Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 634, Mon May 1 10:03:50 2000 UTC	revision 1390, Wed Sep 24 20:34:06 2003 UTC
#	Line 8	Line 8
8	The form of an entry should be:	The form of an entry should be:
9
10	Name:	Name:
11	Date:	Date: yyyy/mm/dd
12	Tag: <post-commit CVS tag>	Tag: <post-commit CVS tag>
13	Description:	Description:
14
15		----------------------------------------------------------------------
16		Name: Matthias Blume (blume (at) tti - c (dot) org)
17		Date: 2003/09/24 15:35:00 CDT
18		Tag: blume-20030924-synsock
19		Description:
20
21		Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
22		but excludes all non-blocking operations.  Defined SOCKET (in Basis)
23		and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
24		implementations of non-blocking operations from CML's Socket
25		structure.
26
27		----------------------------------------------------------------------
28		Name: Matthias Blume (blume (at) tti - c (dot) org)
29		Date: 2003/09/24 15:10:05 CDT
30		Tag: blume-20030924-sockets
31		Description:
32
33		1. Fixed SOCKET API and implementation to match Basis spec.
34		This required changing the internal representation of sockets to one
35		that remembers (for each socket file descriptor) whether it is currently
36		blocking or non-blocking.  This state is maintained lazily (i.e., a system
37		call is made only if the state actually needs to change).
38
39		2. OS-specific details of sockets were moved into separate files, thus
40		making it possible to unify the bulk of the socket implementations
41		between Unix and Win32.
42
43		3. CML's socket API changed accordingly.
44		(Note that we need to remove non-blocking functions from this API
45		since they are redundant in the case of CML!)
46
47		4. CML's socket implementation now makes use of non-blocking functions
48		provided by Basis, thus removing all OS-dependent code from this part
49		of CML.
50
51		5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
52
53		----------------------------------------------------------------------
54		Name: Matthias Blume (blume (at) tti - c (dot) org)
55		Date: 2003/09/22 12:10:00 CDT
56		Tag: blume-20030922-110_43_2
57		Description:
58
59		Made a new interim version and bootfiles for developer's bootstrapping
60		convenience.
61
62		110.43.2 -- NEW BOOTFILES
63
64		----------------------------------------------------------------------
65		Name: Matthias Blume (blume (at) tti - c (dot) org)
66		Date: 2003/09/19 15:55:00 CDT
67		Tag: blume-20030919-cmdir
68		Description:
69
70		1. new-install.sh -> install.sh
71		2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
72		3. tweaked installer so that another name instead of .cm can be chosen
73		at install time (by setting the CM_DIR_ARC environment variable
74		during installation); once installation is complete, the name is
75		fixed
76
77		----------------------------------------------------------------------
78		Name: Matthias Blume (blume (at) tti - c (dot) org)
79		Date: 2003/09/18 16:00:00 CDT
80		Tag: blume-20030918-110_43_1
81		Description:
82
83		Made a new interim version and bootfiles for developer's bootstrapping
84		convenience.
85
86		110.43.1 -- NEW BOOTFILES
87
88		----------------------------------------------------------------------
89		Name: Matthias Blume (blume (at) tti - c (dot) org)
90		Date: 2003/09/18 15:20:00 CDT
91		Tag: blume-20030918-misc
92		Description:
93
94		1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
95		this seems to be controversial at the moment) and used those in
96		Posix.ProcEnv.times.
97
98		2. Added Time.{from,to}Nanoseconds to Time.
99
100		3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
101		For example, fromLargeInt never needs to look at more than 3 "big
102		digits" to get its 53 bits of precision.
103
104		----------------------------------------------------------------------
105		Name: Matthias Blume (blume (at) tti - c (dot) org)
106		Date: 2003/09/17 16:30:00 CDT
107		Tag: blume-20030917-real32-slices
108		Description:
109
110		Added an entry to the primitive environment
111		(compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
112		added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
113
114		Removed some of the "magic" constants in real64.sml and replaced them
115		with code that generates these values from their corresponding
116		integer counterparts.
117
118		Made all(?) the slice-related changes to the Basis and made everything
119		compile again...
120
121		----------------------------------------------------------------------
122		Name: Matthias Blume (blume (at) tti - c (dot) org)
123		Date: 2003/09/15 17:45:00 CDT
124		Tag: blume-20030915-rbase
125		Description:
126
127		Fixed bug in Real.fromLargeInt.
128
129		----------------------------------------------------------------------
130		Name: Matthias Blume (blume (at) tti - c (dot) org)
131		Date: 2003/09/13 18:11:00 CDT
132		Tag: blume-20030913-libinstall
133		Description:
134
135		Minor bugfix in config/libinstall (set anchor with path to
136		standalone tool after installing it, otherwise libraries that
137		need ml-lex or ml-yacc won't compile the first time the installer
138		runs).
139
140		----------------------------------------------------------------------
141		Name: Matthias Blume (blume (at) tti - c (dot) org)
142		Date: 2003/09/12 11:45:00 CDT
143		Tag: blume-20030912-various
144		Description:
145
146		- fixed bug in Real.toLargeInt
147		- fixed bug in Posix.ProcEnv.times
148		- changed inputLine functions to return an option
149		- minor installer improvements / bugfixes
150		- changed default @SMLalloc parameter for x86/celeron to 64k
151
152		----------------------------------------------------------------------
153		Name: Matthias Blume (blume (at) tti - c (dot) org)
154		Date: 2003/09/09 22:00:00 CDT
155		Tag: Release_110_43
156		Description:
157
158		New working release 110.43.  New bootfiles.
159
160		----------------------------------------------------------------------
161		Name: Matthias Blume (blume (at) tti - c (dot) org)
162		Date: 2003/09/09 19:20:00 CDT
163		Tag: blume-20030909-installer
164		Description:
165
166		Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
167		Modified config/install.bat to take advantage of it.  Also modified
168		config/install.sh (and called it config/new-install.sh) to take advantage
169		of it on Unix systems.  (The SML code is (supposed to be) platform-
170		independent.)
171
172		The installer can now install everything under Win32
173		as well as under *nix as long as it compiles.
174
175		Other changes:
176
177		- made CML compile again under Win32
178		- made eXene compile under Win32 (by providing a fake structure UnixSock
179		and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
180		- fixed a bug in nowhere: it assumed that type OS.Process.status is the
181		same as type int;  under Win32 it isn't
182		- fixed some slice-related problems in the win32-specific parts of CML
183		- added a functor argument "sameVol" to os-path-fn.sml in the Basis
184		(under Win32, the volume name is case-insensitive, and the
185		OS.Path code compares volume names for equality)
186
187		----------------------------------------------------------------------
188		Name: Matthias Blume (blume (at) tti - c (dot) org)
189		Date: 2003/09/08 11:55:00 CDT
190		Tag: blume-20030908-fullpath
191		Description:
192
193		Made Win32 version of OS.FileSys.fullPath return current directory
194		when given an empty string.  This is what the spec says, and incidentally,
195		CM depends on it.  (CM otherwise goes into an infinite loop in certain
196		cases when presented with the name of a non-existing .cm file.)
197
198		----------------------------------------------------------------------
199		Name: Matthias Blume (blume (at) tti - c (dot) org)
200		Date: 2003/09/04 16:30:00 CDT
201		Tag: blume-20030905-slices-etc
202		Description:
203
204		1. Changed interface to vectors and arrays in Basis to match
205		(draft) Basis spec.
206		2. Added signatures and implementations of slices according to
207		Basis spec.
208		3. Edited source code throughout the system to make it compile again
209		under 1. and 2.  (In some cases code had to be added to have it
210		match the new signatures.)
211		4. MLRISC should be backward-compatible: the copies of the originals
212		of files that needed to change under 3. were retained, the .cm files
213		check the compiler version number and use old versions when
214		appropriate.
215		5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
216		dirstream -> string option (in accordance with Basis spec).
217		6. When generating code that counts lines, ml-lex used function
218		CharVector.foldli, taking advantage of its old interface.
219		This has been replaced with the corresponding code from
220		CharVectorSlice.  (html-lex must be re-lexed!)
221		7. BitArray in smlnj-lib/Util has been extended/modified to match the
222		new MONO_ARRAY signature.  (Do we need BitArraySlice?)
223		8. Removed temporary additions (fromInternal, toInternal) from the
224		(now obsolete) IntInf in smlnj-lib/Util.
225		9. Cleaned up structure Byte.
226		10. Added localOffset, scan, and fromString to Date (according to spec).
227		Cleaned/corrected implementation of Date.
228		(Still need to check for correctness; implement better canonicalizeDate.)
229		11. Added "scan" to signature IEEE_REAL.
230		12. Some improvements to IntInf [in particular: efficiency-hack for
231		mod and rem when second operand is 2 (for parity checks).]
232		13. Changed representation of type Time.time, using a single IntInf.int
233		value counting microseconds.  This considerably simplified the
234		implementation of structure Time.  We now support negative time
235		values; scan and fromString handle signs.
236		14. Functor PrimIO now takes two additional arguments (VectorSlice and
237		ArraySlice).
238
239		----------------------------------------------------------------------
240		Name: Matthias Blume (blume (at) tti - c (dot) org)
241		Date: 2003/08/28 17:00:00 CDT
242		Tag: blume-20030828-intinf
243		Description:
244
245		This is a major update which comes with a version number bump
246		(110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
247		and an implementation of IntInf in the Basis.
248
249		There are a fairly large number of related changes and updates throughout
250		the system:
251
252		Basis:
253		- Implemented IntInf.
254		- Made LargeInt a projection of IntInf (by filtering through INTEGER).
255		- Added some missing Real64 operations, most notably Real.toLargeInt.
256		- Added FixedInt as a synonym for Int32.
257
258		compiler:
259		* Added support for a built-in intinf type.
260		- literals
261		- pattern matching
262		- conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
263		- overloading on literals and operations
264
265		This required adding a primitive type intinf, some additional
266		primops, and implementations for several non-trivial intinf
267		operations in Core.  (The intinf type is completely abstract
268		to the compiler; all operations get delegated back to the Core.)
269
270		* Intinf equality is handled by polyequal.  However, the compiler
271		does not print its usual warning in this case (since polyequal
272		is the right thing to do there).
273
274		* Improved the organization of structure InlineT.
275
276		* A word about conversion primops:
277		If conversions involving intinf do not cancel out during
278		CPS contract, then the compiler must insert calls to Core functions.
279		Since all core access must be resolved already during the FLINT
280		translate phase, it would be too late a the time of CPS contract
281		to add new Core calls.  For this reason, conversion primops
282		for intinf carry two arguments:  1. the numeric argument that
283		they are supposed to convert, and 2. the Core function that
284		can help with this conversion if necessary.  If CPS contract
285		eliminates a primop, then the associated Core function becomes
286		dead and goes away.  Intinf conversion primops that do not get
287		eliminated by CPS contract get rewritten into calls of their
288		core functions by a separate, new phase.
289
290		interactive system:
291		- Control.Print.intinfDepth controls max length of intinf constants
292		being printed. (Analogous to Control.Print.stringDepth.)
293		- Cleanup in printutil and pputil: got rid of unused stuff and
294		duplicates; replaced some of the code with code that makes better
295		use of library functionality.
296
297		CM:
298		Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
299		are no longer silent.
300
301		CKIT:
302		Fixed mismatched uses of Int32 and LargeInt.  I always decided
303		in favor of LargeInt -- which is now the same as IntInf.
304		CKIT-knowledgable people should check whether this is what's
305		intended and otherwise change things back to using Int32 or
306		FixedInt.
307
308		Throughout the code:
309		Started using IntInf.int literals and built-in operations
310		(e.g., comparison with 0) where this seems appropriate.
311
312
313		----------------------------------------------------------------------
314		Name: Dave MacQueen (dbm@cs.uchicago.edu)
315		Date: 2003/08/13 11:36:00 CDT
316		Tag: dbm-20030813-mcz-merge1
317		Description:
318
319		Merging changes from the mcz-branch development branch into trunk.
320		These changes involve replacement of the emulated old prettyprinter
321		interface with direct use of the SML/NJ Lib PP library, and fixing
322		of a couple of bugs (895, 1186) relating to error messages.  A new
323		prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
324		has been added.
325
326		----------------------------------------------------------------------
327		Name: Matthias Blume (blume (at) tti - c (dot) org)
328		Date: 2003/08/11 15:45:00 CDT
329		Tag: blume-20030811-windows
330		Description:
331
332		Version number bumped to 110.42.9.  NEW BOOTFILES!!!
333
334		http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
335
336		This patch restores SML/NJ's ability to run under win32.  There are a
337		number of changes, including fixes for several bugs that had gone
338		unnoticed until now:
339
340		- uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
341		trying to run the win32 version from within cygwin.)
342
343		- There are a number of simple .bat scripts that substitute for their
344		corresponding Unix shell-scripts.  (See below.)
345
346		- The internals of ml-build have been modified slightly.  The main
347		difference is that instead of calling ".link-sml" (or link-sml.bat)
348		using OS.Process.system, the ML process delegates this task back
349		to the script.  Otherwise problems arise in mixed environments such
350		as Cygwin where scripts look and work like Unix scripts, but
351		where OS.Process.system cannot run them.
352
353		- In CM, the srcpath pickler used native pathname syntax -- which
354		is incorrect in the case of cross-compilation.  The new pickle format
355		is independent of platform-specific naming conventions.
356
357		- Path configuration files (such as lib/pathconfig) can now choose
358		between native and standard syntax.  Placing a line of the form
359
360		standard!
361
362		into the file causes all subsequent paths to be interpreted using
363		CM standard pathname syntax (= Unix conventions); a line
364
365		native!
366
367		switches back to native style.  This was needed so that
368		path config files can be written portably, see src/system/pathconfig.
369
370		- Runtime system:
371
372		- win32-filesys.c:  get_file_time and set_file_time now
373		access modification time, not creation time.
374
375		- I/O code made aware of new array representation.
376
377		- Bug fixes in X86.prim.masm.
378
379		- src/system/makeml made aware of win32. (For use under cygwin
380		and other Unix-environments for windows.)
381
382		- In Basis, fixed off-by-one error in win32-io.sml (function vecF)
383		which caused BinIO.inputAll to fail consistently.
384
385		.bat scripts:
386
387		Windows .bat scripts assume that SMLNJ_HOME is defined.
388
389		- sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
390		applications (sml, ml-yacc, ml-lex).
391		- ml-build.bat: analogous to ml-build.
392		- config\install.bat:  Analogous to config/install.sh.  This requires
393		that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
394		(nmake etc. must be on the path, and vcvars32 must have been run.)
395		Moreover, sources for ml-lex and ml-yacc need to exist under src,
396		and the bootfile hierarchy must have been unpacked under
397		sml.boot.x86-win32.
398		The script is very primitive and does a poor job at error checking.
399		It only installs the base system, ml-lex, and ml-yacc.  No other
400		libraries are being installed (i.e., you get only those that
401		are part of the compiler.)
402		- link-sml.bat:  analogous to .link-sml, but not currently used
403
404		Unrelated bug fixes:
405
406		- ml-nlffigen now exports structures ST_* corresponding to incomplete
407		types.
408		- Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
409		otherwise.)
410
411		----------------------------------------------------------------------
412		Name: Dave MacQueen (macqueen@cs.uchicago.edu)
413		Date: 2003/06/17
414		Tag: macqueen-20030617-bug895
415		Description:
416
417		Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
418		Tag will be used for new development branch (mcz-branch) for
419		use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
420
421		----------------------------------------------------------------------
422		Name: Matthias Blume (blume (at) tti - c (dot) org)
423		Date: 2003/05/27 16:55:00 CDT
424		Tag: blume-20030527-polyeq
425		Description:
426
427		Tried to eliminated most cases of polymorphic equality.
428
429		----------------------------------------------------------------------
430		Name: Matthias Blume (blume (at) tti - c (dot) org)
431		Date: 2003/05/21 17:45:00 CDT
432		Tag: blume-20030517-complete
433		Description:
434
435		Two changes:
436
437		1. Added a flag for controlling whether non-exhaustive bindings will
438		be treated as errors (default is false).
439		2. Cleaned up the *entire* source tree so that CMB.make goes through
440		without a single non-exhaustive match- or bind warning.
