    Finally, John Reppy's loop analysis in MLRISC, further lowers the
1351    probability of the 'then' branch because of the loop in the else
1352    branch.
1353    
1354    
1355    Graphical Viewing:
1356    ------------------
1357    I merely plugged in Allen's graphical viewers into the compiler. The
1358    additional code is not much. At the top level, saying:
1359    
1360            Control.MLRISC.getFlag "cfg-graphical-view" := true;
1361    
1362    will display the graphical view of the control flow graph just before
1363    back-patching.  daVinci must be in your path for this to work. If
1364    daVinci is not available, then the default viewer can be changed
1365    using:
1366    
1367            Control.MLRISC.getString "viewer"
1368    
1369    which can be set to "dot" or "vcg" for the corresponding viewers. Of
1370    course, these viewers must be in your path.
1371    
1372    The above will display the compilation unit at the level of clusters,
1373    many of which are small, boring, and un-interesting. Also setting:
1374    
1375            Control.MLRISC.getInt "cfg-graphical-view_size"
1376    
1377    will display clusters that are larger than the value set by the above.
1378    
1379    
1380    ----------------------------------------------------------------------
1381    Name: Matthias Blume
1382    Date: 2002/03/21 22:20:00 EST
1383    Tag: blume-20020321-kmp-bugfix
1384    Description:
1385    
1386    Changed the interface to the KMP routine in PreString and fixed
1387    a minor bug in one place where it was used.
1388    
1389    ----------------------------------------------------------------------
1390    Name: Allen Leung
1391    Date: 2002/03/21 20:30:00 EST
1392    Tag: leunga-20020321-cfg
1393    Description:
1394    
1395      Fixed a potential problem in cfg edge splitting.
1396    
1397    ----------------------------------------------------------------------
1398    Name: Allen Leung
1399    Date: 2002/03/21 17:15:00 EST
1400    Tag: leunga-20020321-x86-fp-cfg
1401    Description:
1402    
1403      1. Recoded the buggy parts of x86-fp.
1404    
1405         a. All the block reordering code has been removed.
1406            We now depend on the block placement phases to do this work.
1407    
1408         b. Critical edge splitting code has been simplified and moved into the
1409            CFG modules, as where they belong.
1410    
1411         Both of these were quite buggy and complex.  The code is now much, much
1412         simpler.
1413    
1414      2. X86 backend.
1415    
1416         a. Added instructions for 64-bit support.  Instruction selection for
1417            64-bit has not been committed, however, since that
1418            requires changes to MLTREE which haven't been approved by
1419            Lal and John.
1420    
1421         b. Added support for FUCOMI and FUCOMIP when generating code for
1422            PentiumPro and above.  We only generate these instructions in
1423            the fast-fp mode.
1424    
1425         c. Added cases for JP and JNP in X86FreqProps.
1426    
1427      3. CFG
1428    
1429         CFG now has a bunch of methods for edge splitting and merging.
1430    
1431      4. Machine description.
1432    
1433         John's simplification of MLTREE_BASIS.fcond broke a few machine
1434         description things:
1435    
1436         rtl-build.{sig,sml} and hppa.mdl fixed.
1437    
1438         NOTE: the machine description stuff in the repository is still broken.
1439               Again, I can't put my fixes in because that involves
1440               changes to MLTREE.
1441    
1442    ----------------------------------------------------------------------
1443    Name: Matthias Blume
1444    Date: 2002/03/20 15:55:00 EST
1445    Tag: blume-20020320-kmp
1446    Description:
1447    
1448    Implemented Knuth-Morris-Pratt string matching in PreString and used
1449    it for String.isSubstring, Substring.isSubstring, and
1450    Substring.position.
1451    
1452    (Might need some stress-testing.  Simple examples worked fine.)
1453    
1454    ----------------------------------------------------------------------
1455    Name: Matthias Blume
1456    Date: 2002/03/19 16:37:00 EST
1457    Tag: blume-20020319-witnesses
1458    Description:
1459    
1460    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1461    
1462    This implements a generic mechanism for changing constness qualifiers
1463    anywhere within big C types without resorting to outright "casts".
1464    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1465    modify the constness at the outermost level.)
1466    The implementation of "convert" is based on the idea of "witness"
1467    values -- values that are not used by the operation but whose types
1468    "testify" to their applicability.  On the implementation side, "convert"
1469    is simply a projection (returning its second curried argument).  With
1470    cross-module inlining, it should not result in any machine code being
1471    generated.
1472    
1473    ----------------------------------------------------------------------
1474    Name: Matthias Blume
1475    Date: 2002/03/15 16:40:00 EST
1476    Tag: blume-20020315-basis
1477    Description:
1478    
1479    Provided (preliminary?) implementations for
1480    
1481      {String,Substring}.{concatWith,isSuffix,isSubstring}
1482    
1483    and
1484    
1485      Substring.full
1486    
1487    Those are in the Basis spec but they were missing in SML/NJ.
1488    
1489    ----------------------------------------------------------------------
1490    Name: Matthias Blume
1491    Date: 2002/03/14 21:30:00 EST
1492    Tag: blume-20020314-controls
1493    Description:
1494    
1495    Controls:
1496    ---------
1497    
1498    1. Factored out the recently-added Controls : CONTROLS stuff and put
1499       it into its own library $/controls-lib.cm.  The source tree for
1500       this is under src/smlnj-lib/Controls.
1501    
1502    2. Changed the names of types and functions in this interface, so they
1503       make a bit more "sense":
1504    
1505          module -> registry
1506          'a registry -> 'a group
1507    
1508    3. The interface now deals in ref cells only.  The getter/setter interface
1509       is (mostly) gone.
1510    
1511    4. Added a function that lets one register an already-existing ref cell.
1512    
1513    5. Made the corresponding modifications to the rest of the code so that
1514       everything compiles again.
1515    
1516    6. Changed the implementation of Controls.MLRISC back to something closer
1517       to the original.  In particular, this module (and therefore MLRISC)
1518       does not depend on Controls.  There now is some link-time code in
1519       int-sys.sml that registers the MLRISC controls with the Controls
1520       module.
1521    
1522    CM:
1523    ---
1524    
1525      * One can now specify the lambda-split aggressiveness in init.cmi.
1526    
1527    ----------------------------------------------------------------------
1528    Name: Allen Leung
1529    Date: 2002/03/13 17:30:00 EST
1530    Tag: leunga-20020313-x86-fp-unary
1531    Description:
1532    
1533    Bug fix for:
1534    
1535    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1536    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1537    > - fun f(x,(y,z)) = Real.~ y;
1538    > [autoloading]
1539    > [autoloading done]
1540    >       fchsl   (%eax), 184(%esp)
1541    > Error: MLRisc bug: X86MCEmitter.emitInstr
1542    >
1543    > uncaught exception Error
1544    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1545    
1546    The problem was that the code generator did not generate any fp registers
1547    in this case, and the ra didn't know that it needed to run the X86FP phase to
1548    translate the pseudo fp instruction.   This only happened with unary fp
1549    operators in certain situations.
1550    
1551    ----------------------------------------------------------------------
1552    Name: Matthias Blume
1553    Date: 2002/03/13 14:00:00 EST
1554    Tag: blume-20020313-overload-etc
1555    Description:
1556    
1557    1. Added _overload as a synonym for overload for backward compatibility.
1558       (Control.overloadKW must be true for either version to be accepted.)
1559    
1560    2. Fixed bug in install script that caused more things to be installed
1561       than what was requested in config/targets.
1562    
1563    3. Made CM aware of the (_)overload construct so that autoloading
1564       works.
1565    
1566    ----------------------------------------------------------------------
1567    Name: Matthias Blume
1568    Date: 2002/03/12 22:03:00 EST
1569    Tag: blume-20020312-url
1570    Description:
1571    
1572    Forgot to update BOOT and srcarchiveurl.
1573    
1574    ----------------------------------------------------------------------
1575    Name: Matthias Blume
1576    Date: 2002/03/12 17:30:00 EST
1577    Tag: blume-20020312-version110392
1578    Description:
1579    
1580    Yet another version number bump (because of small changes to the
1581    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1582    
1583    Changes:
1584    
1585      The new pid generation scheme described a few weeks ago was overly
1586      complicated.  I implemented a new mechanism that is simpler and
1587      provides a bit more "stability":  Once CM has seen a compilation
1588      unit, it keeps its identity constant (as long as you do not delete
1589      those crucial CM/GUID/* files).  This means that when you change
1590      an interface, compile, then go back to the old interface, and
1591      compile again, you arrive at the original pid.
1592    
1593      There now also is a mechanism that instructs CM to use the plain
1594      environment hash as a module's pid (effectively making its GUID
1595      the empty string).  For this, "noguid" must be specified as an
1596      option to the .sml file in question within its .cm file.
1597      This is most useful for code that is being generated by tools such
1598      as ml-nlffigen (because during development programmers tend to
1599      erase the tool's entire output directory tree including CM's cached
1600      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1601      revert to the old, broken behavior of SML/NJ, but in specific cases
1602      where there is no danger of interface confusion, its use is ok
1603      (I think).
1604    
1605      ml-nlffigen by default generates "noguid" annotations.  They can be
1606      turned off by specifying -guid in its command line.
1607    
1608    ----------------------------------------------------------------------
1609    Name: Lal George
1610    Date: 2002/03/12 12 14:42:36 EST
1611    Tag: george-20020312-frequency-computation
1612    Description:
1613    
1614    Integrated jump chaining and static block frequency into the
1615    compiler. More details and numbers later.
1616    
1617    ----------------------------------------------------------------------
1618    Name: Lal George
1619    Date: 2002/03/11 11 22:38:53 EST
1620    Tag: george-20020311-jump-chain-elim
1621    Description:
1622    
1623    Tested the jump chain elimination on all architectures (except the
1624    hppa).  This is on by default right now and is profitable for the
1625    alpha and x86, however, it may not be profitable for the sparc and ppc
1626    when compiling the compiler.
1627    
1628    The gc test will typically jump to a label at the end of the cluster,
1629    where there is another jump to an external cluster containing the actual
1630    code to invoke gc. This is to allow factoring of common gc invocation
1631    sequences. That is to say, we generate:
1632    
1633            f:
1634               testgc
1635               ja   L1      % jump if above to L1
1636    
1637            L1:
1638               jmp L2
1639    
1640    
1641    After jump chain elimination the 'ja L1' instructions is converted to
1642    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1643    up being implemented in their long form (if L2 is far away) using:
1644    
1645            jbe     L3      % jump if below or equal to L3
1646            jmp     L2
1647         L3:
1648            ...
