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 1078, Tue Feb 19 21:26:48 2002 UTC revision 1507, Thu Jun 17 23:16:43 2004 UTC
# Line 13  Line 13 
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume (blume (at) tti - c (dot) org)
17    Date: 2004/06/17 18:15:00 CDT
18    Tag: Release_110_46
19    Description:
20    
21    New working version (110.46).  NEW BOOTFILES!
22    
23    ----------------------------------------------------------------------
24    Name: Matthias Blume (blume (at) tti - c (dot) org)
25    Date: 2004/06/17 17:20:00 CDT
26    Tag: blume-20040617-timer-unix
27    Description:
28    
29    Changed the interface of structures Timer and Unix to match the most
30    recent Basis spec.
31    
32    In the case of Unix there still seems to be an open/weird issue:
33    
34       The {text,bin}{In,Out}streamOf functions are supposed to create
35       fresh streams whenever they are called -- as opposed to have them
36       return the same stream every time.  This design is supposed to
37       prevent space leaks caused by proc values hanging on to streams.
38    
39       The reap function, on the other hand, is supposed to close the
40       streams.  This cannot be done without having a handle on the
41       stream in proc after all...
42    
43    I took the liberty to implement the following stopgap solution:
44    
45      The proc value hangs on to the most recently created stream(s).
46      Reap closes those.  If either or both of the two streams hadn't
47      been created at all yet, then reap will close the corresponding
48      file descriptors directly.
49    
50    PS: I don't understand the original space leak argument anymore.  If
51    a proc hangs on to the imperative stream, then I/O operations on those
52    will advance the state of the cached stream and avoid the space leak.
53    
54    ----------------------------------------------------------------------
55    Name: Matthias Blume (blume (at) tti - c (dot) org)
56    Date: 2004/05/28 16:45:00 CDT
57    Tag: blume-20040528-basis
58    Description:
59    
60    Added signature PACK_REAL and exported functor PrimIO.
61    
62    ----------------------------------------------------------------------
63    Name: Matthias Blume (blume (at) tti - c (dot) org)
64    Date: 2004/05/25 16:00:00 CDT
65    Tag: blume-20040525-group-owner
66    Description:
67    
68    CM now ignores (but still accepts) the "owner" information in group
69    descriptions.  The owner of a group is its next enclosing
70    library. Each group must have a unique owner.  (There is a virtual
71    "toplevel" library that owns groups which are not nested within a real
72    library.)  Previously, each group had to explicitly declare its owner,
73    and CM would check that such a declaration is correct.  The new scheme
74    is to have CM check that for each group there is precisely one owning
75    library.
76    
77    The advantage of the new scheme is that the programmer no longer needs
78    to maintain the somewhat annoying owner information.  The downside is
79    that CM cannot enforce the ownership rule across multiple runs of
80    CM.make.  Fortunately, enclosing the same group in two different
81    libraries A and B which are not part of the same program does not
82    cause real problems.
83    
84    ----------------------------------------------------------------------
85    Name: Matthias Blume (blume (at) tti - c (dot) org)
86    Date: 2004/05/20 16:00:00 CDT
87    Tag: blume-20040520-win32
88    Description:
89    
90    Made the win32 version work again.  (Strangely, a misplaced comma had
91    slipped into win32-process.c which prevented the runtime from being
92    compiled correctly.)
93    
94    Also, included a minor addition to ml-build.bat analogous to what was
95    done in blume-20040519-ml-build.
96    
97    ----------------------------------------------------------------------
98    Name: Matthias Blume (blume (at) tti - c (dot) org)
99    Date: 2004/05/19 22:10:00 CDT
100    Tag: blume-20040519-ml-build
101    Description:
102    
103    Arranged for ml-build to clean up after itself a little bit better.
104    The script generates a temporary SML source file and compiles it using
105    CM, so CM generates metadata (GUID, SKEL, objectfile) for it.  It now
106    gets rid of those at the end, so they don't accumulate under .cm.
107    
108    This required a minor change to install.sh because the name of the
109    metadata directory (default: .cm) is actually configurable at
110    installation time.
111    
112    ----------------------------------------------------------------------
113    Name: Matthias Blume (blume (at) tti - c (dot) org)
114    Date: 2004/05/18 15:50:00 CDT
115    Tag: blume-20040518-mkreader
116    Description:
117    
118    Added Posix.IO.mk{Bin,Text}{Reader,Writer} by lifting their respective
119    implementations from internal modules PosixBinPrimIO and PosixTextPrimIO.
120    
121    ----------------------------------------------------------------------
122    Name: Matthias Blume (blume (at) tti - c (dot) org)
123    Date: 2004/05/11 14:35:00 CDT
124    Tag: blume-20040511-win32sock
125    Description:
126    
127    Added previously missing support for many socket-related functions
128    under win32.  Thanks to David Hansel <hansel@reactive-systems.com>
129    for the voluminous patch!
130    
131    (I have not tested this patch under win32 yet.)
132    
133    Here is David's e-mail:
134    
135    Hi,
136    
137    Attached to this email you find a diff against sml/nj 110.45
138    that will enable socket support under Windows.
139    
140    To apply the patch (using unix or cygwin)
141    1) gunzip runtime.diff.gz
142    2) "cd" into "src/runtime" in the source tree of a fresh
143       110.45 installation.
144    3) patch -p 1 < [your/path/to]runtime.diff
145    
146    The code compiles fine but has NOT yet been extensively tested.
147    I only ran a few tests for basic socket client functionality
148    (which worked fine).  Especially the functions that use ioctl
149    are not tested at all and might not work (see below).
150    
151    I implemented this since we want to move to a newer version of sml/nj
152    but need socket support in order to use it.  This is the first time I
153    even had a look at the sml/nj source,  so please review my changes
154    before making this part of the distribution!  Here are a few issues
155    that I think might be better for someone to solve who is more
156    familiar with the sml/nj source (and socket programming):
157    
158    - getnetbyaddr.c and getnetbyname.c will raise a "not implemented"
159      exception since I could not figure out what the windows equivalent
160      of these functions is
161    
162    - In sockets-osdep.h there are a some #include statements that are
163      only used in a few files that include sockets-osdep.h
164    
165    - In smlnj-sock-lib.c, function init_fn() calls WSAStartup() but
166      does not process its return value since I don't know how to
167      report an error upwards.
168    
169    - It would probably be good to have a call to WSACleanup() when
170      the library is unloaded (if there is such a possibility).
171      Otherwise I think Windows will take care of this automatically
172      when the process finishes.
173    
174    - I used ioctlsocket() as a replacement for ioctl() but I have
175      no idea if that is actually the proper replacement on Windows.
176    
177    - All these issues are marked in the code by "FIXME" comments.
178    
179    We use sml/nj extensively in our products and are quite happy
180    with it.  I hope this contribution will help you.
181    
182    Keep up the good work!
183    
184    David
185    
186    ----------------------------------------------------------------------
187    Name: Matthias Blume (blume (at) tti - c (dot) org)
188    Date: 2004/05/11 14:20:00 CDT
189    Tag: blume-20040511-installml
190    Description:
191    
192    Fixed two bugs in installml script.  (Thanks to Vesa A. Norrman for
193    the patch.)
194    
195    ----------------------------------------------------------------------
196    Name: Matthias Blume (blume (at) tti - c (dot) org)
197    Date: 2004/05/11 14:05:00 CDT
198    Tag: blume-20040511-nlffi-netbsd
199    Description:
200    
201    Added support for nlffi under netbsd.  (Thanks to Vesa A. Norrman for
202    the patch.)
203    
204    ----------------------------------------------------------------------
205    Name: Matthias Blume (blume (at) tti - c (dot) org)
206    Date: 2004/05/11 12:05:00 CDT
207    Tag: blume-20040511-exports
208    Description:
209    
210    As per request by Adam Chlipala <adam@hcoop.net>, extended various
211    export lists in compiler-related .cm-files.
212    
213    ----------------------------------------------------------------------
214    Name: Matthias Blume (blume (at) tti - c (dot) org)
215    Date: 2004/05/11 11:35:00 CDT
216    Tag: blume-20040511-allsource
217    Description:
218    
219    The installer now honors the "src-smlnj" target again, although its meaning
220    has changed from "all sources required for the compiler" to "all sources
221    the installer knows about".  In other words, if you enable "src-smlnj"
222    in the "targets" file, then the installer will pull in sources for
223    everything.  (Notice that this refers to source code only.  Compiled
224    code is still only installed for modules that were requested explicitly
225    or which are required for other modules that were requested explicitly.)
226    
227    ----------------------------------------------------------------------
228    Name: Matthias Blume (blume (at) tti - c (dot) org)
229    Date: 2004/04/23 17:40:00 CDT
230    Tag: blume-20040423-ieee-scan
231    Description:
232    
233    Fixed IEEEReal.scan (and .fromString) so that if there is an overflow
234    in the exponent calculation we get INF or ZERO (depending on the mantissa
235    and the sign of the exponent).
236    
237    ----------------------------------------------------------------------
238    Name: Matthias Blume (blume (at) tti - c (dot) org)
239    Date: 2004/04/23 10:40:00 CDT
240    Tag: blume-20040423-ml-build
241    Description:
242    
243    The ml-build script now terminates with a non-0 status when something
244    goes wrong.
245    
246    ----------------------------------------------------------------------
247    Name: Matthias Blume (blume (at) tti - c (dot) org)
248    Date: 2004/04/22 16:35:00 CDT
249    Tag: blume-20040422-Option
250    Description:
251    
252    Made exception Option to be the same as exception Option.Option
253    (as it should be).
254    
255    ----------------------------------------------------------------------
256    Name: Allen Leung (leunga (at) reservoir (dot) com)
257    Date: 2004/03/19 14:40:00 EST
258    Tag: leunga-20040319-cygwin-nlffi
259    Description:
260    
261    Fixed the runtime so that ml-nlffi-lib runs on the cygwin version
262    of SML/NJ.  The problem is that
263    
264       lib = dlopen(NULL, ...)
265       f   = dlsym(lib, "malloc");
266    
267    does not work on Windows unless we explicitly export symbols
268    such as 'malloc' during linking.  We fixed this by explicitly
269    exporting the required symbols with the magic gcc incantation:
270    
271        -Wl,--export-all cygwin.def
272    
273    where cygwin.def is a file containing all the symbols that
274    we wish to export.
275    
276    I suspect this is a Windows problem and we'll have to
277    do the same (somehow with windows compilers) when
278    we build the native win32 version with the system
279    calls LoadLibrary/GetProcAddress.