441
442		----------------------------------------------------------------------
443		Name: Matthias Blume (blume (at) tti - c (dot) org)
444		Date: 2003/05/17 10:20:00 CDT
445		Tag: blume-20030517-absyn
446		Description:
447
448		1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
449
450		This mainly affects the quality of error messages.  However, some
451		of the code is now more straightforward than before. (Treatment of
452		the above four constructs in translate.sml is much simpler than
453		the "macro-expansion" that was going on before.  Plus, the mach-
454		compiler no longer gets invoked just to be able to compile an
455		if-expression.)
456
457		2. The ErrorMsg.Error exception is now caught and absorbed by the
458		interactive loop.
459
460		----------------------------------------------------------------------
461		Name: Allen Leung
462		Date: 2003/05/16 13:05:00 CDT
463		Tag: leunga-20030516-cygwin-runtime
464		Description:
465
466		Ported the runtime system to cygwin, which uses the unix
467		x86-unix bin files.  Missing/buggy features:
468
469		o getnetbyname, getnetbyaddr: these functions seem to be missing in
470		the Cygwin library.
471		o Ctrl-C handling may be flaky.
472		o Windows system calls and Windows I/O are not supported.
473
474		A new set of binfiles is located at:
475
476		http://www.dorsai.org/~leunga/boot.x86-unix.tgz
477
478		This is only needed for bootstrapping the cygwin version of smlnj.
479		Other x86 versions can use the existing binfiles.
480
481		----------------------------------------------------------------------
482		Name: Matthias Blume
483		Date: 2003/04/08 15:42:00 CDT
484		Tag: blume-20030408-listpair
485		Description:
486
487		1. Added a target 'mlrisc' to installer.
488
489		2. Added missing elements to structure ListPair.
490
491		----------------------------------------------------------------------
492		Name: Allen Leung
493		Date: 2003/01/07 10:40:00 EST
494		Tag: leunga-20030107-int-rem
495		Description:
496
497		Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
498		The arguments to the SUBL instruction were swapped.
499
500		----------------------------------------------------------------------
501		Name: Matthias Blume
502		Date: 2002/12/12 16:25:00 EST
503		Tag: blume-20021212-risc-ra
504		Description:
505
506		Fixed a serious bug in the rewrite code for FP spilling/reloading that
507		sent the RA into an infinite loop when floating point registers get
508		spilled.  (Because of this bug, e.g., nucleic stopped compiling between
509		110.37 and 110.38.)
510		There was another set of potential problems related to the handling of
511		MLRISC annotations (but those did not yet cause real problems, apparently).
512
513		----------------------------------------------------------------------
514		Name: Matthias Blume
515		Date: 2002/12/06 22:40:00 EST
516		Tag: blume-20021206-cm-fileid
517		Description:
518
519		Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
520		This fixes the problem of CM getting confused by files that suddenly
521		change their identity (e.g., by getting unlinked and recreated by some
522		text editor such as vi).  There might be a better/cheaper/cleaner way
523		of doing this, but for now this will have to do.
524
525		----------------------------------------------------------------------
526		Name: Matthias Blume
527		Date: 2002/10/28 09:50:00 EST
528		Tag: blume-20021028-typecheck
529		Description:
530
531		Exported structure Typecheck from $smlnj/viscomp/core.cm.
532
533		----------------------------------------------------------------------
534		Name: Matthias Blume
535		Date: 2002/10/17 09:10:00 EDT
536		Tag: Release_110_42
537		Description:
538
539		In good old tradition, there has been a slight hiccup so that we have
540		to patch 110.42 after the fact.   The old release tag has been replaced
541		(see below).
542
543		The change solves a problem with two competing approaches the
544		configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
545		each other's way.
546
547		This change only affects the runtime system code and the installer script.
548		(No new bootfiles.)
549
550		----------------------------------------------------------------------
551		Name: Matthias Blume
552		Date: 2002/10/16 12:00:00 EDT
553		Tag: Release_110_42_removed
554		Description:
555
556		New working release.  New bootfiles.
557
558		----------------------------------------------------------------------
559		Name: Matthias Blume
560		Date: 2002/10/10 13:10:00 EDT
561		Tag: blume-20021010-ppc-divs
562		Description:
563
564		The mltree operator DIVS must be implemented with an overflow check on
565		the PPC because the hardware indicates divide-by-zero using "overflow" as
566		well.
567
568		----------------------------------------------------------------------
569		Name: Matthias Blume
570		Date: 2002/07/23 11:45:00
571		Tag: blume-20020723-smlnj-home
572		Description:
573
574		Sml now senses the SMLNJ_HOME environment variable.  If this is set,
575		then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
576		CM_PATHCONFIG is also set), the path configuration file is assumed
577		to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
578		the entire tree to some other place and everything will "just work".
579
580		(Companion commands such as ml-build and ml-makedepend also sense this
581		variable.)
582
583		----------------------------------------------------------------------
584		Name: Matthias Blume
585		Date: 2002/07/12 21:19:00 EDT
586		Tag: blume-20020712-liveness
587		Description:
588
589		Exported two useful "step" functions from liveness module (MLRISC).
590
591		----------------------------------------------------------------------
592		Name: Matthias Blume
593		Date: 2002/07/05 16:00 EDT
594		Tag: Release_110_41
595		Description:
596
597		New working release.  New bootfiles.
598
599		----------------------------------------------------------------------
600		Name: Matthias Blume
601		Date: 2002/07/05 10:25:00 EDT
602		Tag: blume-20020705-btimp
603		Description:
604
605		Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
606		other clients can set up backtracing support.
607
608		----------------------------------------------------------------------
609		Name: Matthias Blume
610		Date: 2002/06/25 17:23:00 EDT
611		Tag: blume-20020625-fpmax
612		Description:
613
614		Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
615		case.  (The sense of the isNaN test was reversed -- which made min and
616		max always return their first argument.)
617
618		----------------------------------------------------------------------
619		Name: Matthias Blume
620		Date: 2002/06/11
621		Tag: blume-20020611-unixpath
622		Description:
623
624		Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
625
626		----------------------------------------------------------------------
627		Name: Matthias Blume
628		Date: 2002/06/10 16:35:00 EDT
629		Tag: blume-20020610-ieeereal
630		Description:
631
632		I back-ported my implementation of IEEEReal.fromString from the
633		idlbasis-devel branch so that we can test it.
634
635		Another small change is that ppDec tries to give more information
636		than just "<sig>" in the case of functors.  However, this code is
637		broken in some mysterious way if the functor's body's signature
638		has not been declared by ascription but gets inferred from the
639		implementation.  This needs fixing...
640
641		----------------------------------------------------------------------
642		Name: Matthias Blume
643		Date: 2002/05/31
644		Tag: blume-20020531-btrace-mode
645		Description:
646
647		Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
648		the wayside when I switched over to using Controls everywhere.)
649
650		----------------------------------------------------------------------
651		Name: Lal George
652		Date: 2002/05/23 12:21:40 EDT
653		Tag: george-20020523-visual-labels
654		Description:
655
656		Labels are now displayed in the graphical output to make
657		the fall-through and target blocks obvious.
658
659		----------------------------------------------------------------------
660		Name: Matthias Blume
661		Date: 2002/05/22 11:03:00 EDT
662		Tag: blume-20020522-shrink
663		Description:
664
665		John tweaked yesterday's fix for 1131 to handle an out-of-memory
666		situation that comes up when allocating huge arrays.
667
668		----------------------------------------------------------------------
669		Name: Matthias Blume
670		Date: 2002/05/21 16:00:00 EDT
671		Tag: Release_110_40
672		Description:
673
674		New working release (110.40).  New bootfiles.
675
676		[Also: John Reppy fixed GC bug 1131.]
677
678		----------------------------------------------------------------------
679		Name: Matthias Blume
680		Date: 2002/05/21 12:35:00 EDT
681		Tag: blume-20020521-cmdoc
682		Description:
683
684		CM documentation update.
685
686		----------------------------------------------------------------------
687		Name: Matthias Blume
688		Date: 2002/05/21 10:55:00 EDT
689		Tag: blume-20020521-misc
690		Description:
691
692		- John tweaked runtime to be silent on heap export (except when
693		GC messages are on).
694		- I added a few more things (cross-compiling versions of CMB) to
695		config/preloads (as suggestions).
696
697		----------------------------------------------------------------------
698		Name: Matthias Blume
699		Date: 2002/05/20 22:25:00 EDT
700		Tag: blume-20020520-controls
701		Description:
702
703		- Added ControlUtil structure to control-lib.cm.
704		- Use it throughout.
705		- Used Controls facility to define MLRISC controls (as opposed to
706		registering MLRISC control ref cells with Controls after the fact)
707		- Fixed messed-up controls priorities.
708
709		* Removed again all the stuff from config/preloads that one wouldn't
710		be able to preload at the time the initial heap image is built.
711		(Many libraries, e.g., CML, do not exist yet at this time.  The
712		only libraries that can be preloaded via config/preloads are those
713		that come bundled with the bootfiles.)
714
715		----------------------------------------------------------------------
716		Name: Matthias Blume
717		Date: 2002/05/20 10:59:00 EDT
718		Tag: blume-20020520-preloads
719		Description:
720
721		Added a lot of commented-out suggestions for things to be included
722		in config/preloads.
723
724		----------------------------------------------------------------------
725		Name: Allen Leung
726		Date: 2002/05/18 14:20:00 EDT
727		Tag: leunga-20020518-mdl
728		Description:
729
730		o Made the mdl tool stuff compile and run again.
731		o I've disabled all the stuff that depends on RTL specifications; they
732		are all badly broken anyway.
733
734		----------------------------------------------------------------------
735		Name: Matthias Blume
736		Date: 2002/05/17 16:49:00 EDT
737		Tag: blume-20020517-controls
738		Description:
739
740		0. John Reppy made several modifications to the SML/NJ library.
741		In particular, there is a shiny new controls-lib.cm.
742
743		1. Pushed new controls interface through compiler so that everything
744		compiles again.
745
746		2. Added FormatComb and FORMAT_COMB to the CML version of the
747		SML/NJ library (so that CML compiles again).
748
749		3. Modified init scripts because XXX_DEFAULT environment variables
750		are no longer with us.  (Boot-time initialization is now done
751		using the same environment variables that are also used for
752		startup-time initialization of controls.)
753
754		----------------------------------------------------------------------
755		Name: Lal George
756		Date: 2002/05/15 09:20:10 EDT
757		Tag: george-20020515-pseudo-op-decls
758		Description:
759
760		All pseudo-ops emitted before the first segment declaration
761		such as TEXT, DATA, and BSS directives are assumed to be global
762		declarations and are emitted first in the assembly file. This is
763		useful in a number of situations where one has pseudo-ops that are not
764		specific to any segment, and also works around the constraint that one
765		cannot have client pseudo-ops in the TEXT segment.
766
767		Because no segment is associated with these declarations it is
768		an error to allocate any space or objects before the first segment
769		directive and an exception will be raised. However, we cannot make
770		this check for client pseudo-ops.
771
772		These top level declarations are a field in the CFG graph_info.
773		In theory you can continue to add to this field after the CFG has been
774		built -- provided you know what you are doing;-)
775
776		----------------------------------------------------------------------
777		Name: Matthias Blume
778		Date: 2002/05/13 16:40:00 EDT
779		Tag: blume-20020513-pp-etc
780		Description:
781
782		A few minor bugfixes:
783
784		- Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
785		(Bogus printouts for redefined bindings still occur.  Compiler
786		bug should no longer occur now.  We need to redo the prettyprinter
787		from scratch.)
788
789		- CM pathname printer now also adds escape sequences for ( and )
790
791		- commend and docu fixes for ml-nlffi
792
793		----------------------------------------------------------------------
794		Name: Matthias Blume
795		Date: 2002/05/10 16:40:00 EDT
796		Tag: blume-20020510-erg-textio
797		Description:
798
799		Applied the following bugfix provided by Emden Gansner:
800
801		Output is corrupted when outputSubstr is used rather than output.
802
803		The problem occurs when a substring
804
805		ss = (s, dataStart, dataLen)
806
807		where dataStart > 0, fills a stream buffer with avail bytes left.
808		avail bytes of s, starting at index dataStart, are copied into the
809		buffer, the buffer is flushed, and then the remaining dataLen-avail
810		bytes of ss are copied into the beginning of the buffer. Instead of
811		starting this copy at index dataStart+avail in s, the current code
812		starts the copy at index avail.
813
814		Fix:
815		In text-io-fn.sml, change line 695 from
816		val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
817		to
818		val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
819
820		----------------------------------------------------------------------
821		Name: Matthias Blume
822		Date: 2002/04/12 13:55:00 EDT
823		Tag: blume-20020412-assyntax
824		Description:
825
826		1. Grabbed newer assyntax.h from the XFree86 project.
827		2. Fiddled with how to compile X86.prim.asm without warnings.
828		3. (Very) Minor cleanup in CM.
829
830		----------------------------------------------------------------------
831		Name: Matthias Blume
832		Date: 2002/04/01 (no joke!) 17:07:00 EST
833		Tag: blume-20020401-x86div
834		Description:
835
836		Added full support for div/mod/rem/quot on the x86, using the machine
837		instruction's two results (without clumsily recomputing the remainder)
838		directly where appropriate.
839
840		Some more extensive power-of-two support was added to the x86 instruction
841		selector (avoiding expensive divs, mods, and muls where they can be
842		replaced with cheaper shifts and masks).  However, this sort of thing
843		ought to be done earlier, e.g., within the CPS optimizer so that
844		all architectures benefit from it.
845
846		The compiler compiles to a fixed point, but changes might be somewhat
847		fragile nevertheless.  Please, report any strange things that you might
848		see wrt. div/mod/quot/rem...
849
850		----------------------------------------------------------------------
851		Name: Matthias Blume
852		Date: 2002/03/29 17:22:00
853		Tag: blume-20020329-div
854		Description:
855
856		Fixed my broken div/mod logic.  Unfortunately, this means that the
857		inline code for div/mod now has one more comparison than before.
858		Fast paths (quotient > 0 or remainder = 0) are not affected, though.
859		The problem was with quotient = 0, because that alone does not tell
860		us which way the rounding went.  One then has to look at whether
861		remainder and divisor have the same sign...  :(
862
863		Anyway, I replaced the bootfiles with fresh ones...
864
865		----------------------------------------------------------------------
866		Name: Matthias Blume
867		Date: 2002/03/29 14:10:00 EST
868		Tag: blume-20020329-inlprims
869		Description:
870
871		NEW BOOTFILES!!!    Version number bumped to 110.39.3.
872
873		Primops have changed. This means that the bin/boot-file formats have
874		changed as well.
875
876		To make sure that there is no confusion, I made a new version.
877
878
879		CHANGES:
880
881		* removed REMT from mltree (remainder should never overflow).
882
883		* added primops to deal with divisions of all flavors to the frontend
884
885		* handled these primops all the way through so they map to their respective
886		MLRISC support
887
888		* used these primops in the implementation of Int, Int32, Word, Word32
889
890		* removed INLDIV, INLMOD, and INLREM as they are no longer necessary
891
892		* parameterized INLMIN, INLMAX, and INLABS by a numkind
893
894		* translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
895		floating point
896
897		* used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
898		and Real (but Real.abs maps to a separate floating-point-only primop)
899
900
901		TODO items:
902
903		* Hacked Alpha32 instruction selection, disabling the selection of REMx
904		instructions because the machine instruction encoder cannot handle
905		them.  (Hppa, PPC, and Sparc instruction selection did not handle
906		REM in the first place, and REM is supported by the x86 machine coder.)
907
908		* Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
909		selection phase.  (The two can be streamlined because the hardware
910		delivers both quotient and remainder at the same time anyway.)
911
912		* Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
913		(Currently the behavior is inconsistent both across architectures and
914		wrt. the draft Basis spec.)
915
916		* Word8 should eventually be handled natively, too.
917
918		* There seems to be one serious bug in mltree-gen.sml.  It appears, though,
919		as if there currently is no execution path that could trigger it in
920		SML/NJ.  (The assumptions underlying functions arith and promotable do not
921		hold for things like multiplication and division.)