1649    
1650    
1651    For large compilation units L2  may be far away.
1652    
1653    
1654    ----------------------------------------------------------------------
1655    Name: Matthias Blume
1656    Date: 2002/03/11 13:30:00 EST
1657    Tag: blume-20020311-mltreeeval
1658    Description:
1659    
1660    A functor parameter was missing.
1661    
1662    ----------------------------------------------------------------------
1663    Name: Allen Leung
1664    Date: 2002/03/11 10:30:00 EST
1665    Tag: leunga-20020311-runtime-string0
1666    Description:
1667    
1668       The representation of the empty string now points to a
1669    legal null terminated C string instead of unit.  It is now possible
1670    to convert an ML string into C string with InlineT.CharVector.getData.
1671    This compiles into one single machine instruction.
1672    
1673    ----------------------------------------------------------------------
1674    Name: Allen Leung
1675    Date: 2002/03/10 23:55:00 EST
1676    Tag: leunga-20020310-x86-call
1677    Description:
1678    
1679       Added machine generation for CALL instruction (relative displacement mode)
1680    
1681    ----------------------------------------------------------------------
1682    Name: Matthias Blume
1683    Date: 2002/03/08 16:05:00
1684    Tag: blume-20020308-entrypoints
1685    Description:
1686    
1687    Version number bumped to 110.39.1.  NEW BOOTFILES!
1688    
1689    Entrypoints: non-zero offset into a code object where execution should begin.
1690    
1691    - Added the notion of an entrypoint to CodeObj.
1692    - Added reading/writing of entrypoint info to Binfile.
1693    - Made runtime system bootloader aware of entrypoints.
1694    - Use the address of the label of the first function given to mlriscGen
1695      as the entrypoint.  This address is currently always 0, but it will
1696      not be 0 once we turn on block placement.
1697    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1698      with entry points) from mlriscGen.
1699    
1700    ----------------------------------------------------------------------
1701    Name: Allen Leung
1702    Date: 2002/03/07 20:45:00 EST
1703    Tag: leunga-20020307-x86-cmov
1704    Description:
1705    
1706       Bug fixes for CMOVcc on x86.
1707    
1708       1. Added machine code generation for CMOVcc
1709       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1710       3. CMOVcc cannot have an immediate operand as argument.
1711    
1712    ----------------------------------------------------------------------
1713    Name: Matthias Blume
1714    Date: 2002/03/07 16:15:00 EST
1715    Tag: blume-20020307-controls
1716    Description:
1717    
1718    This is a very large but mostly boring patch which makes (almost)
1719    every tuneable compiler knob (i.e., pretty much everything under
1720    Control.* plus a few other things) configurable via both the command
1721    line and environment variables in the style CM did its configuration
1722    until now.
1723    
1724    Try starting sml with '-h' (or, if you are brave, '-H')
1725    
1726    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1727    implements the underlying generic mechanism.
1728    
1729    The interface to some of the existing such facilities has changed somewhat.
1730    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1731    (The getFoo interface is still there for backward-compatibility, but its
1732    use is deprecated.)
1733    
1734    The ml-build script passes -Cxxx=yyy command-line arguments through so
1735    that one can now twiddle the compiler settings when using this "batch"
1736    compiler.
1737    
1738    TODO items:
1739    
1740    We should go through and throw out all controls that are no longer
1741    connected to anything.  Moreover, we should go through and provide
1742    meaningful (and correct!) documentation strings for those controls
1743    that still are connected.
1744    
1745    Currently, multiple calls to Controls.new are accepted (only the first
1746    has any effect).  Eventually we should make sure that every control
1747    is being made (via Controls.new) exactly once.  Future access can then
1748    be done using Controls.acc.
1749    
1750    Finally, it would probably be a good idea to use the getter-setter
1751    interface to controls rather than ref cells.  For the time being, both
1752    styles are provided by the Controls module, but getter-setter pairs are
1753    better if thread-safety is of any concern because they can be wrapped.
1754    
1755    *****************************************
1756    
1757    One bug fix: The function blockPlacement in three of the MLRISC
1758    backpatch files used to be hard-wired to one of two possibilities at
1759    link time (according to the value of the placementFlag).  But (I
1760    think) it should rather sense the flag every time.
1761    
1762    *****************************************
1763    
1764    Other assorted changes (by other people who did not supply a HISTORY entry):
1765    
1766    1. the cross-module inliner now works much better (Monnier)
1767    2. representation of weights, frequencies, and probabilities in MLRISC
1768       changed in preparation of using those for weighted block placement
1769       (Reppy, George)
1770    
1771    ----------------------------------------------------------------------
1772    Name: Lal George
1773    Date: 2002/03/07 14:44:24 EST 2002
1774    Tag: george-20020307-weighted-block-placement
1775    
1776    Tested the weighted block placement optimization on all architectures
1777    (except the hppa) using AMPL to generate the block and edge frequencies.
1778    Changes were required in the machine properties to correctly
1779    categorize trap instructions. There is an MLRISC flag
1780    "weighted-block-placement" that can be used to enable weighted block
1781    placement, but this will be ineffective without block/edge
1782    frequencies (coming soon).
1783    
1784    
1785    ----------------------------------------------------------------------
1786    Name: Lal George
1787    Date: 2002/03/05 17:24:48 EST
1788    Tag: george-20020305-linkage-cluster
1789    
1790    In order to support the block placement optimization, a new cluster
1791    is generated as the very first cluster (called the linkage cluster).
1792    It contains a single jump to the 'real' entry point for the compilation
1793    unit. Block placement has no effect on the linkage cluster itself, but
1794    all the other clusters  have full freedom in the manner in which they
1795    reorder blocks or functions.
1796    
1797    On the x86 the typical linkage code that is generated is:
1798       ----------------------
1799            .align 2
1800       L0:
1801            addl    $L1-L0, 72(%esp)
1802            jmp     L1
1803    
1804    
1805            .align  2
1806       L1:
1807       ----------------------
1808    
1809    72(%esp) is the memory location for the stdlink register. This
1810    must contain the address of the CPS function being called. In the
1811    above example, it contains the address of  L0; before
1812    calling L1 (the real entry point for the compilation unit), it
1813    must contain the address for L1, and hence
1814    
1815            addl $L1-L0, 72(%esp)
1816    
1817    I have tested this on all architectures except the hppa.The increase
1818    in code size is of course negligible
1819    
1820    ----------------------------------------------------------------------
1821    Name: Allen Leung
1822    Date: 2002/03/03 13:20:00 EST
1823    Tag: leunga-20020303-mlrisc-tools
1824    
1825      Added #[ ... ] expressions to mlrisc tools
1826    
1827    ----------------------------------------------------------------------
1828    Name: Matthias Blume
1829    Date: 2002/02/27 12:29:00 EST
1830    Tag: blume-20020227-cdebug
1831    Description:
1832    
1833    - made types in structure C and C_Debug to be equal
1834    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1835    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1836    
1837    ----------------------------------------------------------------------
1838    Name: Matthias Blume
1839    Date: 2002/02/26 12:00:00 EST
1840    Tag: blume-20020226-ffi
1841    Description:
1842    
1843    1. Fixed a minor bug in CM's "noweb" tool:
1844       If numbering is turned off, then truly don't number (i.e., do not
1845       supply the -L option to noweb).  The previous behavior was to supply
1846       -L'' -- which caused noweb to use the "default" line numbering scheme.
1847       Thanks to Chris Richards for pointing this out (and supplying the fix).
1848    
1849    2. Once again, I reworked some aspects of the FFI:
1850    
1851       A. The incomplete/complete type business:
1852    
1853       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1854         gone!
1855       - ML types representing an incomplete type are now *equal* to
1856         ML types representing their corresponding complete types (just like
1857         in C).  This is still safe because ml-nlffigen will not generate
1858         RTTI for incomplete types, nor will it generate functions that
1859         require access to such RTTI.   But when ML code generated from both
1860         incomplete and complete versions of the C type meet, the ML types
1861         are trivially interoperable.
1862    
1863         NOTE:  These changes restore the full generality of the translation
1864         (which was previously lost when I eliminated functorization)!
1865    
1866       B. Enum types:
1867    
1868       - Structure C now has a type constructor "enum" that is similar to
1869         how the "su" constructor works.  However, "enum" is not a phantom
1870         type because each "T enum" has values (and is isomorphic to
1871         MLRep.Signed.int).
1872       - There are generic access operations for enum objects (using
1873         MLRep.Signed.int).
1874       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1875         * The structure contains the definition of type "mlrep" (the ML-side
1876         representation type of the enum).  Normally, mlrep is the same
1877         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1878         then mlrep will be defined as a datatype -- thus facilitating
1879         pattern matching on mlrep values.
1880         ("-ec" will be suppressed if there are duplicate values in an
1881          enumeration.)
1882         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1883         will be generated for each C enum constant xxx.
1884         * Conversion functions m2i and i2m convert between mlrep and
1885         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1886         * Coversion functions c and ml convert between mlrep and "tag enum".
1887         * Access functions (get/set) fetch and store mlrep values.
1888       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1889         enumerations are merged into one single enumeration represented by
1890         structure E_'.
1891    
1892    ----------------------------------------------------------------------
1893    Name: Allen Leung
1894    Date: 2002/02/25 04:45:00 EST
1895    Tag: leunga-20020225-cps-spill
1896    
1897    This is a new implementation of the CPS spill phase.
1898    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1899    In case of problems, replace it with the old file spill.sml
1900    
1901    The current compiler runs into some serious performance problems when
1902    constructing a large record.  This can happen when we try to compile a
1903    structure with many items.  Even a very simple structure like the following
1904    makes the compiler slow down.
1905    
1906        structure Foo = struct
1907           val x_1 = 0w1 : Word32.int
1908           val x_2 = 0w2 : Word32.int
1909           val x_3 = 0w3 : Word32.int
1910           ...