280    
281    ----------------------------------------------------------------------
282    Name: Matthias Blume (blume (at) tti - c (dot) org)
283    Date: 2004/03/04 16:35:00 CST
284    Tag: blume-20040304-intinf-fmt
285    Description:
286    
287    Fixed problem with IntInf.fmt (sign would show up on the right instead
288    of on the left for BIN, OCT, and HEX).
289    
290    ----------------------------------------------------------------------
291    Name: Matthias Blume (blume (at) tti - c (dot) org)
292    Date: 2004/03/04 11:25:00 CST
293    Tag: blume-20040304-symlinks
294    Description:
295    
296    Fixed problem with installer script (unix only) where bin/ml-yacc and
297    friends pointed (via symlinks) to absolute locations instead of just
298    .run-sml.  This was reported by Vesa A Norrman.
299    
300    ----------------------------------------------------------------------
301    Name: Matthias Blume (blume (at) tti - c (dot) org)
302    Date: 2004/02/13 14:50:00 CST
303    Tag: Release_110_45
304    Description:
305    
306    New working version (110.45). New bootfiles.
307    
308    ----------------------------------------------------------------------
309    Name: Matthias Blume (blume (at) tti - c (dot) org)
310    Date: 2004/01/26 15:15:15 CST
311    Tag: blume-20040126-toplevel
312    Description:
313    
314    Improved handling of exceptions at the interactive toplevel.
315    
316    ----------------------------------------------------------------------
317    Name: Matthias Blume (blume (at) tti - c (dot) org)
318    Date: 2004/01/26 11:25:00
319    Tag: blume-20040126-app
320    Description:
321    
322    Type of top-level "app" corrected.
323    Added code for setting vp_limitPtrMask to Win32-specific runtime.
324    
325    ----------------------------------------------------------------------
326    Name: Matthias Blume (blume (at) tti - c (dot) org)
327    Date: 2003/11/18 17:10 CST
328    Tag: blume-20031118-basis-fiddle
329    Description:
330    
331    - changed Timer interface to what might become the spec
332    - POSIX_FLAGS -> BIT_FLAGS according to spec
333    - some other minor discrepancies wrt. spec eliminated
334    
335    ----------------------------------------------------------------------
336    Name: Matthias Blume (blume (at) tti - c (dot) org)
337    Date: 2003/11/06 12:00:00 CST
338    Tag: Release_110_44
339    Description:
340    
341    New working version (110.44). New bootfiles.
342    
343    ----------------------------------------------------------------------
344    Name: Matthias Blume (blume (at) tti - c (dot) org)
345    Date: 2003/11/04 11:50:00 CST
346    Tag: blume-20031104-move-libraries
347    Description:
348    
349    Eliminated the "dont_move_libraries" directive in config/targets.
350    (The mechanism was broken and could not be fixed easily.  Moreover,
351    there does not seem to be any reason not to move all libraries into
352    lib during installation.  I originally implemented this directive as a
353    backward-compatibility feature when I first introduced the new CM.
354    Now that things have been stable for a long time and going back to the
355    old CM is not an option, there is no reason to keep it around.)
356    
357    ----------------------------------------------------------------------
358    Name: Matthias Blume (blume (at) tti - c (dot) org)
359    Date: 2003/11/03 16:00:00 CST
360    Tag: blume-20031103-installdir
361    Description:
362    
363    Made installer honor INSTALLDIR variable again.  (Thanks to Chris
364    Richards for pointing out the problem and providing the solution.)
365    
366    ----------------------------------------------------------------------
367    Name: Matthias Blume (blume (at) tti - c (dot) org)
368    Date: 2003/10/01 17:05:00 CDT
369    Tag: blume-20031001-lal-mlrisc
370    Description:
371    
372    MLRISC bug fix from Lal.
373    
374    ----------------------------------------------------------------------
375    Name: Matthias Blume (blume (at) tti - c (dot) org)
376    Date: 2003/09/30 16:10:00 CDT
377    Tag: blume-20030930-primio-bat
378    Description:
379    
380    1. Added openVector, nullRd, and nullWr to PRIM_IO.
381    2. Improved .bat files (for Win32 port) to make things work under Win95.
382       (thanks to Aaron S. Hawley for this one)
383    
384    ----------------------------------------------------------------------
385    Name: Matthias Blume (blume (at) tti - c (dot) org)
386    Date: 2003/09/26 16:05:00 CDT
387    Tag: blume-20030926-wrappriv
388    Description:
389    
390    Added missing wrapper for privilege "primitive" in $smlnj/viscomp/core.cm.
391    
392    ----------------------------------------------------------------------
393    Name: Matthias Blume (blume (at) tti - c (dot) org)
394    Date: 2003/09/26 15:00:00 CDT
395    Tag: blume-20030926-110_43_3
396    Description:
397    
398    - additional cleanup
399    - version number bump, NEW BOOTFILES
400    
401    ----------------------------------------------------------------------
402    Name: Matthias Blume (blume (at) tti - c (dot) org)
403    Date: 2003/09/26 12:00:00 CDT
404    Tag: blume-20030926-ppautoload
405    Description:
406    
407    I modified the read-eval-print loop so that the autoloader gets
408    invoked whenever the prettyprinter tries to look up a symbol that
409    is not currently defined in the toplevel environment but which
410    appears in CM's autoload registry.  As a result, we see far fewer of
411    those ?.Foo.Bar.xxx names in the prettyprinter's output.
412    
413    In addition to this I tried to clean up some pieces of the Basis
414    implementation (e.g., Socket, Word8Array) in order to prevent other
415    instances of these ?.Foo.Bar.xxx names from being printed.
416    
417    The mechanism that picks names for types still needs some work, though.
418    (Right now it seems that if there is a type A.t which is defined to
419    be B.u, but B is unavailable at toplevel, then A.t gets printed as
420    "?.B.u" although the perhaps more sensible solution would be to use
421    "A.t" in this case.  In other words, the prettyprinter should follow
422    a chain of DEFtycs not farther than there are corresponding toplevel
423    names in the current environment.)
424    
425    ----------------------------------------------------------------------
426    Name: Matthias Blume (blume (at) tti - c (dot) org)
427    Date: 2003/09/24 16:31:00 CDT
428    Tag: blume-20030924-installer
429    Description:
430    
431    Another installer tweak:  All the ML code for the installer is now
432    compiled during CMB.make and put into a little library called
433    $smlnj/installer.cm.  The installation then simply invokes
434    
435       sml -m $smlnj/installer.cm
436    
437    and everything happens automagically.
438    
439    Win32:  ML code senses value of environment variable SMLNJ_HOME.
440    Unix:   ML code senses values of environment variables ROOT, CONFIGDIR,
441            and BINDIR.
442    
443    The new scheme guarantees that the ML code responsible for the installation
444    is in sync with the APIs of the main system.  Also, the installer is
445    somewhat faster because the installer script is precompiled.
446    
447    ----------------------------------------------------------------------
448    Name: Matthias Blume (blume (at) tti - c (dot) org)
449    Date: 2003/09/24 15:35:00 CDT
450    Tag: blume-20030924-synsock
451    Description:
452    
453    Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
454    but excludes all non-blocking operations.  Defined SOCKET (in Basis)
455    and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
456    implementations of non-blocking operations from CML's Socket
457    structure.
458    
459    ----------------------------------------------------------------------
460    Name: Matthias Blume (blume (at) tti - c (dot) org)
461    Date: 2003/09/24 15:10:05 CDT
462    Tag: blume-20030924-sockets
463    Description:
464    
465    1. Fixed SOCKET API and implementation to match Basis spec.
466       This required changing the internal representation of sockets to one
467       that remembers (for each socket file descriptor) whether it is currently
468       blocking or non-blocking.  This state is maintained lazily (i.e., a system
469       call is made only if the state actually needs to change).
470    
471    2. OS-specific details of sockets were moved into separate files, thus
472       making it possible to unify the bulk of the socket implementations
473       between Unix and Win32.
474    
475    3. CML's socket API changed accordingly.
476       (Note that we need to remove non-blocking functions from this API
477       since they are redundant in the case of CML!)
478    
479    4. CML's socket implementation now makes use of non-blocking functions
480       provided by Basis, thus removing all OS-dependent code from this part
481       of CML.
482    
483    5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
484    
485    ----------------------------------------------------------------------
486    Name: Matthias Blume (blume (at) tti - c (dot) org)
487    Date: 2003/09/22 12:10:00 CDT
488    Tag: blume-20030922-110_43_2
489    Description:
490    
491    Made a new interim version and bootfiles for developer's bootstrapping
492    convenience.
493    
494       110.43.2 -- NEW BOOTFILES
495    
496    ----------------------------------------------------------------------
497    Name: Matthias Blume (blume (at) tti - c (dot) org)
498    Date: 2003/09/19 15:55:00 CDT
499    Tag: blume-20030919-cmdir
500    Description:
501    
502    1. new-install.sh -> install.sh
503    2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
504    3. tweaked installer so that another name instead of .cm can be chosen
505       at install time (by setting the CM_DIR_ARC environment variable
506       during installation); once installation is complete, the name is
507       fixed
508    
509    ----------------------------------------------------------------------
510    Name: Matthias Blume (blume (at) tti - c (dot) org)
511    Date: 2003/09/18 16:00:00 CDT
512    Tag: blume-20030918-110_43_1
513    Description:
514    
515    Made a new interim version and bootfiles for developer's bootstrapping
516    convenience.
517    
518       110.43.1 -- NEW BOOTFILES
519    
520    ----------------------------------------------------------------------
521    Name: Matthias Blume (blume (at) tti - c (dot) org)
522    Date: 2003/09/18 15:20:00 CDT
523    Tag: blume-20030918-misc
524    Description:
525    
526    1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
527       this seems to be controversial at the moment) and used those in
528       Posix.ProcEnv.times.
529    
530    2. Added Time.{from,to}Nanoseconds to Time.
531    
532    3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
533       For example, fromLargeInt never needs to look at more than 3 "big
534       digits" to get its 53 bits of precision.
535    
536    ----------------------------------------------------------------------
537    Name: Matthias Blume (blume (at) tti - c (dot) org)
538    Date: 2003/09/17 16:30:00 CDT
539    Tag: blume-20030917-real32-slices
540    Description:
541    
542    Added an entry to the primitive environment
543    (compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
544    added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
545    
546    Removed some of the "magic" constants in real64.sml and replaced them
547    with code that generates these values from their corresponding
548    integer counterparts.
549    
550    Made all(?) the slice-related changes to the Basis and made everything
551    compile again...