922
923		----------------------------------------------------------------------
924		Name: Matthias Blume
925		Date: 2002/03/27 16:27:00 EST
926		Tag: blume-20020327-mlrisc-divisions
927		Description:
928
929		Added support for all four division operations (ML's div, mod, quot,
930		and rem) to MLRISC.  In the course of doing so, I also rationalized
931		the naming (no more annoying switch-around of DIV and QUOT), by
932		parameterizing the operation by div_rounding_mode (which can be either
933		DIV_TO_ZERO or DIV_TO_NEGINF).
934
935		The generic MLTreeGen functor takes care of compiling all four
936		operations down to only round-to-zero div.
937
938		Missing pieces:
939
940		* Doing something smarter than relying on MLTreeGen on architectures
941		like, e.g., the x86 where hardware division delivers both quotient and
942		remainder at the same time.  With this, the implementation of the
943		round-to-neginf operations could be further streamlined.
944
945		* Remove inlining support for div/mod/rem from the frontend and replace it
946		with primops that get carried through to the backend.  Do this for all
947		int and word types.
948
949		----------------------------------------------------------------------
950		Name: Matthias Blume
951		Date: 2002/03/25 17:25:00 EST
952		Tag: blume-20020325-divmod
953		Description:
954
955		I improved (hopefully without breaking them) the implementation of Int.div,
956		Int.mod, and Int.rem.   For this, the code in translate.sml now takes
957		advantage of the following observations:
958
959		Let  q = x quot y      r = x rem y
960		d = x div  y      m = x mod y
961
962		where "quot" is the round-to-zero version of integer division that
963		hardware usually provides.  Then we have:
964
965		r = x - q * y        where neither the * nor the - will overflow
966		d = if q >= 0 orelse x = q * y then q else q - 1
967		where neither the * nor the - will overflow
968		m = if q >= 0 orelse r = 0 then r else r + y
969		where the + will not overflow
970
971		This results in substantial simplification of the generated code.
972		The following table shows the number of CFG nodes and edges generated
973		for
974		fun f (x, y) = x OPER y
975		(* with OPER \in div, mod, quot, rem *)
976
977
978		OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
979		--------------------------------------------------------
980		div |         24 |         39 |         12 |         16
981		mod |         41 |         71 |         12 |         16
982		quot |          8 |         10 |          8 |         10
983		rem |         10 |         14 |          8 |         10
984
985
986		----------------------------------------------------------------------
987		Name: Matthias Blume
988		Date: 2002/03/25 22:06:00 EST
989		Tag: blume-20020325-cprotobug
990		Description:
991
992		Fixed a bug in cproto (c prototype decoder).
993
994		----------------------------------------------------------------------
995		Name: Matthias Blume
996		Date: 2002/03/25 16:00:00 EST
997		Tag: blume-20020325-raw-primops
998		Description:
999
1000		I did some cleanup to Allen's new primop code and
1001		replaced yesterday's bootfiles with new ones.
1002		(But they are stored in the same place.)
1003
1004		----------------------------------------------------------------------
1005		Name: Matthias Blume
1006		Date: 2002/03/24 22:40:00 EST
1007		Tag: blume-20020324-bootfiles
1008		Description:
1009
1010		Made the bootfiles that Allen asked for.
1011
1012		----------------------------------------------------------------------
1013		Name: Allen Leung
1014		Date: 2002/03/23 15:50:00 EST
1015		Tag: leunga-20020323-flint-cps-rcc-primops
1016		Description:
1017
1018		1. Changes to FLINT primops:
1019
1020		(* make a call to a C-function;
1021		* The primop carries C function prototype information and specifies
1022		* which of its (ML-) arguments are floating point. C prototype
1023		* information is for use by the backend, ML information is for
1024		* use by the CPS converter. *)
1025		| RAW_CCALL of { c_proto: CTypes.c_proto,
1026		ml_args: ccall_type list,
1027		ml_res_opt: ccall_type option,
1028		reentrant : bool
1029		} option
1030		(* Allocate uninitialized storage on the heap.
1031		* The record is meant to hold short-lived C objects, i.e., they
1032		* are not ML pointers.  With the tag, the representation is
1033		* the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
1034		*)
1035		| RAW_RECORD of {tag:bool,sz:int}
1036		and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
1037
1038		2.  These CPS primops are now overloaded:
1039
1040		rawload of {kind:numkind}
1041		rawstore of {kind:numkind}
1042
1043		The one argument form is:
1044
1045		rawload {kind} address
1046
1047		The two argument form is:
1048
1049		rawload {kind} [ml object, byte-offset]
1050
1051		3. RAW_CCALL/RCC now takes two extra arguments:
1052
1053		a. The first is whether the C call is reentrant, i.e., whether
1054		ML state should be saved and restored.
1055		b. The second argument is a string argument specifying the name of
1056		library and the C function.
1057
1058		These things are currently not handled in the code generator, yet.
1059
1060		4. In CProto,
1061
1062		An encoding type of "bool" means "ml object" and is mapped into
1063		C prototype of PTR.  Note that "bool" is different than "string",
1064		even though "string" is also mapped into PTR, because "bool"
1065		is assigned an CPS type of BOGt, while "string" is assigned INT32t.
1066
1067		5. Pickler/unpicker
1068
1069		Changed to handle RAW_RECORD and newest RAW_CCALL
1070
1071		6. MLRiscGen,
1072
1073		1. Changed to handle the new rawload/rawstore/rawrecord operators.
1074		2. Code for handling C Calls has been moved to a new module CPSCCalls,
1075		in the file CodeGen/cpscompile/cps-c-calls.sml
1076
1077		7. Added the conditional move operator
1078
1079		condmove of branch
1080
1081		to cps.  Generation of this is still buggy so it is currently
1082		disabled.
1083
1084		----------------------------------------------------------------------
1085		Name: Lal George
1086		Date: 2002/03/22 14:18:25 EST
1087		Tag: george-20020322-cps-branch-prob
1088		Description:
1089
1090		Implemented the Ball-Larus branch prediction-heuristics, and
1091		incorporated graphical viewers for control flow graphs.
1092
1093		Ball-Larus Heuristics:
1094		---------------------
1095		See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
1096
1097		By design it uses the Dempster-Shafer theory for combining
1098		probabilities.  For example, in the function:
1099
1100		fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
1101
1102		the ball-larus heuristics predicts that the n=0 is unlikely
1103		(OH-heuristic), and the 'then' branch is unlikely because of the
1104		RH-heuristic -- giving the 'then' branch an even lower combined
1105		probability using the Dempster-Shafer theory.
1106
1107		Finally, John Reppy's loop analysis in MLRISC, further lowers the
1108		probability of the 'then' branch because of the loop in the else
1109		branch.
1110
1111
1112		Graphical Viewing:
1113		------------------
1114		I merely plugged in Allen's graphical viewers into the compiler. The
1115		additional code is not much. At the top level, saying:
1116
1117		Control.MLRISC.getFlag "cfg-graphical-view" := true;
1118
1119		will display the graphical view of the control flow graph just before
1120		back-patching.  daVinci must be in your path for this to work. If
1121		daVinci is not available, then the default viewer can be changed
1122		using:
1123
1124		Control.MLRISC.getString "viewer"
1125
1126		which can be set to "dot" or "vcg" for the corresponding viewers. Of
1127		course, these viewers must be in your path.
1128
1129		The above will display the compilation unit at the level of clusters,
1130		many of which are small, boring, and un-interesting. Also setting:
1131
1132		Control.MLRISC.getInt "cfg-graphical-view_size"
1133
1134		will display clusters that are larger than the value set by the above.
1135
1136
1137		----------------------------------------------------------------------
1138		Name: Matthias Blume
1139		Date: 2002/03/21 22:20:00 EST
1140		Tag: blume-20020321-kmp-bugfix
1141		Description:
1142
1143		Changed the interface to the KMP routine in PreString and fixed
1144		a minor bug in one place where it was used.
1145
1146		----------------------------------------------------------------------
1147		Name: Allen Leung
1148		Date: 2002/03/21 20:30:00 EST
1149		Tag: leunga-20020321-cfg
1150		Description:
1151
1152		Fixed a potential problem in cfg edge splitting.
1153
1154		----------------------------------------------------------------------
1155		Name: Allen Leung
1156		Date: 2002/03/21 17:15:00 EST
1157		Tag: leunga-20020321-x86-fp-cfg
1158		Description:
1159
1160		1. Recoded the buggy parts of x86-fp.
1161
1162		a. All the block reordering code has been removed.
1163		We now depend on the block placement phases to do this work.
1164
1165		b. Critical edge splitting code has been simplified and moved into the
1166		CFG modules, as where they belong.
1167
1168		Both of these were quite buggy and complex.  The code is now much, much
1169		simpler.
1170
1171		2. X86 backend.
1172
1173		a. Added instructions for 64-bit support.  Instruction selection for
1174		64-bit has not been committed, however, since that
1175		requires changes to MLTREE which haven't been approved by
1176		Lal and John.
1177
1178		b. Added support for FUCOMI and FUCOMIP when generating code for
1179		PentiumPro and above.  We only generate these instructions in
1180		the fast-fp mode.
1181
1182		c. Added cases for JP and JNP in X86FreqProps.
1183
1184		3. CFG
1185
1186		CFG now has a bunch of methods for edge splitting and merging.
1187
1188		4. Machine description.
1189
1190		John's simplification of MLTREE_BASIS.fcond broke a few machine
1191		description things:
1192
1193		rtl-build.{sig,sml} and hppa.mdl fixed.
1194
1195		NOTE: the machine description stuff in the repository is still broken.
1196		Again, I can't put my fixes in because that involves
1197		changes to MLTREE.
1198
1199		----------------------------------------------------------------------
1200		Name: Matthias Blume
1201		Date: 2002/03/20 15:55:00 EST
1202		Tag: blume-20020320-kmp
1203		Description:
1204
1205		Implemented Knuth-Morris-Pratt string matching in PreString and used
1206		it for String.isSubstring, Substring.isSubstring, and
1207		Substring.position.
1208
1209		(Might need some stress-testing.  Simple examples worked fine.)
1210
1211		----------------------------------------------------------------------
1212		Name: Matthias Blume
1213		Date: 2002/03/19 16:37:00 EST
1214		Tag: blume-20020319-witnesses
1215		Description:
1216
1217		Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
1218
1219		This implements a generic mechanism for changing constness qualifiers
1220		anywhere within big C types without resorting to outright "casts".
1221		(So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
1222		modify the constness at the outermost level.)
1223		The implementation of "convert" is based on the idea of "witness"
1224		values -- values that are not used by the operation but whose types
1225		"testify" to their applicability.  On the implementation side, "convert"
1226		is simply a projection (returning its second curried argument).  With
1227		cross-module inlining, it should not result in any machine code being
1228		generated.
1229
1230		----------------------------------------------------------------------
1231		Name: Matthias Blume
1232		Date: 2002/03/15 16:40:00 EST
1233		Tag: blume-20020315-basis
1234		Description:
1235
1236		Provided (preliminary?) implementations for
1237
1238		{String,Substring}.{concatWith,isSuffix,isSubstring}
1239
1240		and
1241
1242		Substring.full
1243
1244		Those are in the Basis spec but they were missing in SML/NJ.
1245
1246		----------------------------------------------------------------------
1247		Name: Matthias Blume
1248		Date: 2002/03/14 21:30:00 EST
1249		Tag: blume-20020314-controls
1250		Description:
1251
1252		Controls:
1253		---------
1254
1255		1. Factored out the recently-added Controls : CONTROLS stuff and put
1256		it into its own library $/controls-lib.cm.  The source tree for
1257		this is under src/smlnj-lib/Controls.
1258
1259		2. Changed the names of types and functions in this interface, so they
1260		make a bit more "sense":
1261
1262		module -> registry
1263		'a registry -> 'a group
1264
1265		3. The interface now deals in ref cells only.  The getter/setter interface
1266		is (mostly) gone.
1267
1268		4. Added a function that lets one register an already-existing ref cell.
1269
1270		5. Made the corresponding modifications to the rest of the code so that
1271		everything compiles again.
1272
1273		6. Changed the implementation of Controls.MLRISC back to something closer
1274		to the original.  In particular, this module (and therefore MLRISC)
1275		does not depend on Controls.  There now is some link-time code in
1276		int-sys.sml that registers the MLRISC controls with the Controls
1277		module.
1278
1279		CM:
1280		---
1281
1282		* One can now specify the lambda-split aggressiveness in init.cmi.
1283
1284		----------------------------------------------------------------------
1285		Name: Allen Leung
1286		Date: 2002/03/13 17:30:00 EST
1287		Tag: leunga-20020313-x86-fp-unary
1288		Description:
1289
1290		Bug fix for:
1291
1292		> leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
1293		> Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
1294		> - fun f(x,(y,z)) = Real.~ y;
1295		> [autoloading]
1296		> [autoloading done]
1297		>       fchsl   (%eax), 184(%esp)
1298		> Error: MLRisc bug: X86MCEmitter.emitInstr
1299		>
1300		> uncaught exception Error
1301		>   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
1302
1303		The problem was that the code generator did not generate any fp registers
1304		in this case, and the ra didn't know that it needed to run the X86FP phase to
1305		translate the pseudo fp instruction.   This only happened with unary fp
1306		operators in certain situations.
1307
1308		----------------------------------------------------------------------
1309		Name: Matthias Blume
1310		Date: 2002/03/13 14:00:00 EST
1311		Tag: blume-20020313-overload-etc
1312		Description:
1313
1314		1. Added _overload as a synonym for overload for backward compatibility.
1315		(Control.overloadKW must be true for either version to be accepted.)
1316
1317		2. Fixed bug in install script that caused more things to be installed
1318		than what was requested in config/targets.
1319
1320		3. Made CM aware of the (_)overload construct so that autoloading
1321		works.
1322
1323		----------------------------------------------------------------------
1324		Name: Matthias Blume
1325		Date: 2002/03/12 22:03:00 EST
1326		Tag: blume-20020312-url
1327		Description:
1328
1329		Forgot to update BOOT and srcarchiveurl.
1330
1331		----------------------------------------------------------------------
1332		Name: Matthias Blume
1333		Date: 2002/03/12 17:30:00 EST
1334		Tag: blume-20020312-version110392
1335		Description:
1336
1337		Yet another version number bump (because of small changes to the
1338		binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
1339
1340		Changes:
1341
1342		The new pid generation scheme described a few weeks ago was overly
1343		complicated.  I implemented a new mechanism that is simpler and
1344		provides a bit more "stability":  Once CM has seen a compilation
1345		unit, it keeps its identity constant (as long as you do not delete
1346		those crucial CM/GUID/* files).  This means that when you change
1347		an interface, compile, then go back to the old interface, and
1348		compile again, you arrive at the original pid.
1349
1350		There now also is a mechanism that instructs CM to use the plain
1351		environment hash as a module's pid (effectively making its GUID
1352		the empty string).  For this, "noguid" must be specified as an
1353		option to the .sml file in question within its .cm file.
1354		This is most useful for code that is being generated by tools such
1355		as ml-nlffigen (because during development programmers tend to
1356		erase the tool's entire output directory tree including CM's cached
1357		GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
1358		revert to the old, broken behavior of SML/NJ, but in specific cases
1359		where there is no danger of interface confusion, its use is ok
1360		(I think).
1361
1362		ml-nlffigen by default generates "noguid" annotations.  They can be
1363		turned off by specifying -guid in its command line.
1364
1365		----------------------------------------------------------------------
1366		Name: Lal George
1367		Date: 2002/03/12 12 14:42:36 EST
1368		Tag: george-20020312-frequency-computation
1369		Description:
1370
1371		Integrated jump chaining and static block frequency into the
1372		compiler. More details and numbers later.
1373
1374		----------------------------------------------------------------------
1375		Name: Lal George
1376		Date: 2002/03/11 11 22:38:53 EST
1377		Tag: george-20020311-jump-chain-elim
1378		Description:
1379
1380		Tested the jump chain elimination on all architectures (except the
1381		hppa).  This is on by default right now and is profitable for the
1382		alpha and x86, however, it may not be profitable for the sparc and ppc
1383		when compiling the compiler.
1384
1385		The gc test will typically jump to a label at the end of the cluster,
1386		where there is another jump to an external cluster containing the actual
1387		code to invoke gc. This is to allow factoring of common gc invocation
1388		sequences. That is to say, we generate:
1389
1390		f:
1391		testgc
1392		ja   L1      % jump if above to L1
1393
1394		L1:
1395		jmp L2
1396
1397
1398		After jump chain elimination the 'ja L1' instructions is converted to
1399		'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
1400		up being implemented in their long form (if L2 is far away) using:
1401
1402		jbe     L3      % jump if below or equal to L3
1403		jmp     L2
1404		L3:
1405		...