1911           val x_N = 0wN : Word32.int
1912        end
1913    
1914    The following table shows the compile time, from N=1000 to N=4000,
1915    with the old compiler:
1916    
1917    N
1918    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1919           MLRISC ra                               0.06u  0.00s  0.05g
1920              (spills = 0 reloads = 0)
1921           TOTAL                                   0.63u  0.07s  0.21g
1922    
1923    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1924           MLRISC ra                               5.68u  0.59s  3.93g
1925              (spills = 0 reloads = 0)
1926           TOTAL                                   14.71u  0.99s  4.81g
1927    
1928    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1929           MLRISC ra                               5.54u  0.65s  3.91g
1930              (spills = 543 reloads = 1082)
1931           TOTAL                                   65.40u  3.13s  6.00g
1932    
1933    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1934           MLRISC ra                               0.80u  0.10s  0.55g
1935              (spills = 42 reloads = 84)
1936           TOTAL                                   129.42u  5.10s  4.13g
1937    
1938    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1939           MLRISC ra                               2.69u  0.27s  1.38g
1940              (spills = 62 reloads = 124)
1941           TOTAL                                   682.48u  19.61s  13.99g
1942    
1943    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1944           MLRISC ra                               4.96u  0.27s  2.72g
1945              (spills = 85 reloads = 170)
1946           TOTAL                                   2375.26u  57.21s  48.00g
1947    
1948    As you can see the old cps spill module suffers from some serious
1949    performance problem.  But since I cannot decipher the old code fully,
1950    instead of patching the problems up, I'm reimplementing it
1951    with a different algorithm.  The new code is more modular,
1952    smaller when compiled, and substantially faster
1953    (O(n log n) time and O(n) space).  Timing of the new spill module:
1954    
1955    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1956          MLRISC ra                               0.25u  0.02s  0.15g
1957             (spills=1 reloads=3)
1958          TOTAL                                   7.74u  0.34s  1.62g
1959    
1960    Implementation details:
1961    
1962    As far as I can tell, the purpose of the CPS spill module is to make sure the
1963    number of live variables at any program point (the bandwidth)
1964    does not exceed a certain limit, which is determined by the
1965    size of the spill area.
1966    
1967    When the bandwidth is too large, we decrease the register pressure by
1968    packing live variables into spill records.  How we achieve this is
1969    completely different than what we did in the old code.
1970    
1971    First, there is something about the MLRiscGen code generator
1972    that we should be aware of:
1973    
1974    o MLRiscGen performs code motion!
1975    
1976       In particular, it will move floating point computations and
1977       address computations involving only the heap pointer to
1978       their use sites (if there is only a single use).
1979       What this means is that if we have a CPS record construction
1980       statement
1981    
1982           RECORD(k,vl,w,e)
1983    
1984       we should never count the new record address w as live if w
1985       has only one use (which is often the case).
1986    
1987       We should do something similar to floating point, but the transformation
1988       there is much more complex, so I won't deal with that.
1989    
1990    Secondly, there are now two new cps primops at our disposal:
1991    
1992     1. rawrecord of record_kind option
1993        This pure operator allocates some uninitialized storage from the heap.
1994        There are two forms:
1995    
1996         rawrecord NONE [INT n]  allocates a tagless record of length n
1997         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1998                                     and initializes the tag.
1999    
2000     2. rawupdate of cty
2001          rawupdate cty (v,i,x)
2002          Assigns to x to the ith component of record v.
2003          The storelist is not updated.
2004    
2005    We use these new primops for both spilling and increment record construction.
2006    
2007     1. Spilling.
2008    
2009        This is implemented with a linear scan algorithm (but generalized
2010        to trees).  The algorithm will create a single spill record at the
2011        beginning of the cps function and use rawupdate to spill to it,
2012        and SELECT or SELp to reload from it.  So both spills and reloads
2013        are fine-grain operations.  In contrast, in the old algorithm
2014        "spills" have to be bundled together in records.
2015    
2016        Ideally, we should sink the spill record construction to where
2017        it is needed.  We can even split the spill record into multiple ones
2018        at the places where they are needed.  But CPS is not a good
2019        representation for global code motion, so I'll keep it simple and
2020        am not attempting this.
2021    
2022     2. Incremental record construction (aka record splitting).
2023    
2024        Long records with many component values which are simulatenously live
2025        (recall that single use record addresses are not considered to
2026         be live) are constructed with rawrecord and rawupdate.
2027        We allocate space on the heap with rawrecord first, then gradually
2028        fill it in with rawupdate.  This is the technique suggested to me
2029        by Matthias.
2030    
2031        Some restrictions on when this is applicable:
2032        1. It is not a VECTOR record.  The code generator currently does not handle
2033           this case. VECTOR record uses double indirection like arrays.
2034        2. All the record component values are defined in the same "basic block"
2035           as the record constructor.  This is to prevent speculative
2036           record construction.
2037    
2038    ----------------------------------------------------------------------
2039    Name: Allen Leung
2040    Date: 2002/02/22 01:02:00 EST
2041    Tag: leunga-20020222-mlrisc-tools
2042    
2043    Minor bug fixes in the parser and rewriter
2044    
2045    ----------------------------------------------------------------------
2046    Name: Allen Leung
2047    Date: 2002/02/21 20:20:00 EST
2048    Tag: leunga-20020221-peephole
2049    
2050    Regenerated the peephole files.  Some contained typos in the specification
2051    and some didn't compile because of pretty printing bugs in the old version
2052    of 'nowhere'.
2053    
2054    ----------------------------------------------------------------------
2055    Name: Allen Leung
2056    Date: 2002/02/19 20:20:00 EST
2057    Tag: leunga-20020219-mlrisc-tools
2058    Description:
2059    
2060       Minor bug fixes to the mlrisc-tools library:
2061    
2062       1.  Fixed up parsing colon suffixed keywords
2063       2.  Added the ability to shut the error messages up
2064       3.  Reimplemented the pretty printer and fixed up/improved
2065           the pretty printing of handle and -> types.
2066       4.  Fixed up generation of literal symbols in the nowhere tool.
2067       5.  Added some SML keywords to to sml.sty
2068    
2069    ----------------------------------------------------------------------
2070    Name: Matthias Blume
2071    Date: 2002/02/19 16:20:00 EST
2072    Tag: blume-20020219-cmffi
2073    Description:
2074    
2075    A wild mix of changes, some minor, some major:
2076    
2077    * All C FFI-related libraries are now anchored under $c:
2078        $/c.cm      --> $c/c.cm
2079        $/c-int.cm  --> $c/internals/c-int.cm
2080        $/memory.cm --> $c/memory/memory.cm
2081    
2082    * "make" tool (in CM) now treats its argument pathname slightly
2083      differently:
2084        1. If the native expansion is an absolute name, then before invoking
2085           the "make" command on it, CM will apply OS.Path.mkRelative
2086           (with relativeTo = OS.FileSys.getDir()) to it.
2087        2. The argument will be passed through to subsequent phases of CM
2088           processing without "going native".  In particular, if the argument
2089           was an anchored path, then "make" will not lose track of that anchor.
2090    
2091    * Compiler backends now "know" their respective C calling conventions
2092      instead of having to be told about it by ml-nlffigen.  This relieves
2093      ml-nlffigen from one of its burdens.
2094    
2095    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
2096    
2097    * Export C_DEBUG and C_Debug from $c/c.cm.
2098    
2099    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
2100      subtyping relationship between incomplete pointers and their complete
2101      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
2102      with the convention of instantiating 'o with ('t, 'c) obj whenever
2103      the pointer target type is complete.  In the incomplete case, 'o
2104      will be instantiated with some "'c iobj" -- a type obtained by
2105      using one of the functors PointerToIncompleteType or PointerToCompleteType.
2106    
2107      Operations that work on both incomplete and complete pointer types are
2108      typed as taking an 'o ptr while operations that require the target to
2109      be known are typed as taking some ('t, 'c) obj ptr.
2110    
2111      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
2112      where void is an eqtype without any values.  This makes it possible
2113      to work on voidptr values using functions meant to operate on light
2114      incomplete pointers.
2115    
2116    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
2117      been vastly simplified.
2118    
2119    ----------------------------------------------------------------------
2120    Name: Matthias Blume
2121    Date: 2002/02/19 10:48:00 EST
2122    Tag: blume-20020219-pqfix
2123    Description:
2124    
2125    Applied Chris Okasaki's bug fix for priority queues.
2126    
2127    ----------------------------------------------------------------------
2128    Name: Matthias Blume
2129    Date: 2002/02/15 17:05:00
2130    Tag: Release_110_39
2131    Description:
2132    
2133    Last-minute retagging is becoming a tradition... :-(
2134    
2135    This is the working release 110.39.
2136    
2137    ----------------------------------------------------------------------
2138    Name: Matthias Blume
2139    Date: 2002/02/15 16:00:00 EST
2140    Tag: Release_110_39-orig
2141    Description:
2142    
2143    Working release 110.39.  New bootfiles.
2144    
2145    (Update: There was a small bug in the installer so it wouldn't work
2146    with all shells.  So I retagged. -Matthias)
2147    
2148    ----------------------------------------------------------------------
2149    Name: Matthias Blume
2150    Date: 2002/02/15 14:17:00 EST
2151    Tag: blume-20020215-showbindings
2152    Description:
2153    
2154    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
2155    the latter can be useful for exploring what bindings are available at
2156    the interactive prompt.  (The first function returns only the list
2157    of symbols that are really bound, the second prints those but also the
2158    ones that CM's autoloading mechanism knows about.)
2159    
2160    ----------------------------------------------------------------------
2161    Name: Matthias Blume
2162    Date: 2002/02/15 12:08:00 EST
2163    Tag: blume-20020215-iptrs
2164    Description:
2165    
2166    Two improvements to ml-nlffigen:
2167    
2168      1. Write files only if they do not exist or if their current contents
2169         do not coincide with what's being written.  (That is, avoid messing
2170         with the time stamps unless absolutely necessary.)
2171    
2172      2. Implement a "repository" mechanism for generated files related
2173         to "incomplete pointer types".   See the README file for details.
2174    
2175    ----------------------------------------------------------------------
2176    Name: Matthias Blume
2177    Date: 2002/02/14 11:50:00 EST
2178    Tag: blume-20020214-quote
2179    Description:
2180    
2181    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
2182    because of the new and improved tag generation scheme.  (Thanks to Allen
2183    Leung for pointing it out.)
2184    
2185    ----------------------------------------------------------------------
2186    Name: Lal George
2187    Date: 2002/02/14 09:55:27 EST 2002
2188    Tag: george-20020214-isabelle-bug
2189    Description:
2190    
2191    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
2192    of Isabelle on the x86.