552    
553    ----------------------------------------------------------------------
554    Name: Matthias Blume (blume (at) tti - c (dot) org)
555    Date: 2003/09/15 17:45:00 CDT
556    Tag: blume-20030915-rbase
557    Description:
558    
559    Fixed bug in Real.fromLargeInt.
560    
561    ----------------------------------------------------------------------
562    Name: Matthias Blume (blume (at) tti - c (dot) org)
563    Date: 2003/09/13 18:11:00 CDT
564    Tag: blume-20030913-libinstall
565    Description:
566    
567    Minor bugfix in config/libinstall (set anchor with path to
568    standalone tool after installing it, otherwise libraries that
569    need ml-lex or ml-yacc won't compile the first time the installer
570    runs).
571    
572    ----------------------------------------------------------------------
573    Name: Matthias Blume (blume (at) tti - c (dot) org)
574    Date: 2003/09/12 11:45:00 CDT
575    Tag: blume-20030912-various
576    Description:
577    
578     - fixed bug in Real.toLargeInt
579     - fixed bug in Posix.ProcEnv.times
580     - changed inputLine functions to return an option
581     - minor installer improvements / bugfixes
582     - changed default @SMLalloc parameter for x86/celeron to 64k
583    
584    ----------------------------------------------------------------------
585    Name: Matthias Blume (blume (at) tti - c (dot) org)
586    Date: 2003/09/09 22:00:00 CDT
587    Tag: Release_110_43
588    Description:
589    
590        New working release 110.43.  New bootfiles.
591    
592    ----------------------------------------------------------------------
593    Name: Matthias Blume (blume (at) tti - c (dot) org)
594    Date: 2003/09/09 19:20:00 CDT
595    Tag: blume-20030909-installer
596    Description:
597    
598    Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
599    Modified config/install.bat to take advantage of it.  Also modified
600    config/install.sh (and called it config/new-install.sh) to take advantage
601    of it on Unix systems.  (The SML code is (supposed to be) platform-
602    independent.)
603    
604    The installer can now install everything under Win32
605    as well as under *nix as long as it compiles.
606    
607    Other changes:
608    
609      - made CML compile again under Win32
610      - made eXene compile under Win32 (by providing a fake structure UnixSock
611        and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
612      - fixed a bug in nowhere: it assumed that type OS.Process.status is the
613        same as type int;  under Win32 it isn't
614      - fixed some slice-related problems in the win32-specific parts of CML
615      - added a functor argument "sameVol" to os-path-fn.sml in the Basis
616        (under Win32, the volume name is case-insensitive, and the
617        OS.Path code compares volume names for equality)
618    
619    ----------------------------------------------------------------------
620    Name: Matthias Blume (blume (at) tti - c (dot) org)
621    Date: 2003/09/08 11:55:00 CDT
622    Tag: blume-20030908-fullpath
623    Description:
624    
625    Made Win32 version of OS.FileSys.fullPath return current directory
626    when given an empty string.  This is what the spec says, and incidentally,
627    CM depends on it.  (CM otherwise goes into an infinite loop in certain
628    cases when presented with the name of a non-existing .cm file.)
629    
630    ----------------------------------------------------------------------
631    Name: Matthias Blume (blume (at) tti - c (dot) org)
632    Date: 2003/09/04 16:30:00 CDT
633    Tag: blume-20030905-slices-etc
634    Description:
635    
636      1. Changed interface to vectors and arrays in Basis to match
637         (draft) Basis spec.
638      2. Added signatures and implementations of slices according to
639         Basis spec.
640      3. Edited source code throughout the system to make it compile again
641         under 1. and 2.  (In some cases code had to be added to have it
642         match the new signatures.)
643      4. MLRISC should be backward-compatible: the copies of the originals
644         of files that needed to change under 3. were retained, the .cm files
645         check the compiler version number and use old versions when
646         appropriate.
647      5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
648         dirstream -> string option (in accordance with Basis spec).
649      6. When generating code that counts lines, ml-lex used function
650         CharVector.foldli, taking advantage of its old interface.
651         This has been replaced with the corresponding code from
652         CharVectorSlice.  (html-lex must be re-lexed!)
653      7. BitArray in smlnj-lib/Util has been extended/modified to match the
654         new MONO_ARRAY signature.  (Do we need BitArraySlice?)
655      8. Removed temporary additions (fromInternal, toInternal) from the
656         (now obsolete) IntInf in smlnj-lib/Util.
657      9. Cleaned up structure Byte.
658     10. Added localOffset, scan, and fromString to Date (according to spec).
659         Cleaned/corrected implementation of Date.
660         (Still need to check for correctness; implement better canonicalizeDate.)
661     11. Added "scan" to signature IEEE_REAL.
662     12. Some improvements to IntInf [in particular: efficiency-hack for
663         mod and rem when second operand is 2 (for parity checks).]
664     13. Changed representation of type Time.time, using a single IntInf.int
665         value counting microseconds.  This considerably simplified the
666         implementation of structure Time.  We now support negative time
667         values; scan and fromString handle signs.
668     14. Functor PrimIO now takes two additional arguments (VectorSlice and
669         ArraySlice).
670    
671    ----------------------------------------------------------------------
672    Name: Matthias Blume (blume (at) tti - c (dot) org)
673    Date: 2003/08/28 17:00:00 CDT
674    Tag: blume-20030828-intinf
675    Description:
676    
677    This is a major update which comes with a version number bump
678    (110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
679    and an implementation of IntInf in the Basis.
680    
681    There are a fairly large number of related changes and updates throughout
682    the system:
683    
684    Basis:
685            - Implemented IntInf.
686            - Made LargeInt a projection of IntInf (by filtering through INTEGER).
687            - Added some missing Real64 operations, most notably Real.toLargeInt.
688            - Added FixedInt as a synonym for Int32.
689    
690    compiler:
691            * Added support for a built-in intinf type.
692              - literals
693              - pattern matching
694              - conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
695              - overloading on literals and operations
696    
697            This required adding a primitive type intinf, some additional
698            primops, and implementations for several non-trivial intinf
699            operations in Core.  (The intinf type is completely abstract
700            to the compiler; all operations get delegated back to the Core.)
701    
702            * Intinf equality is handled by polyequal.  However, the compiler
703            does not print its usual warning in this case (since polyequal
704            is the right thing to do there).
705    
706            * Improved the organization of structure InlineT.
707    
708            * A word about conversion primops:
709            If conversions involving intinf do not cancel out during
710            CPS contract, then the compiler must insert calls to Core functions.
711            Since all core access must be resolved already during the FLINT
712            translate phase, it would be too late a the time of CPS contract
713            to add new Core calls.  For this reason, conversion primops
714            for intinf carry two arguments:  1. the numeric argument that
715            they are supposed to convert, and 2. the Core function that
716            can help with this conversion if necessary.  If CPS contract
717            eliminates a primop, then the associated Core function becomes
718            dead and goes away.  Intinf conversion primops that do not get
719            eliminated by CPS contract get rewritten into calls of their
720            core functions by a separate, new phase.
721    
722    interactive system:
723            - Control.Print.intinfDepth controls max length of intinf constants
724            being printed. (Analogous to Control.Print.stringDepth.)
725            - Cleanup in printutil and pputil: got rid of unused stuff and
726            duplicates; replaced some of the code with code that makes better
727            use of library functionality.
728    
729    CM:
730            Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
731            are no longer silent.
732    
733    CKIT:
734            Fixed mismatched uses of Int32 and LargeInt.  I always decided
735            in favor of LargeInt -- which is now the same as IntInf.
736            CKIT-knowledgable people should check whether this is what's
737            intended and otherwise change things back to using Int32 or
738            FixedInt.
739    
740    Throughout the code:
741            Started using IntInf.int literals and built-in operations
742            (e.g., comparison with 0) where this seems appropriate.
743    
744    
745    ----------------------------------------------------------------------
746    Name: Dave MacQueen (dbm@cs.uchicago.edu)
747    Date: 2003/08/13 11:36:00 CDT
748    Tag: dbm-20030813-mcz-merge1
749    Description:
750    
751    Merging changes from the mcz-branch development branch into trunk.
752    These changes involve replacement of the emulated old prettyprinter
753    interface with direct use of the SML/NJ Lib PP library, and fixing
754    of a couple of bugs (895, 1186) relating to error messages.  A new
755    prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
756    has been added.
757    
758    ----------------------------------------------------------------------
759    Name: Matthias Blume (blume (at) tti - c (dot) org)
760    Date: 2003/08/11 15:45:00 CDT
761    Tag: blume-20030811-windows
762    Description:
763    
764    Version number bumped to 110.42.9.  NEW BOOTFILES!!!
765    
766        http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
767    
768    This patch restores SML/NJ's ability to run under win32.  There are a
769    number of changes, including fixes for several bugs that had gone
770    unnoticed until now:
771    
772      - uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
773        trying to run the win32 version from within cygwin.)
774    
775      - There are a number of simple .bat scripts that substitute for their
776        corresponding Unix shell-scripts.  (See below.)
777    
778      - The internals of ml-build have been modified slightly.  The main
779        difference is that instead of calling ".link-sml" (or link-sml.bat)
780        using OS.Process.system, the ML process delegates this task back
781        to the script.  Otherwise problems arise in mixed environments such
782        as Cygwin where scripts look and work like Unix scripts, but
783        where OS.Process.system cannot run them.
784    
785      - In CM, the srcpath pickler used native pathname syntax -- which
786        is incorrect in the case of cross-compilation.  The new pickle format
787        is independent of platform-specific naming conventions.
788    
789      - Path configuration files (such as lib/pathconfig) can now choose
790        between native and standard syntax.  Placing a line of the form
791    
792            standard!
793    
794        into the file causes all subsequent paths to be interpreted using
795        CM standard pathname syntax (= Unix conventions); a line
796    
797            native!
798    
799        switches back to native style.  This was needed so that
800        path config files can be written portably, see src/system/pathconfig.
801    
802      - Runtime system:
803    
804          - win32-filesys.c:  get_file_time and set_file_time now
805            access modification time, not creation time.
806    
807          - I/O code made aware of new array representation.
808    
809          - Bug fixes in X86.prim.masm.
810    
811       - src/system/makeml made aware of win32. (For use under cygwin
812            and other Unix-environments for windows.)
813    
814       - In Basis, fixed off-by-one error in win32-io.sml (function vecF)
815         which caused BinIO.inputAll to fail consistently.
816    
817    .bat scripts:
818    
819       Windows .bat scripts assume that SMLNJ_HOME is defined.
820    
821       - sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
822         applications (sml, ml-yacc, ml-lex).
823       - ml-build.bat: analogous to ml-build.