1406
1407
1408		For large compilation units L2  may be far away.
1409
1410
1411		----------------------------------------------------------------------
1412		Name: Matthias Blume
1413		Date: 2002/03/11 13:30:00 EST
1414		Tag: blume-20020311-mltreeeval
1415		Description:
1416
1417		A functor parameter was missing.
1418
1419		----------------------------------------------------------------------
1420		Name: Allen Leung
1421		Date: 2002/03/11 10:30:00 EST
1422		Tag: leunga-20020311-runtime-string0
1423		Description:
1424
1425		The representation of the empty string now points to a
1426		legal null terminated C string instead of unit.  It is now possible
1427		to convert an ML string into C string with InlineT.CharVector.getData.
1428		This compiles into one single machine instruction.
1429
1430		----------------------------------------------------------------------
1431		Name: Allen Leung
1432		Date: 2002/03/10 23:55:00 EST
1433		Tag: leunga-20020310-x86-call
1434		Description:
1435
1436		Added machine generation for CALL instruction (relative displacement mode)
1437
1438		----------------------------------------------------------------------
1439		Name: Matthias Blume
1440		Date: 2002/03/08 16:05:00
1441		Tag: blume-20020308-entrypoints
1442		Description:
1443
1444		Version number bumped to 110.39.1.  NEW BOOTFILES!
1445
1446		Entrypoints: non-zero offset into a code object where execution should begin.
1447
1448		- Added the notion of an entrypoint to CodeObj.
1449		- Added reading/writing of entrypoint info to Binfile.
1450		- Made runtime system bootloader aware of entrypoints.
1451		- Use the address of the label of the first function given to mlriscGen
1452		as the entrypoint.  This address is currently always 0, but it will
1453		not be 0 once we turn on block placement.
1454		- Removed the linkage cluster code (which was The Other Way(tm) of dealing
1455		with entry points) from mlriscGen.
1456
1457		----------------------------------------------------------------------
1458		Name: Allen Leung
1459		Date: 2002/03/07 20:45:00 EST
1460		Tag: leunga-20020307-x86-cmov
1461		Description:
1462
1463		Bug fixes for CMOVcc on x86.
1464
1465		1. Added machine code generation for CMOVcc
1466		2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1467		3. CMOVcc cannot have an immediate operand as argument.
1468
1469		----------------------------------------------------------------------
1470		Name: Matthias Blume
1471		Date: 2002/03/07 16:15:00 EST
1472		Tag: blume-20020307-controls
1473		Description:
1474
1475		This is a very large but mostly boring patch which makes (almost)
1476		every tuneable compiler knob (i.e., pretty much everything under
1477		Control.* plus a few other things) configurable via both the command
1478		line and environment variables in the style CM did its configuration
1479		until now.
1480
1481		Try starting sml with '-h' (or, if you are brave, '-H')
1482
1483		To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1484		implements the underlying generic mechanism.
1485
1486		The interface to some of the existing such facilities has changed somewhat.
1487		For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1488		(The getFoo interface is still there for backward-compatibility, but its
1489		use is deprecated.)
1490
1491		The ml-build script passes -Cxxx=yyy command-line arguments through so
1492		that one can now twiddle the compiler settings when using this "batch"
1493		compiler.
1494
1495		TODO items:
1496
1497		We should go through and throw out all controls that are no longer
1498		connected to anything.  Moreover, we should go through and provide
1499		meaningful (and correct!) documentation strings for those controls
1500		that still are connected.
1501
1502		Currently, multiple calls to Controls.new are accepted (only the first
1503		has any effect).  Eventually we should make sure that every control
1504		is being made (via Controls.new) exactly once.  Future access can then
1505		be done using Controls.acc.
1506
1507		Finally, it would probably be a good idea to use the getter-setter
1508		interface to controls rather than ref cells.  For the time being, both
1509		styles are provided by the Controls module, but getter-setter pairs are
1510		better if thread-safety is of any concern because they can be wrapped.
1511
1512		*****************************************
1513
1514		One bug fix: The function blockPlacement in three of the MLRISC
1515		backpatch files used to be hard-wired to one of two possibilities at
1516		link time (according to the value of the placementFlag).  But (I
1517		think) it should rather sense the flag every time.
1518
1519		*****************************************
1520
1521		Other assorted changes (by other people who did not supply a HISTORY entry):
1522
1523		1. the cross-module inliner now works much better (Monnier)
1524		2. representation of weights, frequencies, and probabilities in MLRISC
1525		changed in preparation of using those for weighted block placement
1526		(Reppy, George)
1527
1528		----------------------------------------------------------------------
1529		Name: Lal George
1530		Date: 2002/03/07 14:44:24 EST 2002
1531		Tag: george-20020307-weighted-block-placement
1532
1533		Tested the weighted block placement optimization on all architectures
1534		(except the hppa) using AMPL to generate the block and edge frequencies.
1535		Changes were required in the machine properties to correctly
1536		categorize trap instructions. There is an MLRISC flag
1537		"weighted-block-placement" that can be used to enable weighted block
1538		placement, but this will be ineffective without block/edge
1539		frequencies (coming soon).
1540
1541
1542		----------------------------------------------------------------------
1543		Name: Lal George
1544		Date: 2002/03/05 17:24:48 EST
1545		Tag: george-20020305-linkage-cluster
1546
1547		In order to support the block placement optimization, a new cluster
1548		is generated as the very first cluster (called the linkage cluster).
1549		It contains a single jump to the 'real' entry point for the compilation
1550		unit. Block placement has no effect on the linkage cluster itself, but
1551		all the other clusters  have full freedom in the manner in which they
1552		reorder blocks or functions.
1553
1554		On the x86 the typical linkage code that is generated is:
1555		----------------------
1556		.align 2
1557		L0:
1558		addl    $L1-L0, 72(%esp)
1559		jmp     L1
1560
1561
1562		.align  2
1563		L1:
1564		----------------------
1565
1566		72(%esp) is the memory location for the stdlink register. This
1567		must contain the address of the CPS function being called. In the
1568		above example, it contains the address of  L0; before
1569		calling L1 (the real entry point for the compilation unit), it
1570		must contain the address for L1, and hence
1571
1572		addl $L1-L0, 72(%esp)
1573
1574		I have tested this on all architectures except the hppa.The increase
1575		in code size is of course negligible
1576
1577		----------------------------------------------------------------------
1578		Name: Allen Leung
1579		Date: 2002/03/03 13:20:00 EST
1580		Tag: leunga-20020303-mlrisc-tools
1581
1582		Added #[ ... ] expressions to mlrisc tools
1583
1584		----------------------------------------------------------------------
1585		Name: Matthias Blume
1586		Date: 2002/02/27 12:29:00 EST
1587		Tag: blume-20020227-cdebug
1588		Description:
1589
1590		- made types in structure C and C_Debug to be equal
1591		- got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1592		- there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1593
1594		----------------------------------------------------------------------
1595		Name: Matthias Blume
1596		Date: 2002/02/26 12:00:00 EST
1597		Tag: blume-20020226-ffi
1598		Description:
1599
1600		1. Fixed a minor bug in CM's "noweb" tool:
1601		If numbering is turned off, then truly don't number (i.e., do not
1602		supply the -L option to noweb).  The previous behavior was to supply
1603		-L'' -- which caused noweb to use the "default" line numbering scheme.
1604		Thanks to Chris Richards for pointing this out (and supplying the fix).
1605
1606		2. Once again, I reworked some aspects of the FFI:
1607
1608		A. The incomplete/complete type business:
1609
1610		- Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1611		gone!
1612		- ML types representing an incomplete type are now *equal* to
1613		ML types representing their corresponding complete types (just like
1614		in C).  This is still safe because ml-nlffigen will not generate
1615		RTTI for incomplete types, nor will it generate functions that
1616		require access to such RTTI.   But when ML code generated from both
1617		incomplete and complete versions of the C type meet, the ML types
1618		are trivially interoperable.
1619
1620		NOTE:  These changes restore the full generality of the translation
1621		(which was previously lost when I eliminated functorization)!
1622
1623		B. Enum types:
1624
1625		- Structure C now has a type constructor "enum" that is similar to
1626		how the "su" constructor works.  However, "enum" is not a phantom
1627		type because each "T enum" has values (and is isomorphic to
1628		MLRep.Signed.int).
1629		- There are generic access operations for enum objects (using
1630		MLRep.Signed.int).
1631		- ml-nlffigen will generate a structure E_foo for each "enum foo".
1632		* The structure contains the definition of type "mlrep" (the ML-side
1633		representation type of the enum).  Normally, mlrep is the same
1634		as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1635		then mlrep will be defined as a datatype -- thus facilitating
1636		pattern matching on mlrep values.
1637		("-ec" will be suppressed if there are duplicate values in an
1638		enumeration.)
1639		* Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1640		will be generated for each C enum constant xxx.
1641		* Conversion functions m2i and i2m convert between mlrep and
1642		MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1643		* Coversion functions c and ml convert between mlrep and "tag enum".
1644		* Access functions (get/set) fetch and store mlrep values.
1645		- By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1646		enumerations are merged into one single enumeration represented by
1647		structure E_'.
1648
1649		----------------------------------------------------------------------
1650		Name: Allen Leung
1651		Date: 2002/02/25 04:45:00 EST
1652		Tag: leunga-20020225-cps-spill
1653
1654		This is a new implementation of the CPS spill phase.
1655		The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1656		In case of problems, replace it with the old file spill.sml
1657
1658		The current compiler runs into some serious performance problems when
1659		constructing a large record.  This can happen when we try to compile a
1660		structure with many items.  Even a very simple structure like the following
1661		makes the compiler slow down.
1662
1663		structure Foo = struct
1664		val x_1 = 0w1 : Word32.int
1665		val x_2 = 0w2 : Word32.int
1666		val x_3 = 0w3 : Word32.int
1667		...
1668		val x_N = 0wN : Word32.int
1669		end
1670
1671		The following table shows the compile time, from N=1000 to N=4000,
1672		with the old compiler:
1673
1674		N
1675		1000   CPS 100 spill                           0.04u  0.00s  0.00g
1676		MLRISC ra                               0.06u  0.00s  0.05g
1677		(spills = 0 reloads = 0)
1678		TOTAL                                   0.63u  0.07s  0.21g
1679
1680		1100   CPS 100 spill                           8.25u  0.32s  0.64g
1681		MLRISC ra                               5.68u  0.59s  3.93g
1682		(spills = 0 reloads = 0)
1683		TOTAL                                   14.71u  0.99s  4.81g
1684
1685		1500   CPS 100 spill                           58.55u  2.34s  1.74g
1686		MLRISC ra                               5.54u  0.65s  3.91g
1687		(spills = 543 reloads = 1082)
1688		TOTAL                                   65.40u  3.13s  6.00g
1689
1690		2000   CPS 100 spill                           126.69u  4.84s  3.08g
1691		MLRISC ra                               0.80u  0.10s  0.55g
1692		(spills = 42 reloads = 84)
1693		TOTAL                                   129.42u  5.10s  4.13g
1694
1695		3000   CPS 100 spill                           675.59u  19.03s  11.64g
1696		MLRISC ra                               2.69u  0.27s  1.38g
1697		(spills = 62 reloads = 124)
1698		TOTAL                                   682.48u  19.61s  13.99g
1699
1700		4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1701		MLRISC ra                               4.96u  0.27s  2.72g
1702		(spills = 85 reloads = 170)
1703		TOTAL                                   2375.26u  57.21s  48.00g
1704
1705		As you can see the old cps spill module suffers from some serious
1706		performance problem.  But since I cannot decipher the old code fully,
1707		instead of patching the problems up, I'm reimplementing it
1708		with a different algorithm.  The new code is more modular,
1709		smaller when compiled, and substantially faster
1710		(O(n log n) time and O(n) space).  Timing of the new spill module:
1711
1712		4000  CPS 100 spill                           0.02u  0.00s  0.00g
1713		MLRISC ra                               0.25u  0.02s  0.15g
1714		(spills=1 reloads=3)
1715		TOTAL                                   7.74u  0.34s  1.62g
1716
1717		Implementation details:
1718
1719		As far as I can tell, the purpose of the CPS spill module is to make sure the
1720		number of live variables at any program point (the bandwidth)
1721		does not exceed a certain limit, which is determined by the
1722		size of the spill area.
1723
1724		When the bandwidth is too large, we decrease the register pressure by
1725		packing live variables into spill records.  How we achieve this is
1726		completely different than what we did in the old code.
1727
1728		First, there is something about the MLRiscGen code generator
1729		that we should be aware of:
1730
1731		o MLRiscGen performs code motion!
1732
1733		In particular, it will move floating point computations and
1734		address computations involving only the heap pointer to
1735		their use sites (if there is only a single use).
1736		What this means is that if we have a CPS record construction
1737		statement
1738
1739		RECORD(k,vl,w,e)
1740
1741		we should never count the new record address w as live if w
1742		has only one use (which is often the case).
1743
1744		We should do something similar to floating point, but the transformation
1745		there is much more complex, so I won't deal with that.
1746
1747		Secondly, there are now two new cps primops at our disposal:
1748
1749		1. rawrecord of record_kind option
1750		This pure operator allocates some uninitialized storage from the heap.
1751		There are two forms:
1752
1753		rawrecord NONE [INT n]  allocates a tagless record of length n
1754		rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1755		and initializes the tag.
1756
1757		2. rawupdate of cty
1758		rawupdate cty (v,i,x)
1759		Assigns to x to the ith component of record v.
1760		The storelist is not updated.
1761
1762		We use these new primops for both spilling and increment record construction.
1763
1764		1. Spilling.
1765
1766		This is implemented with a linear scan algorithm (but generalized
1767		to trees).  The algorithm will create a single spill record at the
1768		beginning of the cps function and use rawupdate to spill to it,
1769		and SELECT or SELp to reload from it.  So both spills and reloads
1770		are fine-grain operations.  In contrast, in the old algorithm
1771		"spills" have to be bundled together in records.
1772
1773		Ideally, we should sink the spill record construction to where
1774		it is needed.  We can even split the spill record into multiple ones
1775		at the places where they are needed.  But CPS is not a good
1776		representation for global code motion, so I'll keep it simple and
1777		am not attempting this.
1778
1779		2. Incremental record construction (aka record splitting).
1780
1781		Long records with many component values which are simulatenously live
1782		(recall that single use record addresses are not considered to
1783		be live) are constructed with rawrecord and rawupdate.
1784		We allocate space on the heap with rawrecord first, then gradually
1785		fill it in with rawupdate.  This is the technique suggested to me
1786		by Matthias.
1787
1788		Some restrictions on when this is applicable:
1789		1. It is not a VECTOR record.  The code generator currently does not handle
1790		this case. VECTOR record uses double indirection like arrays.
1791		2. All the record component values are defined in the same "basic block"
1792		as the record constructor.  This is to prevent speculative
1793		record construction.
1794
1795		----------------------------------------------------------------------
1796		Name: Allen Leung
1797		Date: 2002/02/22 01:02:00 EST
1798		Tag: leunga-20020222-mlrisc-tools
1799
1800		Minor bug fixes in the parser and rewriter
1801
1802		----------------------------------------------------------------------
1803		Name: Allen Leung
1804		Date: 2002/02/21 20:20:00 EST
1805		Tag: leunga-20020221-peephole
1806
1807		Regenerated the peephole files.  Some contained typos in the specification
1808		and some didn't compile because of pretty printing bugs in the old version
1809		of 'nowhere'.
1810
1811		----------------------------------------------------------------------
1812		Name: Allen Leung
1813		Date: 2002/02/19 20:20:00 EST
1814		Tag: leunga-20020219-mlrisc-tools
1815		Description:
1816
1817		Minor bug fixes to the mlrisc-tools library:
1818
1819		1.  Fixed up parsing colon suffixed keywords
1820		2.  Added the ability to shut the error messages up
1821		3.  Reimplemented the pretty printer and fixed up/improved
1822		the pretty printing of handle and -> types.
1823		4.  Fixed up generation of literal symbols in the nowhere tool.