2193    
2194    From Allen:
2195    -----------
2196     I've found the problem:
2197    
2198         in ra-core.sml, I use the counter "blocked" to keep track of the
2199         true number of elements in the freeze queue.  When the counter goes
2200         to zero, I skip examining the queue.  But I've messed up the
2201         bookkeeping in combine():
2202    
2203             else ();
2204             case !ucol of
2205               PSEUDO => (if !cntv > 0 then
2206                     (if !cntu > 0 then blocked := !blocked - 1 else ();
2207                                        ^^^^^^^^^^^^^^^^^^^^^^^
2208                      moveu := mergeMoveList(!movev, !moveu)
2209                     )
2210                  else ();
2211    
2212         combine() is called to coalesce two nodes u and v.
2213         I think I was thinking that if the move counts of u and v are both
2214         greater than zero then after they are coalesced then one node is
2215         removed from the freeze queue.  Apparently I was thinking that
2216         both u and v are of low degree, but that's clearly not necessarily true.
2217    
2218    
2219    02/12/2002:
2220        Here's the patch.  HOL now compiles.
2221    
2222        I don't know how this impact on performance (compile
2223        time or runtime).  This bug caused the RA (especially on the x86)
2224        to go thru the potential spill phase when there are still nodes on the
2225        freeze queue.
2226    
2227    
2228    
2229    
2230    ----------------------------------------------------------------------
2231    Name: Matthias Blume
2232    Date: 2002/02/13 22:40:00 EST
2233    Tag: blume-20020213-fptr-rtti
2234    Description:
2235    
2236    Fixed a bug in ml-nlffigen that was introduced with one of the previous
2237    updates.
2238    
2239    ----------------------------------------------------------------------
2240    Name: Matthias Blume
2241    Date: 2002/02/13 16:41:00 EST
2242    Tag: blume-20020213-cmlpq
2243    Description:
2244    
2245    Added new priority queue export symbols (which have just been added to
2246    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
2247    not compile and the installer would choke.)
2248    
2249    ----------------------------------------------------------------------
2250    Name: Matthias Blume
2251    Date: 2002/02/13 16:15:00 EST
2252    Tag: blume-20020213-various
2253    Description:
2254    
2255    1. More tweaks to ml-nlffigen:
2256    
2257       - better internal datastructures (resulting in slight speedup)
2258       - "-match" option requires exact match
2259       - "localized" gensym counters (untagged structs/unions nested within
2260         other structs/unions or within typedefs get a fresh counter; their
2261         tag will be prefixed by a concatenation of their parents' tags)
2262       - bug fixes (related to calculation of transitive closure of types
2263         to be included in the output)
2264    
2265    2. Minor Basis updates:
2266    
2267       - added implementations for List.collate and Option.app
2268    
2269    ----------------------------------------------------------------------
2270    Name: Matthias Blume
2271    Date: 2002/02/11 15:55:00 EST
2272    Tag: blume-20020211-gensym
2273    Description:
2274    
2275    Added a "-gensym" option to command line of ml-nlffigen.  This can be
2276    used to specify a "stem" -- a string that is inserted in all "gensym'd"
2277    names (ML structure names that correspond to unnamed C structs, unions,
2278    and enums), so that separate runs of ml-nlffigen do not clash.
2279    
2280    ----------------------------------------------------------------------
2281    Name: Matthias Blume
2282    Date: 2002/02/11 12:05:00 EST
2283    Tag: blume-20020211-gensml
2284    Description:
2285    
2286    A quick fix for a problem with GenSML (in the pgraph-util library):
2287    Make generation of toplevel "local" optional.  (Strictly speaking,
2288    signature definitions within "local" are not legal SML.)
2289    
2290    Other than that: updates to INSTALL and cm/TODO.
2291    
2292    ----------------------------------------------------------------------
2293    Name: Matthias Blume
2294    Date: 2002/02/08 15:00:00 EST
2295    Tag: blume-20020208-uniquepid
2296    Description:
2297    
2298    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
2299    
2300    1. The installer (config/install.sh) has gotten smarter:
2301    
2302         - Configuration options are a bit easier to specify now
2303           (in config/targets).
2304         - Bug in recognizing .tar.bz2 files fixed.
2305         - Installer automatically resolves dependencies between
2306           configuration options (e.g., if you ask for eXene, you will
2307           also get cml -- regardless whether you asked for it or not).
2308         - Installer can run in "quieter mode" by setting the environment
2309           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
2310           "completely silent", though.
2311         - Build HashCons library as part of smlnj-lib.
2312    
2313    2. A new scheme for assigning persistent identifiers to compilation
2314       units (and, by extension, to types etc.) has been put into place.
2315       This fixes a long-standing bug where types and even dynamic values
2316       can get internally confused, thereby compromising type safety
2317       (abstraction) and dynamic correctness.  See
2318    
2319         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
2320    
2321       for an example of how things could go wrong until now.
2322    
2323       The downside of the new scheme is that pids are not quite as
2324       persistent as they used to be: CM will generate a fresh pid
2325       for every compilation unit that it thinks it sees for the first
2326       time.  That means that if you compile starting from a clean, fresh
2327       source tree at two different times, you end up with different
2328       binaries.
2329    
2330       Cutoff recompilation, however, has not been compromised because
2331       CM keeps pid information in special caches between runs.
2332    
2333    ----------------------------------------------------------------------
2334    Name: Lal George
2335    Date: 2002/02/07 15:34:13 EST 2002
2336    Tag: <none>
2337    Description:
2338    
2339    Compilers that generate assembly code may produce  global labels
2340    whose value is resolved at link time. The various peephole optimization
2341    modules did not take this in account.
2342    
2343    TODO. The Labels.addrOf function should really return an option
2344    type so that clients are forced to deal with this issue, rather
2345    than an exception being raised.
2346    
2347    ----------------------------------------------------------------------
2348    Name: Lal George
2349    Date: 2002/02/06 13:55:02 EST
2350    Tag: george-20020206-ra-breakup
2351    Description:
2352    
2353    1. A bug fix from Allen.
2354    
2355        A typo causes extra fstp %st(0)'s to be generated at compensation
2356        edges, which might cause stack underflow traps at runtime.  This
2357        occurs in fft where there are extraneous fstps right before the 'into'
2358        trap instruction (in this case they are harmless since none of the
2359        integers overflow.)
2360    
2361    2. Pulled out various utility modules that were embedded in the modules
2362       of the register allocator. I need these modules for other purposes, but
2363       they are not complete enough to put into a library (just yet).
2364    ----------------------------------------------------------------------
2365    Name: Matthias Blume
2366    Date: 2002/01/31 16:05:00 EST
2367    Tag: blume-20020131-sparc-ccalls
2368    Description:
2369    
2370    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
2371       of extra stack space by mistake.  Fixed.
2372    
2373    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
2374    
2375    ----------------------------------------------------------------------
2376    Name: Allen Leung
2377    Date: 2002/01/30
2378    Tag: leunga-20020130-nowhere-bug-fix
2379    Description:
2380    
2381       MLRISC bug fixes:
2382       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
2383       2. MachineInt.fromString was negating its value.
2384    
2385    ----------------------------------------------------------------------
2386    Name: Matthias Blume
2387    Date: 2002/01/29
2388    Tag: blume-20020129-INSTALL
2389    Description:
2390    
2391    - Added somewhat detailed installation instructions (file INSTALL).
2392    - Fixed curl-detection bug in config/install.sh.
2393    - It is now possible to select the URL getter using the URLGETTER
2394      environment variable:
2395    
2396          not set / "unknown"      --> automatic detection (script tries wget,
2397                                       curl, and lynx)
2398          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
2399                                       how to properly invoke them)
2400          other                    --> use $URLGETTER directly, it must take
2401                                       precisely two command-line arguments
2402                                       (source URL and destination file name)
2403    
2404    ----------------------------------------------------------------------
2405    Name: Matthias Blume
2406    Date: 2002/01/28
2407    Tag: blume-20020128-sparc-ccalls
2408    Description:
2409    
2410    - Fixed problem with calculation of "used" registers in sparc-c-calls.
2411    - Make use of the allocParam argument in sparc-c-calls.
2412    
2413    ----------------------------------------------------------------------
2414    Name: Matthias Blume
2415    Date: 2002/01/28
2416    Tag: blume-20020128-allocParam
2417    Description:
2418    
2419    John Reppy:  Changes c-calls API to accept client-callback for
2420    allocating extra stack space.
2421    me: Corresponding changes to mlriscGen (using a dummy argument that
2422        does not change the current behavior).
2423    
2424    ----------------------------------------------------------------------
2425    Name: Matthias Blume
2426    Date: 2002/01/28 12:00:00
2427    Tag: Release_110_38
2428    Description:
2429    
2430    This time for real!!!
2431    
2432    ----------------------------------------------------------------------
2433    Name: Matthias Blume
2434    Date: 2002/01/28 10:56:00 EST
2435    Tag: blume-20020128-retraction
2436    Description:
2437    
2438    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
2439       with blume-Release_110_38-retracted.)
2440    
2441    1. Fixed a problem with incorrect rounding modes in real64.sml.
2442       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
2443    
2444    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
2445       and enums fixed.  The naming of corresponding ML identifiers should
2446       now be consistent again.
2447    
2448    ----------------------------------------------------------------------
2449    Name: Allen Leung
2450    Date: 2002/01/27
2451    Tag: leunga-20020127-nowhere
2452    Description:
2453    
2454       Added a target called nowhere in the configuration scripts.
2455       Enabling this will build the MLRISC 'nowhere' tool (for translating
2456       programs with where-clauses into legal SML code) during installation.
2457    
2458    ----------------------------------------------------------------------
2459    Name: Matthias Blume
2460    Date: 2002/01/25 21:27:00 EST
2461    Tag: blume-Release_110_38-retracted
2462    Description:
2463    
2464    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
2465    
2466    README will be added later.
2467    
2468    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
2469    (Release_110_38) removed.  Reason: Last-minute bug fixes.
2470    
2471    ----------------------------------------------------------------------
2472    Name: Matthias Blume
2473    Date: 2002/01/25
2474    Tag: blume-20020125-ffi
2475    Description:
2476    
2477    A large number of tweaks and improvements to ml-nlffi-lib and
2478    ml-nlffigen:
2479    
2480       - ML represenation types have been streamlined
2481       - getter and setter functions work with concrete values, not abstract
2482         ones where possible
2483       - ml-nlffigen command line more flexible (see README file there)
2484       - some bugs have been fixed (hopefully)
2485    
2486    ----------------------------------------------------------------------
2487    Name: Lal George
2488    Date: 2002/01/24
2489    Tag: george-20020124-risc-ra-interface
2490    Description:
2491    
2492       There is a dramatic simplification in the interface to the
2493       register allocator for RISC architectures as a result of making
2494       parallel copy instructions explicit.