824       - config\install.bat:  Analogous to config/install.sh.  This requires
825         that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
826         (nmake etc. must be on the path, and vcvars32 must have been run.)
827         Moreover, sources for ml-lex and ml-yacc need to exist under src,
828         and the bootfile hierarchy must have been unpacked under
829         sml.boot.x86-win32.
830         The script is very primitive and does a poor job at error checking.
831         It only installs the base system, ml-lex, and ml-yacc.  No other
832         libraries are being installed (i.e., you get only those that
833         are part of the compiler.)
834       - link-sml.bat:  analogous to .link-sml, but not currently used
835    
836    Unrelated bug fixes:
837    
838      - ml-nlffigen now exports structures ST_* corresponding to incomplete
839        types.
840      - Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
841        otherwise.)
842    
843    ----------------------------------------------------------------------
844    Name: Dave MacQueen (macqueen@cs.uchicago.edu)
845    Date: 2003/06/17
846    Tag: macqueen-20030617-bug895
847    Description:
848    
849    Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
850    Tag will be used for new development branch (mcz-branch) for
851    use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
852    
853    ----------------------------------------------------------------------
854    Name: Matthias Blume (blume (at) tti - c (dot) org)
855    Date: 2003/05/27 16:55:00 CDT
856    Tag: blume-20030527-polyeq
857    Description:
858    
859    Tried to eliminated most cases of polymorphic equality.
860    
861    ----------------------------------------------------------------------
862    Name: Matthias Blume (blume (at) tti - c (dot) org)
863    Date: 2003/05/21 17:45:00 CDT
864    Tag: blume-20030517-complete
865    Description:
866    
867    Two changes:
868    
869      1. Added a flag for controlling whether non-exhaustive bindings will
870         be treated as errors (default is false).
871      2. Cleaned up the *entire* source tree so that CMB.make goes through
872         without a single non-exhaustive match- or bind warning.
873    
874    ----------------------------------------------------------------------
875    Name: Matthias Blume (blume (at) tti - c (dot) org)
876    Date: 2003/05/17 10:20:00 CDT
877    Tag: blume-20030517-absyn
878    Description:
879    
880    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
881    
882       This mainly affects the quality of error messages.  However, some
883       of the code is now more straightforward than before. (Treatment of
884       the above four constructs in translate.sml is much simpler than
885       the "macro-expansion" that was going on before.  Plus, the mach-
886       compiler no longer gets invoked just to be able to compile an
887       if-expression.)
888    
889    2. The ErrorMsg.Error exception is now caught and absorbed by the
890       interactive loop.
891    
892    ----------------------------------------------------------------------
893    Name: Allen Leung
894    Date: 2003/05/16 13:05:00 CDT
895    Tag: leunga-20030516-cygwin-runtime
896    Description:
897    
898        Ported the runtime system to cygwin, which uses the unix
899    x86-unix bin files.  Missing/buggy features:
900    
901       o getnetbyname, getnetbyaddr: these functions seem to be missing in
902         the Cygwin library.
903       o Ctrl-C handling may be flaky.
904       o Windows system calls and Windows I/O are not supported.
905    
906       A new set of binfiles is located at:
907    
908         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
909    
910       This is only needed for bootstrapping the cygwin version of smlnj.
911       Other x86 versions can use the existing binfiles.
912    
913    ----------------------------------------------------------------------
914    Name: Matthias Blume
915    Date: 2003/04/08 15:42:00 CDT
916    Tag: blume-20030408-listpair
917    Description:
918    
919    1. Added a target 'mlrisc' to installer.
920    
921    2. Added missing elements to structure ListPair.
922    
923    ----------------------------------------------------------------------
924    Name: Allen Leung
925    Date: 2003/01/07 10:40:00 EST
926    Tag: leunga-20030107-int-rem
927    Description:
928    
929    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
930    The arguments to the SUBL instruction were swapped.
931    
932    ----------------------------------------------------------------------
933    Name: Matthias Blume
934    Date: 2002/12/12 16:25:00 EST
935    Tag: blume-20021212-risc-ra
936    Description:
937    
938    Fixed a serious bug in the rewrite code for FP spilling/reloading that
939    sent the RA into an infinite loop when floating point registers get
940    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
941    110.37 and 110.38.)
942    There was another set of potential problems related to the handling of
943    MLRISC annotations (but those did not yet cause real problems, apparently).
944    
945    ----------------------------------------------------------------------
946    Name: Matthias Blume
947    Date: 2002/12/06 22:40:00 EST
948    Tag: blume-20021206-cm-fileid
949    Description:
950    
951    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
952    This fixes the problem of CM getting confused by files that suddenly
953    change their identity (e.g., by getting unlinked and recreated by some
954    text editor such as vi).  There might be a better/cheaper/cleaner way
955    of doing this, but for now this will have to do.
956    
957    ----------------------------------------------------------------------
958    Name: Matthias Blume
959    Date: 2002/10/28 09:50:00 EST
960    Tag: blume-20021028-typecheck
961    Description:
962    
963    Exported structure Typecheck from $smlnj/viscomp/core.cm.
964    
965    ----------------------------------------------------------------------
966    Name: Matthias Blume
967    Date: 2002/10/17 09:10:00 EDT
968    Tag: Release_110_42
969    Description:
970    
971    In good old tradition, there has been a slight hiccup so that we have
972    to patch 110.42 after the fact.   The old release tag has been replaced
973    (see below).
974    
975    The change solves a problem with two competing approaches the
976    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
977    each other's way.
978    
979    This change only affects the runtime system code and the installer script.
980    (No new bootfiles.)
981    
982    ----------------------------------------------------------------------
983    Name: Matthias Blume
984    Date: 2002/10/16 12:00:00 EDT
985    Tag: Release_110_42_removed
986    Description:
987    
988    New working release.  New bootfiles.
989    
990    ----------------------------------------------------------------------
991    Name: Matthias Blume
992    Date: 2002/10/10 13:10:00 EDT
993    Tag: blume-20021010-ppc-divs
994    Description:
995    
996    The mltree operator DIVS must be implemented with an overflow check on
997    the PPC because the hardware indicates divide-by-zero using "overflow" as
998    well.
999    
1000    ----------------------------------------------------------------------
1001    Name: Matthias Blume
1002    Date: 2002/07/23 11:45:00
1003    Tag: blume-20020723-smlnj-home
1004    Description:
1005    
1006    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
1007    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
1008    CM_PATHCONFIG is also set), the path configuration file is assumed
1009    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
1010    the entire tree to some other place and everything will "just work".
1011    
1012    (Companion commands such as ml-build and ml-makedepend also sense this
1013    variable.)
1014    
1015    ----------------------------------------------------------------------
1016    Name: Matthias Blume
1017    Date: 2002/07/12 21:19:00 EDT
1018    Tag: blume-20020712-liveness
1019    Description:
1020    
1021    Exported two useful "step" functions from liveness module (MLRISC).
1022    
1023    ----------------------------------------------------------------------
1024    Name: Matthias Blume
1025    Date: 2002/07/05 16:00 EDT
1026    Tag: Release_110_41
1027    Description:
1028    
1029    New working release.  New bootfiles.
1030    
1031    ----------------------------------------------------------------------
1032    Name: Matthias Blume
1033    Date: 2002/07/05 10:25:00 EDT
1034    Tag: blume-20020705-btimp
1035    Description:
1036    
1037    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
1038    other clients can set up backtracing support.
1039    
1040    ----------------------------------------------------------------------
1041    Name: Matthias Blume
1042    Date: 2002/06/25 17:23:00 EDT
1043    Tag: blume-20020625-fpmax
1044    Description:
1045    
1046    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
1047    case.  (The sense of the isNaN test was reversed -- which made min and
1048    max always return their first argument.)
1049    
1050    ----------------------------------------------------------------------
1051    Name: Matthias Blume
1052    Date: 2002/06/11
1053    Tag: blume-20020611-unixpath
1054    Description:
1055    
1056    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
1057    
1058    ----------------------------------------------------------------------
1059    Name: Matthias Blume
1060    Date: 2002/06/10 16:35:00 EDT
1061    Tag: blume-20020610-ieeereal
1062    Description:
1063    
1064    I back-ported my implementation of IEEEReal.fromString from the
1065    idlbasis-devel branch so that we can test it.
1066    
1067    Another small change is that ppDec tries to give more information
1068    than just "<sig>" in the case of functors.  However, this code is
1069    broken in some mysterious way if the functor's body's signature
1070    has not been declared by ascription but gets inferred from the
1071    implementation.  This needs fixing...
1072    
1073    ----------------------------------------------------------------------
1074    Name: Matthias Blume
1075    Date: 2002/05/31
1076    Tag: blume-20020531-btrace-mode
1077    Description:
1078    
1079    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
1080    the wayside when I switched over to using Controls everywhere.)
1081    
1082    ----------------------------------------------------------------------
1083    Name: Lal George
1084    Date: 2002/05/23 12:21:40 EDT
1085    Tag: george-20020523-visual-labels
1086    Description:
1087    
1088    Labels are now displayed in the graphical output to make
1089    the fall-through and target blocks obvious.
1090    
1091    ----------------------------------------------------------------------
1092    Name: Matthias Blume
1093    Date: 2002/05/22 11:03:00 EDT
1094    Tag: blume-20020522-shrink
1095    Description:
1096    
1097    John tweaked yesterday's fix for 1131 to handle an out-of-memory
1098    situation that comes up when allocating huge arrays.
1099    
1100    ----------------------------------------------------------------------
1101    Name: Matthias Blume
1102    Date: 2002/05/21 16:00:00 EDT
1103    Tag: Release_110_40
1104    Description:
1105    
1106    New working release (110.40).  New bootfiles.
1107    
1108    [Also: John Reppy fixed GC bug 1131.]
1109    
1110    ----------------------------------------------------------------------
1111    Name: Matthias Blume
1112    Date: 2002/05/21 12:35:00 EDT
1113    Tag: blume-20020521-cmdoc
1114    Description:
1115    
1116    CM documentation update.
1117    
1118    ----------------------------------------------------------------------
1119    Name: Matthias Blume
1120    Date: 2002/05/21 10:55:00 EDT
1121    Tag: blume-20020521-misc
1122    Description:
1123    
1124    - John tweaked runtime to be silent on heap export (except when
1125      GC messages are on).
1126    - I added a few more things (cross-compiling versions of CMB) to
1127      config/preloads (as suggestions).
1128    
1129    ----------------------------------------------------------------------
1130    Name: Matthias Blume
1131    Date: 2002/05/20 22:25:00 EDT
1132    Tag: blume-20020520-controls
1133    Description:
1134    
1135    - Added ControlUtil structure to control-lib.cm.