1824		5.  Added some SML keywords to to sml.sty
1825
1826		----------------------------------------------------------------------
1827		Name: Matthias Blume
1828		Date: 2002/02/19 16:20:00 EST
1829		Tag: blume-20020219-cmffi
1830		Description:
1831
1832		A wild mix of changes, some minor, some major:
1833
1834		* All C FFI-related libraries are now anchored under $c:
1835		$/c.cm      --> $c/c.cm
1836		$/c-int.cm  --> $c/internals/c-int.cm
1837		$/memory.cm --> $c/memory/memory.cm
1838
1839		* "make" tool (in CM) now treats its argument pathname slightly
1840		differently:
1841		1. If the native expansion is an absolute name, then before invoking
1842		the "make" command on it, CM will apply OS.Path.mkRelative
1843		(with relativeTo = OS.FileSys.getDir()) to it.
1844		2. The argument will be passed through to subsequent phases of CM
1845		processing without "going native".  In particular, if the argument
1846		was an anchored path, then "make" will not lose track of that anchor.
1847
1848		* Compiler backends now "know" their respective C calling conventions
1849		instead of having to be told about it by ml-nlffigen.  This relieves
1850		ml-nlffigen from one of its burdens.
1851
1852		* The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1853
1854		* Export C_DEBUG and C_Debug from $c/c.cm.
1855
1856		* C type encoding in ml-nlffi-lib has been improved to model the conceptual
1857		subtyping relationship between incomplete pointers and their complete
1858		counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1859		with the convention of instantiating 'o with ('t, 'c) obj whenever
1860		the pointer target type is complete.  In the incomplete case, 'o
1861		will be instantiated with some "'c iobj" -- a type obtained by
1862		using one of the functors PointerToIncompleteType or PointerToCompleteType.
1863
1864		Operations that work on both incomplete and complete pointer types are
1865		typed as taking an 'o ptr while operations that require the target to
1866		be known are typed as taking some ('t, 'c) obj ptr.
1867
1868		voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1869		where void is an eqtype without any values.  This makes it possible
1870		to work on voidptr values using functions meant to operate on light
1871		incomplete pointers.
1872
1873		* As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1874		been vastly simplified.
1875
1876		----------------------------------------------------------------------
1877		Name: Matthias Blume
1878		Date: 2002/02/19 10:48:00 EST
1879		Tag: blume-20020219-pqfix
1880		Description:
1881
1882		Applied Chris Okasaki's bug fix for priority queues.
1883
1884		----------------------------------------------------------------------
1885		Name: Matthias Blume
1886		Date: 2002/02/15 17:05:00
1887		Tag: Release_110_39
1888		Description:
1889
1890		Last-minute retagging is becoming a tradition... :-(
1891
1892		This is the working release 110.39.
1893
1894		----------------------------------------------------------------------
1895		Name: Matthias Blume
1896		Date: 2002/02/15 16:00:00 EST
1897		Tag: Release_110_39-orig
1898		Description:
1899
1900		Working release 110.39.  New bootfiles.
1901
1902		(Update: There was a small bug in the installer so it wouldn't work
1903		with all shells.  So I retagged. -Matthias)
1904
1905		----------------------------------------------------------------------
1906		Name: Matthias Blume
1907		Date: 2002/02/15 14:17:00 EST
1908		Tag: blume-20020215-showbindings
1909		Description:
1910
1911		Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1912		the latter can be useful for exploring what bindings are available at
1913		the interactive prompt.  (The first function returns only the list
1914		of symbols that are really bound, the second prints those but also the
1915		ones that CM's autoloading mechanism knows about.)
1916
1917		----------------------------------------------------------------------
1918		Name: Matthias Blume
1919		Date: 2002/02/15 12:08:00 EST
1920		Tag: blume-20020215-iptrs
1921		Description:
1922
1923		Two improvements to ml-nlffigen:
1924
1925		1. Write files only if they do not exist or if their current contents
1926		do not coincide with what's being written.  (That is, avoid messing
1927		with the time stamps unless absolutely necessary.)
1928
1929		2. Implement a "repository" mechanism for generated files related
1930		to "incomplete pointer types".   See the README file for details.
1931
1932		----------------------------------------------------------------------
1933		Name: Matthias Blume
1934		Date: 2002/02/14 11:50:00 EST
1935		Tag: blume-20020214-quote
1936		Description:
1937
1938		Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1939		because of the new and improved tag generation scheme.  (Thanks to Allen
1940		Leung for pointing it out.)
1941
1942		----------------------------------------------------------------------
1943		Name: Lal George
1944		Date: 2002/02/14 09:55:27 EST 2002
1945		Tag: george-20020214-isabelle-bug
1946		Description:
1947
1948		Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1949		of Isabelle on the x86.
1950
1951		From Allen:
1952		-----------
1953		I've found the problem:
1954
1955		in ra-core.sml, I use the counter "blocked" to keep track of the
1956		true number of elements in the freeze queue.  When the counter goes
1957		to zero, I skip examining the queue.  But I've messed up the
1958		bookkeeping in combine():
1959
1960		else ();
1961		case !ucol of
1962		PSEUDO => (if !cntv > 0 then
1963		(if !cntu > 0 then blocked := !blocked - 1 else ();
1964		^^^^^^^^^^^^^^^^^^^^^^^
1965		moveu := mergeMoveList(!movev, !moveu)
1966		)
1967		else ();
1968
1969		combine() is called to coalesce two nodes u and v.
1970		I think I was thinking that if the move counts of u and v are both
1971		greater than zero then after they are coalesced then one node is
1972		removed from the freeze queue.  Apparently I was thinking that
1973		both u and v are of low degree, but that's clearly not necessarily true.
1974
1975
1976		02/12/2002:
1977		Here's the patch.  HOL now compiles.
1978
1979		I don't know how this impact on performance (compile
1980		time or runtime).  This bug caused the RA (especially on the x86)
1981		to go thru the potential spill phase when there are still nodes on the
1982		freeze queue.
1983
1984
1985
1986
1987		----------------------------------------------------------------------
1988		Name: Matthias Blume
1989		Date: 2002/02/13 22:40:00 EST
1990		Tag: blume-20020213-fptr-rtti
1991		Description:
1992
1993		Fixed a bug in ml-nlffigen that was introduced with one of the previous
1994		updates.
1995
1996		----------------------------------------------------------------------
1997		Name: Matthias Blume
1998		Date: 2002/02/13 16:41:00 EST
1999		Tag: blume-20020213-cmlpq
2000		Description:
2001
2002		Added new priority queue export symbols (which have just been added to
2003		smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
2004		not compile and the installer would choke.)
2005
2006		----------------------------------------------------------------------
2007		Name: Matthias Blume
2008		Date: 2002/02/13 16:15:00 EST
2009		Tag: blume-20020213-various
2010		Description:
2011
2012		1. More tweaks to ml-nlffigen:
2013
2014		- better internal datastructures (resulting in slight speedup)
2015		- "-match" option requires exact match
2016		- "localized" gensym counters (untagged structs/unions nested within
2017		other structs/unions or within typedefs get a fresh counter; their
2018		tag will be prefixed by a concatenation of their parents' tags)
2019		- bug fixes (related to calculation of transitive closure of types
2020		to be included in the output)
2021
2022		2. Minor Basis updates:
2023
2024		- added implementations for List.collate and Option.app
2025
2026		----------------------------------------------------------------------
2027		Name: Matthias Blume
2028		Date: 2002/02/11 15:55:00 EST
2029		Tag: blume-20020211-gensym
2030		Description:
2031
2032		Added a "-gensym" option to command line of ml-nlffigen.  This can be
2033		used to specify a "stem" -- a string that is inserted in all "gensym'd"
2034		names (ML structure names that correspond to unnamed C structs, unions,
2035		and enums), so that separate runs of ml-nlffigen do not clash.
2036
2037		----------------------------------------------------------------------
2038		Name: Matthias Blume
2039		Date: 2002/02/11 12:05:00 EST
2040		Tag: blume-20020211-gensml
2041		Description:
2042
2043		A quick fix for a problem with GenSML (in the pgraph-util library):
2044		Make generation of toplevel "local" optional.  (Strictly speaking,
2045		signature definitions within "local" are not legal SML.)
2046
2047		Other than that: updates to INSTALL and cm/TODO.
2048
2049		----------------------------------------------------------------------
2050		Name: Matthias Blume
2051		Date: 2002/02/08 15:00:00 EST
2052		Tag: blume-20020208-uniquepid
2053		Description:
2054
2055		0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
2056
2057		1. The installer (config/install.sh) has gotten smarter:
2058
2059		- Configuration options are a bit easier to specify now
2060		(in config/targets).
2061		- Bug in recognizing .tar.bz2 files fixed.
2062		- Installer automatically resolves dependencies between
2063		configuration options (e.g., if you ask for eXene, you will
2064		also get cml -- regardless whether you asked for it or not).
2065		- Installer can run in "quieter mode" by setting the environment
2066		variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
2067		"completely silent", though.
2068		- Build HashCons library as part of smlnj-lib.
2069
2070		2. A new scheme for assigning persistent identifiers to compilation
2071		units (and, by extension, to types etc.) has been put into place.
2072		This fixes a long-standing bug where types and even dynamic values
2073		can get internally confused, thereby compromising type safety
2074		(abstraction) and dynamic correctness.  See
2075
2076		http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
2077
2078		for an example of how things could go wrong until now.
2079
2080		The downside of the new scheme is that pids are not quite as
2081		persistent as they used to be: CM will generate a fresh pid
2082		for every compilation unit that it thinks it sees for the first
2083		time.  That means that if you compile starting from a clean, fresh
2084		source tree at two different times, you end up with different
2085		binaries.
2086
2087		Cutoff recompilation, however, has not been compromised because
2088		CM keeps pid information in special caches between runs.
2089
2090		----------------------------------------------------------------------
2091		Name: Lal George
2092		Date: 2002/02/07 15:34:13 EST 2002
2093		Tag: <none>
2094		Description:
2095
2096		Compilers that generate assembly code may produce  global labels
2097		whose value is resolved at link time. The various peephole optimization
2098		modules did not take this in account.
2099
2100		TODO. The Labels.addrOf function should really return an option
2101		type so that clients are forced to deal with this issue, rather
2102		than an exception being raised.
2103
2104		----------------------------------------------------------------------
2105		Name: Lal George
2106		Date: 2002/02/06 13:55:02 EST
2107		Tag: george-20020206-ra-breakup
2108		Description:
2109
2110		1. A bug fix from Allen.
2111
2112		A typo causes extra fstp %st(0)'s to be generated at compensation
2113		edges, which might cause stack underflow traps at runtime.  This
2114		occurs in fft where there are extraneous fstps right before the 'into'
2115		trap instruction (in this case they are harmless since none of the
2116		integers overflow.)
2117
2118		2. Pulled out various utility modules that were embedded in the modules
2119		of the register allocator. I need these modules for other purposes, but
2120		they are not complete enough to put into a library (just yet).
2121		----------------------------------------------------------------------
2122		Name: Matthias Blume
2123		Date: 2002/01/31 16:05:00 EST
2124		Tag: blume-20020131-sparc-ccalls
2125		Description:
2126
2127		1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
2128		of extra stack space by mistake.  Fixed.
2129
2130		2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
2131
2132		----------------------------------------------------------------------
2133		Name: Allen Leung
2134		Date: 2002/01/30
2135		Tag: leunga-20020130-nowhere-bug-fix
2136		Description:
2137
2138		MLRISC bug fixes:
2139		1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
2140		2. MachineInt.fromString was negating its value.
2141
2142		----------------------------------------------------------------------
2143		Name: Matthias Blume
2144		Date: 2002/01/29
2145		Tag: blume-20020129-INSTALL
2146		Description:
2147
2148		- Added somewhat detailed installation instructions (file INSTALL).
2149		- Fixed curl-detection bug in config/install.sh.
2150		- It is now possible to select the URL getter using the URLGETTER
2151		environment variable:
2152
2153		not set / "unknown"      --> automatic detection (script tries wget,
2154		curl, and lynx)
2155		"wget" / "curl" / "lynx" --> use the specified program (script "knows"
2156		how to properly invoke them)
2157		other                    --> use $URLGETTER directly, it must take
2158		precisely two command-line arguments
2159		(source URL and destination file name)
2160
2161		----------------------------------------------------------------------
2162		Name: Matthias Blume
2163		Date: 2002/01/28
2164		Tag: blume-20020128-sparc-ccalls
2165		Description:
2166
2167		- Fixed problem with calculation of "used" registers in sparc-c-calls.
2168		- Make use of the allocParam argument in sparc-c-calls.
2169
2170		----------------------------------------------------------------------
2171		Name: Matthias Blume
2172		Date: 2002/01/28
2173		Tag: blume-20020128-allocParam
2174		Description:
2175
2176		John Reppy:  Changes c-calls API to accept client-callback for
2177		allocating extra stack space.
2178		me: Corresponding changes to mlriscGen (using a dummy argument that
2179		does not change the current behavior).
2180
2181		----------------------------------------------------------------------
2182		Name: Matthias Blume
2183		Date: 2002/01/28 12:00:00
2184		Tag: Release_110_38
2185		Description:
2186
2187		This time for real!!!
2188
2189		----------------------------------------------------------------------
2190		Name: Matthias Blume
2191		Date: 2002/01/28 10:56:00 EST
2192		Tag: blume-20020128-retraction
2193		Description:
2194
2195		0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
2196		with blume-Release_110_38-retracted.)
2197
2198		1. Fixed a problem with incorrect rounding modes in real64.sml.
2199		(Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
2200
2201		2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
2202		and enums fixed.  The naming of corresponding ML identifiers should
2203		now be consistent again.
2204
2205		----------------------------------------------------------------------
2206		Name: Allen Leung
2207		Date: 2002/01/27
2208		Tag: leunga-20020127-nowhere
2209		Description:
2210
2211		Added a target called nowhere in the configuration scripts.
2212		Enabling this will build the MLRISC 'nowhere' tool (for translating
2213		programs with where-clauses into legal SML code) during installation.
2214
2215		----------------------------------------------------------------------
2216		Name: Matthias Blume
2217		Date: 2002/01/25 21:27:00 EST
2218		Tag: blume-Release_110_38-retracted
2219		Description:
2220
2221		Call it a (working) release!  Version is 110.38. Bootfiles are ready.
2222
2223		README will be added later.
2224
2225		!!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
2226		(Release_110_38) removed.  Reason: Last-minute bug fixes.
2227
2228		----------------------------------------------------------------------
2229		Name: Matthias Blume
2230		Date: 2002/01/25
2231		Tag: blume-20020125-ffi
2232		Description:
2233
2234		A large number of tweaks and improvements to ml-nlffi-lib and
2235		ml-nlffigen:
2236
2237		- ML represenation types have been streamlined
2238		- getter and setter functions work with concrete values, not abstract
2239		ones where possible
2240		- ml-nlffigen command line more flexible (see README file there)
2241		- some bugs have been fixed (hopefully)
2242
2243		----------------------------------------------------------------------
2244		Name: Lal George
2245		Date: 2002/01/24
2246		Tag: george-20020124-risc-ra-interface
2247		Description:
2248
2249		There is a dramatic simplification in the interface to the
2250		register allocator for RISC architectures as a result of making
2251		parallel copy instructions explicit.
2252
2253		----------------------------------------------------------------------
2254		Name: Matthias Blume
2255		Date: 2002/01/22
2256		Tag: blume-20020122-x86-ccalls
2257		Description:
2258
2259		Bug fix for c-calls on x86 (having to do with how char- and
2260		short-arguments are being handled).
2261
2262		----------------------------------------------------------------------
2263		Name: Matthias Blume
2264		Date: 2002/01/21
2265		Tag: blume-20020121-ff
2266		Description:
2267
2268		Another day of fiddling with the FFI...
2269
2270		1. Bug fix/workaround:  CKIT does not complain about negative array
2271		dimensions, so ml-nlffigen has to guard itself against this possibility.
2272		(Otherwise a negative dimension would send it into an infinite loop.)
2273
2274		2. Some of the abstract types (light objects, light pointers, most "base"
2275		types) in structure C are now eqtypes.
2276
2277		3. Added constructors and test functions for NULL function pointers.
2278
2279		----------------------------------------------------------------------
2280		Name: Matthias Blume
2281		Date: 2002/01/18
2282		Tag: blume-20020118-ready-for-new-release
2283		Description:
2284
2285		Made config/srcarchiveurl point to a new place.  (Will provide boot
2286		files shortly.)