2495    
2496    ----------------------------------------------------------------------
2497    Name: Matthias Blume
2498    Date: 2002/01/22
2499    Tag: blume-20020122-x86-ccalls
2500    Description:
2501    
2502    Bug fix for c-calls on x86 (having to do with how char- and
2503    short-arguments are being handled).
2504    
2505    ----------------------------------------------------------------------
2506    Name: Matthias Blume
2507    Date: 2002/01/21
2508    Tag: blume-20020121-ff
2509    Description:
2510    
2511    Another day of fiddling with the FFI...
2512    
2513    1. Bug fix/workaround:  CKIT does not complain about negative array
2514       dimensions, so ml-nlffigen has to guard itself against this possibility.
2515       (Otherwise a negative dimension would send it into an infinite loop.)
2516    
2517    2. Some of the abstract types (light objects, light pointers, most "base"
2518       types) in structure C are now eqtypes.
2519    
2520    3. Added constructors and test functions for NULL function pointers.
2521    
2522    ----------------------------------------------------------------------
2523    Name: Matthias Blume
2524    Date: 2002/01/18
2525    Tag: blume-20020118-ready-for-new-release
2526    Description:
2527    
2528    Made config/srcarchiveurl point to a new place.  (Will provide boot
2529    files shortly.)
2530    
2531    Maybe we christen this to be 110.38?
2532    
2533    ----------------------------------------------------------------------
2534    Name: Matthias Blume
2535    Date: 2002/01/18
2536    Tag: blume-20020118-more-ffifiddle
2537    Description:
2538    
2539    Today's FFI fiddling:
2540    
2541      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
2542        C.Set.* that use concrete (MLRep.*) arguments and results instead
2543        of abstract ones.
2544    
2545      - Provided word-style bit operations etc. for "int" representation
2546        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
2547        and Long.
2548    
2549    ----------------------------------------------------------------------
2550    Name: Matthias Blume
2551    Date: 2002/01/18
2552    Tag: blume-20020118-use-x86-fp
2553    Description:
2554    
2555    Now that x86-fast-fp seems to be working, I turned it back on again
2556    by default.  (Seems to work fine now, even with the FFI.)
2557    
2558    Other than that, I added some documentation about the FFI to
2559    src/ml-nlffigen/README and updated the FFI test examples in
2560    src/ml-nlffi-lib/Tests/*.
2561    
2562    ----------------------------------------------------------------------
2563    Name: Allen Leung
2564    Date: 2002/01/17
2565    Tag: leunga-20020117-x86-fast-fp-call
2566    Description:
2567    
2568       1. Fixed a problem with handling return fp values when x86's fast fp
2569          mode is turned on.
2570    
2571       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
2572          of %f32.
2573    
2574       3. Added a constructor INT32lit to the ast of MLRISC tools.
2575    
2576    ----------------------------------------------------------------------
2577    Name: Matthias Blume
2578    Date: 2002/01/16
2579    Tag: blume-20020116-ffifiddle
2580    Description:
2581    
2582    More fiddling with the FFI interface:
2583    
2584     - Make constness 'c instead of rw wherever possible.  This eliminates
2585       the need for certain explicit coercions.  (However, due to ML's
2586       value polymorphism, there will still be many cases where explicit
2587       coercions are necessary.  Phantom types are not the whole answer
2588       to modeling a subtyping relationship in ML.)
2589    
2590     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
2591    
2592     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
2593       clumsy workaround for varargs problem.)
2594    
2595    ----------------------------------------------------------------------
2596    Name: Lal George
2597    Date: 2002/01/15
2598    Tag: <none>
2599    Description:
2600    
2601    1. Since COPY instructions are no longer native to the architecture,
2602       a generic functor can be used to implement the expandCopies function.
2603    
2604    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
2605       TEXT segment.
2606    
2607    ----------------------------------------------------------------------
2608    Name: Matthias Blume
2609    Date: 2002/01/15
2610    Tag: blume-20020115-ffiupdates
2611    Description:
2612    
2613    1. Fix for bug resulting in single-precision float values being returned
2614       incorrectly from FFI calls.
2615    
2616    2. Small modifications to C FFI API:
2617    
2618        - memory-allocation routines return straight objects (no options)
2619          and raise an exception in out-of-memory situations
2620        - unsafe extensions to cast between function pointers and pointers
2621          from/to ints
2622        - added structure C_Debug as an alternative to structure C where
2623          pointer-dereferencing (|*| and |*!) always check for null-pointers
2624        - added open_lib' to DynLinkage;  open_lib' works like open_lib
2625          but also takes a (possibly empty) list of existing library handles
2626          that the current library depends on
2627    
2628    ----------------------------------------------------------------------
2629    Name: Matthias Blume
2630    Date: 2002/01/10
2631    Tag: blume-20020110-newffigen
2632    Description:
2633    
2634    1. Updates to portable graph code.
2635    
2636    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
2637       more scalable now so that even huge interfaces such as the one
2638       for GTK compile in finite time and space. :-)
2639       See src/ml-nlffigen/README for details on what's new.
2640    
2641    ----------------------------------------------------------------------
2642    Name: Lal George
2643    Date: 2001/01/09 14:31:35 EST 2002
2644    Tag: george-20011206-rm-native-copy
2645    Description:
2646    
2647            Removed the native COPY and FCOPY instructions
2648            from all the architectures and replaced it with the
2649            explicit COPY instruction from the previous commit.
2650    
2651            It is now possible to simplify many of the optimizations
2652            modules that manipulate copies. This has not been
2653            done in this change.
2654    
2655    ----------------------------------------------------------------------
2656    Name: Lal George
2657    Date: 2001/12/06 16:50:13 EST 2001
2658    Tag: george-20011206-mlrisc-instruction
2659    Description:
2660    
2661    Changed the representation of instructions from being fully abstract
2662    to being partially concrete. That is to say:
2663    
2664      from
2665            type instruction
2666    
2667      to
2668            type instr                              (* machine instruction *)
2669    
2670            datatype instruction =
2671                LIVE of {regs: C.cellset, spilled: C.cellset}
2672              | KILL of {regs: C.cellset, spilled: C.cellset}
2673              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
2674              | ANNOTATION of {i: instruction, a: Annotations.annotation}
2675              | INSTR of instr
2676    
2677    This makes the handling of certain special instructions that appear on
2678    all architectures easier and uniform.
2679    
2680    LIVE and KILL say that a list of registers are live or killed at the
2681    program point where they appear. No spill code is generated when an
2682    element of the 'regs' field is spilled, but the register is moved to
2683    the 'spilled' (which is present, more for debugging than anything else).
2684    
2685    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
2686    We used to generate:
2687    
2688            DEFFREG f1
2689            f1 := f2 + f3
2690            trapb
2691    
2692    but now generate:
2693    
2694            f1 := f2 + f3
2695            trapb
2696            LIVE {regs=[f1,f2,f3], spilled=[]}
2697    
2698    Furthermore, the DEFFREG (hack) required that all floating point instruction
2699    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
2700    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
2701    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
2702    this architecture).
2703    
2704    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
2705    all the architectures. This will result in further simplification of the
2706    register allocator that must be aware of them for coalescing purposes, and
2707    will also simplify certain aspects of the machine description that provides
2708    callbacks related to parallel copies.
2709    
2710    ANNOTATION should be obvious, and now INSTR represents the honest to God
2711    machine instruction set!
2712    
2713    The <arch>/instructions/<arch>Instr.sml files define certain utility
2714    functions for making porting easier -- essentially converting upper case
2715    to lower case. All machine instructions (of type instr) are in upper case,
2716    and the lower case form generates an MLRISC instruction. For example on
2717    the alpha we have:
2718    
2719      datatype instr =
2720         LDA of {r:cell, b:cell, d:operand}
2721       | ...
2722    
2723      val lda : {r:cell, b:cell, d:operand} -> instruction
2724        ...
2725    
2726    where lda is just (INSTR o LDA), etc.
2727    
2728    ----------------------------------------------------------------------
2729    Name: Matthias Blume
2730    Date: 2001/11/22 21:40:00 EST
2731    Tag: Release_110_37
2732    Description:
2733    
2734    Release 110.37.  This time for real.
2735    
2736    ----------------------------------------------------------------------
2737    Name: Matthias Blume
2738    Date: 2001/11/21 16:35:00 EST
2739    Tag: blume-20011121-foot-in-mouth
2740    Description:
2741    
2742    Removed the "Release_110_37" tag because of a serious bug.
2743    This will be re-tagged once the bug is fixed.
2744    
2745    ----------------------------------------------------------------------
2746    Name: Matthias Blume
2747    Date: 2001/11/21 16:14:00 EST
2748    Tag: blume-20011121-forgottenfile
2749    Description:
2750    
2751    Forgot to add a file.  (Just a .tex-file -- part of
2752    the CM manual source.)
2753    
2754    ----------------------------------------------------------------------
2755    Name: Matthias Blume
2756    Date: 2001/11/21 16:10:00 EST
2757    Tag: blume-20011121-invalid_110_37
2758    Description:
2759    
2760    Note: I removed the original tag "Release_110_37" from this commit
2761          because we found a serious bug in all non-x86 backends.
2762          - Matthias
2763    
2764    1. Modifications to the SML/NJ code generator and to the runtime system
2765       so that code object name strings are directly inserted into code
2766       objects at code generation time.  The only business the runtime system
2767       has with this is now to read the name strings on occasions.
2768       (The encoding of the name string has also changed somewhat.)
2769    
2770    2. CM now implements a simple "set calculus" for specifying export lists.
2771       In particular, it is now possible to refer to the export lists of
2772       other libraries/groups/sources and form unions as well as differences.
2773       See the latest CM manual for details.
2774    
2775    3. An separate notion of "proxy" libraries has again be eliminated from
2776       CM's model.  (Proxy libraries are now simply a special case of using
2777       the export list calculus.)
2778    
2779    4. Some of the existing libraries now take advantage of the new set
2780       calculus.
2781       (Notice that not all libraries have been converted because some
2782       of the existing .cm-files are supposed to be backward compatible
2783       with 110.0.x.)
2784    
2785    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2786       "General.exnMessage"!  The former relies on a certain hook to be
2787       initialized, and that often does not happen in the stand-alone case.)
2788    
2789    ----------------------------------------------------------------------
2790    Name: Lal George
2791    Date: 2001/11/21  13:56:18 EST
2792    Tag: george-2001121-pseudo-ops
2793    Description:
2794    
2795      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2796      ought to never be any question of incompatabilities with
2797      pseudo-op syntax expected by host assemblers.
2798    
2799      For now, only modules supporting GAS syntax are implemented
2800      but more should follow, such as MASM, and vendor assembler
2801      syntax, e.g. IBM as, Sun as, etc.
2802    
2803    ----------------------------------------------------------------------
2804    Name: Matthias Blume
2805    Date: 2001/11/14 11:52:00 EST
2806    Tag: blume-20011114-srcname
2807    Description:
2808    
2809    1. Routed the name of the current source file to mlriscgen where it
2810       should be directly emitted into the code object.  (This last part
2811       is yet to be done.)
2812    
2813    2. Some cleanup of the pgraph code to make it match the proposal that
2814       I put out the other day.  (The proposal notwithstanding, things are
2815       still in flux here.)
2816    
2817    ----------------------------------------------------------------------
2818    Name: Lal George
2819    Date: 2001/11/14 09:44:04 EST
2820    Tag:
2821    Description:
2822    
2823      Fix for a backpatching bug reported by Allen.
2824    
2825      Because the boundary between short and long span-dependent
2826      instructions is +/- 128, there are an astounding number of
2827      span-dependent instructions whose size is over estimated.
2828    
2829      Allen came up with the idea of letting the size of span
2830      dependent instructions be non-monotonic, for a maxIter
2831      number of times, after which the size must be monotonically
2832      increasing.
2833    
2834      This table shows the number of span-dependent instructions
2835      whose size was over-estimated as a function of maxIter, for the
2836      file Parse/parse/ml.grm.sml:
2837    
2838         maxIter            # of instructions:
2839            10                      687
2840            20                      438
2841            30                      198
2842            40                        0
2843    
2844      In compiling the compiler, there is no significant difference in
2845      compilation speed between maxIter=10 and maxIter=40. Actually,
2846      my measurements showed that maxIter=40 was a tad faster than
2847      maxIter=10! Also 96% of the  files in the compiler reach a fix
2848      point within 13 iterations, so fixing maxIter at 40, while high,
2849      is okay.
2850    
2851    ----------------------------------------------------------------------
2852    Name: Matthias Blume
2853    Date: 2001/10/31 15:25:00 EST
2854    Tag: blume-20011031-pgraph
2855    Description:
2856    
2857    CKIT:
2858    * Changed the "Function" constructor of type Ast.ctype to carry optional
2859      argument identifiers.
2860    * Changed the return type of TypeUtil.getFunction accordingly.
2861    * Type equality ignores the argument names.
2862    * TypeUtil.composite tries to preserve argument names but gives up quickly
2863      if there is a mismatch.
2864    
2865    installation script:
2866    * attempts to use "curl" if available (unless "wget" is available as well)
2867    
2868    CM:
2869    * has an experimental implementation of "portable graphs" which I will
2870      soon propose as an implementation-independent library format
2871    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2872    
2873    NLFFI-LIB:
2874    * some cleanup (all cosmetic)
2875    
2876    NLFFIGEN:
2877    * temporarily disabled the mechanism that suppresses ML output for
2878      C definitions whose identifiers start with an underscore character
2879    * generate val bindings for enum constants
2880    * user can request that only one style (light or heavy) is being used;
2881      default is to use both (command-line arguments: -heavy and -light)
2882    * fixed bug in handling of function types involving incomplete pointers
2883    * generate ML entry points that take record arguments (i.e., using
2884      named arguments) for C functions that have a prototype with named
2885      arguments
2886      (see changes to CKIT)
2887    
2888    ----------------------------------------------------------------------
2889    Name: Allen Leung
2890    Date: 2001/10/27 20:34:00 EDT
2891    Tag: leunga-20011027-x86-fast-fp-call
2892    Description:
2893    
2894       Fixed the bug described in blume-20010920-slowfp.
2895    
2896       The fix involves
2897          1. generating FCOPYs in FSTP in ia32-svid
2898          2. marking a CALL with the appropriate annotation
2899    
2900    ----------------------------------------------------------------------
2901    Name: Matthias Blume
2902    Date: 2001/10/16 11:32:00 EDT
2903    Tag: blume-20011016-netbsd
2904    Description:
2905    
2906    Underscore patch from Chris Richards (fixing problem with compiling
2907    runtime system under recent NetBSD).
2908    
2909    ----------------------------------------------------------------------
2910    Name: Allen Leung
2911    Date: 2001/10/12 17:18:32 EDT 2001
2912    Tag: leung-20011012-x86-printflowgraph
2913    Description:
2914    
2915    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2916    
2917    ----------------------------------------------------------------------
2918    Name: Lal George
2919    Date: 2001/10/11 23:51:34 EDT
2920    Tag: george-20011011-too-many-instrs
2921    Description:
2922    
2923    The representation of a program point never expected to see more
2924    than 65536 instructions in a basic block!
2925    
2926    ----------------------------------------------------------------------
2927    Name: Lal George
2928    Date: 2001/10/09 09:41:37 EDT
2929    Tag: george-20011008-mlrisc-labels
2930    Description:
2931    
2932    Changed the machine description files to support printing of
2933    local and global labels in assembly code, based on host assembler
2934    conventions.
2935    
2936    ----------------------------------------------------------------------
2937    Name: Matthias Blume
2938    Date: 2001/09/25 15:25:00 EDT
2939    Tag: blume-20010925-exninfo
2940    Description:
2941    
2942    I provided a non-hook implementation of exnName (at the toplevel) and
2943    made the "dummy" implementation of exnMessage (at the toplevel) more
2944    useful: if nothing gets "hooked in", then at least you are going to
2945    see the exception name and a message indicating why you don't see more.
2946    
2947    [For the time being, programs that need exnMessage and want to use
2948    ml-build should either use General.exnMessage (strongly recommended) or
2949    refer to structure General at some other point so that CM sees a
2950    static dependency.]
2951    
2952    [Similar remarks go for "print" and "use":  If you want to use their
2953    functionality in stand-alone programs generated by ml-build, then use
2954    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2955    
2956    ----------------------------------------------------------------------
2957    Name: Matthias Blume
2958    Date: 2001/09/20 17:28:00 EDT
2959    Tag: blume-20010920-slowfp
2960    Description:
2961    
2962    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2963    
2964    ----------------------------------------------------------------------
2965    Name: Matthias Blume
2966    Date: 2001/09/20 17:20:00 EDT
2967    Tag: blume-20010920-canonicalpaths
2968    Description:
2969    
2970    0. Updated the BOOT file (something that I forgot to do earlier).
2971    
2972    1. Small internal change to CM so that it avoids "/../" in filenames
2973       as much as possible (but only where it is safe).
2974    
2975    2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2976       that arguments that contain delimiters are passed through correctly.
2977       This change also means that all "special" arguments of the form
2978       @SMLxxx... must come first.
2979    
2980    3. Changed install script to put relative anchor names for tool commands
2981       into pathconfig.
2982    
2983    ----------------------------------------------------------------------
2984    Name: Matthias Blume
2985    Date: 2001/09/18 15:35:00 EDT
2986    Tag: blume-20010918-readme11036
2987    Description:
2988    
2989    Added README files.
2990    
2991    ----------------------------------------------------------------------
2992    Name: Matthias Blume
2993    Date: 2001/09/18 11:45:00 EDT
2994    Tag: Release_110_36 (retag)
2995    Description:
2996    
2997    Fixed mistake in config/preloads. Retagged as 110.36.
2998    
2999    ----------------------------------------------------------------------
3000    Name: Matthias Blume
3001    Date: 2001/09/18 09:40:00 EDT
3002    Tag: Release_110_36_orig (tag changed)
3003    Description:
3004    
3005    New version (110.36).  New bootfiles.
3006    
3007    ----------------------------------------------------------------------
3008    Name: Matthias Blume
3009    Date: 2001/09/14 16:15:00 EDT
3010    Tag: blume-20010914-x86fastfp
3011    Description:
3012    
3013    John committed some changes that Allen made, in particular a (hopefully)
3014    correctly working version of the x86-fp module.
3015    
3016    I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
3017    flag to "true".  Everything seems to compile to a fixpoint ok, and
3018    "mandelbrot" speeds up by about 15%.
3019    
3020    ----------------------------------------------------------------------
3021    Name: Matthias Blume
3022    Date: 2001/09/13 11:20:00 EDT
3023    Tag: blume-20010913-minimal
3024    Description:
3025    
3026    1. Stefan Monnier's patch to fix a miscompilation problem that
3027       was brought to light by John Reppy's work on Moby.
3028    
3029    2. Implemented a minimal "structure Compiler" that contains just
3030       "version" and "architecture".  The minimal version will be
3031       available when the full version is not.  This is for backward-
3032       compatibility with code that wants to test Compiler.version.
3033    
3034    ----------------------------------------------------------------------
3035    Name: Matthias Blume
3036    Date: 2001/08/28 14:03:00 EDT
3037    Tag: blume-20010828-ml-lex
3038    Description:
3039    
3040    Fix for bug 1581, received from Neophytos Michael.
3041    
3042    ----------------------------------------------------------------------
3043    Name: Matthias Blume
3044    Date: 2001/08/27 11:20:00 EDT
3045    Tag: blume-20010827-readme11035
3046    Description:
3047    
3048    Fleshed out the README file for 110.35.
3049    
3050    ----------------------------------------------------------------------
3051    Name: Matthias Blume
3052    Date: 2001/08/24 17:10:00 EDT
3053    Tag: Release_110_35
3054    Description:
3055    
3056    New version number (110.35).  New bootfiles.
3057    
3058    ----------------------------------------------------------------------
3059    Name: Lal George
3060    Date: 2001/08/24 13:47:18 EDT 2001
3061    Tag: george-20010824-MLRISC-graphs
3062    Description:
3063    
3064     removed clusters from MLRISC completely and replaced with graphs.
3065    
3066    ----------------------------------------------------------------------
3067    Name: Matthias Blume
3068    Date: 2001/08/23 17:50:00 EDT
3069    Tag: blume-20010823-toplevel
3070    Description:
3071    
3072    - some reorganization of the code that implements various kinds of
3073      environments in the compiler (static, dynamic, symbolic, combined)
3074    - re-implemented the EnvRef module so that evalStream works properly
3075      (if the stream contains references to "use", "CM.make", etc.)
3076    - cleaned up evalloop.sml and interact.sml (but they need more cleaning)
3077    
3078    ----------------------------------------------------------------------
3079    Name: Matthias Blume
3080    Date: 2001/08/20 15:50 EDT
3081    Tag: blume20010820-slipup
3082    Description:
3083    
3084    I forgot to commit a few files.  Here they are...
3085    
3086    ----------------------------------------------------------------------
3087    Name: Matthias Blume
3088    Date: 2001/08/20 15:35:00 EDT
3089    Tag: blume-20010820-debugprof
3090    Description:
3091    
3092    !!!! NEW BOOTFILES !!!!
3093    
3094    This is another round of reorganizing the compiler sources.  This
3095    time the main goal was to factor out all the "instrumentation"
3096    passes (for profiling and backtracing) into their own library.
3097    The difficulty was to do it in such a way that it does not depend
3098    on elaborate.cm but only on elabdata.cm.
3099    
3100    Therefore there have been further changes to both elaborate.cm and
3101    elabdata.cm -- more "generic" things have been moved from the former
3102    to the latter.  As a result, I was forced to split the assignment
3103    of numbers indicating "primtyc"s into two portions: SML-generic and
3104    SML/NJ-specific.  Since it would have been awkward to maintain,
3105    I bit the bullet and actually _changed_ the mapping between these
3106    numbers and primtycs.  The bottom line of this is that you need
3107    a new set of bin- and bootfiles.
3108    
3109    I have built new bootfiles for all architectures, so doing a fresh
3110    checkout and config/install.sh should be all you need.
3111    
3112    The newly created library's name is
3113    
3114        $smlnj/viscomp/debugprof.cm
3115    
3116    and its sources live under
3117    
3118        src/compiler/DebugProf
3119    
3120    ----------------------------------------------------------------------
3121    Name: Matthias Blume
3122    Date: 2001/08/15 17:15:00 EDT
3123    Tag: blume-20010815-compreorg
3124    Description:
3125    
3126    This is a first cut at reorganizing the CM libraries that make up the
3127    core of the compiler.  The idea is to separate out pieces that could
3128    be used independently by tools, e.g., the parser, the typechecker, etc.
3129    
3130    The current status is a step in this direction, but it is not quite
3131    satisfactory yet.  Expect more changes in the future.
3132    
3133    Here is the current (new) organization...
3134    
3135        What used to be $smlnj/viscomp/core.cm is now divided into
3136        six CM libraries:
3137    
3138             $smlnj/viscomp/basics.cm
3139                           /parser.cm
3140                           /elabdata.cm
3141                           /elaborate.cm
3142                           /execute.cm
3143                           /core.cm
3144    
3145        The CM files for these libraries live under src/system/smlnj/viscomp.
3146        All these libraries are proxy libraries that contain precisely
3147        one CM library component.  Here are the locations of the components
3148        (all within the src/compiler tree):
3149    
3150             Basics/basics.cm
3151             Parse/parser.cm
3152             ElabData/elabdata.cm
3153             Elaborator/elaborate.cm
3154             Execution/execute.cm
3155             core.cm
3156    
3157         [This organization is the same that has been used already
3158         for a while for the architecture-specific parts of the visible
3159         compiler and for the old version of core.cm.]
3160    
3161         As you will notice, many source files have been moved from their
3162         respective original locations to a new home in one of the above
3163         subtrees.
3164    
3165         The division of labor between the new libraries is the following:
3166    
3167             basics.cm:
3168                - Simple, basic definitions that pertain to many (or all) of
3169                  the other libraries.
3170             parser.cm:
3171                - The SML parser, producing output of type Ast.dec.
3172                - The type family for Ast is also defined and exported here.
3173             elabdata.cm:
3174                - The datatypes that describe input and output of the elaborator.
3175                  This includes types, absyn, and static environments.
3176             elaborator.cm:
3177                - The SML/NJ type checker and elaborator.
3178                  This maps an Ast.dec (with a given static environment) to
3179                  an Absyn.dec (with a new static environment).
3180                - This libraries implements certain modules that used to be
3181                  structures as functors (to remove dependencies on FLINT).
3182             execute.cm:
3183                - Everything having to do with executing binary code objects.
3184                - Dynamic environments.
3185             core.cm:
3186                - SML/NJ-specific instantiations of the elaborator and MLRISC.
3187                - Top-level modules.
3188                - FLINT (this should eventually become its own library)
3189    
3190    Notes:
3191    
3192    I am not 100% happy with the way I separated the elaborator (and its
3193    data structures) from FLINT.  Two instances of the same problem:
3194    
3195        1. Data structures contain certain fields that carry FLINT-specific
3196           information.  I hacked around this using exn and the property list
3197           module from smlnj-lib.  But the fact that there are middle-end
3198           specific fields around at all is a bit annoying.
3199    
3200        2. The elaborator calculates certain FLINT-related information.  I tried
3201           to make this as abstract as I could using functorization, but, again,
3202           the fact that the elaborator has to perform calculations on behalf
3203           of the middle-end at all is not nice.
3204    
3205        3. Having to used exn and property lists is unfortunate because it
3206           weakens type checking.  The other alternative (parameterizing
3207           nearly *everything*) is not appealing, though.
3208    
3209    I removed the "rebinding =" warning hack because due to the new organization
3210    it was awkward to maintain it.  As a result, the compiler now issues some of
3211    these warnings when compiling init.cmi during bootstrap compilation. On
3212    the plus side, you also get a warning when you do, for example:
3213       val op = = Int32.+
3214    which was not the case up to now.
3215    
3216    I placed "assign" and "deref" into the _Core structure so that the
3217    code that deals with the "lazy" keyword can find them there.  This
3218    removes the need for having access to the primitive environment
3219    during elaboration.
3220    
3221    ----------------------------------------------------------------------
3222    Name: Matthias Blume
3223    Date: 2001/08/13
3224    Tag: blume-20010813-closures
3225    Description:
3226    
3227    This fix was sent to us by Zhong Shao.  It is supposed to improve the
3228    performance of certain loops by avoiding needless closure allocation.
3229    
3230    ----------------------------------------------------------------------
3231    Name: Lal George
3232    Date: 2001/07/31 10:03:23 EDT 2001
3233    Tag: george-20010731-x86-fmalloc
3234    Description: Fixed bug in x86 calls
3235    
3236        There was a bug where call instructions would mysteriously
3237        vanish. The call instruction had to be one that returned
3238        a floating point value.
3239    
3240    ----------------------------------------------------------------------
3241    Name: Lal George
3242    Date: 2001/07/19 16:36:29 EDT 2001
3243    Tag: george-20010719-simple-cells
3244    Description:
3245    
3246    I have dramatically simplified the interface for CELLS in MLRISC.
3247    
3248    In summary, the cells interface is broken up into three parts:
3249    
3250      1. CellsBasis : CELLS_BASIS
3251    
3252            CellsBasis is a top level structure and common for all
3253            architectures.  it contains the definitions of basic datatypes
3254            and utility  functions over these types.
3255    
3256      2. functor Cells() : CELLS
3257    
3258            Cells generates an interface for CELLS that incorporates the
3259            specific resources on the target architecture, such as the
3260            presence of special register classes, their number and size,
3261            and various useful substructures.
3262    
3263      3. <ARCH>CELLS
3264    
3265            e.g. SparcCells: SPARCCELLS
3266    
3267            <ARCH>CELLS usually contains additional bindings for special
3268            registers  on the architecture, such as:
3269    
3270                    val r0 : cell           (* register zero *)
3271                    val y : cell            (* Y register *)
3272                    val psr : cell          (* processor status register *)
3273                    ...
3274    
3275            The structure returned by applying the Cells functor is opened
3276            in this interface.
3277    
3278    The main implication of all this is that the datatypes for cells is
3279    split between CellsBasis and CELLS -- a fairly simple change for user
3280    code.
3281    
3282    In the old scheme the CELLS interface had a definitional binding of
3283    the form:
3284    
3285            signature CELLS = sig
3286    
3287               structure CellsBasis = CellsBasis
3288    
3289               ...
3290    
3291            end
3292    
3293    With all the sharing constraints that goes on in MLRISC, this old
3294    design  quickly leads to errors such as:
3295    
3296            "structure definition spec inside of sharing ... "
3297    
3298    
3299    and appears to require an unacceptable amount of sharing and where
3300    constraint hackery.
3301    
3302    I think this error message (the interaction of definitional specs and
3303    sharing) requires more explanation on our web page.
3304    
3305    ----------------------------------------------------------------------
3306    Name: Matthias Blume
3307    Date: 2001/07/19 15:00:00 EDT
3308    Tag: blume-20010719-libreorg
3309    Description:
3310    
3311    This update puts together a fairly extensive but straightforward change
3312    to the way the libraries that implement the interactive system are
3313    organized:
3314    
3315       The biggest change is the elimination of structure Compiler.  As a
3316       replacement for this structure, there is now a CM library
3317       (known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
3318       that exports all the substructures of the original structure Compiler
3319       directly.  So instead of saying Compiler.Foo.bar one now simply
3320       says Foo.bar.  (The CM libraries actually export a collection of
3321       structures that is richer than the collection of substructures of
3322       structure Compiler.)
3323    
3324       To make the transition smooth, there is a separate library called
3325       $smlnj/compiler/compiler.cm which puts together and exports the
3326       original structure Compiler (or at least something very close to it).
3327    
3328       There are five members of the original structure Compiler
3329       that are not exported directly but which instead became members
3330       of a new structure Backend (described by signature BACKEND).  These are:
3331       structure Profile (: PROFILE), structure Compile (: COMPILE), structure
3332       Interact (: INTERACT), structure Machine (: MACHINE), and val
3333       architecture (: string).
3334    
3335       Structure Compiler.Version has become structure CompilerVersion.
3336    
3337       Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
3338       by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
3339       or x86, respectively.
3340       Each of these exports the same frontend structures that
3341       $smlnj/compiler.cm exports.  But they do not have a structure Backend
3342       and instead export some structure <Arch>Backend where <Arch> is Alpha32,
3343       Hppa, PPC, Sparc, or X86, respectively.
3344    
3345       Library $smlnj/compiler/all.cm exports the union of the exports of
3346       $smlnj/compiler/<arch>.cm
3347    
3348       There are no structures <Arch>Compiler anymore, use
3349       $smlnj/compiler/<arch>.cm instead.
3350    
3351       Library host-compiler-0.cm is gone.  Instead, the internal library
3352       that instantiates CM is now called cm0.cm.  Selection of the host
3353       compiler (backend) is no longer done here but. (Responsibility for it
3354       now lies with $smlnj/compiler/current.cm.  This seems to be more
3355       logical.)
3356    
3357       Many individual files have been moved or renamed.  Some files have
3358       been split into multiple files, and some "dead" files have been deleted.
3359    
3360    Aside from these changes to library organization, there are also changes
3361    to the way the code itself is organized:
3362    
3363       Structure Binfile has been re-implemented in such a way that it no
3364       longer needs any knowledge of the compiler.  It exclusively deals
3365       with the details of binfile layout.  It no longer invokes the
3366       compiler (for the purpose of creating new prospective binfile
3367       content), and it no longer has any knowledge of how to interpret
3368       pickles.
3369    
3370       Structure Compile (: COMPILE) has been stripped down to the bare
3371       essentials of compilation.  It no longer deals with linking/execution.
3372       The interface has been cleaned up considerably.
3373    
3374       Utility routines for dealing with linking and execution have been
3375       moved into their own substructures.
3376    
3377       (The ultimate goal of these changes is to provide a light-weight
3378       binfile loader/linker (at least for, e.g., stable libraries) that
3379       does not require CM or the compiler to be present.)
3380    
3381    CM documentation has been updated to reflect the changes to library
3382    organization.
3383    
3384    ----------------------------------------------------------------------
3385    Name: Matthias Blume
3386    Date: 2001/07/10 17:30:00 EDT
3387    Tag: Release_110_34
3388    Description:
3389    
3390    Minor tweak to 110.34 (re-tagged):
3391    
3392      - README.html file added to CVS repository
3393      - runtime compiles properly under FreeBSD 3.X and 4.X
3394    
3395    ----------------------------------------------------------------------
3396    Name: Matthias Blume
3397    Date: 2001/07/10 17:30:00 EDT
3398    Tag: Release_110_34
3399    Description:
3400    
3401    New version number (110.34). New bootfiles.
3402    
3403    ----------------------------------------------------------------------
3404    Name: Matthias Blume
3405    Date: 2001/07/09 16:00:00 EDT
3406    Tag: blume-20010709-more-varargs
3407    Description:
3408    
3409    I changed the handling of varargs in ml-nlffigen again:
3410    The ellipsis ... will now simply be ignored (with an accompanying warning).
3411    
3412    The immediate effect is that you can actually call a varargs function
3413    from ML -- but you can't actually supply any arguments beyond the ones
3414    specified explicitly.  (For example, you can call printf with its format
3415    string, but you cannot pass additional arguments.)
3416    
3417    This behavior is only marginally more useful than the one before, but
3418    it has the advantage that a function or, more importantly, a function
3419    type never gets dropped on the floor, thus avoiding follow-up problems with
3420    other types that refer to the offending one.
3421    
3422    ----------------------------------------------------------------------
3423    Name: Matthias Blume
3424    Date: 2001/07/09 11:25:00 EDT
3425    Tag: blume-20010709-varargs
3426    Description:
3427    
3428    1. ckit-lib.cm now exports structure Error
3429    2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
3430       with a warning accompanied by a source location.  Moreover, it
3431       merely skips the offending function or type and proceeds with the
3432       rest of its work.u  As a result, one can safely feed C code containing
3433       "..." to ml-nlffigen.
3434    3. There are some internal improvements to CM, providing slightly
3435       more general string substitutions in the tools subsystem.
3436    
3437    ----------------------------------------------------------------------
3438    Name: Matthias Blume
3439    Date: 2001/06/27 15:10:00 EDT
3440    Tag: blume-20010627-concur
3441    Description:
3442    
3443    Fixed a small bug in CM's handling of parallel compilation.
3444    (You could observe the bug by Control-C-interrupting an ordinary
3445    CMB.make or CM.stabilize and then attaching some compile servers.
3446    The result was that all of a sudden the previously interrupted
3447    compilation would continue on its own.  This was because of
3448    an over-optimization: CM did not bother to clean out certain queues
3449    when no servers were attached "anyway", resulting in the contents
3450    of these queues to grab control when new servers did get attached.)
3451    
3452    There is also another minor update to the CM manual.
3453    
3454    ----------------------------------------------------------------------
3455    Name: Matthias Blume
3456    Date: 2001/06/26 16:15:00 EDT
3457    Tag: blume-20010626-cmdoc
3458    Description:
3459    
3460    Minor typo fixed in CM manual (syntax diagram for libraries).
3461    
3462    ----------------------------------------------------------------------
3463    Name: Matthias Blume
3464    Date: 2001/06/25 22:55:00 EDT
3465    Tag: blume-20010625-x86pc
3466    Description:
3467    
3468    Fixed a nasty bug in the X86 assembly code that caused signal
3469    handlers to fail (crash) randomly.
3470    
3471    ----------------------------------------------------------------------
3472    Name: Matthias Blume
3473    Date: 2001/06/25 12:05:00 EDT
3474    Tag: blume-20010625-nlffigen
3475    Description:
3476    
3477    This update fixes a number of minor bugs in ml-nlffigen as reported by
3478    Nick Carter <nbc@andrew.cmu.edu>.
3479    
3480      1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
3481      2. Default names for generated files are now derived from the name of
3482         the C file *without its directory*.  In particular, this causes generated
3483         files to be placed locally even if the C file is in some system directory.
3484      3. Default names for generated signatures and structures are also derived
3485         from the C file name without its directory.  This avoids silly things
3486         like "structure GL/GL".
3487         (Other silly names are still possible because ml-nlffigen does not do
3488          a thorough check of whether generated names are legal ML identifiers.
3489          When in doubt, use command line arguments to force particular names.)
3490    
3491    ----------------------------------------------------------------------
3492    Name: Matthias Blume
3493    Date: 2001/06/21 12:25:00 EDT
3494    Tag: blume-20010621-eXene
3495    Description:
3496    
3497    eXene now compiles and (sort of) works again.
3498    
3499    The library name (for version > 110.33) is $/eXene.cm.
3500    
3501    I also added an new example in src/eXene/examples/nbody.  See the
3502    README file there for details.
3503    
3504    ----------------------------------------------------------------------
3505    Name: Matthias Blume
3506    Date: 2001/06/20 16:40:00 EDT
3507    Tag: blume-20010620-cml
3508    Description:
3509    
3510    CML now compiles and works again.
3511    
3512    Libraries (for version > 110.33):
3513    
3514      $cml/cml.cm            Main CML library.
3515      $cml/basis.cm          CML's version of $/basis.cm.
3516      $cml/cml-internal.cm   Internal helper library.
3517      $cml/core-cml.cm       Internal helper library.
3518      $cml-lib/trace-cml.cm  Tracing facility.
3519      $cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
3520    
3521    The installer (config/install.sh) has been taught how to properly
3522    install this stuff.
3523    
3524    ----------------------------------------------------------------------
3525    Name: Matthias Blume
3526    Date: 2001/06/19 17:55:00 EDT
3527    Tag: blume-20010619-instantiate
3528    Description:
3529    
3530    This un-breaks the fix for bug 1432.
3531    (The bug was originally fixed in 110.9 but I broke it again some
3532    time after that.)
3533    
3534    ----------------------------------------------------------------------
3535    Name: Matthias Blume
3536    Date: 2001/06/19 17:25:00 EDT
3537    Tag: blume-20010619-signals
3538    Description:
3539    
3540    This should (hopefully) fix the long-standing signal handling bug.
3541    (The runtime system was constructing a continuation record with an
3542    incorrect descriptor which would cause the GC to drop data on the floor...)
3543    
3544    ----------------------------------------------------------------------
3545    Name: Matthias Blume
3546    Date: 2001/06/15 15:05:00 EDT
3547    Tag: blume-20010615-moresparc
3548    Description:
3549    
3550    Here is a short late-hour update related to Sparc c-calls:
3551    
3552     -- made handling of double-word arguments a bit smarter
3553    
3554     -- instruction selection phase tries to collapse certain clumsily
3555        constructed ML-Trees; typical example:
3556    
3557            ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
3558    
3559        This currently has no further impact on SML/NJ since mlriscGen does
3560        not seem to generate such patterns in the first place, and c-calls
3561        (which did generate them in the beginning) has meanwhile been fixed
3562        so as to avoid them as well.
3563    
3564    ----------------------------------------------------------------------
3565    Name: Matthias Blume
3566    Date: 2001/06/15 15:05:00 EDT
3567    Tag: blume-20010615-sparc
3568    Description:
3569    
3570    The purpose of this update is to provide an implementation of NLFFI
3571    on Sparc machines.
3572    
3573    Here are the changes in detail:
3574    
3575       * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
3576       the Sparc implementation of the c-calls API.
3577       * The Sparc backend of SML/NJ has been modified to uniformely use %fp
3578       for accessing the ML frame.  Thus, we have a real frame pointer and
3579       can freely modify %sp without need for an omit-frame-ptr phase.
3580       The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
3581       this case.
3582       * ml-nlffigen has been taught to produce code for different architectures
3583       and calling conventions.
3584       * In a way similar to what was done in the x86 case, the Sparc
3585       backend uses its own specific extension to mltree.  (For example,
3586       it needs to be able to generate UNIMP instructions which are part
3587       of the calling convention.)
3588       * ml-nlffi-lib was reorganized to make it more modular (in particular,
3589       to make it easier to plug in new machine- and os-dependent parts).
3590    
3591    There are some other fairly unrelated bug fixes and cleanups as well:
3592    
3593       * I further hacked the .cm files for MLRISC tools (like MDLGen) so
3594       that they properly share their libraries with existing SML/NJ libraries.
3595       * I fixed a minor cosmetic bug in CM, supressing certain spurious
3596       follow-up error messages.
3597       * Updates to CM/CMB documentation.
3598    
3599    TODO items:
3600    
3601       * MLRISC should use a different register as its asmTemp on the Sparc.
3602         (The current %o2 is a really bad choice because it is part of the
3603         calling conventions, so things might interfere in unexpected ways.)
3604    
3605    ----------------------------------------------------------------------
3606    Name: Matthias Blume
3607    Date: 2001/06/07
3608    Tag: blume-20010607-calls
3609    Description:
3610    
3611    A number of internal changes related to C calls and calling conventions:
3612