1136    - Use it throughout.
1137    - Used Controls facility to define MLRISC controls (as opposed to
1138      registering MLRISC control ref cells with Controls after the fact)
1139    - Fixed messed-up controls priorities.
1140    
1141    * Removed again all the stuff from config/preloads that one wouldn't
1142      be able to preload at the time the initial heap image is built.
1143      (Many libraries, e.g., CML, do not exist yet at this time.  The
1144       only libraries that can be preloaded via config/preloads are those
1145       that come bundled with the bootfiles.)
1146    
1147    ----------------------------------------------------------------------
1148    Name: Matthias Blume
1149    Date: 2002/05/20 10:59:00 EDT
1150    Tag: blume-20020520-preloads
1151    Description:
1152    
1153    Added a lot of commented-out suggestions for things to be included
1154    in config/preloads.
1155    
1156    ----------------------------------------------------------------------
1157    Name: Allen Leung
1158    Date: 2002/05/18 14:20:00 EDT
1159    Tag: leunga-20020518-mdl
1160    Description:
1161    
1162    o Made the mdl tool stuff compile and run again.
1163    o I've disabled all the stuff that depends on RTL specifications; they
1164      are all badly broken anyway.
1165    
1166    ----------------------------------------------------------------------
1167    Name: Matthias Blume
1168    Date: 2002/05/17 16:49:00 EDT
1169    Tag: blume-20020517-controls
1170    Description:
1171    
1172    0. John Reppy made several modifications to the SML/NJ library.
1173       In particular, there is a shiny new controls-lib.cm.
1174    
1175    1. Pushed new controls interface through compiler so that everything
1176       compiles again.
1177    
1178    2. Added FormatComb and FORMAT_COMB to the CML version of the
1179       SML/NJ library (so that CML compiles again).
1180    
1181    3. Modified init scripts because XXX_DEFAULT environment variables
1182       are no longer with us.  (Boot-time initialization is now done
1183       using the same environment variables that are also used for
1184       startup-time initialization of controls.)
1185    
1186    ----------------------------------------------------------------------
1187    Name: Lal George
1188    Date: 2002/05/15 09:20:10 EDT
1189    Tag: george-20020515-pseudo-op-decls
1190    Description:
1191    
1192            All pseudo-ops emitted before the first segment declaration
1193    such as TEXT, DATA, and BSS directives are assumed to be global
1194    declarations and are emitted first in the assembly file. This is
1195    useful in a number of situations where one has pseudo-ops that are not
1196    specific to any segment, and also works around the constraint that one
1197    cannot have client pseudo-ops in the TEXT segment.
1198    
1199            Because no segment is associated with these declarations it is
1200    an error to allocate any space or objects before the first segment
1201    directive and an exception will be raised. However, we cannot make
1202    this check for client pseudo-ops.
1203    
1204            These top level declarations are a field in the CFG graph_info.
1205    In theory you can continue to add to this field after the CFG has been
1206    built -- provided you know what you are doing;-)
1207    
1208    ----------------------------------------------------------------------
1209    Name: Matthias Blume
1210    Date: 2002/05/13 16:40:00 EDT
1211    Tag: blume-20020513-pp-etc
1212    Description:
1213    
1214    A few minor bugfixes:
1215    
1216      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
1217        (Bogus printouts for redefined bindings still occur.  Compiler
1218        bug should no longer occur now.  We need to redo the prettyprinter
1219        from scratch.)
1220    
1221      - CM pathname printer now also adds escape sequences for ( and )
1222    
1223      - commend and docu fixes for ml-nlffi
1224    
1225    ----------------------------------------------------------------------
1226    Name: Matthias Blume
1227    Date: 2002/05/10 16:40:00 EDT
1228    Tag: blume-20020510-erg-textio
1229    Description:
1230    
1231    Applied the following bugfix provided by Emden Gansner:
1232    
1233        Output is corrupted when outputSubstr is used rather than output.
1234    
1235        The problem occurs when a substring
1236    
1237            ss = (s, dataStart, dataLen)
1238    
1239        where dataStart > 0, fills a stream buffer with avail bytes left.
1240        avail bytes of s, starting at index dataStart, are copied into the
1241        buffer, the buffer is flushed, and then the remaining dataLen-avail
1242        bytes of ss are copied into the beginning of the buffer. Instead of
1243        starting this copy at index dataStart+avail in s, the current code
1244        starts the copy at index avail.
1245    
1246        Fix:
1247        In text-io-fn.sml, change line 695 from
1248             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
1249        to
1250             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
1251    
1252    ----------------------------------------------------------------------
1253    Name: Matthias Blume
1254    Date: 2002/04/12 13:55:00 EDT
1255    Tag: blume-20020412-assyntax
1256    Description:
1257    
1258    1. Grabbed newer assyntax.h from the XFree86 project.
1259    2. Fiddled with how to compile X86.prim.asm without warnings.
1260    3. (Very) Minor cleanup in CM.
1261    
1262    ----------------------------------------------------------------------
1263    Name: Matthias Blume
1264    Date: 2002/04/01 (no joke!) 17:07:00 EST
1265    Tag: blume-20020401-x86div
1266    Description:
1267    
1268    Added full support for div/mod/rem/quot on the x86, using the machine
1269    instruction's two results (without clumsily recomputing the remainder)
1270    directly where appropriate.
1271    
1272    Some more extensive power-of-two support was added to the x86 instruction
1273    selector (avoiding expensive divs, mods, and muls where they can be
1274    replaced with cheaper shifts and masks).  However, this sort of thing
1275    ought to be done earlier, e.g., within the CPS optimizer so that
1276    all architectures benefit from it.
1277    
1278    The compiler compiles to a fixed point, but changes might be somewhat
1279    fragile nevertheless.  Please, report any strange things that you might
1280    see wrt. div/mod/quot/rem...
1281    
1282    ----------------------------------------------------------------------
1283    Name: Matthias Blume
1284    Date: 2002/03/29 17:22:00
1285    Tag: blume-20020329-div
1286    Description:
1287    
1288    Fixed my broken div/mod logic.  Unfortunately, this means that the
1289    inline code for div/mod now has one more comparison than before.
1290    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
1291    The problem was with quotient = 0, because that alone does not tell
1292    us which way the rounding went.  One then has to look at whether
1293    remainder and divisor have the same sign...  :(
1294    
1295    Anyway, I replaced the bootfiles with fresh ones...
1296    
1297    ----------------------------------------------------------------------
1298    Name: Matthias Blume
1299    Date: 2002/03/29 14:10:00 EST
1300    Tag: blume-20020329-inlprims
1301    Description:
1302    
1303    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
1304    
1305    Primops have changed. This means that the bin/boot-file formats have
1306    changed as well.
1307    
1308    To make sure that there is no confusion, I made a new version.
1309    
1310    
1311    CHANGES:
1312    
1313    * removed REMT from mltree (remainder should never overflow).
1314    
1315    * added primops to deal with divisions of all flavors to the frontend
1316    
1317    * handled these primops all the way through so they map to their respective
1318      MLRISC support
1319    
1320    * used these primops in the implementation of Int, Int32, Word, Word32
1321    
1322    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
1323    
1324    * parameterized INLMIN, INLMAX, and INLABS by a numkind
1325    
1326    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
1327      floating point
1328    
1329    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
1330      and Real (but Real.abs maps to a separate floating-point-only primop)
1331    
1332    
1333    TODO items:
1334    
1335    * Hacked Alpha32 instruction selection, disabling the selection of REMx
1336      instructions because the machine instruction encoder cannot handle
1337      them.  (Hppa, PPC, and Sparc instruction selection did not handle
1338      REM in the first place, and REM is supported by the x86 machine coder.)
1339    
1340    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
1341      selection phase.  (The two can be streamlined because the hardware
1342      delivers both quotient and remainder at the same time anyway.)
1343    
1344    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
1345      (Currently the behavior is inconsistent both across architectures and
1346      wrt. the draft Basis spec.)
1347    
1348    * Word8 should eventually be handled natively, too.
1349    
1350    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
1351      as if there currently is no execution path that could trigger it in
1352      SML/NJ.  (The assumptions underlying functions arith and promotable do not
1353      hold for things like multiplication and division.)
1354    
1355    ----------------------------------------------------------------------
1356    Name: Matthias Blume
1357    Date: 2002/03/27 16:27:00 EST
1358    Tag: blume-20020327-mlrisc-divisions
1359    Description:
1360    
1361    Added support for all four division operations (ML's div, mod, quot,
1362    and rem) to MLRISC.  In the course of doing so, I also rationalized
1363    the naming (no more annoying switch-around of DIV and QUOT), by
1364    parameterizing the operation by div_rounding_mode (which can be either
1365    DIV_TO_ZERO or DIV_TO_NEGINF).
1366    
1367    The generic MLTreeGen functor takes care of compiling all four
1368    operations down to only round-to-zero div.
1369    
1370    Missing pieces:
1371    
1372      * Doing something smarter than relying on MLTreeGen on architectures
1373        like, e.g., the x86 where hardware division delivers both quotient and
1374        remainder at the same time.  With this, the implementation of the
1375        round-to-neginf operations could be further streamlined.
1376    
1377      * Remove inlining support for div/mod/rem from the frontend and replace it
1378        with primops that get carried through to the backend.  Do this for all
1379        int and word types.
1380    
1381    ----------------------------------------------------------------------
1382    Name: Matthias Blume
1383    Date: 2002/03/25 17:25:00 EST
1384    Tag: blume-20020325-divmod
1385    Description:
1386    
1387    I improved (hopefully without breaking them) the implementation of Int.div,
1388    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
1389    advantage of the following observations:
1390    
1391      Let  q = x quot y      r = x rem y
1392           d = x div  y      m = x mod y
1393    
1394    where "quot" is the round-to-zero version of integer division that
1395    hardware usually provides.  Then we have:
1396    
1397         r = x - q * y        where neither the * nor the - will overflow
1398         d = if q >= 0 orelse x = q * y then q else q - 1
1399                              where neither the * nor the - will overflow
1400         m = if q >= 0 orelse r = 0 then r else r + y
1401                              where the + will not overflow
1402    
1403    This results in substantial simplification of the generated code.
1404    The following table shows the number of CFG nodes and edges generated
1405    for
1406            fun f (x, y) = x OPER y
1407            (* with OPER \in div, mod, quot, rem *)
1408    
1409    
1410        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
1411        --------------------------------------------------------
1412         div |         24 |         39 |         12 |         16
1413         mod |         41 |         71 |         12 |         16
1414        quot |          8 |         10 |          8 |         10
1415         rem |         10 |         14 |          8 |         10
1416    
1417    
1418    ----------------------------------------------------------------------
1419    Name: Matthias Blume
1420    Date: 2002/03/25 22:06:00 EST
1421    Tag: blume-20020325-cprotobug
1422    Description:
1423    
1424    Fixed a bug in cproto (c prototype decoder).
1425    
1426    ----------------------------------------------------------------------
1427    Name: Matthias Blume
1428    Date: 2002/03/25 16:00:00 EST
1429    Tag: blume-20020325-raw-primops
1430    Description:
1431    
1432    I did some cleanup to Allen's new primop code and
1433    replaced yesterday's bootfiles with new ones.
1434    (But they are stored in the same place.)
1435    
1436    ----------------------------------------------------------------------
1437    Name: Matthias Blume
1438    Date: 2002/03/24 22:40:00 EST
1439    Tag: blume-20020324-bootfiles
1440    Description:
1441    
1442    Made the bootfiles that Allen asked for.
1443    
1444    ----------------------------------------------------------------------
1445    Name: Allen Leung
1446    Date: 2002/03/23 15:50:00 EST
1447    Tag: leunga-20020323-flint-cps-rcc-primops
1448    Description:
1449    
1450      1. Changes to FLINT primops:
1451    
1452        (* make a call to a C-function;
1453         * The primop carries C function prototype information and specifies
1454         * which of its (ML-) arguments are floating point. C prototype
1455         * information is for use by the backend, ML information is for
1456         * use by the CPS converter. *)
1457      | RAW_CCALL of { c_proto: CTypes.c_proto,
1458                       ml_args: ccall_type list,
1459                       ml_res_opt: ccall_type option,
1460                       reentrant : bool
1461                     } option
1462       (* Allocate uninitialized storage on the heap.
1463        * The record is meant to hold short-lived C objects, i.e., they
1464        * are not ML pointers.  With the tag, the representation is
1465        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
1466        *)
1467      | RAW_RECORD of {tag:bool,sz:int}
1468      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
1469    
1470      2.  These CPS primops are now overloaded:
1471    
1472           rawload of {kind:numkind}
1473           rawstore of {kind:numkind}
1474    
1475          The one argument form is:
1476    
1477             rawload {kind} address
1478    
1479          The two argument form is:
1480    
1481             rawload {kind} [ml object, byte-offset]
1482    
1483      3. RAW_CCALL/RCC now takes two extra arguments:
1484    
1485         a. The first is whether the C call is reentrant, i.e., whether
1486            ML state should be saved and restored.
1487         b. The second argument is a string argument specifying the name of
1488            library and the C function.
1489    
1490         These things are currently not handled in the code generator, yet.
1491    
1492      4. In CProto,
1493    
1494         An encoding type of "bool" means "ml object" and is mapped into
1495         C prototype of PTR.  Note that "bool" is different than "string",
1496         even though "string" is also mapped into PTR, because "bool"
1497         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
1498    
1499      5. Pickler/unpicker
1500    
1501         Changed to handle RAW_RECORD and newest RAW_CCALL
1502    
1503      6. MLRiscGen,
1504    
1505         1. Changed to handle the new rawload/rawstore/rawrecord operators.
1506         2. Code for handling C Calls has been moved to a new module CPSCCalls,
1507            in the file CodeGen/cpscompile/cps-c-calls.sml
1508    
1509      7. Added the conditional move operator
1510    
1511             condmove of branch
1512    
1513         to cps.  Generation of this is still buggy so it is currently
1514         disabled.
1515    
1516    ----------------------------------------------------------------------
1517    Name: Lal George
1518    Date: 2002/03/22 14:18:25 EST
1519    Tag: george-20020322-cps-branch-prob
1520    Description:
1521    
1522    Implemented the Ball-Larus branch prediction-heuristics, and
1523    incorporated graphical viewers for control flow graphs.
1524    
1525    Ball-Larus Heuristics:
1526    ---------------------
1527    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1528    
1529    By design it uses the Dempster-Shafer theory for combining
1530    probabilities.  For example, in the function:
1531    
1532        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1533    
1534    the ball-larus heuristics predicts that the n=0 is unlikely
1535    (OH-heuristic), and the 'then' branch is unlikely because of the
1536    RH-heuristic -- giving the 'then' branch an even lower combined
1537    probability using the Dempster-Shafer theory.
1538    
1539    Finally, John Reppy's loop analysis in MLRISC, further lowers the
1540    probability of the 'then' branch because of the loop in the else
1541    branch.
1542    
1543    
1544    Graphical Viewing:
1545    ------------------
1546    I merely plugged in Allen's graphical viewers into the compiler. The
1547    additional code is not much. At the top level, saying:
1548    
1549            Control.MLRISC.getFlag "cfg-graphical-view" := true;
1550    
1551    will display the graphical view of the control flow graph just before
1552    back-patching.  daVinci must be in your path for this to work. If
1553    daVinci is not available, then the default viewer can be changed
1554    using:
1555    
1556            Control.MLRISC.getString "viewer"
1557    
1558    which can be set to "dot" or "vcg" for the corresponding viewers. Of
1559    course, these viewers must be in your path.
1560    
1561    The above will display the compilation unit at the level of clusters,
1562    many of which are small, boring, and un-interesting. Also setting:
1563    
1564            Control.MLRISC.getInt "cfg-graphical-view_size"
1565    
1566    will display clusters that are larger than the value set by the above.
1567    
1568    
1569    ----------------------------------------------------------------------
1570    Name: Matthias Blume
1571    Date: 2002/03/21 22:20:00 EST
1572    Tag: blume-20020321-kmp-bugfix
1573    Description:
1574    
1575    Changed the interface to the KMP routine in PreString and fixed
1576    a minor bug in one place where it was used.
1577    
1578    ----------------------------------------------------------------------
1579    Name: Allen Leung
1580    Date: 2002/03/21 20:30:00 EST
1581    Tag: leunga-20020321-cfg
1582    Description:
1583    
1584      Fixed a potential problem in cfg edge splitting.
1585    
1586    ----------------------------------------------------------------------
1587    Name: Allen Leung
1588    Date: 2002/03/21 17:15:00 EST
1589    Tag: leunga-20020321-x86-fp-cfg
1590    Description:
1591    
1592      1. Recoded the buggy parts of x86-fp.
1593    
1594         a. All the block reordering code has been removed.
1595            We now depend on the block placement phases to do this work.
1596    
1597         b. Critical edge splitting code has been simplified and moved into the
1598            CFG modules, as where they belong.
1599    
1600         Both of these were quite buggy and complex.  The code is now much, much
1601         simpler.
1602    
1603      2. X86 backend.
1604    
1605         a. Added instructions for 64-bit support.  Instruction selection for
1606            64-bit has not been committed, however, since that
1607            requires changes to MLTREE which haven't been approved by
1608            Lal and John.
1609    
1610         b. Added support for FUCOMI and FUCOMIP when generating code for
1611            PentiumPro and above.  We only generate these instructions in
1612            the fast-fp mode.
1613    
1614         c. Added cases for JP and JNP in X86FreqProps.
1615    
1616      3. CFG
1617    
1618         CFG now has a bunch of methods for edge splitting and merging.
1619    
1620      4. Machine description.
1621    
1622         John's simplification of MLTREE_BASIS.fcond broke a few machine
1623         description things:
1624    
1625         rtl-build.{sig,sml} and hppa.mdl fixed.
1626    
1627         NOTE: the machine description stuff in the repository is still broken.
1628               Again, I can't put my fixes in because that involves
1629               changes to MLTREE.
1630    
1631    ----------------------------------------------------------------------
1632    Name: Matthias Blume
1633    Date: 2002/03/20 15:55:00 EST
1634    Tag: blume-20020320-kmp
1635    Description:
1636    
1637    Implemented Knuth-Morris-Pratt string matching in PreString and used
1638    it for String.isSubstring, Substring.isSubstring, and
1639    Substring.position.
1640    
1641    (Might need some stress-testing.  Simple examples worked fine.)
1642    
1643    ----------------------------------------------------------------------
1644    Name: Matthias Blume
1645    Date: 2002/03/19 16:37:00 EST
1646    Tag: blume-20020319-witnesses
1647    Description:
1648    
1649    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1650    
1651    This implements a generic mechanism for changing constness qualifiers
1652    anywhere within big C types without resorting to outright "casts".
1653    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1654    modify the constness at the outermost level.)
1655    The implementation of "convert" is based on the idea of "witness"
1656    values -- values that are not used by the operation but whose types
1657    "testify" to their applicability.  On the implementation side, "convert"
1658    is simply a projection (returning its second curried argument).  With
1659    cross-module inlining, it should not result in any machine code being
1660    generated.
1661    
1662    ----------------------------------------------------------------------
1663    Name: Matthias Blume
1664    Date: 2002/03/15 16:40:00 EST
1665    Tag: blume-20020315-basis
1666    Description:
1667    
1668    Provided (preliminary?) implementations for
1669    
1670      {String,Substring}.{concatWith,isSuffix,isSubstring}
1671    
1672    and
1673    
1674      Substring.full
1675    
1676    Those are in the Basis spec but they were missing in SML/NJ.
1677    
1678    ----------------------------------------------------------------------
1679    Name: Matthias Blume
1680    Date: 2002/03/14 21:30:00 EST
1681    Tag: blume-20020314-controls
1682    Description:
1683    
1684    Controls:
1685    ---------
1686    
1687    1. Factored out the recently-added Controls : CONTROLS stuff and put
1688       it into its own library $/controls-lib.cm.  The source tree for
1689       this is under src/smlnj-lib/Controls.
1690    
1691    2. Changed the names of types and functions in this interface, so they
1692       make a bit more "sense":
1693    
1694          module -> registry
1695          'a registry -> 'a group
1696    
1697    3. The interface now deals in ref cells only.  The getter/setter interface
1698       is (mostly) gone.
1699    
1700    4. Added a function that lets one register an already-existing ref cell.
1701    
1702    5. Made the corresponding modifications to the rest of the code so that
1703       everything compiles again.
1704    
1705    6. Changed the implementation of Controls.MLRISC back to something closer
1706       to the original.  In particular, this module (and therefore MLRISC)
1707       does not depend on Controls.  There now is some link-time code in
1708       int-sys.sml that registers the MLRISC controls with the Controls
1709       module.
1710    
1711    CM:
1712    ---
1713    
1714      * One can now specify the lambda-split aggressiveness in init.cmi.
1715    
1716    ----------------------------------------------------------------------
1717    Name: Allen Leung
1718    Date: 2002/03/13 17:30:00 EST
1719    Tag: leunga-20020313-x86-fp-unary
1720    Description:
1721    
1722    Bug fix for:
1723    
1724    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1725    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1726    > - fun f(x,(y,z)) = Real.~ y;
1727    > [autoloading]
1728    > [autoloading done]
1729    >       fchsl   (%eax), 184(%esp)
1730    > Error: MLRisc bug: X86MCEmitter.emitInstr
1731    >
1732    > uncaught exception Error
1733    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1734    
1735    The problem was that the code generator did not generate any fp registers
1736    in this case, and the ra didn't know that it needed to run the X86FP phase to
1737    translate the pseudo fp instruction.   This only happened with unary fp
1738    operators in certain situations.
1739    
1740    ----------------------------------------------------------------------
1741    Name: Matthias Blume
1742    Date: 2002/03/13 14:00:00 EST
1743    Tag: blume-20020313-overload-etc
1744    Description:
1745    
1746    1. Added _overload as a synonym for overload for backward compatibility.
1747       (Control.overloadKW must be true for either version to be accepted.)
1748    
1749    2. Fixed bug in install script that caused more things to be installed
1750       than what was requested in config/targets.
1751    
1752    3. Made CM aware of the (_)overload construct so that autoloading
1753       works.
1754    
1755    ----------------------------------------------------------------------
1756    Name: Matthias Blume
1757    Date: 2002/03/12 22:03:00 EST
1758    Tag: blume-20020312-url
1759    Description:
1760    
1761    Forgot to update BOOT and srcarchiveurl.
1762    
1763    ----------------------------------------------------------------------
1764    Name: Matthias Blume
1765    Date: 2002/03/12 17:30:00 EST
1766    Tag: blume-20020312-version110392
1767    Description:
1768    
1769    Yet another version number bump (because of small changes to the
1770    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1771    
1772    Changes:
1773    
1774      The new pid generation scheme described a few weeks ago was overly
1775      complicated.  I implemented a new mechanism that is simpler and
1776      provides a bit more "stability":  Once CM has seen a compilation
1777      unit, it keeps its identity constant (as long as you do not delete
1778      those crucial CM/GUID/* files).  This means that when you change
1779      an interface, compile, then go back to the old interface, and
1780      compile again, you arrive at the original pid.
1781    
1782      There now also is a mechanism that instructs CM to use the plain
1783      environment hash as a module's pid (effectively making its GUID
1784      the empty string).  For this, "noguid" must be specified as an
1785      option to the .sml file in question within its .cm file.
1786      This is most useful for code that is being generated by tools such
1787      as ml-nlffigen (because during development programmers tend to
1788      erase the tool's entire output directory tree including CM's cached
1789      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1790      revert to the old, broken behavior of SML/NJ, but in specific cases
1791      where there is no danger of interface confusion, its use is ok
1792      (I think).
1793    
1794      ml-nlffigen by default generates "noguid" annotations.  They can be
1795      turned off by specifying -guid in its command line.
1796    
1797    ----------------------------------------------------------------------
1798    Name: Lal George
1799    Date: 2002/03/12 12 14:42:36 EST
1800    Tag: george-20020312-frequency-computation
1801    Description:
1802    
1803    Integrated jump chaining and static block frequency into the
1804    compiler. More details and numbers later.
1805    
1806    ----------------------------------------------------------------------
1807    Name: Lal George
1808    Date: 2002/03/11 11 22:38:53 EST
1809    Tag: george-20020311-jump-chain-elim
1810    Description:
1811    
1812    Tested the jump chain elimination on all architectures (except the
1813    hppa).  This is on by default right now and is profitable for the
1814    alpha and x86, however, it may not be profitable for the sparc and ppc
1815    when compiling the compiler.
1816    
1817    The gc test will typically jump to a label at the end of the cluster,
1818    where there is another jump to an external cluster containing the actual
1819    code to invoke gc. This is to allow factoring of common gc invocation
1820    sequences. That is to say, we generate:
1821    
1822            f:
1823               testgc
1824               ja   L1      % jump if above to L1
1825    
1826            L1:
1827               jmp L2
1828    
1829    
1830    After jump chain elimination the 'ja L1' instructions is converted to
1831    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1832    up being implemented in their long form (if L2 is far away) using:
1833    
1834            jbe     L3      % jump if below or equal to L3
1835            jmp     L2
1836         L3:
1837            ...
1838    
1839    
1840    For large compilation units L2  may be far away.
1841    
1842    
1843    ----------------------------------------------------------------------
1844    Name: Matthias Blume
1845    Date: 2002/03/11 13:30:00 EST
1846    Tag: blume-20020311-mltreeeval
1847    Description:
1848    
1849    A functor parameter was missing.
1850    
1851    ----------------------------------------------------------------------
1852    Name: Allen Leung
1853    Date: 2002/03/11 10:30:00 EST
1854    Tag: leunga-20020311-runtime-string0
1855    Description:
1856    
1857       The representation of the empty string now points to a
1858    legal null terminated C string instead of unit.  It is now possible
1859    to convert an ML string into C string with InlineT.CharVector.getData.
1860    This compiles into one single machine instruction.
1861    
1862    ----------------------------------------------------------------------
1863    Name: Allen Leung
1864    Date: 2002/03/10 23:55:00 EST
1865    Tag: leunga-20020310-x86-call
1866    Description:
1867    
1868       Added machine generation for CALL instruction (relative displacement mode)
1869    
1870    ----------------------------------------------------------------------
1871    Name: Matthias Blume
1872    Date: 2002/03/08 16:05:00
1873    Tag: blume-20020308-entrypoints
1874    Description:
1875    
1876    Version number bumped to 110.39.1.  NEW BOOTFILES!
1877    
1878    Entrypoints: non-zero offset into a code object where execution should begin.
1879    
1880    - Added the notion of an entrypoint to CodeObj.
1881    - Added reading/writing of entrypoint info to Binfile.
1882    - Made runtime system bootloader aware of entrypoints.
1883    - Use the address of the label of the first function given to mlriscGen
1884      as the entrypoint.  This address is currently always 0, but it will
1885      not be 0 once we turn on block placement.
1886    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1887      with entry points) from mlriscGen.
1888    
1889    ----------------------------------------------------------------------
1890    Name: Allen Leung
1891    Date: 2002/03/07 20:45:00 EST
1892    Tag: leunga-20020307-x86-cmov
1893    Description:
1894    
1895       Bug fixes for CMOVcc on x86.
1896    
1897       1. Added machine code generation for CMOVcc
1898       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1899       3. CMOVcc cannot have an immediate operand as argument.
1900    
1901    ----------------------------------------------------------------------
1902    Name: Matthias Blume
1903    Date: 2002/03/07 16:15:00 EST
1904    Tag: blume-20020307-controls
1905    Description:
1906    
1907    This is a very large but mostly boring patch which makes (almost)
1908    every tuneable compiler knob (i.e., pretty much everything under
1909    Control.* plus a few other things) configurable via both the command
1910    line and environment variables in the style CM did its configuration
1911    until now.
1912    
1913    Try starting sml with '-h' (or, if you are brave, '-H')
1914    
1915    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1916    implements the underlying generic mechanism.
1917    
1918    The interface to some of the existing such facilities has changed somewhat.
1919    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1920    (The getFoo interface is still there for backward-compatibility, but its
1921    use is deprecated.)
1922    
1923    The ml-build script passes -Cxxx=yyy command-line arguments through so
1924    that one can now twiddle the compiler settings when using this "batch"
1925    compiler.
1926    
1927    TODO items:
1928    
1929    We should go through and throw out all controls that are no longer
1930    connected to anything.  Moreover, we should go through and provide
1931    meaningful (and correct!) documentation strings for those controls
1932    that still are connected.
1933    
1934    Currently, multiple calls to Controls.new are accepted (only the first
1935    has any effect).  Eventually we should make sure that every control
1936    is being made (via Controls.new) exactly once.  Future access can then
1937    be done using Controls.acc.
1938    
1939    Finally, it would probably be a good idea to use the getter-setter
1940    interface to controls rather than ref cells.  For the time being, both
1941    styles are provided by the Controls module, but getter-setter pairs are
1942    better if thread-safety is of any concern because they can be wrapped.
1943    
1944    *****************************************
1945    
1946    One bug fix: The function blockPlacement in three of the MLRISC
1947    backpatch files used to be hard-wired to one of two possibilities at
1948    link time (according to the value of the placementFlag).  But (I
1949    think) it should rather sense the flag every time.
1950    
1951    *****************************************
1952    
1953    Other assorted changes (by other people who did not supply a HISTORY entry):
1954    
1955    1. the cross-module inliner now works much better (Monnier)
1956    2. representation of weights, frequencies, and probabilities in MLRISC
1957       changed in preparation of using those for weighted block placement
1958       (Reppy, George)
1959    
1960    ----------------------------------------------------------------------
1961    Name: Lal George
1962    Date: 2002/03/07 14:44:24 EST 2002
1963    Tag: george-20020307-weighted-block-placement
1964    
1965    Tested the weighted block placement optimization on all architectures
1966    (except the hppa) using AMPL to generate the block and edge frequencies.
1967    Changes were required in the machine properties to correctly
1968    categorize trap instructions. There is an MLRISC flag
1969    "weighted-block-placement" that can be used to enable weighted block
1970    placement, but this will be ineffective without block/edge
1971    frequencies (coming soon).
1972    
1973    
1974    ----------------------------------------------------------------------
1975    Name: Lal George
1976    Date: 2002/03/05 17:24:48 EST
1977    Tag: george-20020305-linkage-cluster
1978    
1979    In order to support the block placement optimization, a new cluster
1980    is generated as the very first cluster (called the linkage cluster).
1981    It contains a single jump to the 'real' entry point for the compilation
1982    unit. Block placement has no effect on the linkage cluster itself, but
1983    all the other clusters  have full freedom in the manner in which they
1984    reorder blocks or functions.
1985    
1986    On the x86 the typical linkage code that is generated is:
1987       ----------------------
1988            .align 2
1989       L0:
1990            addl    $L1-L0, 72(%esp)
1991            jmp     L1
1992    
1993    
1994            .align  2
1995       L1:
1996       ----------------------
1997    
1998    72(%esp) is the memory location for the stdlink register. This
1999    must contain the address of the CPS function being called. In the
2000    above example, it contains the address of  L0; before
2001    calling L1 (the real entry point for the compilation unit), it
2002    must contain the address for L1, and hence
2003    
2004            addl $L1-L0, 72(%esp)
2005    
2006    I have tested this on all architectures except the hppa.The increase
2007    in code size is of course negligible
2008    
2009    ----------------------------------------------------------------------
2010    Name: Allen Leung
2011    Date: 2002/03/03 13:20:00 EST
2012    Tag: leunga-20020303-mlrisc-tools
2013    
2014      Added #[ ... ] expressions to mlrisc tools
2015    
2016    ----------------------------------------------------------------------
2017    Name: Matthias Blume
2018    Date: 2002/02/27 12:29:00 EST
2019    Tag: blume-20020227-cdebug
2020    Description:
2021    
2022    - made types in structure C and C_Debug to be equal
2023    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
2024    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
2025    
2026    ----------------------------------------------------------------------
2027    Name: Matthias Blume
2028    Date: 2002/02/26 12:00:00 EST
2029    Tag: blume-20020226-ffi
2030    Description:
2031    
2032    1. Fixed a minor bug in CM's "noweb" tool:
2033       If numbering is turned off, then truly don't number (i.e., do not
2034       supply the -L option to noweb).  The previous behavior was to supply
2035       -L'' -- which caused noweb to use the "default" line numbering scheme.
2036       Thanks to Chris Richards for pointing this out (and supplying the fix).
2037    
2038    2. Once again, I reworked some aspects of the FFI:
2039    
2040       A. The incomplete/complete type business:
2041    
2042       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
2043         gone!
2044       - ML types representing an incomplete type are now *equal* to
2045         ML types representing their corresponding complete types (just like
2046         in C).  This is still safe because ml-nlffigen will not generate
2047         RTTI for incomplete types, nor will it generate functions that
2048         require access to such RTTI.   But when ML code generated from both
2049         incomplete and complete versions of the C type meet, the ML types
2050         are trivially interoperable.
2051    
2052         NOTE:  These changes restore the full generality of the translation
2053         (which was previously lost when I eliminated functorization)!
2054    
2055       B. Enum types:
2056    
2057       - Structure C now has a type constructor "enum" that is similar to
2058         how the "su" constructor works.  However, "enum" is not a phantom
2059         type because each "T enum" has values (and is isomorphic to
2060         MLRep.Signed.int).
2061       - There are generic access operations for enum objects (using
2062         MLRep.Signed.int).
2063       - ml-nlffigen will generate a structure E_foo for each "enum foo".
2064         * The structure contains the definition of type "mlrep" (the ML-side
2065         representation type of the enum).  Normally, mlrep is the same
2066         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
2067         then mlrep will be defined as a datatype -- thus facilitating
2068         pattern matching on mlrep values.
2069         ("-ec" will be suppressed if there are duplicate values in an
2070          enumeration.)
2071         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
2072         will be generated for each C enum constant xxx.
2073         * Conversion functions m2i and i2m convert between mlrep and
2074         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
2075         * Coversion functions c and ml convert between mlrep and "tag enum".
2076         * Access functions (get/set) fetch and store mlrep values.
2077       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
2078         enumerations are merged into one single enumeration represented by
2079         structure E_'.
2080    
2081    ----------------------------------------------------------------------
2082    Name: Allen Leung
2083    Date: 2002/02/25 04:45:00 EST
2084    Tag: leunga-20020225-cps-spill
2085    
2086    This is a new implementation of the CPS spill phase.
2087    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
2088    In case of problems, replace it with the old file spill.sml
2089    
2090    The current compiler runs into some serious performance problems when
2091    constructing a large record.  This can happen when we try to compile a
2092    structure with many items.  Even a very simple structure like the following
2093    makes the compiler slow down.
2094    
2095        structure Foo = struct
2096           val x_1 = 0w1 : Word32.int
2097           val x_2 = 0w2 : Word32.int
2098           val x_3 = 0w3 : Word32.int
2099           ...
2100           val x_N = 0wN : Word32.int
2101        end
2102    
2103    The following table shows the compile time, from N=1000 to N=4000,
2104    with the old compiler:
2105    
2106    N
2107    1000   CPS 100 spill                           0.04u  0.00s  0.00g
2108           MLRISC ra                               0.06u  0.00s  0.05g
2109              (spills = 0 reloads = 0)
2110           TOTAL                                   0.63u  0.07s  0.21g
2111    
2112    1100   CPS 100 spill                           8.25u  0.32s  0.64g
2113           MLRISC ra                               5.68u  0.59s  3.93g
2114              (spills = 0 reloads = 0)
2115           TOTAL                                   14.71u  0.99s  4.81g
2116    
2117    1500   CPS 100 spill                           58.55u  2.34s  1.74g
2118           MLRISC ra                               5.54u  0.65s  3.91g
2119              (spills = 543 reloads = 1082)
2120           TOTAL                                   65.40u  3.13s  6.00g
2121    
2122    2000   CPS 100 spill                           126.69u  4.84s  3.08g
2123           MLRISC ra                               0.80u  0.10s  0.55g
2124              (spills = 42 reloads = 84)
2125           TOTAL                                   129.42u  5.10s  4.13g
2126    
2127    3000   CPS 100 spill                           675.59u  19.03s  11.64g
2128           MLRISC ra                               2.69u  0.27s  1.38g
2129              (spills = 62 reloads = 124)
2130           TOTAL                                   682.48u  19.61s  13.99g
2131    
2132    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
2133           MLRISC ra                               4.96u  0.27s  2.72g
2134              (spills = 85 reloads = 170)
2135           TOTAL                                   2375.26u  57.21s  48.00g
2136    
2137    As you can see the old cps spill module suffers from some serious
2138    performance problem.  But since I cannot decipher the old code fully,
2139    instead of patching the problems up, I'm reimplementing it
2140    with a different algorithm.  The new code is more modular,
2141    smaller when compiled, and substantially faster
2142    (O(n log n) time and O(n) space).  Timing of the new spill module:
2143    
2144    4000  CPS 100 spill                           0.02u  0.00s  0.00g
2145          MLRISC ra                               0.25u  0.02s  0.15g
2146             (spills=1 reloads=3)
2147          TOTAL                                   7.74u  0.34s  1.62g
2148    
2149    Implementation details:
2150    
2151    As far as I can tell, the purpose of the CPS spill module is to make sure the
2152    number of live variables at any program point (the bandwidth)
2153    does not exceed a certain limit, which is determined by the
2154    size of the spill area.
2155    
2156    When the bandwidth is too large, we decrease the register pressure by
2157    packing live variables into spill records.  How we achieve this is
2158    completely different than what we did in the old code.
2159    
2160    First, there is something about the MLRiscGen code generator
2161    that we should be aware of:
2162    
2163    o MLRiscGen performs code motion!
2164    
2165       In particular, it will move floating point computations and
2166       address computations involving only the heap pointer to
2167       their use sites (if there is only a single use).
2168       What this means is that if we have a CPS record construction
2169       statement
2170    
2171           RECORD(k,vl,w,e)
2172    
2173       we should never count the new record address w as live if w
2174       has only one use (which is often the case).
2175    
2176       We should do something similar to floating point, but the transformation
2177       there is much more complex, so I won't deal with that.
2178    
2179    Secondly, there are now two new cps primops at our disposal:
2180    
2181     1. rawrecord of record_kind option
2182        This pure operator allocates some uninitialized storage from the heap.
2183        There are two forms:
2184    
2185         rawrecord NONE [INT n]  allocates a tagless record of length n
2186         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
2187                                     and initializes the tag.
2188    
2189     2. rawupdate of cty
2190          rawupdate cty (v,i,x)
2191          Assigns to x to the ith component of record v.
2192          The storelist is not updated.
2193    
2194    We use these new primops for both spilling and increment record construction.
2195    
2196     1. Spilling.
2197    
2198        This is implemented with a linear scan algorithm (but generalized
2199        to trees).  The algorithm will create a single spill record at the
2200        beginning of the cps function and use rawupdate to spill to it,
2201        and SELECT or SELp to reload from it.  So both spills and reloads
2202        are fine-grain operations.  In contrast, in the old algorithm
2203        "spills" have to be bundled together in records.
2204    
2205        Ideally, we should sink the spill record construction to where
2206        it is needed.  We can even split the spill record into multiple ones
2207        at the places where they are needed.  But CPS is not a good
2208        representation for global code motion, so I'll keep it simple and
2209        am not attempting this.
2210    
2211     2. Incremental record construction (aka record splitting).
2212    
2213        Long records with many component values which are simulatenously live
2214        (recall that single use record addresses are not considered to
2215         be live) are constructed with rawrecord and rawupdate.
2216        We allocate space on the heap with rawrecord first, then gradually
2217        fill it in with rawupdate.  This is the technique suggested to me
2218        by Matthias.
2219    
2220        Some restrictions on when this is applicable:
2221        1. It is not a VECTOR record.  The code generator currently does not handle
2222           this case. VECTOR record uses double indirection like arrays.
2223        2. All the record component values are defined in the same "basic block"
2224           as the record constructor.  This is to prevent speculative
2225           record construction.
2226    
2227    ----------------------------------------------------------------------
2228    Name: Allen Leung
2229    Date: 2002/02/22 01:02:00 EST
2230    Tag: leunga-20020222-mlrisc-tools
2231    
2232    Minor bug fixes in the parser and rewriter
2233    
2234    ----------------------------------------------------------------------
2235    Name: Allen Leung
2236    Date: 2002/02/21 20:20:00 EST
2237    Tag: leunga-20020221-peephole
2238    
2239    Regenerated the peephole files.  Some contained typos in the specification
2240    and some didn't compile because of pretty printing bugs in the old version
2241    of 'nowhere'.
2242    
2243    ----------------------------------------------------------------------
2244    Name: Allen Leung
2245    Date: 2002/02/19 20:20:00 EST
2246    Tag: leunga-20020219-mlrisc-tools
2247    Description:
2248    
2249       Minor bug fixes to the mlrisc-tools library:
2250    
2251       1.  Fixed up parsing colon suffixed keywords
2252       2.  Added the ability to shut the error messages up
2253       3.  Reimplemented the pretty printer and fixed up/improved
2254           the pretty printing of handle and -> types.
2255       4.  Fixed up generation of literal symbols in the nowhere tool.
2256       5.  Added some SML keywords to to sml.sty
2257    
2258    ----------------------------------------------------------------------
2259  Name: Matthias Blume  Name: Matthias Blume
2260  Date: 2002/02/19 16:20:00 EST  Date: 2002/02/19 16:20:00 EST
2261  Tag: blume-20020219-cmffi  Tag: blume-20020219-cmffi

Legend:
Removed from v.1078  
changed lines
  Added in v.1507

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