2287
2288		Maybe we christen this to be 110.38?
2289
2290		----------------------------------------------------------------------
2291		Name: Matthias Blume
2292		Date: 2002/01/18
2293		Tag: blume-20020118-more-ffifiddle
2294		Description:
2295
2296		Today's FFI fiddling:
2297
2298		- Provided a structure CGetSet with "convenient" versions of C.Get.* and
2299		C.Set.* that use concrete (MLRep.*) arguments and results instead
2300		of abstract ones.
2301
2302		- Provided word-style bit operations etc. for "int" representation
2303		types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
2304		and Long.
2305
2306		----------------------------------------------------------------------
2307		Name: Matthias Blume
2308		Date: 2002/01/18
2309		Tag: blume-20020118-use-x86-fp
2310		Description:
2311
2312		Now that x86-fast-fp seems to be working, I turned it back on again
2313		by default.  (Seems to work fine now, even with the FFI.)
2314
2315		Other than that, I added some documentation about the FFI to
2316		src/ml-nlffigen/README and updated the FFI test examples in
2317		src/ml-nlffi-lib/Tests/*.
2318
2319		----------------------------------------------------------------------
2320		Name: Allen Leung
2321		Date: 2002/01/17
2322		Tag: leunga-20020117-x86-fast-fp-call
2323		Description:
2324
2325		1. Fixed a problem with handling return fp values when x86's fast fp
2326		mode is turned on.
2327
2328		2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
2329		of %f32.
2330
2331		3. Added a constructor INT32lit to the ast of MLRISC tools.
2332
2333		----------------------------------------------------------------------
2334		Name: Matthias Blume
2335		Date: 2002/01/16
2336		Tag: blume-20020116-ffifiddle
2337		Description:
2338
2339		More fiddling with the FFI interface:
2340
2341		- Make constness 'c instead of rw wherever possible.  This eliminates
2342		the need for certain explicit coercions.  (However, due to ML's
2343		value polymorphism, there will still be many cases where explicit
2344		coercions are necessary.  Phantom types are not the whole answer
2345		to modeling a subtyping relationship in ML.)
2346
2347		- ro/rw coersions for pointers added.  (Avoids the detour through */&.)
2348
2349		- "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
2350		clumsy workaround for varargs problem.)
2351
2352		----------------------------------------------------------------------
2353		Name: Lal George
2354		Date: 2002/01/15
2355		Tag: <none>
2356		Description:
2357
2358		1. Since COPY instructions are no longer native to the architecture,
2359		a generic functor can be used to implement the expandCopies function.
2360
2361		2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
2362		TEXT segment.
2363
2364		----------------------------------------------------------------------
2365		Name: Matthias Blume
2366		Date: 2002/01/15
2367		Tag: blume-20020115-ffiupdates
2368		Description:
2369
2370		1. Fix for bug resulting in single-precision float values being returned
2371		incorrectly from FFI calls.
2372
2373		2. Small modifications to C FFI API:
2374
2375		- memory-allocation routines return straight objects (no options)
2376		and raise an exception in out-of-memory situations
2377		- unsafe extensions to cast between function pointers and pointers
2378		from/to ints
2379		- added structure C_Debug as an alternative to structure C where
2380		pointer-dereferencing (|*| and |*!) always check for null-pointers
2381		- added open_lib' to DynLinkage;  open_lib' works like open_lib
2382		but also takes a (possibly empty) list of existing library handles
2383		that the current library depends on
2384
2385		----------------------------------------------------------------------
2386		Name: Matthias Blume
2387		Date: 2002/01/10
2388		Tag: blume-20020110-newffigen
2389		Description:
2390
2391		1. Updates to portable graph code.
2392
2393		2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
2394		more scalable now so that even huge interfaces such as the one
2395		for GTK compile in finite time and space. :-)
2396		See src/ml-nlffigen/README for details on what's new.
2397
2398		----------------------------------------------------------------------
2399		Name: Lal George
2400		Date: 2001/01/09 14:31:35 EST 2002
2401		Tag: george-20011206-rm-native-copy
2402		Description:
2403
2404		Removed the native COPY and FCOPY instructions
2405		from all the architectures and replaced it with the
2406		explicit COPY instruction from the previous commit.
2407
2408		It is now possible to simplify many of the optimizations
2409		modules that manipulate copies. This has not been
2410		done in this change.
2411
2412		----------------------------------------------------------------------
2413		Name: Lal George
2414		Date: 2001/12/06 16:50:13 EST 2001
2415		Tag: george-20011206-mlrisc-instruction
2416		Description:
2417
2418		Changed the representation of instructions from being fully abstract
2419		to being partially concrete. That is to say:
2420
2421		from
2422		type instruction
2423
2424		to
2425		type instr                              (* machine instruction *)
2426
2427		datatype instruction =
2428		LIVE of {regs: C.cellset, spilled: C.cellset}
2429		| KILL of {regs: C.cellset, spilled: C.cellset}
2430		| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
2431		| ANNOTATION of {i: instruction, a: Annotations.annotation}
2432		| INSTR of instr
2433
2434		This makes the handling of certain special instructions that appear on
2435		all architectures easier and uniform.
2436
2437		LIVE and KILL say that a list of registers are live or killed at the
2438		program point where they appear. No spill code is generated when an
2439		element of the 'regs' field is spilled, but the register is moved to
2440		the 'spilled' (which is present, more for debugging than anything else).
2441
2442		LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
2443		We used to generate:
2444
2445		DEFFREG f1
2446		f1 := f2 + f3
2447		trapb
2448
2449		but now generate:
2450
2451		f1 := f2 + f3
2452		trapb
2453		LIVE {regs=[f1,f2,f3], spilled=[]}
2454
2455		Furthermore, the DEFFREG (hack) required that all floating point instruction
2456		use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
2457		defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
2458		in a cleaner alpha implementation. (Hopefully, intel will not get rid of
2459		this architecture).
2460
2461		COPYXXX is intended to replace the parallel COPY and FCOPY  available on
2462		all the architectures. This will result in further simplification of the
2463		register allocator that must be aware of them for coalescing purposes, and
2464		will also simplify certain aspects of the machine description that provides
2465		callbacks related to parallel copies.
2466
2467		ANNOTATION should be obvious, and now INSTR represents the honest to God
2468		machine instruction set!
2469
2470		The <arch>/instructions/<arch>Instr.sml files define certain utility
2471		functions for making porting easier -- essentially converting upper case
2472		to lower case. All machine instructions (of type instr) are in upper case,
2473		and the lower case form generates an MLRISC instruction. For example on
2474		the alpha we have:
2475
2476		datatype instr =
2477		LDA of {r:cell, b:cell, d:operand}
2478		| ...
2479
2480		val lda : {r:cell, b:cell, d:operand} -> instruction
2481		...
2482
2483		where lda is just (INSTR o LDA), etc.
2484
2485		----------------------------------------------------------------------
2486		Name: Matthias Blume
2487		Date: 2001/11/22 21:40:00 EST
2488		Tag: Release_110_37
2489		Description:
2490
2491		Release 110.37.  This time for real.
2492
2493		----------------------------------------------------------------------
2494		Name: Matthias Blume
2495		Date: 2001/11/21 16:35:00 EST
2496		Tag: blume-20011121-foot-in-mouth
2497		Description:
2498
2499		Removed the "Release_110_37" tag because of a serious bug.
2500		This will be re-tagged once the bug is fixed.
2501
2502		----------------------------------------------------------------------
2503		Name: Matthias Blume
2504		Date: 2001/11/21 16:14:00 EST
2505		Tag: blume-20011121-forgottenfile
2506		Description:
2507
2508		Forgot to add a file.  (Just a .tex-file -- part of
2509		the CM manual source.)
2510
2511		----------------------------------------------------------------------
2512		Name: Matthias Blume
2513		Date: 2001/11/21 16:10:00 EST
2514		Tag: blume-20011121-invalid_110_37
2515		Description:
2516
2517		Note: I removed the original tag "Release_110_37" from this commit
2518		because we found a serious bug in all non-x86 backends.
2519		- Matthias
2520
2521		1. Modifications to the SML/NJ code generator and to the runtime system
2522		so that code object name strings are directly inserted into code
2523		objects at code generation time.  The only business the runtime system
2524		has with this is now to read the name strings on occasions.
2525		(The encoding of the name string has also changed somewhat.)
2526
2527		2. CM now implements a simple "set calculus" for specifying export lists.
2528		In particular, it is now possible to refer to the export lists of
2529		other libraries/groups/sources and form unions as well as differences.
2530		See the latest CM manual for details.
2531
2532		3. An separate notion of "proxy" libraries has again be eliminated from
2533		CM's model.  (Proxy libraries are now simply a special case of using
2534		the export list calculus.)
2535
2536		4. Some of the existing libraries now take advantage of the new set
2537		calculus.
2538		(Notice that not all libraries have been converted because some
2539		of the existing .cm-files are supposed to be backward compatible
2540		with 110.0.x.)
2541
2542		5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2543		"General.exnMessage"!  The former relies on a certain hook to be
2544		initialized, and that often does not happen in the stand-alone case.)
2545
2546		----------------------------------------------------------------------
2547		Name: Lal George
2548		Date: 2001/11/21  13:56:18 EST
2549		Tag: george-2001121-pseudo-ops
2550		Description:
2551
2552		Implemented a complete redesign of MLRISC pseudo-ops. Now there
2553		ought to never be any question of incompatabilities with
2554		pseudo-op syntax expected by host assemblers.
2555
2556		For now, only modules supporting GAS syntax are implemented
2557		but more should follow, such as MASM, and vendor assembler
2558		syntax, e.g. IBM as, Sun as, etc.
2559
2560		----------------------------------------------------------------------
2561		Name: Matthias Blume
2562		Date: 2001/11/14 11:52:00 EST
2563		Tag: blume-20011114-srcname
2564		Description:
2565
2566		1. Routed the name of the current source file to mlriscgen where it
2567		should be directly emitted into the code object.  (This last part
2568		is yet to be done.)
2569
2570		2. Some cleanup of the pgraph code to make it match the proposal that
2571		I put out the other day.  (The proposal notwithstanding, things are
2572		still in flux here.)
2573
2574		----------------------------------------------------------------------
2575		Name: Lal George
2576		Date: 2001/11/14 09:44:04 EST
2577		Tag:
2578		Description:
2579
2580		Fix for a backpatching bug reported by Allen.
2581
2582		Because the boundary between short and long span-dependent
2583		instructions is +/- 128, there are an astounding number of
2584		span-dependent instructions whose size is over estimated.
2585
2586		Allen came up with the idea of letting the size of span
2587		dependent instructions be non-monotonic, for a maxIter
2588		number of times, after which the size must be monotonically
2589		increasing.
2590
2591		This table shows the number of span-dependent instructions
2592		whose size was over-estimated as a function of maxIter, for the
2593		file Parse/parse/ml.grm.sml:
2594
2595		maxIter            # of instructions:
2596		10                      687
2597		20                      438
2598		30                      198
2599		40                        0
2600
2601		In compiling the compiler, there is no significant difference in
2602		compilation speed between maxIter=10 and maxIter=40. Actually,
2603		my measurements showed that maxIter=40 was a tad faster than
2604		maxIter=10! Also 96% of the  files in the compiler reach a fix
2605		point within 13 iterations, so fixing maxIter at 40, while high,
2606		is okay.
2607
2608		----------------------------------------------------------------------
2609		Name: Matthias Blume
2610		Date: 2001/10/31 15:25:00 EST
2611		Tag: blume-20011031-pgraph
2612		Description:
2613
2614		CKIT:
2615		* Changed the "Function" constructor of type Ast.ctype to carry optional
2616		argument identifiers.
2617		* Changed the return type of TypeUtil.getFunction accordingly.
2618		* Type equality ignores the argument names.
2619		* TypeUtil.composite tries to preserve argument names but gives up quickly
2620		if there is a mismatch.
2621
2622		installation script:
2623		* attempts to use "curl" if available (unless "wget" is available as well)
2624
2625		CM:
2626		* has an experimental implementation of "portable graphs" which I will
2627		soon propose as an implementation-independent library format
2628		* there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2629
2630		NLFFI-LIB:
2631		* some cleanup (all cosmetic)
2632
2633		NLFFIGEN:
2634		* temporarily disabled the mechanism that suppresses ML output for
2635		C definitions whose identifiers start with an underscore character
2636		* generate val bindings for enum constants
2637		* user can request that only one style (light or heavy) is being used;
2638		default is to use both (command-line arguments: -heavy and -light)
2639		* fixed bug in handling of function types involving incomplete pointers
2640		* generate ML entry points that take record arguments (i.e., using
2641		named arguments) for C functions that have a prototype with named
2642		arguments
2643		(see changes to CKIT)
2644
2645		----------------------------------------------------------------------
2646		Name: Allen Leung
2647		Date: 2001/10/27 20:34:00 EDT
2648		Tag: leunga-20011027-x86-fast-fp-call
2649		Description:
2650
2651		Fixed the bug described in blume-20010920-slowfp.
2652
2653		The fix involves
2654		1. generating FCOPYs in FSTP in ia32-svid
2655		2. marking a CALL with the appropriate annotation
2656
2657		----------------------------------------------------------------------
2658		Name: Matthias Blume
2659		Date: 2001/10/16 11:32:00 EDT
2660		Tag: blume-20011016-netbsd
2661		Description:
2662
2663		Underscore patch from Chris Richards (fixing problem with compiling
2664		runtime system under recent NetBSD).
2665
2666		----------------------------------------------------------------------
2667		Name: Allen Leung
2668		Date: 2001/10/12 17:18:32 EDT 2001
2669		Tag: leung-20011012-x86-printflowgraph
2670		Description:
2671
2672		X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2673
2674		----------------------------------------------------------------------
2675		Name: Lal George
2676		Date: 2001/10/11 23:51:34 EDT
2677		Tag: george-20011011-too-many-instrs
2678		Description:
2679
2680		The representation of a program point never expected to see more
2681		than 65536 instructions in a basic block!
2682
2683		----------------------------------------------------------------------
2684		Name: Lal George
2685		Date: 2001/10/09 09:41:37 EDT
2686		Tag: george-20011008-mlrisc-labels
2687		Description:
2688
2689		Changed the machine description files to support printing of
2690		local and global labels in assembly code, based on host assembler
2691		conventions.
2692
2693		----------------------------------------------------------------------
2694		Name: Matthias Blume
2695		Date: 2001/09/25 15:25:00 EDT
2696		Tag: blume-20010925-exninfo
2697		Description:
2698
2699		I provided a non-hook implementation of exnName (at the toplevel) and
2700		made the "dummy" implementation of exnMessage (at the toplevel) more
2701		useful: if nothing gets "hooked in", then at least you are going to
2702		see the exception name and a message indicating why you don't see more.
2703
2704		[For the time being, programs that need exnMessage and want to use
2705		ml-build should either use General.exnMessage (strongly recommended) or
2706		refer to structure General at some other point so that CM sees a
2707		static dependency.]
2708
2709		[Similar remarks go for "print" and "use":  If you want to use their
2710		functionality in stand-alone programs generated by ml-build, then use
2711		TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2712
2713		----------------------------------------------------------------------
2714		Name: Matthias Blume
2715		Date: 2001/09/20 17:28:00 EDT
2716		Tag: blume-20010920-slowfp
2717		Description:
2718
2719		Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2720
2721		----------------------------------------------------------------------
2722		Name: Matthias Blume
2723		Date: 2001/09/20 17:20:00 EDT
2724		Tag: blume-20010920-canonicalpaths
2725		Description:
2726
2727		0. Updated the BOOT file (something that I forgot to do earlier).
2728
2729		1. Small internal change to CM so that it avoids "/../" in filenames
2730		as much as possible (but only where it is safe).
2731
2732		2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2733		that arguments that contain delimiters are passed through correctly.
2734		This change also means that all "special" arguments of the form
2735		@SMLxxx... must come first.
2736
2737		3. Changed install script to put relative anchor names for tool commands
2738		into pathconfig.
2739
2740		----------------------------------------------------------------------
2741		Name: Matthias Blume
2742		Date: 2001/09/18 15:35:00 EDT
2743		Tag: blume-20010918-readme11036
2744		Description:
2745
2746		Added README files.
2747
2748		----------------------------------------------------------------------
2749		Name: Matthias Blume
2750		Date: 2001/09/18 11:45:00 EDT
2751		Tag: Release_110_36 (retag)
2752		Description:
2753
2754		Fixed mistake in config/preloads. Retagged as 110.36.
2755
2756		----------------------------------------------------------------------
2757		Name: Matthias Blume
2758		Date: 2001/09/18 09:40:00 EDT
2759		Tag: Release_110_36_orig (tag changed)
2760		Description:
2761
2762		New version (110.36).  New bootfiles.
2763
2764		----------------------------------------------------------------------
2765		Name: Matthias Blume
2766		Date: 2001/09/14 16:15:00 EDT
2767		Tag: blume-20010914-x86fastfp
2768		Description:
2769
2770		John committed some changes that Allen made, in particular a (hopefully)
2771		correctly working version of the x86-fp module.
2772
2773		I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
2774		flag to "true".  Everything seems to compile to a fixpoint ok, and
2775		"mandelbrot" speeds up by about 15%.
2776
2777		----------------------------------------------------------------------
2778		Name: Matthias Blume
2779		Date: 2001/09/13 11:20:00 EDT
2780		Tag: blume-20010913-minimal
2781		Description:
2782
2783		1. Stefan Monnier's patch to fix a miscompilation problem that
2784		was brought to light by John Reppy's work on Moby.
2785
2786		2. Implemented a minimal "structure Compiler" that contains just
2787		"version" and "architecture".  The minimal version will be
2788		available when the full version is not.  This is for backward-
2789		compatibility with code that wants to test Compiler.version.
2790
2791		----------------------------------------------------------------------
2792		Name: Matthias Blume
2793		Date: 2001/08/28 14:03:00 EDT
2794		Tag: blume-20010828-ml-lex
2795		Description:
2796
2797		Fix for bug 1581, received from Neophytos Michael.
2798
2799		----------------------------------------------------------------------
2800		Name: Matthias Blume
2801		Date: 2001/08/27 11:20:00 EDT
2802		Tag: blume-20010827-readme11035
2803		Description:
2804
2805		Fleshed out the README file for 110.35.
2806
2807		----------------------------------------------------------------------
2808		Name: Matthias Blume
2809		Date: 2001/08/24 17:10:00 EDT
2810		Tag: Release_110_35
2811		Description:
2812
2813		New version number (110.35).  New bootfiles.
2814
2815		----------------------------------------------------------------------
2816		Name: Lal George
2817		Date: 2001/08/24 13:47:18 EDT 2001
2818		Tag: george-20010824-MLRISC-graphs
2819		Description:
2820
2821		removed clusters from MLRISC completely and replaced with graphs.
2822
2823		----------------------------------------------------------------------
2824		Name: Matthias Blume
2825		Date: 2001/08/23 17:50:00 EDT
2826		Tag: blume-20010823-toplevel
2827		Description:
2828
2829		- some reorganization of the code that implements various kinds of
2830		environments in the compiler (static, dynamic, symbolic, combined)
2831		- re-implemented the EnvRef module so that evalStream works properly
2832		(if the stream contains references to "use", "CM.make", etc.)
2833		- cleaned up evalloop.sml and interact.sml (but they need more cleaning)
2834
2835		----------------------------------------------------------------------
2836		Name: Matthias Blume
2837		Date: 2001/08/20 15:50 EDT
2838		Tag: blume20010820-slipup
2839		Description:
2840
2841		I forgot to commit a few files.  Here they are...
2842
2843		----------------------------------------------------------------------
2844		Name: Matthias Blume
2845		Date: 2001/08/20 15:35:00 EDT
2846		Tag: blume-20010820-debugprof
2847		Description:
2848
2849		!!!! NEW BOOTFILES !!!!
2850
2851		This is another round of reorganizing the compiler sources.  This
2852		time the main goal was to factor out all the "instrumentation"
2853		passes (for profiling and backtracing) into their own library.
2854		The difficulty was to do it in such a way that it does not depend
2855		on elaborate.cm but only on elabdata.cm.
2856
2857		Therefore there have been further changes to both elaborate.cm and
2858		elabdata.cm -- more "generic" things have been moved from the former
2859		to the latter.  As a result, I was forced to split the assignment
2860		of numbers indicating "primtyc"s into two portions: SML-generic and
2861		SML/NJ-specific.  Since it would have been awkward to maintain,
2862		I bit the bullet and actually _changed_ the mapping between these
2863		numbers and primtycs.  The bottom line of this is that you need
2864		a new set of bin- and bootfiles.
2865
2866		I have built new bootfiles for all architectures, so doing a fresh
2867		checkout and config/install.sh should be all you need.
2868
2869		The newly created library's name is
2870
2871		$smlnj/viscomp/debugprof.cm
2872
2873		and its sources live under
2874
2875		src/compiler/DebugProf
2876
2877		----------------------------------------------------------------------
2878		Name: Matthias Blume
2879		Date: 2001/08/15 17:15:00 EDT
2880		Tag: blume-20010815-compreorg
2881		Description:
2882
2883		This is a first cut at reorganizing the CM libraries that make up the
2884		core of the compiler.  The idea is to separate out pieces that could
2885		be used independently by tools, e.g., the parser, the typechecker, etc.
2886
2887		The current status is a step in this direction, but it is not quite
2888		satisfactory yet.  Expect more changes in the future.
2889
2890		Here is the current (new) organization...
2891
2892		What used to be $smlnj/viscomp/core.cm is now divided into
2893		six CM libraries:
2894
2895		$smlnj/viscomp/basics.cm
2896		/parser.cm
2897		/elabdata.cm
2898		/elaborate.cm
2899		/execute.cm
2900		/core.cm
2901
2902		The CM files for these libraries live under src/system/smlnj/viscomp.
2903		All these libraries are proxy libraries that contain precisely
2904		one CM library component.  Here are the locations of the components
2905		(all within the src/compiler tree):
2906
2907		Basics/basics.cm
2908		Parse/parser.cm
2909		ElabData/elabdata.cm
2910		Elaborator/elaborate.cm
2911		Execution/execute.cm
2912		core.cm
2913
2914		[This organization is the same that has been used already
2915		for a while for the architecture-specific parts of the visible
2916		compiler and for the old version of core.cm.]
2917
2918		As you will notice, many source files have been moved from their
2919		respective original locations to a new home in one of the above
2920		subtrees.
2921
2922		The division of labor between the new libraries is the following:
2923
2924		basics.cm:
2925		- Simple, basic definitions that pertain to many (or all) of
2926		the other libraries.
2927		parser.cm:
2928		- The SML parser, producing output of type Ast.dec.
2929		- The type family for Ast is also defined and exported here.
2930		elabdata.cm:
2931		- The datatypes that describe input and output of the elaborator.
2932		This includes types, absyn, and static environments.
2933		elaborator.cm:
2934		- The SML/NJ type checker and elaborator.
2935		This maps an Ast.dec (with a given static environment) to
2936		an Absyn.dec (with a new static environment).
2937		- This libraries implements certain modules that used to be
2938		structures as functors (to remove dependencies on FLINT).
2939		execute.cm:
2940		- Everything having to do with executing binary code objects.
2941		- Dynamic environments.
2942		core.cm:
2943		- SML/NJ-specific instantiations of the elaborator and MLRISC.
2944		- Top-level modules.
2945		- FLINT (this should eventually become its own library)
2946
2947		Notes:
2948
2949		I am not 100% happy with the way I separated the elaborator (and its
2950		data structures) from FLINT.  Two instances of the same problem:
2951
2952		1. Data structures contain certain fields that carry FLINT-specific
2953		information.  I hacked around this using exn and the property list
2954		module from smlnj-lib.  But the fact that there are middle-end
2955		specific fields around at all is a bit annoying.
2956
2957		2. The elaborator calculates certain FLINT-related information.  I tried
2958		to make this as abstract as I could using functorization, but, again,
2959		the fact that the elaborator has to perform calculations on behalf
2960		of the middle-end at all is not nice.
2961
2962		3. Having to used exn and property lists is unfortunate because it
2963		weakens type checking.  The other alternative (parameterizing
2964		nearly *everything*) is not appealing, though.
2965
2966		I removed the "rebinding =" warning hack because due to the new organization
2967		it was awkward to maintain it.  As a result, the compiler now issues some of
2968		these warnings when compiling init.cmi during bootstrap compilation. On
2969		the plus side, you also get a warning when you do, for example:
2970		val op = = Int32.+
2971		which was not the case up to now.
2972
2973		I placed "assign" and "deref" into the _Core structure so that the
2974		code that deals with the "lazy" keyword can find them there.  This
2975		removes the need for having access to the primitive environment
2976		during elaboration.
2977
2978		----------------------------------------------------------------------
2979		Name: Matthias Blume
2980		Date: 2001/08/13
2981		Tag: blume-20010813-closures
2982		Description:
2983
2984		This fix was sent to us by Zhong Shao.  It is supposed to improve the
2985		performance of certain loops by avoiding needless closure allocation.
2986
2987		----------------------------------------------------------------------
2988		Name: Lal George
2989		Date: 2001/07/31 10:03:23 EDT 2001
2990		Tag: george-20010731-x86-fmalloc
2991		Description: Fixed bug in x86 calls
2992
2993		There was a bug where call instructions would mysteriously
2994		vanish. The call instruction had to be one that returned
2995		a floating point value.
2996
2997		----------------------------------------------------------------------
2998		Name: Lal George
2999		Date: 2001/07/19 16:36:29 EDT 2001
3000		Tag: george-20010719-simple-cells
3001		Description:
3002
3003		I have dramatically simplified the interface for CELLS in MLRISC.
3004
3005		In summary, the cells interface is broken up into three parts:
3006
3007		1. CellsBasis : CELLS_BASIS
3008
3009		CellsBasis is a top level structure and common for all
3010		architectures.  it contains the definitions of basic datatypes
3011		and utility  functions over these types.
3012
3013		2. functor Cells() : CELLS
3014
3015		Cells generates an interface for CELLS that incorporates the
3016		specific resources on the target architecture, such as the
3017		presence of special register classes, their number and size,
3018		and various useful substructures.
3019
3020		3. <ARCH>CELLS
3021
3022		e.g. SparcCells: SPARCCELLS
3023
3024		<ARCH>CELLS usually contains additional bindings for special
3025		registers  on the architecture, such as:
3026
3027		val r0 : cell           (* register zero *)
3028		val y : cell            (* Y register *)
3029		val psr : cell          (* processor status register *)
3030		...
3031
3032		The structure returned by applying the Cells functor is opened
3033		in this interface.
3034
3035		The main implication of all this is that the datatypes for cells is
3036		split between CellsBasis and CELLS -- a fairly simple change for user
3037		code.
3038
3039		In the old scheme the CELLS interface had a definitional binding of
3040		the form:
3041
3042		signature CELLS = sig
3043
3044		structure CellsBasis = CellsBasis
3045
3046		...
3047
3048		end
3049
3050		With all the sharing constraints that goes on in MLRISC, this old
3051		design  quickly leads to errors such as:
3052
3053		"structure definition spec inside of sharing ... "
3054
3055
3056		and appears to require an unacceptable amount of sharing and where
3057		constraint hackery.
3058
3059		I think this error message (the interaction of definitional specs and
3060		sharing) requires more explanation on our web page.
3061
3062		----------------------------------------------------------------------
3063		Name: Matthias Blume
3064		Date: 2001/07/19 15:00:00 EDT
3065		Tag: blume-20010719-libreorg
3066		Description:
3067
3068		This update puts together a fairly extensive but straightforward change
3069		to the way the libraries that implement the interactive system are
3070		organized:
3071
3072		The biggest change is the elimination of structure Compiler.  As a
3073		replacement for this structure, there is now a CM library
3074		(known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
3075		that exports all the substructures of the original structure Compiler
3076		directly.  So instead of saying Compiler.Foo.bar one now simply
3077		says Foo.bar.  (The CM libraries actually export a collection of
3078		structures that is richer than the collection of substructures of
3079		structure Compiler.)
3080
3081		To make the transition smooth, there is a separate library called
3082		$smlnj/compiler/compiler.cm which puts together and exports the
3083		original structure Compiler (or at least something very close to it).
3084
3085		There are five members of the original structure Compiler
3086		that are not exported directly but which instead became members
3087		of a new structure Backend (described by signature BACKEND).  These are:
3088		structure Profile (: PROFILE), structure Compile (: COMPILE), structure
3089		Interact (: INTERACT), structure Machine (: MACHINE), and val
3090		architecture (: string).
3091
3092		Structure Compiler.Version has become structure CompilerVersion.
3093
3094		Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
3095		by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
3096		or x86, respectively.
3097		Each of these exports the same frontend structures that
3098		$smlnj/compiler.cm exports.  But they do not have a structure Backend
3099		and instead export some structure <Arch>Backend where <Arch> is Alpha32,
3100		Hppa, PPC, Sparc, or X86, respectively.
3101
3102		Library $smlnj/compiler/all.cm exports the union of the exports of
3103		$smlnj/compiler/<arch>.cm
3104
3105		There are no structures <Arch>Compiler anymore, use
3106		$smlnj/compiler/<arch>.cm instead.
3107
3108		Library host-compiler-0.cm is gone.  Instead, the internal library
3109		that instantiates CM is now called cm0.cm.  Selection of the host
3110		compiler (backend) is no longer done here but. (Responsibility for it
3111		now lies with $smlnj/compiler/current.cm.  This seems to be more
3112		logical.)
3113
3114		Many individual files have been moved or renamed.  Some files have
3115		been split into multiple files, and some "dead" files have been deleted.
3116
3117		Aside from these changes to library organization, there are also changes
3118		to the way the code itself is organized:
3119
3120		Structure Binfile has been re-implemented in such a way that it no
3121		longer needs any knowledge of the compiler.  It exclusively deals
3122		with the details of binfile layout.  It no longer invokes the
3123		compiler (for the purpose of creating new prospective binfile
3124		content), and it no longer has any knowledge of how to interpret
3125		pickles.
3126
3127		Structure Compile (: COMPILE) has been stripped down to the bare
3128		essentials of compilation.  It no longer deals with linking/execution.
3129		The interface has been cleaned up considerably.
3130
3131		Utility routines for dealing with linking and execution have been
3132		moved into their own substructures.
3133
3134		(The ultimate goal of these changes is to provide a light-weight
3135		binfile loader/linker (at least for, e.g., stable libraries) that
3136		does not require CM or the compiler to be present.)
3137
3138		CM documentation has been updated to reflect the changes to library
3139		organization.
3140
3141		----------------------------------------------------------------------
3142		Name: Matthias Blume
3143		Date: 2001/07/10 17:30:00 EDT
3144		Tag: Release_110_34
3145		Description:
3146
3147		Minor tweak to 110.34 (re-tagged):
3148
3149		- README.html file added to CVS repository
3150		- runtime compiles properly under FreeBSD 3.X and 4.X
3151
3152		----------------------------------------------------------------------
3153		Name: Matthias Blume
3154		Date: 2001/07/10 17:30:00 EDT
3155		Tag: Release_110_34
3156		Description:
3157
3158		New version number (110.34). New bootfiles.
3159
3160		----------------------------------------------------------------------
3161		Name: Matthias Blume
3162		Date: 2001/07/09 16:00:00 EDT
3163		Tag: blume-20010709-more-varargs
3164		Description:
3165
3166		I changed the handling of varargs in ml-nlffigen again:
3167		The ellipsis ... will now simply be ignored (with an accompanying warning).
3168
3169		The immediate effect is that you can actually call a varargs function
3170		from ML -- but you can't actually supply any arguments beyond the ones
3171		specified explicitly.  (For example, you can call printf with its format
3172		string, but you cannot pass additional arguments.)
3173
3174		This behavior is only marginally more useful than the one before, but
3175		it has the advantage that a function or, more importantly, a function
3176		type never gets dropped on the floor, thus avoiding follow-up problems with
3177		other types that refer to the offending one.
3178
3179		----------------------------------------------------------------------
3180		Name: Matthias Blume
3181		Date: 2001/07/09 11:25:00 EDT
3182		Tag: blume-20010709-varargs
3183		Description:
3184
3185		1. ckit-lib.cm now exports structure Error
3186		2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
3187		with a warning accompanied by a source location.  Moreover, it
3188		merely skips the offending function or type and proceeds with the
3189		rest of its work.u  As a result, one can safely feed C code containing
3190		"..." to ml-nlffigen.
3191		3. There are some internal improvements to CM, providing slightly
3192		more general string substitutions in the tools subsystem.
3193
3194		----------------------------------------------------------------------
3195		Name: Matthias Blume
3196		Date: 2001/06/27 15:10:00 EDT
3197		Tag: blume-20010627-concur
3198		Description:
3199
3200		Fixed a small bug in CM's handling of parallel compilation.
3201		(You could observe the bug by Control-C-interrupting an ordinary
3202		CMB.make or CM.stabilize and then attaching some compile servers.
3203		The result was that all of a sudden the previously interrupted
3204		compilation would continue on its own.  This was because of
3205		an over-optimization: CM did not bother to clean out certain queues
3206		when no servers were attached "anyway", resulting in the contents
3207		of these queues to grab control when new servers did get attached.)
3208
3209		There is also another minor update to the CM manual.
3210
3211		----------------------------------------------------------------------
3212		Name: Matthias Blume
3213		Date: 2001/06/26 16:15:00 EDT
3214		Tag: blume-20010626-cmdoc
3215		Description:
3216
3217		Minor typo fixed in CM manual (syntax diagram for libraries).
3218
3219		----------------------------------------------------------------------
3220		Name: Matthias Blume
3221		Date: 2001/06/25 22:55:00 EDT
3222		Tag: blume-20010625-x86pc
3223		Description:
3224
3225		Fixed a nasty bug in the X86 assembly code that caused signal
3226		handlers to fail (crash) randomly.
3227
3228		----------------------------------------------------------------------
3229		Name: Matthias Blume
3230		Date: 2001/06/25 12:05:00 EDT
3231		Tag: blume-20010625-nlffigen
3232		Description:
3233
3234		This update fixes a number of minor bugs in ml-nlffigen as reported by
3235		Nick Carter <nbc@andrew.cmu.edu>.
3236
3237		1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
3238		2. Default names for generated files are now derived from the name of
3239		the C file *without its directory*.  In particular, this causes generated
3240		files to be placed locally even if the C file is in some system directory.
3241		3. Default names for generated signatures and structures are also derived
3242		from the C file name without its directory.  This avoids silly things
3243		like "structure GL/GL".
3244		(Other silly names are still possible because ml-nlffigen does not do
3245		a thorough check of whether generated names are legal ML identifiers.
3246		When in doubt, use command line arguments to force particular names.)
3247
3248		----------------------------------------------------------------------
3249		Name: Matthias Blume
3250		Date: 2001/06/21 12:25:00 EDT
3251		Tag: blume-20010621-eXene
3252		Description:
3253
3254		eXene now compiles and (sort of) works again.
3255
3256		The library name (for version > 110.33) is $/eXene.cm.
3257
3258		I also added an new example in src/eXene/examples/nbody.  See the
3259		README file there for details.
3260
3261		----------------------------------------------------------------------
3262		Name: Matthias Blume
3263		Date: 2001/06/20 16:40:00 EDT
3264		Tag: blume-20010620-cml
3265		Description:
3266
3267		CML now compiles and works again.
3268
3269		Libraries (for version > 110.33):
3270
3271		$cml/cml.cm            Main CML library.
3272		$cml/basis.cm          CML's version of $/basis.cm.
3273		$cml/cml-internal.cm   Internal helper library.
3274		$cml/core-cml.cm       Internal helper library.
3275		$cml-lib/trace-cml.cm  Tracing facility.
3276		$cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
3277
3278		The installer (config/install.sh) has been taught how to properly
3279		install this stuff.
3280
3281		----------------------------------------------------------------------
3282		Name: Matthias Blume
3283		Date: 2001/06/19 17:55:00 EDT
3284		Tag: blume-20010619-instantiate
3285		Description:
3286
3287		This un-breaks the fix for bug 1432.
3288		(The bug was originally fixed in 110.9 but I broke it again some
3289		time after that.)
3290
3291		----------------------------------------------------------------------
3292		Name: Matthias Blume
3293		Date: 2001/06/19 17:25:00 EDT
3294		Tag: blume-20010619-signals
3295		Description:
3296
3297		This should (hopefully) fix the long-standing signal handling bug.
3298		(The runtime system was constructing a continuation record with an
3299		incorrect descriptor which would cause the GC to drop data on the floor...)
3300
3301		----------------------------------------------------------------------
3302		Name: Matthias Blume
3303		Date: 2001/06/15 15:05:00 EDT
3304		Tag: blume-20010615-moresparc
3305		Description:
3306
3307		Here is a short late-hour update related to Sparc c-calls:
3308
3309		-- made handling of double-word arguments a bit smarter
3310
3311		-- instruction selection phase tries to collapse certain clumsily
3312		constructed ML-Trees; typical example:
3313
3314		ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
3315
3316		This currently has no further impact on SML/NJ since mlriscGen does
3317		not seem to generate such patterns in the first place, and c-calls
3318		(which did generate them in the beginning) has meanwhile been fixed
3319		so as to avoid them as well.
3320
3321		----------------------------------------------------------------------
3322		Name: Matthias Blume
3323		Date: 2001/06/15 15:05:00 EDT
3324		Tag: blume-20010615-sparc
3325		Description:
3326
3327		The purpose of this update is to provide an implementation of NLFFI
3328		on Sparc machines.
3329
3330		Here are the changes in detail:
3331
3332		* src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
3333		the Sparc implementation of the c-calls API.
3334		* The Sparc backend of SML/NJ has been modified to uniformely use %fp
3335		for accessing the ML frame.  Thus, we have a real frame pointer and
3336		can freely modify %sp without need for an omit-frame-ptr phase.
3337		The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
3338		this case.
3339		* ml-nlffigen has been taught to produce code for different architectures
3340		and calling conventions.
3341		* In a way similar to what was done in the x86 case, the Sparc
3342		backend uses its own specific extension to mltree.  (For example,
3343		it needs to be able to generate UNIMP instructions which are part
3344		of the calling convention.)
3345		* ml-nlffi-lib was reorganized to make it more modular (in particular,
3346		to make it easier to plug in new machine- and os-dependent parts).
3347
3348		There are some other fairly unrelated bug fixes and cleanups as well:
3349
3350		* I further hacked the .cm files for MLRISC tools (like MDLGen) so
3351		that they properly share their libraries with existing SML/NJ libraries.
3352		* I fixed a minor cosmetic bug in CM, supressing certain spurious
3353		follow-up error messages.
3354		* Updates to CM/CMB documentation.
3355
3356		TODO items:
3357
3358		* MLRISC should use a different register as its asmTemp on the Sparc.
3359		(The current %o2 is a really bad choice because it is part of the
3360		calling conventions, so things might interfere in unexpected ways.)
3361
3362		----------------------------------------------------------------------
3363		Name: Matthias Blume
3364		Date: 2001/06/07
3365		Tag: blume-20010607-calls
3366		Description:
3367
3368		A number of internal changes related to C calls and calling conventions:
3369
3370		1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
3371		number of bytes popped implicitly (by the callee).  In most cases
3372		this field is 0 but on x86/win32 it is some non-zero value.  This
3373		is information provided for the benefit of the "omit-frameptr" pass.
3374		2. The CALL instruction on the x86 carries a similar "pops" field.
3375		The instruction selection phase copies its value from the ML-Tree
3376		CALL statement.
3377		3. On all other architectures, the instruction selection phase checks
3378		whether "pops=0" and complains if not.
3379		4. The c-calls implementation for x86 now accepts two calling conventions:
3380		"ccall" and "stdcall".  When "ccall" is selected, the caller cleans
3381		up after the call and pops is set to 0.  For "stdcall", the caller
3382		does nothing, leaving the cleanup to the callee; pops is set to
3383		the number of bytes that were pushed onto the stack.
3384		5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
3385		distinguish between "ccall" and "stdcall".
3386		6. The UNIMP instruction has been added to the supported Sparc instruction
3387		set. (This is needed for implementing the official C calling convention
3388		on this architecture.)
3389		7. I fixed some of the .cm files under src/MLRISC/Tools to make them
3390		work with the latest CM.
3391
3392		----------------------------------------------------------------------
3393		Name: Matthias Blume
3394		Date: 2001/06/05 15:10:00 EDT
3395		Tag: blume-20010605-cm-index
3396		Description:
3397
3398		0. The "lambdasplit" parameter for class "sml" in CM has been documented.
3399
3400		1. CM can now generate "index files".  These are human-readable files
3401		that list on a per-.cm-file basis each toplevel symbol defined or
3402		imported.  The location of the index file for
3403		<p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
3404		To enable index-file generation, set CM.Control.generate_index to true
3405		or export an environment-symbol: export CM_GENERATE_INDEX=true.
3406
3407		The CM manual has been updated accordingly.
3408
3409		2. I made some slight modifications to the c-calls API in MLRISC.
3410
3411		a) There is now a callback to support saving/restoring of
3412		dedicated but caller-save registers around the actual call
3413		instruction.
3414		b) One can optionally specify a comment-annotation for the
3415		call instruction.
3416
3417		3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
3418		(For example, the comment annotation shows the C prototype of
3419		the function being called.)
3420
3421		----------------------------------------------------------------------
3422		Name: Matthias Blume
3423		Date: 2001/06/01 13:30:00 EDT
3424		Tag: blume-20010601-nlffi-cleanup
3425		Description:
3426
3427		This is mostly a cleanup of MLFFI stuff:
3428
3429		- some signature files have been put into a more exposed place
3430		- the ugly 'f type parameter is gone (simplifies types tremendously!)
3431		- ml-nlffigen changed accordingly
3432		- tutorial updated
3433
3434		Other changes:
3435
3436		- author's affiliation in CM manual(s) updated
3437		- some more recognized keywords added to Allen's sml.sty
3438
3439		----------------------------------------------------------------------
3440		Name: Matthias Blume
3441		Date: 2001/05/25 15:30:00 EDT
3442		Tag: blume-20010525-iptr
3443		Description:
3444
3445		- put the official 110.33-README (as it appears on the ftp server) under
3446		CVS
3447		- fixed a small bug related to incomplete pointer types in
3448		ml-nlffigen
3449		- small cosmetic change to the ml-nlffi-lib's "arr" type constructor
3450		(it does not need the 'f type parameter)
3451
3452		----------------------------------------------------------------------
3453		Name: Matthias Blume
3454		Date: 2001/05/23 14:30:00 EDT
3455		Tag: Release_110_33
3456		Description:
3457
3458		New version number (110.33).  New bootfiles.
3459
3460		----------------------------------------------------------------------
3461		Name: Matthias Blume
3462		Date: 2001/05/22 18:06:00 EDT
3463		Tag: blume-20010522-targets
3464		Description:
3465
3466		Made install.sh use file config/targets.customized if it exists, falling
3467		back to config/targets if it doesn't.  This way one can have a customized
3468		version of the targets file without touching the "real thing", thus
3469		eliminating the constant fear of accidentally checking something bogus
3470		back into the CVS repository...  (File config/targets.customized must
3471		not be added to the repository!)
3472
3473		----------------------------------------------------------------------
3474		Name: Matthias Blume
3475		Date: 2001/05/22 16:30:00 EDT
3476		Tag: blume-20010522-minitut
3477		Description:
3478
3479		1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
3480		struct returns.
3481		2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
3482		incomplete, preliminary documentation for NLFFI.
3483
3484		----------------------------------------------------------------------
3485		Name: Matthias Blume
3486		Date: 2001/05/14 11:30:00 EDT
3487		Tag: blume-20010514-script
3488		Description:
3489
3490		Some bugs in install script fixed.
3491
3492		In addition to that I also made a slight change to the NLFFI API:
3493		Functors generated by ml-nlffigen now take the dynamic library as a
3494		straight functor argument, not as a suspended one.  (The original
3495		functor code used to force the suspension right away anyway, so there
3496		was nothing gained by this complication of the interface.)
3497
3498		----------------------------------------------------------------------
3499		Name: Matthias Blume
3500		Date: 2001/05/11 14:35:00 EDT
3501		Tag: blume-20010511-ml-nlffi
3502		Description:
3503
3504		I finally took the plunge and added my new FFI code to the main
3505		repository.  For x86-linux it is now ready for prime-time.
3506
3507		There are two new subdirectories of "src":
3508
3509		- ml-nlffi-lib:
3510		The utility library for programs using the FFI interface.
3511		Here is the implementation of $/c.cm and its associated low-level
3512		partners $/c-int.cm and $/memory.cm.
3513		- ml-nlffigen:
3514		A stand-alone program for generating ML glue code from C source
3515		code.
3516
3517		Building ml-nlffigen requires $/ckit-lib.cm.
3518
3519		The config/install.sh script has been updates to do the Right Thing
3520		(hopefully).
3521
3522		Notice that the source tree for the C-Kit will not be put under "src"
3523		but directly under the installation root directory.  (This is the
3524		structure that currently exists on the CVS server when you check out
3525		module "sml".)  Fortunately, config/install.sh knows about this oddity.
3526
3527		Bugs: No documentation yet.
3528
3529		----------------------------------------------------------------------
3530		Name: Matthias Blume
3531		Date: 2001/05/09 16:35:00 EDT
3532		Tag: blume-20010509-cpscontract
3533		Description:
3534
3535		Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
3536		wrapper/unwrapper elimination did not decrement usage counts and some
3537		dead variables got overlooked by the dead-up logic.)
3538
3539		----------------------------------------------------------------------
3540		Name: Lal George
3541		Date: 2001/05/08  17:26:09 EDT
3542		Tag: george-20010508-omit-frameptr
3543		Description:
3544
3545		Changes to implement the omit-frame-pointer optimization to support
3546		raw C calls. For now, there is only support on the Intel x86, but
3547		other architectures will follow as more experience is gained with this.
3548
3549
3550		----------------------------------------------------------------------
3551		Name: Matthias Blume
3552		Date: 2001/05/07 14:40:00 EDT
3553		Tag: blume-20010507-proxies
3554		Description:
3555
3556		I made into "proxy libraries" all libraries that qualify for such a
3557		change.  (A qualifying library is a library that has another library or
3558		groups as its sole member and repeats that member's export list
3559		verbatim.  A proxy library avoids this repetition by omitting its export
3560		list, effectively inheriting the list that its (only) member exports.
3561		See the CM manual for more explanation.)
3562		The main effect is that explicit export lists for these libraries
3563		do not have to be kepts in sync, making maintenance a bit easier.
3564
3565		I also added copyright notices to many .cm-files.
3566
3567		Last but not least, I made a new set of bootfiles.
3568
3569		----------------------------------------------------------------------
3570		Name: Matthias Blume
3571		Date: 2001/05/04 17:00:00 EDT
3572		Tag: blume-20010504-cm-lsplit
3573		Description:
3574
3575		0. John merged pending changes to $/smlnj-lib.cm
3576
3577		1. Allen's previous change accidentally backed out of one of Lal's
3578		earlier changes.  I undid this mistake (re-introducing Lal's change).
3579
3580		2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
3581		within the compiler where applicable.  There is some code simplification
3582		because of that.
3583
3584		3. The "split" phase (in FLINT) is now part of the default list of phases.
3585		Compiler.Control.LambdaSplitting.* can be used to globally control the
3586		lambda-splitting (cross-module-inlining) engine.  In addition to that,
3587		it can now also be controlled on a per-source basis: CM has been taught
3588		a new tool parameter applicable to ML source files.
3589
3590		- To turn lambda-splitting off completely:
3591		local open Compiler.Control.LambdaSplitting in
3592		val _ = set Off
3593		end
3594		- To make "no lambda-splitting" the global default (but allow per-source
3595		overriding); this is the initial setting:
3596		local open Compiler.Control.LambdaSplitting in
3597		val _ = set (Default NONE)
3598		end
3599		- To make "lambda-splitting with aggressiveness a" the global default
3600		(and allow per-source overriding):
3601		local open Compiler.Control.LambdaSplitting in
3602		val _ = set (Default (SOME a))
3603		end
3604
3605		- To turn lambda-splitting off for a given ML souce file (say: a.sml)
3606		write (in the respective .cm-file):
3607		a.sml (lambdasplitting:off)
3608		- To turn lambda-splitting for a.sml on with minimal aggressiveness: