Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 1067, Fri Feb 15 17:08:17 2002 UTC	revision 1759, Thu Feb 24 22:46:30 2005 UTC
#	Line 13	Line 13
13	Description:	Description:
14
15	----------------------------------------------------------------------	----------------------------------------------------------------------
16		Name: Matthias Blume (blume (at) tti - c (dot) org)
17		Date: 2005/02/24 16:50:00 CST
18		Tag: blume-20050224-cml-atom
19		Description:
20
21		Simpler and at the same time more general implementation of structure
22		Atom in CML.
23
24		----------------------------------------------------------------------
25		Name: Matthias Blume (blume (at) tti - c (dot) org)
26		Date: 2005/02/15 17:35:00 CST
27		Tag: blume-20050215-tools
28		Description:
29
30		Created new "tools" directory under "src" and moved "TraceDebugProf"
31		there.
32
33		----------------------------------------------------------------------
34		Name: Matthias Blume (blume (at) tti - c (dot) org)
35		Date: 2005/02/10 17:55:00 CST
36		Tag: blume-20050210-longlong
37		Description:
38
39		Implemented "long long" arguments and results for NLFFI.  (Only the
40		PPC/MacOS implementation is complete, the other backends still need to
41		be updated.)
42
43		----------------------------------------------------------------------
44		Name: Matthias Blume (blume (at) tti - c (dot) org)
45		Date: 2005/01/24 17:40:00 CST
46		Tag: blume-20050124-mlyacc
47		Description:
48
49		Minor cleanup in ML-Yacc rule printing mechanism.  This should fix a
50		problem with certain "as" patterns which previously got rendered
51		using incorrect syntax.
52
53		----------------------------------------------------------------------
54		Name: Matthias Blume (blume (at) tti - c (dot) org)
55		Date: 2005/01/18 12:00:00 CST
56		Tag: blume-20050118-profile
57		Description:
58
59		Made time profiling code (interrupt handler) in runtime system aware
60		of new array representation.
61
62		----------------------------------------------------------------------
63		Name: Matthias Blume (blume (at) tti - c (dot) org)
64		Date: 2005/01/14 18:00:00 CST
65		Tag: blume-20050114-heap2exec
66		Description:
67
68		Implemented new (but still experimental) heap2exec facility.  This is
69		tested under Mac OS X and should work under Linux (will test shortly).
70		It will probably also work on the Sparc (will test some time later).
71
72		- removed old "HACKED_STANDALONE" hack from runtime
73
74		To be able to test this, uncomment the request for "heap2asm" in
75		config/targets prior to installation.  (Notice that this is different
76		from "heap2exec" mentioned below.  Not a typo.)
77
78		To perform an actual test, run the command
79
80		$ bin/heap2exec heapfile execfile
81
82		(You can put heap2exec on your shell's path.)
83
84		For example, run
85
86		$ bin/heap2exec bin/.heap/ml-yacc.ppc-darwin mly
87
88		This will create a standalone executable called "mly" which you can
89		then invoke directly as a command.
90
91		----------------------------------------------------------------------
92		Name: Matthias Blume (blume (at) tti - c (dot) org)
93		Date: 2005/01/07 11:44:00 CST
94		Tag: blume-20050107-mlstring
95		Description:
96
97		fixed off-by-one error in ML_STRING macro (globals.c)
98
99		----------------------------------------------------------------------
100		Name: Matthias Blume (blume (at) tti - c (dot) org)
101		Date: 2004/12/23 18:00:00 CST
102		Tag: blume-20041223-santa
103		Description:
104
105		Made ml-build script "smarter" (but only very little).
106
107		----------------------------------------------------------------------
108		Name: Matthias Blume (blume (at) tti - c (dot) org)
109		Date: 2004/12/21 15:05:00 CST
110		Tag: blume-20041221-longlong
111		Description:
112
113		* Implemented access to signed and unsigned long long data in NLFFI.
114		(The parameter-passing part of the picture has not complete. But
115		data structure access seems to work.)
116
117		* Fixed CM's incorrect assumption that the PPC is little-endian.
118		(On the Mac, it is big-endian.  And that's currently our only
119		PPC platform.)
120
121		----------------------------------------------------------------------
122		Name: Matthias Blume (blume (at) tti - c (dot) org)
123		Date: 2004/12/21 12:50:00 CST
124		Tag: blume-20041221-memory
125		Description:
126
127		Some cleanup in the $c/memory.cm library: separated some concerns by
128		moving allocation code and memory access code each into their own
129		files.
130
131		----------------------------------------------------------------------
132		Name: Allen Leung (leunga (at) reservoir (dot) com)
133		Date: 2004/12/17 16:12:00 EST
134		Tag: leunga-20041217-cygwin-smlnj-home
135		Description:
136
137		The Unix I/O library of SML/NJ on cygwin does not understand
138		Windows style pathname, so problems arise when SMLNJ_HOME is set
139		to a Windows style pathname.  _run-sml now converts SMLNJ_HOME
140		to a POSIX pathname on cygwin.
141
142		----------------------------------------------------------------------
143		Name: Matthias Blume (blume (at) tti - c (dot) org)
144		Date: 2004/12/16 13:35:00 CST
145		Tag: Release_110_52
146		Description:
147
148		Last-minute changes incorporated into 110.52.  Release tag moved.
149
150		The changes:
151		- HashString.hashString' -> HashString.hashSubstring
152		- bug fix in UnivariateStats
153
154		----------------------------------------------------------------------
155		Name: Matthias Blume (blume (at) tti - c (dot) org)
156		Date: 2004/12/15 23:40:00 CST
157		Tag: blume-20041215-hashSubstring
158		Description:
159
160		- HashString.hashString' -> HashString.hashSubstring
161		- corresponding changes in atom.sml
162		- "de-compressed" (aka. un-obfuscated) code for UnivariateStats and
163		added some comments
164
165		----------------------------------------------------------------------
166		Name: Matthias Blume (blume (at) tti - c (dot) org)
167		Date: 2004/12/15 15:30:00 CST
168		Tag: (Release_110_52)
169		Description:
170
171		New working version (110.52).  NEW BOOTFILES!
172
173		----------------------------------------------------------------------
174		Name: Matthias Blume (blume (at) tti - c (dot) org)
175		Date: 2004/12/15 12:45:00 CST
176		Tag: blume-20041215-spaces
177		Description:
178
179		More on the space problem (this time for Win32).
180
181		----------------------------------------------------------------------
182		Name: Matthias Blume (blume (at) tti - c (dot) org)
183		Date: 2004/12/14 17:30:00 CST
184		Tag: blume-20041214-spaces
185		Description:
186
187		Hacked some of the scripts (in particular: the installer) to cope with
188		spaces in filenames a bit better.  But beware: the current "solution"
189		is likely still full of bugs and inherently incomplete.  (We need to
190		do away with those shell scripts for a comprehensive solution.)
191
192		----------------------------------------------------------------------
193		Name: Matthias Blume (blume (at) tti - c (dot) org)
194		Date: 2004/12/13 14:45:00 CST
195		Tag: blume-20041213-ml-makedepend
196		Description:
197
198		Fixed bug in code for ml-makedepend.
199
200		----------------------------------------------------------------------
201		Name: Matthias Blume (blume (at) tti - c (dot) org)
202		Date: 2004/12/09 16:30:00 CST
203		Tag: blume-20041209-statistics
204		Description:
205
206		Added two simple but potentially useful statistics modules to SML/NJ Library.
207		(See CHANGES file there.)
208
209		----------------------------------------------------------------------
210		Name: Matthias Blume (blume (at) tti - c (dot) org)
211		Date: 2004/12/01 16:50:00 CST
212		Tag: blume-20041201-atom
213		Description:
214
215		smlnj-lib:
216
217		Added function HashString.hashString' for substrings.
218		Hand-inlined CharVector.foldl into HashString (for speed).
219		Modified implementation of structure Atom to avoid extracting
220		strings from substrings unless necessary.
221		(Also see CHANGES file for smlnj-lib.)
222
223		----------------------------------------------------------------------
224		Name: Matthias Blume (blume (at) tti - c (dot) org)
225		Date: 2004/11/24 22:15:00 CST
226		Tag: blume-20041124-cml
227		Description:
228
229		Made sure CML compiles when Position = Int64.
230
231		----------------------------------------------------------------------
232		Name: Matthias Blume (blume (at) tti - c (dot) org)
233		Date: 2004/11/24 14:30:00 CST
234		Tag: blume-20041124-position
235		Description:
236
237		The compiler can now be compiled in a mode that makes structure Position
238		equal to Int64.  The default, however, is unchanged (Position = Int31)
239		for the time being.
240
241		To enable 64-bit positions, use the following procedure:
242
243		1. Start sml
244		2. Autoload $smlnj/cmb.cm (if not already autoloaded)
245		3. Type
246		#set (CMB.symval "USE_64_BIT_POSITIONS") (SOME 1);
247		4. Run CMB.make() as usual.
248
249		This is barely tested.  The only test so far was a little SML program
250		counting the number of characters in an 8-gigabyte file by
251		reading it character-by-character.  That test was successful.
252
253		In support of 64-bit positions, a number of new functions have been
254		added to the runtime system.
255
256		----------------------------------------------------------------------
257		Name: Matthias Blume (blume (at) tti - c (dot) org)
258		Date: 2004/11/23 14:45:00 CST
259		Tag: blume-20041123-useFile
260		Description:
261
262		Fixed a problem with unhelpful error messages related to problems with
263		.cm- or .sml files that appear as part of the sml command line.
264
265		----------------------------------------------------------------------
266		Name: Matthias Blume (blume (at) tti - c (dot) org)
267		Date: 2004/11/18 15:40:00 CST
268		Tag: Release_110_51
269		Description:
270
271		New working version (110.51).  NEW BOOTFILES!
272
273		----------------------------------------------------------------------
274		Name: Matthias Blume (blume (at) tti - c (dot) org)
275		Date: 2004/11/18 15:35:00 CST
276		Tag: <none>
277		Description:
278
279		Enabled dlopen and friends for FreeBSD (as recommended by Johannes 5
280		Joemann).
281
282		----------------------------------------------------------------------
283		Name: Allen Leung (leunga (at) reservoir (dot) com)
284		Date: 2004/11/17 16:05:21 EST 2004
285		Tag: leunga-20041117-mlrisc-live-kill
286		Description:
287
288		Added support for MLTree constructs LIVE and KILL
289		to all the architectures.
290
291		----------------------------------------------------------------------
292		Name: Matthias Blume (blume (at) tti - c (dot) org)
293		Date: 2004/11/13 00:20:00 CST
294		Tag: blume-20041113-versiontool
295		Description:
296
297		- Stripped down the versiontool: It now only handles the version number.
298		The date string is generated at bootstrap time (during makeml).
299
300		- In a previous commit, fixed a minor issue with how polyequal is being
301		translated.  In particular, the code now "looks through" abstractions.
302		This results in slightly fewer polyEqual warnings and hopefully slightly
303		more efficient code.  Important examples for where this matters are
304		the new int64 and word64 types.
305
306		----------------------------------------------------------------------
307		Name: Matthias Blume (blume (at) tti - c (dot) org)
308		Date: 2004/11/12 00:30:00 CST
309		Tag: blume-20041112-int64
310		Description:
311
312		Structure Int64 fully hooked in.  (The implementation is not very
313		efficient, though.)
314
315		----------------------------------------------------------------------
316		Name: Matthias Blume (blume (at) tti - c (dot) org)
317		Date: 2004/11/11 17:30:00 CST
318		Tag: blume-20041111-more64
319		Description:
320
321		All the pieces of Word64 are now there, with the exception of the
322		conversions from and to LargeWord. (Eventually these need to be identities,
323		but for the time being they don't even make sense because LargeWord is
324		32-bit wide.)
325
326		Also started to add similar support for Int64, but major pieces of that
327		are still missing.
328
329		----------------------------------------------------------------------
330		Name: Matthias Blume (blume (at) tti - c (dot) org)
331		Date: 2004/11/11 00:15:00 CST
332		Tag: blume-20041111-word64
333		Description:
334
335		Structure Word64 is now (almost) complete, word literals and patterns
336		seem to work.  There are a few odd pieces missing.  In particular,
337		I didn't do the {from,to}LargeWord parts because LargeWord is still Word32
338		at the moment.
339
340		Making Word64 official would mean that LargeWord becomes Word64.  But
341		this requires extreme care because most word-word conversions have to
342		go through LargeWord, so making a mistake means loss of efficiency or
343		worse.  Eventually there will be a solution similar to (but actually
344		simpler than) what I did with IntInf.
345
346		----------------------------------------------------------------------
347		Name: Matthias Blume (blume (at) tti - c (dot) org)
348		Date: 2004/11/10 18:12:00 CST
349		Tag: blume-20041110-64bit
350		Description:
351
352		More 64-bit hacking (but still not even half-way there yet).
353		Also, some assorted improvements to the handling of 8-bit words.
354
355		----------------------------------------------------------------------
356		Name: Matthias Blume (blume (at) tti - c (dot) org)
357		Date: 2004/11/09 17:50:00 CST
358		Tag: <none>
359		Description:
360
361		Started adding some infrastructure for supporting 64-bit int- and
362		word-types.  (Still in its very early stages.)
363
364		----------------------------------------------------------------------
365		Name: Matthias Blume (blume (at) tti - c (dot) org)
366		Date: 2004/10/28 10:45:00 CDT
367		Tag: Release_110_50
368		Description:
369
370		New working version (110.50).  NEW BOOTFILES!
371
372		=====================
373
374		Also:
375
376		- Changend config/srcarchiveurl from a file just
377		containing the URL string into a file containing
378		shell script code.  The code has access to the $VERSION variable.
379		- Made corresponding changes to config/install.sh and config/unpack.
380		- Default contents of config/srcarchiveurl uses $VERSION and
381		normally does not have to be edited to reflect a version change.
382
383		(As a result, a version change can be done by just editing
384		config/version, the rest is now automatic.)
385
386		----------------------------------------------------------------------
387		Name: Matthias Blume (blume (at) tti - c (dot) org)
388		Date: 2004/10/27 17:50:00 CDT
389		Tag: blume-20041027-btrace-msg
390		Description:
391
392		BackTrace.monitor now also reports the source of the exception that
393		triggered the trace.
394
395		----------------------------------------------------------------------
396		Name: Matthias Blume (blume (at) tti - c (dot) org)
397		Date: 2004/10/27 17:20:00 CDT
398		Tag: blume-20041027-x86-c-calls
399		Description:
400
401		This is the HISTORY entry for two earlier commits, both concerning
402		the x86 c-calls code in MLRISC:
403
404		- added a missing LOAD in the code that deals with struct arguments
405		- made sure the caller does not add the wrong number of bytes to the
406		stack pointer after a call of a function returning a struct
407		(the callee already pops the implicit argument which points to
408		the space reserved for the result)
409
410		----------------------------------------------------------------------
411		Name: Allen Leung (leunga (at) reservoir (dot) com)
412		Date: 2004/10/24 14:00:00 EST
413		Tag: leunga-20041024-x86-gas-fucomip
414		Description:
415
416		John discovered a bug in the syntax of fucomip.
417		The opcodes FU?COMIP? have been changed to
418
419		fu?comip? %st(i), %st
420
421		----------------------------------------------------------------------
422		Name: Matthias Blume (blume (at) tti - c (dot) org)
423		Date: 2004/10/20 15:06:00 CDT
424		Tag: blume-20041020-standalone-backtrace
425		Description:
426
427		Added a mechanism for getting back-trace information from standalone
428		programs.  Here is how it works:
429
430		1. The part of the program from which you want to get backtrace
431		information (usually the whole program) should be wrapped with
432		BackTrace.monitor.  This is a (unit->'a)->'a function, and your
433		main program could be modified from something like
434
435		fun main (pgm, args) = ...
436
437		to
438
439		fun main (pgm, args) = BackTrace.monitor (fn () => ...)
440
441		2. To be able to access BackTrace.monitor, you have to add
442		library $smlnj-tdp/plugins.cm to the .cm file that contains your
443		main function.
444
445		3. Remove all compiled code (i.e., all the .cm/ subdirectories that
446		CM might have created in the past for your project).
447
448		4. Build the system using this command line:
449
450		ml-build -Ctdp.instrument=true \$smlnj-tdp/back-trace.cm \
451		myprog.cm MyProg.main myprog
452
453		instead of the usual
454
455		ml-build myprog.cm MyProg.main myprog
456
457		I changed a library name:
458
459		$/trace-debug-profile.cm --> $smlnj-tdp/plugins.cm
460
461		New libraries:
462
463		$smlnj-tdp/back-trace.cm
464		-- when loaded causes the back-trace plugin to be installed
465		$smlnj-tdp/coverage.cm
466		-- when loaded causes the coverage plugin to be installed
467
468		----------------------------------------------------------------------
469		Name: Matthias Blume (blume (at) tti - c (dot) org)
470		Date: 2004/10/18 16:45:00 CDT
471		Tag: blume-20041018-groupowner
472		Description:
473
474		Added an "obsolete" warning for the "group owner" syntax to CM's parser.
475
476		Eliminated group owner specs from .cm files throughout the source tree.
477
478		----------------------------------------------------------------------
479		Name: Matthias Blume (blume (at) tti - c (dot) org)
480		Date: 2004/10/15 15:45:00 CDT
481		Tag: blume-20041015-coverage
482		Description:
483
484		* Test coverage tool added!
485		* Further reorganization of tracing-, debugging-, and profiling support:
486
487		- moved original BTImp -- now called BackTrace -- into a separate
488		library called $/trace-debug-profile.cm
489		- eliminated all mentions of BTrace from SMLofNJ.Internals
490		- only the instrumentation mechanism is now left in the compiler proper
491		- BackTrace module is a plugin which is NOT plugged in by default
492		- Coverage module is another such plugin
493
494		To get the benefits of any of these plugin modules, the code in
495		question must be compiled with tdp instrumentation turned on.  This
496		can be done by setting SMLofNJ.Internals.TDP.mode to true.
497		(The ref cell is also controlled via the -Ctdp.instrument=... switch.)
498
499		Plugins are selected at link time.  (Pre-compiled instrumented code
500		can be re-loaded with different plugins in effect.)  When an
501		instrumented module is linked, whatever plugins are at that time
502		enabled will come into effect for that module.
503
504		To enable the back-trace plugin, load library $/trace-debug-profile.cm
505		and invoke BackTrace.install() (e.g., from the interactive prompt).
506		To enable the coverage plugin, load the same library and invoke
507		Coverage.install().
508
509		Back-traces are generated automatically on uncaught exceptions and
510		when the code in question explicitly invokes BackTrace.trigger().
511
512		Coverage (and execution frequency-) information must be queried
513		explicitly by calling Coverage.not_covered and Coverage.hot_spots.
514
515		----------------------------------------------------------------------
516		Name: Matthias Blume (blume (at) tti - c (dot) org)
517		Date: 2004/10/14 17:40:00 CDT
518		Tag: blume-20041014-tdp-core
519		Description:
520
521		Snapshot of a significant overhaul of how the trace/debug/profile support
522		is hooked into the system (specifically: Core and SMLofNJ.Internals).
523
524		----------------------------------------------------------------------
525		Name: Matthias Blume (blume (at) tti - c (dot) org)
526		Date: 2004/10/13 16:34:00 CDT
527		Tag: blume-20041013-tdp
528		Description:
529
530		Some rationalization of names:
531
532		structure BTrace -> structure TDPInstrument
533		etc.
534
535		This is is preparation of using the original back-trace
536		instrumentation for other purposes.  "TDP" stands for
537		Trace/Debug/Profile.
538
539		The control flag controlling whether instrumentation is on or off is now
540		registered under a different name, so instead of running sml as
541
542		sml -Cinstrument.btrace-mode=true
543
544		one has to say
545
546		sml -Ctdp.instrument=true
547
548		----------------------------------------------------------------------
549		Name: Matthias Blume (blume (at) tti - c (dot) org)
550		Date: 2004/10/11 16:37:00 CDT
551		Tag: blume-20041011-regions
552		Description:
553
554		Made some minor modifications to elabcore.sml to have source regions
555		be propagated more tightly -- resulting in better (i.e., smaller)
556		regions being reported in error- and debug messages.
557
558		----------------------------------------------------------------------
559		Name: Matthias Blume (blume (at) tti - c (dot) org)
560		Date: 2004/10/08 22:50:00 CDT
561		Tag: blume-20041008-cmkw
562		Description:
563
564		Fixed handling of keywords in .cm files: After seeing "is" the lexer
565		treats subsequent occurrences of "group", "library", "source", "is",
566		"*", and "-" as ordinary identifiers rather than keywords.
567
568		Most seriously, this fixes a problem with CM's "shell" tool.  The tool
569		is supposed to accept a tool argument called "source", but this did
570		not work because of the clash with the keyword.
571
572		----------------------------------------------------------------------
573		Name: Matthias Blume (blume (at) tti - c (dot) org)
574		Date: 2004/10/07 16:00:00 CDT
575		Tag: blume-20041007-cleanup
576		Description:
577
578		Assorted cleanup work:
579
580		- got rid of intstrmap in favor of using the library's
581		hash table implementation
582		- threw out most of the pathnames stuff, as it was not used anyway
583		- simplified tokentable implementation
584		- fixed some minor spelling errors
585
586		----------------------------------------------------------------------
587		Name: Matthias Blume (blume (at) tti - c (dot) org)
588		Date: 2004/10/06 15:15:15 CDT
589		Tag: blume-20041006-handler
590		Description:
591
592		Cleaned up the absyn to reflect the invariant that HANDLE always
593		carries a FNexp as part of the type definition.  This eliminates some
594		superfluous sanity checks at runtime down the road.
595
596		Some minor cleanup of the btrace code.
597
598		----------------------------------------------------------------------
599		Name: Matthias Blume (blume (at) tti - c (dot) org)
600		Date: 2004/10/01 10:20:30 CDT
601		Tag: blume-20041001-slave
602		Description:
603
604		Added hack to make slave mode work in the presence of the version
605		tool.  (Still, since the master does two passes over the code for
606		CMB.make, the release number gets bumped twice when slaves are
607		attached.  I don't know if this is worth fixing...)
608
609		----------------------------------------------------------------------
610		Name: Matthias Blume (blume (at) tti - c (dot) org)
611		Date: 2004/09/30 10:55:00 CDT
612		Tag: blume-20040930-version
613		Description:
614
615		* Moved the "version" magic into its own little library under
616		src/system/smlnj/internal.  This avoids expensive reconstruction of
617		a stable src/compiler/core.cm.
618
619		* At the same time, structure CompilerVersion is now known as
620		structure SMLNJVersion.
621
622		* Arranged for the version tool to NOT kick in when rebuilding the system
623		(makeml -rebuild, fixpt).  Otherwise one would never reach a fixpoint.
624		Also, loading the versiontool does not work when rebuilding the system
625		because CM is not properly initialized at that time.
626
627		----------------------------------------------------------------------
628		Name: Matthias Blume (blume (at) tti - c (dot) org)
629		Date: 2004/09/29 14:00:00 CDT
630		Tag: blume-20040929-autoversion
631		Description:
632
633		Implemented some CM magic to have
634		file src/compiler/TopLevel/main/version.sml
635		generated automagically.
636		The version is taken from two files: config/version and config/release.
637		The first is expected to contain a two-part version number such as 110.49.
638		The second should contain a single number, but it may be missing.
639
640		If the environment variable VERSIONTOOL_BUMP_RELEASE is defined at the
641		time the version tool is loaded (which is the first time you say CMB.make),
642		then the tool will increment the value stored in config/release every
643		time CMB.make is invoked.
644
645		The binfile format is now insensitive to anything beyond the first
646		two components of a version number, so bumping the release does not render
647		binfiles incompatible.  Auto-bumping can be used to keep track of versions
648		during development without invalidating existing binfiles.
649
650		In any case, every CMB.make updates the date information in version.sml.
651		(This is the date that is printed in the banner.)
652
653		----------------------------------------------------------------------
654		Name: Matthias Blume (blume (at) tti - c (dot) org)
655		Date: 2004/09/28 10:53:00 CDT
656		Tag: blume-20040928-controls
657		Description:
658
659		Some cleanup of the controls code.
660
661		----------------------------------------------------------------------
662		Name: Matthias Blume (blume (at) tti - c (dot) org)
663		Date: 2004/09/27 22:08:00 CDT
664		Tag: blume-20040927-controls
665		Description:
666
667		Added two pieces of functionality to the Controls interface:
668
669		1.   val save'restore: 'a control -> unit -> unit
670
671		grabs the current value of the control in stage 1 and restores it
672		in stage 2.
673
674		2.   val set' : 'a control * 'a -> unit -> unit
675
676		stores the given value into the control in stage 2 (i.e., delayed)
677		but does all error checking in stage 1.
678		(This is for string controls that need to do parse their argument --
679		something that might fail.  In some cases, notably in CM, one
680		already knows the intended argument but wants to delay the actual
681		assignment until a time when error recovery would be more difficult.)
682
683		Changed the handling of controls in tool arguments to classes "sml" and
684		"lazysml":
685		- use Controls.save'restore as a more robust way of restoring the
686		old value (in particular: without having to re-parse the string)
687		- use controls to handle the "overload" keyword in the init group
688		(I believe this change actually fixes a long-standing obscure bug.)
689
690		----------------------------------------------------------------------
691		Name: Matthias Blume (blume (at) tti - c (dot) org)
692		Date: 2004/09/27 17:00:00 CDT
693		Tag: blume-20040927-lazysml
694		Description:
695
696		Added a new tool class called "lazysml" to CM's tool chest.  The only
697		difference to "sml" is that compilation is done with Control.lazysml
698		set to true.  A source of class "lazysml" is automatically recognized
699		by a file name suffix of ".lml".
700
701		In addition to the above feature, the original class "sml" now also
702		supports a tool argument "lazy" which has the same effect.  As a
703		result, the following three lines are equivalent:
704
705		foo.sml : lazysml
706		foo.sml : sml (lazy)
707		foo.sml (lazy)
708
709		The setting goes into effect both during parsing and during
710		compilation.  The original setting is restored right after parsing and
711		after compilation, respectively.
712
713		In addition to all the above, there is also a general mechanism to set
714		ANY of the "controls" that are available at the command line via
715		"-C..." on a per-sml-file basis.  The same rules that apply for "lazy"
716		apply as well. (In fact, "lazy" is implemented as a special case of
717		the general mechanism.)
718
719		The .cm file syntax uses a new keyword tool argument called "with".
720		There are several ways of indicating the desired settings:
721
722		foo.sml (with:parser.quotations=true)
723		foo.sml (with:(name:parser.quotations value:true))
724		foo.sml (with:(name:name1 value:value1 name:name2 value:value2 ...))
725		foo.sml (with:(name1=value1 name2=value2 ...))
726		foo.sml (with:(name1=value1 name:name2 value:value2 name3=value3 ...))
727
728		etc.
729
730		Another possible abbreviation is to leave out the =v or value:v part
731		if the name refers to a boolean control (in which case the value is
732		taken to be true).  Thus, one could get lazy sml also by saying:
733
734		foo.sml (with:parser.lazy-keyword=true)
735		foo.sml (with:parser.lazy-keyword)
736		foo.sml (with:(name:parser.lazy-keyword value:true))
737		foo.sml (with:(name:parser.lazy-keyword))
738
739		----------------------------------------------------------------------
740		Name: Matthias Blume (blume (at) tti - c (dot) org)
741		Date: 2004/09/24 16:22:00 CDT
742		Tag: blume-20040924-ppc-long-branch
743		Description:
744
745		Turned message about "emiting long form of branch" off by default.
746		Added a control flag to turn it back on when desired.
747
748		----------------------------------------------------------------------
749		Name: Matthias Blume (blume (at) tti - c (dot) org)
750		Date: 2004/09/24 16:05:00 CDT
751		Tag: blume-20040924-rounding
752		Description:
753
754		Applied patch for setting rounding modes under Mac OS X.  Thanks to
755		Melissa O'Neill for providing the code!
756
757		----------------------------------------------------------------------
758		Name: Matthias Blume (blume (at) tti - c (dot) org)
759		Date: 2004/09/23 17:30:00 CDT
760		Tag: blume-20040923-envvars
761		Description:
762
763		1. Changed definition of type ControlRegistry.registry_tree to
764		include control_info (i.e., the name of the controlling environment
765		variable).
766
767		2. Added command-line flags -e and -E to print the names of environment
768		variables that can be used to control internal settings.
769		(This uses the new API mentioned in 1.)
770
771		----------------------------------------------------------------------
772		Name: Matthias Blume (blume (at) tti - c (dot) org)
773		Date: 2004/09/13 16:50:00 CDT
774		Tag: Release_110_49
775		Description:
776
777		New working version (110.49).  NEW BOOTFILES!
778
779		----------------------------------------------------------------------
780		Name: Matthias Blume
781		Date: 2004/09/13 16:20:00 CDT
782		Tag: blume-20040913-config-mlrisc
783		Description:
784
785		Put target "mlrisc" back into the default list.
786		(There is no harm in having it, and some users have expressed their
787		wish to have "mlrisc" included by default.)
788
789		----------------------------------------------------------------------
790		Name: John Reppy
791		Date: 2004/09/13
792		Tag: jhr-20040913-signals
793		Description:
794
795		Fixed the signal masking code to properly nest mask/unmask operations
796		on a per-signal basis.
797
798		----------------------------------------------------------------------
799		Name: Matthias Blume (blume (at) tti - c (dot) org)
800		Date: 2004/09/08 13:20:00 CDT
801		Tag: blume-20040908-heap-magic
802		Description:
803
804		Bumped the heap macig number to 0x09082004 to account for the changed
805		layout of the ML frame under MacOS X.
806
807		----------------------------------------------------------------------
808		Name: Allen Leung (leunga (at) reservoir (dot) com)
809		Date: 2004/09/03 11:26:00 EST
810		Tag: leunga-20040903-cygwin-install
811		Description:
812
813		Added a patch to _arch-n-opsys to enable the Cygwin runtime.
814		The Cygwin runtime is turned on by setting the environment
815		variable SMLNJ_CYGWIN_RUNTIME to 1.
816
817		----------------------------------------------------------------------
818		Name: Matthias Blume (blume (at) tti - c (dot) org)
819		Date: 2004/08/31 17:15:00 CDT
820		Tag: blume-20040831-core
821		Description:
822
823		Added some exports to src/compiler/core.cm upon request by J. Joemann.
824
825		----------------------------------------------------------------------
826		Name: Matthias Blume (blume (at) tti - c (dot) org)
827		Date: 2004/08/30 17:55:00 CDT
828		Tag: blume-20040830-installer
829		Description:
830
831		Upon request by Johannes Joemann:
832
833		- improved ML code of installer to fall back to coping when renaming
834		fails (i.e., when source and target are on different file systems);
835		the code compiles but has yet to be tested in anger
836		- removed mlrisc from list of default targets (config/targets)
837
838		----------------------------------------------------------------------
839		Name: Matthias Blume (blume (at) tti - c (dot) org)
840		Date: 2004/08/27 17:20:00 CDT
841		Tag: blume-20040827-ptreql
842		Description:
843
844		Added ptreql primop to structure InlineT (upon request from Larry
845		Paulson).
846
847		----------------------------------------------------------------------
848		Name: Allen Leung (leunga (at) reservoir (dot) com)
849		Date: 2004/08/15 21:21:00 EST
850		Tag: leunga-110_48-udgraph
851		Description:
852
853		Another bug fix from Carl Hauser:
854
855		diff /net/niflab/smlnj48/src/MLRISC/graphs/udgraph.sml udgraph.sml
856		> 48c48
857		> <              | rmv((e as (k,_))::es,L) = rmv(es,if k = i then es else
858		> e::L)
859		> ---
860		> >              | rmv((e as (k,_))::es,L) = rmv(es,if k = i then L else e::L)
861		> Without this, any deletion of an edge in an undirected graph does severe
862		> violence to the graph.
863
864		----------------------------------------------------------------------
865		Name: Allen Leung (leunga (at) reservoir (dot) com)
866		Date: 2004/08/10 23:35:00 EST
867		Tag: leunga-110_48-ppc
868		Description:
869
870		The IBM/MacOS syntax switch on PPC was incorrectly swapped.
871
872		----------------------------------------------------------------------
873		Name: Matthias Blume (blume (at) tti - c (dot) org)
874		Date: 2004/08/10 12:00:00 CDT
875		Tag: Release_110_48
876		Description:
877
878		New working version (110.48).  NEW BOOTFILES!
879
880		----------------------------------------------------------------------
881		Name: Allen Leung (leunga (at) reservoir (dot) com)
882		Date: 2004/08/09 12:21:00 EST
883		Tag: leunga-110_47-dijsktra
884		Description:
885
886		Bug fix from Carl Hauser:
887
888		single_source_shortest_paths in dijkstra.sml was observed to get wrong
889		answers (by comparing to single_source_shortest_paths in bellman-ford.sml).
890
891		The problem is that following the expression A.update(dist,s,Num.zero)
892		it is necessary to update the priority queue using Q.decreaseWeight(Q,s).
893
894		----------------------------------------------------------------------
895		Name: Matthias Blume (blume (at) tti - c (dot) org)
896		Date: 2004/08/06 18:10:00 CDT
897		Tag: blume-20040806-cmdline
898		Description:
899
900		Fiddled with handling of command-line options:
901
902		* sml now quits after processing the command line
903		if -H, -S, -h<n>, or -s<n> appears as the last
904		command-line argument
905		* a new option -q terminates the session when encountered on
906		the command line; subsequent arguments will be ignored
907		* bug fixes: short (erroneous) arguments are no longer ignored
908		completely
909
910		----------------------------------------------------------------------
911		Name: Allen Leung (leunga (at) reservoir (dot) com)
912		Date: 2004/08/04 18:17:00 EST
913		Tag: leunga-110_47-ppc-ibm-asm
914		Description:
915
916		- Added minimal IBM assembly syntax support for PowerPC.
917
918		- Cygwin: manually changed the file cygwin.def.  Some exported symbols have
919		been altered in the runtime.  We need an automatic way to keep the file
920		in sync.
921
922		----------------------------------------------------------------------
923		Name: Matthias Blume (blume (at) tti - c (dot) org)
924		Date: 2004/08/04 14:00:00 CDT
925		Tag: Release_110_47
926		Description:
927
928		New working version (110.47).  NEW BOOTFILES!
929
930		----------------------------------------------------------------------
931		Name: Matthias Blume (blume (at) tti - c (dot) org)
932		Date: 2004/08/03 14:25:00 CDT
933		Tag: blume-20040803-callingconv
934		Description:
935
936		Added low-level support for choosing C calling conventions by
937		twiddling the type of rawccall.  (See
938		src/compiler/Semant/types/cproto.sml for details.)
939
940		----------------------------------------------------------------------
941		Name: Matthias Blume (blume (at) tti - c (dot) org)
942		Date: 2004/08/02 15:55:00 CDT
943		Tag: blume-20040802-backout
944		Description:
945
946		Backed out of change to win32-filesys.c.  The earlier patch to
947		get_file_time caused CM to produce files with the wrong time stamp.
948
949		----------------------------------------------------------------------
950		Name: Matthias Blume (blume (at) tti - c (dot) org)
951		Date: 2004/08/02 14:45:00 CDT
952		Tag: blume-20040802-nlffi-win32
953		Description:
954
955		Added NLFFI support for Win32, adapted from a patch provided by David
956		Hansel.  This is currently completely untested.  Also, the issue
957		concerning stdcall vs. ccall is still unresolved.
958
959		----------------------------------------------------------------------
960		Name: Matthias Blume (blume (at) tti - c (dot) org)
961		Date: 2004/07/30 17:55:00 CDT
962		Tag: blume-20040730-various
963		Description:
964
965		Gearing up towards 110.47...
966
967		- various minor bugfixes to ml-nlffigen
968		- a beginning of a manual for nlffi
969
970		- eliminated 'export name=value' in config/install.sh as this does
971		not work with certain versions of /bin/sh
972		(Thanks to David King at Motorola for catching this.)
973
974		- several bugfixes provided or suggested by David Hansel at Reactive Systems:
975		- added a test for tm==NULL to gmtime.c and localtime.c
976		- applied patch for incorrect GetFileTime under win32
977		- toSeconds -> toMilliseconds in Win32/win32-process.sml
978
979		----------------------------------------------------------------------
980		Name: Matthias Blume (blume (at) tti - c (dot) org)
981		Date: 2004/07/21 18:20:00 CDT
982		Tag: blume-20040721-nlffigen
983		Description:
984
985		- Fixed minor issue in ml-nlffigen: Now generate structure T_foo
986		for a typedef to an incomplete type, but leave out the "typ" member.
987		(This is just for consistency.)
988		- Started to produce what is supposed to become better (i.e., comprehensive)
989		documentation of what ml-nlffigen does and produces.
990
991		----------------------------------------------------------------------
992		Name: Matthias Blume (blume (at) tti - c (dot) org)
993		Date: 2004/07/14 16:25:00 CDT
994		Tag: blume-20040714-union
995		Description:
996
997		Added C_UNION to c-calls/c-types.sml and updated the machinery
998		(ml-nlffigen, cproto.sml) that conveys C function interface
999		information to the code generator.
1000
1001		However, the actual architecture-specific implementation of function
1002		arguments and results that are C unions is still not implemented.
1003
1004		----------------------------------------------------------------------
1005		Name: Allen Leung (leunga (at) reservoir (dot) com)
1006		Date: 2004/07/14 14:38:00 EST
1007		Tag: leunga-110_46_1-ppc-lwzu
1008		Description:
1009
1010		Added these instructions to the PowerPC architecture:
1011		LBZU(X), LHZU(X), LWZU(X),
1012		STWU(X), STFDU, STFSU
1013
1014		etc...
1015
1016		Note: I haven't added their instruction encoding into the description.
1017
1018		----------------------------------------------------------------------
1019		Name: Allen Leung (leunga (at) reservoir (dot) com)
1020		Date: 2004/07/13 15:04:00 EST
1021		Tag: leunga-110_46_1-ppc-lwarx
1022		Description:
1023
1024		Added the two instructions LWARX and STWCX to the PowerPC
1025		instruction set.
1026
1027		A (untested) rewrite of loop-structure.sml.  The old version
1028		is completely broken.
1029
1030		----------------------------------------------------------------------
1031		Name: Matthias Blume (blume (at) tti - c (dot) org)
1032		Date: 2004/07/13 13:50:00 CDT
1033		Tag: blume-20040713-nlffi
1034		Description:
1035
1036		- use paramAlloc to report c-calls with too many arguments
1037		(for PPC version where parameter area is pre-allocated)
1038		- added ccall_maxargspace to machspec (to implement the above)
1039		- made "make" commend in CM's "make" tool configurable
1040		- added option (default: on) for passing the name of the SML/NJ's "bin"
1041		directory to "make"; the call looks like this:
1042
1043		make <options> SMLNJ_BINDIR=<dir> <target>
1044
1045		This can be used by the Makefile to, e.g., pick the "right" version
1046		of ml-nlffigen.
1047		- minor code tweaks
1048
1049		----------------------------------------------------------------------
1050		Name: Matthias Blume (blume (at) tti - c (dot) org)
1051		Date: 2004/07/12 22:50:00 CDT
1052		Tag: blume-110_46_1-macosx-nlffi
1053		Description:
1054
1055		NLFFI under Mac OS X now working (sort of).  This is largely untested,
1056		though.
1057
1058		Note:  1. You have to make a new, clean build of the runtime system.
1059		2. There are new BOOTFILES, you have to use them!
1060		(Doing the bootstrap process yourself would be *very* painful!
1061		If you absolutely have to do it, build the system under
1062		a different architecture and then cross-compile.)
1063
1064		Version bumped to 110.46.1 to account for runtime data format changes.
1065
1066		----------------------------------------------------------------------
1067		Name: Matthias Blume (blume (at) tti - c (dot) org)
1068		Date: 2004/06/18 14:30:00 CDT
1069		Tag: blume-20040618-unix
1070		Description:
1071
1072		Changed the implementation of structure Unix so that the same stream
1073		is returned every time one of the {text,bin}{In,Out}streamOf functions
1074		is invoked on the same proc.  This is not what the spec currently
1075		says -- although IMO it arguably should.  (See discussion below.)
1076
1077		----------------------------------------------------------------------
1078		Name: Matthias Blume (blume (at) tti - c (dot) org)
1079		Date: 2004/06/17 18:15:00 CDT
1080		Tag: Release_110_46
1081		Description:
1082
1083		New working version (110.46).  NEW BOOTFILES!
1084
1085		----------------------------------------------------------------------
1086		Name: Matthias Blume (blume (at) tti - c (dot) org)
1087		Date: 2004/06/17 17:20:00 CDT
1088		Tag: blume-20040617-timer-unix
1089		Description:
1090
1091		Changed the interface of structures Timer and Unix to match the most
1092		recent Basis spec.
1093
1094		In the case of Unix there still seems to be an open/weird issue:
1095
1096		The {text,bin}{In,Out}streamOf functions are supposed to create
1097		fresh streams whenever they are called -- as opposed to have them
1098		return the same stream every time.  This design is supposed to
1099		prevent space leaks caused by proc values hanging on to streams.
1100
1101		The reap function, on the other hand, is supposed to close the
1102		streams.  This cannot be done without having a handle on the
1103		stream in proc after all...
1104
1105		I took the liberty to implement the following stopgap solution:
1106
1107		The proc value hangs on to the most recently created stream(s).
1108		Reap closes those.  If either or both of the two streams hadn't
1109		been created at all yet, then reap will close the corresponding
1110		file descriptors directly.
1111
1112		PS: I don't understand the original space leak argument anymore.  If
1113		a proc hangs on to the imperative stream, then I/O operations on those
1114		will advance the state of the cached stream and avoid the space leak.
1115
1116		----------------------------------------------------------------------
1117		Name: Matthias Blume (blume (at) tti - c (dot) org)
1118		Date: 2004/05/28 16:45:00 CDT
1119		Tag: blume-20040528-basis
1120		Description:
1121
1122		Added signature PACK_REAL and exported functor PrimIO.
1123
1124		----------------------------------------------------------------------
1125		Name: Matthias Blume (blume (at) tti - c (dot) org)
1126		Date: 2004/05/25 16:00:00 CDT
1127		Tag: blume-20040525-group-owner
1128		Description:
1129
1130		CM now ignores (but still accepts) the "owner" information in group
1131		descriptions.  The owner of a group is its next enclosing
1132		library. Each group must have a unique owner.  (There is a virtual
1133		"toplevel" library that owns groups which are not nested within a real
1134		library.)  Previously, each group had to explicitly declare its owner,
1135		and CM would check that such a declaration is correct.  The new scheme
1136		is to have CM check that for each group there is precisely one owning
1137		library.
1138
1139		The advantage of the new scheme is that the programmer no longer needs
1140		to maintain the somewhat annoying owner information.  The downside is
1141		that CM cannot enforce the ownership rule across multiple runs of
1142		CM.make.  Fortunately, enclosing the same group in two different
1143		libraries A and B which are not part of the same program does not
1144		cause real problems.
1145
1146		----------------------------------------------------------------------
1147		Name: Matthias Blume (blume (at) tti - c (dot) org)
1148		Date: 2004/05/20 16:00:00 CDT
1149		Tag: blume-20040520-win32
1150		Description:
1151
1152		Made the win32 version work again.  (Strangely, a misplaced comma had
1153		slipped into win32-process.c which prevented the runtime from being
1154		compiled correctly.)
1155
1156		Also, included a minor addition to ml-build.bat analogous to what was
1157		done in blume-20040519-ml-build.
1158
1159		----------------------------------------------------------------------
1160		Name: Matthias Blume (blume (at) tti - c (dot) org)
1161		Date: 2004/05/19 22:10:00 CDT
1162		Tag: blume-20040519-ml-build
1163		Description:
1164
1165		Arranged for ml-build to clean up after itself a little bit better.
1166		The script generates a temporary SML source file and compiles it using
1167		CM, so CM generates metadata (GUID, SKEL, objectfile) for it.  It now
1168		gets rid of those at the end, so they don't accumulate under .cm.
1169
1170		This required a minor change to install.sh because the name of the
1171		metadata directory (default: .cm) is actually configurable at
1172		installation time.
1173
1174		----------------------------------------------------------------------
1175		Name: Matthias Blume (blume (at) tti - c (dot) org)
1176		Date: 2004/05/18 15:50:00 CDT
1177		Tag: blume-20040518-mkreader
1178		Description:
1179
1180		Added Posix.IO.mk{Bin,Text}{Reader,Writer} by lifting their respective
1181		implementations from internal modules PosixBinPrimIO and PosixTextPrimIO.
1182
1183		----------------------------------------------------------------------
1184		Name: Matthias Blume (blume (at) tti - c (dot) org)
1185		Date: 2004/05/11 14:35:00 CDT
1186		Tag: blume-20040511-win32sock
1187		Description:
1188
1189		Added previously missing support for many socket-related functions
1190		under win32.  Thanks to David Hansel <hansel@reactive-systems.com>
1191		for the voluminous patch!
1192
1193		(I have not tested this patch under win32 yet.)
1194
1195		Here is David's e-mail:
1196
1197		Hi,
1198
1199		Attached to this email you find a diff against sml/nj 110.45
1200		that will enable socket support under Windows.
1201
1202		To apply the patch (using unix or cygwin)
1203		1) gunzip runtime.diff.gz
1204		2) "cd" into "src/runtime" in the source tree of a fresh
1205		110.45 installation.
1206		3) patch -p 1 < [your/path/to]runtime.diff
1207
1208		The code compiles fine but has NOT yet been extensively tested.
1209		I only ran a few tests for basic socket client functionality
1210		(which worked fine).  Especially the functions that use ioctl
1211		are not tested at all and might not work (see below).
1212
1213		I implemented this since we want to move to a newer version of sml/nj
1214		but need socket support in order to use it.  This is the first time I
1215		even had a look at the sml/nj source,  so please review my changes
1216		before making this part of the distribution!  Here are a few issues
1217		that I think might be better for someone to solve who is more
1218		familiar with the sml/nj source (and socket programming):
1219
1220		- getnetbyaddr.c and getnetbyname.c will raise a "not implemented"
1221		exception since I could not figure out what the windows equivalent
1222		of these functions is
1223
1224		- In sockets-osdep.h there are a some #include statements that are
1225		only used in a few files that include sockets-osdep.h
1226
1227		- In smlnj-sock-lib.c, function init_fn() calls WSAStartup() but
1228		does not process its return value since I don't know how to
1229		report an error upwards.
1230
1231		- It would probably be good to have a call to WSACleanup() when
1232		the library is unloaded (if there is such a possibility).
1233		Otherwise I think Windows will take care of this automatically
1234		when the process finishes.
1235
1236		- I used ioctlsocket() as a replacement for ioctl() but I have
1237		no idea if that is actually the proper replacement on Windows.
1238
1239		- All these issues are marked in the code by "FIXME" comments.
1240
1241		We use sml/nj extensively in our products and are quite happy
1242		with it.  I hope this contribution will help you.
1243
1244		Keep up the good work!
1245
1246		David
1247
1248		----------------------------------------------------------------------
1249		Name: Matthias Blume (blume (at) tti - c (dot) org)
1250		Date: 2004/05/11 14:20:00 CDT
1251		Tag: blume-20040511-installml
1252		Description:
1253
1254		Fixed two bugs in installml script.  (Thanks to Vesa A. Norrman for
1255		the patch.)
1256
1257		----------------------------------------------------------------------
1258		Name: Matthias Blume (blume (at) tti - c (dot) org)
1259		Date: 2004/05/11 14:05:00 CDT
1260		Tag: blume-20040511-nlffi-netbsd
1261		Description:
1262
1263		Added support for nlffi under netbsd.  (Thanks to Vesa A. Norrman for
1264		the patch.)
1265
1266		----------------------------------------------------------------------
1267		Name: Matthias Blume (blume (at) tti - c (dot) org)
1268		Date: 2004/05/11 12:05:00 CDT
1269		Tag: blume-20040511-exports
1270		Description:
1271
1272		As per request by Adam Chlipala <adam@hcoop.net>, extended various
1273		export lists in compiler-related .cm-files.
1274
1275		----------------------------------------------------------------------
1276		Name: Matthias Blume (blume (at) tti - c (dot) org)
1277		Date: 2004/05/11 11:35:00 CDT
1278		Tag: blume-20040511-allsource
1279		Description:
1280
1281		The installer now honors the "src-smlnj" target again, although its meaning
1282		has changed from "all sources required for the compiler" to "all sources
1283		the installer knows about".  In other words, if you enable "src-smlnj"
1284		in the "targets" file, then the installer will pull in sources for
1285		everything.  (Notice that this refers to source code only.  Compiled
1286		code is still only installed for modules that were requested explicitly
1287		or which are required for other modules that were requested explicitly.)
1288
1289		----------------------------------------------------------------------
1290		Name: Matthias Blume (blume (at) tti - c (dot) org)
1291		Date: 2004/04/23 17:40:00 CDT
1292		Tag: blume-20040423-ieee-scan
1293		Description:
1294
1295		Fixed IEEEReal.scan (and .fromString) so that if there is an overflow
1296		in the exponent calculation we get INF or ZERO (depending on the mantissa
1297		and the sign of the exponent).
1298
1299		----------------------------------------------------------------------
1300		Name: Matthias Blume (blume (at) tti - c (dot) org)
1301		Date: 2004/04/23 10:40:00 CDT
1302		Tag: blume-20040423-ml-build
1303		Description:
1304
1305		The ml-build script now terminates with a non-0 status when something
1306		goes wrong.
1307
1308		----------------------------------------------------------------------
1309		Name: Matthias Blume (blume (at) tti - c (dot) org)
1310		Date: 2004/04/22 16:35:00 CDT
1311		Tag: blume-20040422-Option
1312		Description:
1313
1314		Made exception Option to be the same as exception Option.Option
1315		(as it should be).
1316
1317		----------------------------------------------------------------------
1318		Name: Allen Leung (leunga (at) reservoir (dot) com)
1319		Date: 2004/03/19 14:40:00 EST
1320		Tag: leunga-20040319-cygwin-nlffi
1321		Description:
1322
1323		Fixed the runtime so that ml-nlffi-lib runs on the cygwin version
1324		of SML/NJ.  The problem is that
1325
1326		lib = dlopen(NULL, ...)
1327		f   = dlsym(lib, "malloc");
1328
1329		does not work on Windows unless we explicitly export symbols
1330		such as 'malloc' during linking.  We fixed this by explicitly
1331		exporting the required symbols with the magic gcc incantation:
1332
1333		-Wl,--export-all cygwin.def
1334
1335		where cygwin.def is a file containing all the symbols that
1336		we wish to export.
1337
1338		I suspect this is a Windows problem and we'll have to
1339		do the same (somehow with windows compilers) when
1340		we build the native win32 version with the system
1341		calls LoadLibrary/GetProcAddress.
1342
1343		----------------------------------------------------------------------
1344		Name: Matthias Blume (blume (at) tti - c (dot) org)
1345		Date: 2004/03/04 16:35:00 CST
1346		Tag: blume-20040304-intinf-fmt
1347		Description:
1348
1349		Fixed problem with IntInf.fmt (sign would show up on the right instead
1350		of on the left for BIN, OCT, and HEX).
1351
1352		----------------------------------------------------------------------
1353		Name: Matthias Blume (blume (at) tti - c (dot) org)
1354		Date: 2004/03/04 11:25:00 CST
1355		Tag: blume-20040304-symlinks
1356		Description:
1357
1358		Fixed problem with installer script (unix only) where bin/ml-yacc and
1359		friends pointed (via symlinks) to absolute locations instead of just
1360		.run-sml.  This was reported by Vesa A Norrman.
1361
1362		----------------------------------------------------------------------
1363		Name: Matthias Blume (blume (at) tti - c (dot) org)
1364		Date: 2004/02/13 14:50:00 CST
1365		Tag: Release_110_45
1366		Description:
1367
1368		New working version (110.45). New bootfiles.
1369
1370		----------------------------------------------------------------------
1371		Name: Matthias Blume (blume (at) tti - c (dot) org)
1372		Date: 2004/01/26 15:15:15 CST
1373		Tag: blume-20040126-toplevel
1374		Description:
1375
1376		Improved handling of exceptions at the interactive toplevel.
1377
1378		----------------------------------------------------------------------
1379		Name: Matthias Blume (blume (at) tti - c (dot) org)
1380		Date: 2004/01/26 11:25:00
1381		Tag: blume-20040126-app
1382		Description:
1383
1384		Type of top-level "app" corrected.
1385		Added code for setting vp_limitPtrMask to Win32-specific runtime.
1386
1387		----------------------------------------------------------------------
1388		Name: Matthias Blume (blume (at) tti - c (dot) org)
1389		Date: 2003/11/18 17:10 CST
1390		Tag: blume-20031118-basis-fiddle
1391		Description:
1392
1393		- changed Timer interface to what might become the spec
1394		- POSIX_FLAGS -> BIT_FLAGS according to spec
1395		- some other minor discrepancies wrt. spec eliminated
1396
1397		----------------------------------------------------------------------
1398		Name: Matthias Blume (blume (at) tti - c (dot) org)
1399		Date: 2003/11/06 12:00:00 CST
1400		Tag: Release_110_44
1401		Description:
1402
1403		New working version (110.44). New bootfiles.
1404
1405		----------------------------------------------------------------------
1406		Name: Matthias Blume (blume (at) tti - c (dot) org)
1407		Date: 2003/11/04 11:50:00 CST
1408		Tag: blume-20031104-move-libraries
1409		Description:
1410
1411		Eliminated the "dont_move_libraries" directive in config/targets.
1412		(The mechanism was broken and could not be fixed easily.  Moreover,
1413		there does not seem to be any reason not to move all libraries into
1414		lib during installation.  I originally implemented this directive as a
1415		backward-compatibility feature when I first introduced the new CM.
1416		Now that things have been stable for a long time and going back to the
1417		old CM is not an option, there is no reason to keep it around.)
1418
1419		----------------------------------------------------------------------
1420		Name: Matthias Blume (blume (at) tti - c (dot) org)
1421		Date: 2003/11/03 16:00:00 CST
1422		Tag: blume-20031103-installdir
1423		Description:
1424
1425		Made installer honor INSTALLDIR variable again.  (Thanks to Chris
1426		Richards for pointing out the problem and providing the solution.)
1427
1428		----------------------------------------------------------------------
1429		Name: Matthias Blume (blume (at) tti - c (dot) org)
1430		Date: 2003/10/01 17:05:00 CDT
1431		Tag: blume-20031001-lal-mlrisc
1432		Description:
1433
1434		MLRISC bug fix from Lal.
1435
1436		----------------------------------------------------------------------
1437		Name: Matthias Blume (blume (at) tti - c (dot) org)
1438		Date: 2003/09/30 16:10:00 CDT
1439		Tag: blume-20030930-primio-bat
1440		Description:
1441
1442		1. Added openVector, nullRd, and nullWr to PRIM_IO.
1443		2. Improved .bat files (for Win32 port) to make things work under Win95.
1444		(thanks to Aaron S. Hawley for this one)
1445
1446		----------------------------------------------------------------------
1447		Name: Matthias Blume (blume (at) tti - c (dot) org)
1448		Date: 2003/09/26 16:05:00 CDT
1449		Tag: blume-20030926-wrappriv
1450		Description:
1451
1452		Added missing wrapper for privilege "primitive" in $smlnj/viscomp/core.cm.
1453
1454		----------------------------------------------------------------------
1455		Name: Matthias Blume (blume (at) tti - c (dot) org)
1456		Date: 2003/09/26 15:00:00 CDT
1457		Tag: blume-20030926-110_43_3
1458		Description:
1459
1460		- additional cleanup
1461		- version number bump, NEW BOOTFILES
1462
1463		----------------------------------------------------------------------
1464		Name: Matthias Blume (blume (at) tti - c (dot) org)
1465		Date: 2003/09/26 12:00:00 CDT
1466		Tag: blume-20030926-ppautoload
1467		Description:
1468
1469		I modified the read-eval-print loop so that the autoloader gets
1470		invoked whenever the prettyprinter tries to look up a symbol that
1471		is not currently defined in the toplevel environment but which
1472		appears in CM's autoload registry.  As a result, we see far fewer of
1473		those ?.Foo.Bar.xxx names in the prettyprinter's output.
1474
1475		In addition to this I tried to clean up some pieces of the Basis
1476		implementation (e.g., Socket, Word8Array) in order to prevent other
1477		instances of these ?.Foo.Bar.xxx names from being printed.
1478
1479		The mechanism that picks names for types still needs some work, though.
1480		(Right now it seems that if there is a type A.t which is defined to
1481		be B.u, but B is unavailable at toplevel, then A.t gets printed as
1482		"?.B.u" although the perhaps more sensible solution would be to use
1483		"A.t" in this case.  In other words, the prettyprinter should follow
1484		a chain of DEFtycs not farther than there are corresponding toplevel
1485		names in the current environment.)
1486
1487		----------------------------------------------------------------------
1488		Name: Matthias Blume (blume (at) tti - c (dot) org)
1489		Date: 2003/09/24 16:31:00 CDT
1490		Tag: blume-20030924-installer
1491		Description:
1492
1493		Another installer tweak:  All the ML code for the installer is now
1494		compiled during CMB.make and put into a little library called
1495		$smlnj/installer.cm.  The installation then simply invokes
1496
1497		sml -m $smlnj/installer.cm
1498
1499		and everything happens automagically.
1500
1501		Win32:  ML code senses value of environment variable SMLNJ_HOME.
1502		Unix:   ML code senses values of environment variables ROOT, CONFIGDIR,
1503		and BINDIR.
1504
1505		The new scheme guarantees that the ML code responsible for the installation
1506		is in sync with the APIs of the main system.  Also, the installer is
1507		somewhat faster because the installer script is precompiled.
1508
1509		----------------------------------------------------------------------
1510		Name: Matthias Blume (blume (at) tti - c (dot) org)
1511		Date: 2003/09/24 15:35:00 CDT
1512		Tag: blume-20030924-synsock
1513		Description:
1514
1515		Added a signature SYNCHRONOUS_SOCKET to basis.cm.  This is like SOCKET
1516		but excludes all non-blocking operations.  Defined SOCKET (in Basis)
1517		and CML_SOCKET in terms of SYNCHRONOUS_SOCKET.  Removed superfluous
1518		implementations of non-blocking operations from CML's Socket
1519		structure.
1520
1521		----------------------------------------------------------------------
1522		Name: Matthias Blume (blume (at) tti - c (dot) org)
1523		Date: 2003/09/24 15:10:05 CDT
1524		Tag: blume-20030924-sockets
1525		Description:
1526
1527		1. Fixed SOCKET API and implementation to match Basis spec.
1528		This required changing the internal representation of sockets to one
1529		that remembers (for each socket file descriptor) whether it is currently
1530		blocking or non-blocking.  This state is maintained lazily (i.e., a system
1531		call is made only if the state actually needs to change).
1532
1533		2. OS-specific details of sockets were moved into separate files, thus
1534		making it possible to unify the bulk of the socket implementations
1535		between Unix and Win32.
1536
1537		3. CML's socket API changed accordingly.
1538		(Note that we need to remove non-blocking functions from this API
1539		since they are redundant in the case of CML!)
1540
1541		4. CML's socket implementation now makes use of non-blocking functions
1542		provided by Basis, thus removing all OS-dependent code from this part
1543		of CML.
1544
1545		5. Changed Real64.precision from 52 to 53.  Minor cleanup in Real64 code.
1546
1547		----------------------------------------------------------------------
1548		Name: Matthias Blume (blume (at) tti - c (dot) org)
1549		Date: 2003/09/22 12:10:00 CDT
1550		Tag: blume-20030922-110_43_2
1551		Description:
1552
1553		Made a new interim version and bootfiles for developer's bootstrapping
1554		convenience.
1555
1556		110.43.2 -- NEW BOOTFILES
1557
1558		----------------------------------------------------------------------
1559		Name: Matthias Blume (blume (at) tti - c (dot) org)
1560		Date: 2003/09/19 15:55:00 CDT
1561		Tag: blume-20030919-cmdir
1562		Description:
1563
1564		1. new-install.sh -> install.sh
1565		2. changed default CM "metadata" directory name to ".cm" (instead of "CM")
1566		3. tweaked installer so that another name instead of .cm can be chosen
1567		at install time (by setting the CM_DIR_ARC environment variable
1568		during installation); once installation is complete, the name is
1569		fixed
1570
1571		----------------------------------------------------------------------
1572		Name: Matthias Blume (blume (at) tti - c (dot) org)
1573		Date: 2003/09/18 16:00:00 CDT
1574		Tag: blume-20030918-110_43_1
1575		Description:
1576
1577		Made a new interim version and bootfiles for developer's bootstrapping
1578		convenience.
1579
1580		110.43.1 -- NEW BOOTFILES
1581
1582		----------------------------------------------------------------------
1583		Name: Matthias Blume (blume (at) tti - c (dot) org)
1584		Date: 2003/09/18 15:20:00 CDT
1585		Tag: blume-20030918-misc
1586		Description:
1587
1588		1. Exported fractionsPerSecond etc. from TimeImp (but not from Time as
1589		this seems to be controversial at the moment) and used those in
1590		Posix.ProcEnv.times.
1591
1592		2. Added Time.{from,to}Nanoseconds to Time.
1593
1594		3. Improved Real.{from,to}LargeInt by avoiding needless calculations.
1595		For example, fromLargeInt never needs to look at more than 3 "big
1596		digits" to get its 53 bits of precision.
1597
1598		----------------------------------------------------------------------
1599		Name: Matthias Blume (blume (at) tti - c (dot) org)
1600		Date: 2003/09/17 16:30:00 CDT
1601		Tag: blume-20030917-real32-slices
1602		Description:
1603
1604		Added an entry to the primitive environment
1605		(compiler/Semant/statenv/prim.sml) for int32->real64 conversion and
1606		added code to compiler/CodeGen/main/mlriscGen.sml to implement it.
1607
1608		Removed some of the "magic" constants in real64.sml and replaced them
1609		with code that generates these values from their corresponding
1610		integer counterparts.
1611
1612		Made all(?) the slice-related changes to the Basis and made everything
1613		compile again...
1614
1615		----------------------------------------------------------------------
1616		Name: Matthias Blume (blume (at) tti - c (dot) org)
1617		Date: 2003/09/15 17:45:00 CDT
1618		Tag: blume-20030915-rbase
1619		Description:
1620
1621		Fixed bug in Real.fromLargeInt.
1622
1623		----------------------------------------------------------------------
1624		Name: Matthias Blume (blume (at) tti - c (dot) org)
1625		Date: 2003/09/13 18:11:00 CDT
1626		Tag: blume-20030913-libinstall
1627		Description:
1628
1629		Minor bugfix in config/libinstall (set anchor with path to
1630		standalone tool after installing it, otherwise libraries that
1631		need ml-lex or ml-yacc won't compile the first time the installer
1632		runs).
1633
1634		----------------------------------------------------------------------
1635		Name: Matthias Blume (blume (at) tti - c (dot) org)
1636		Date: 2003/09/12 11:45:00 CDT
1637		Tag: blume-20030912-various
1638		Description:
1639
1640		- fixed bug in Real.toLargeInt
1641		- fixed bug in Posix.ProcEnv.times
1642		- changed inputLine functions to return an option
1643		- minor installer improvements / bugfixes
1644		- changed default @SMLalloc parameter for x86/celeron to 64k
1645
1646		----------------------------------------------------------------------
1647		Name: Matthias Blume (blume (at) tti - c (dot) org)
1648		Date: 2003/09/09 22:00:00 CDT
1649		Tag: Release_110_43
1650		Description:
1651
1652		New working release 110.43.  New bootfiles.
1653
1654		----------------------------------------------------------------------
1655		Name: Matthias Blume (blume (at) tti - c (dot) org)
1656		Date: 2003/09/09 19:20:00 CDT
1657		Tag: blume-20030909-installer
1658		Description:
1659
1660		Rewrote large parts of config/install.sh in SML (config/libinstall.sml).
1661		Modified config/install.bat to take advantage of it.  Also modified
1662		config/install.sh (and called it config/new-install.sh) to take advantage
1663		of it on Unix systems.  (The SML code is (supposed to be) platform-
1664		independent.)
1665
1666		The installer can now install everything under Win32
1667		as well as under *nix as long as it compiles.
1668
1669		Other changes:
1670
1671		- made CML compile again under Win32
1672		- made eXene compile under Win32 (by providing a fake structure UnixSock
1673		and by using OS.Process.getEnv instead of Posix.ProcEnv.getenv)
1674		- fixed a bug in nowhere: it assumed that type OS.Process.status is the
1675		same as type int;  under Win32 it isn't
1676		- fixed some slice-related problems in the win32-specific parts of CML
1677		- added a functor argument "sameVol" to os-path-fn.sml in the Basis
1678		(under Win32, the volume name is case-insensitive, and the
1679		OS.Path code compares volume names for equality)
1680
1681		----------------------------------------------------------------------
1682		Name: Matthias Blume (blume (at) tti - c (dot) org)
1683		Date: 2003/09/08 11:55:00 CDT
1684		Tag: blume-20030908-fullpath
1685		Description:
1686
1687		Made Win32 version of OS.FileSys.fullPath return current directory
1688		when given an empty string.  This is what the spec says, and incidentally,
1689		CM depends on it.  (CM otherwise goes into an infinite loop in certain
1690		cases when presented with the name of a non-existing .cm file.)
1691
1692		----------------------------------------------------------------------
1693		Name: Matthias Blume (blume (at) tti - c (dot) org)
1694		Date: 2003/09/04 16:30:00 CDT
1695		Tag: blume-20030905-slices-etc
1696		Description:
1697
1698		1. Changed interface to vectors and arrays in Basis to match
1699		(draft) Basis spec.
1700		2. Added signatures and implementations of slices according to
1701		Basis spec.
1702		3. Edited source code throughout the system to make it compile again
1703		under 1. and 2.  (In some cases code had to be added to have it
1704		match the new signatures.)
1705		4. MLRISC should be backward-compatible: the copies of the originals
1706		of files that needed to change under 3. were retained, the .cm files
1707		check the compiler version number and use old versions when
1708		appropriate.
1709		5. Changed type of OS.FileSys.readDir and Posix.FileSys.readdir to
1710		dirstream -> string option (in accordance with Basis spec).
1711		6. When generating code that counts lines, ml-lex used function
1712		CharVector.foldli, taking advantage of its old interface.
1713		This has been replaced with the corresponding code from
1714		CharVectorSlice.  (html-lex must be re-lexed!)
1715		7. BitArray in smlnj-lib/Util has been extended/modified to match the
1716		new MONO_ARRAY signature.  (Do we need BitArraySlice?)
1717		8. Removed temporary additions (fromInternal, toInternal) from the
1718		(now obsolete) IntInf in smlnj-lib/Util.
1719		9. Cleaned up structure Byte.
1720		10. Added localOffset, scan, and fromString to Date (according to spec).
1721		Cleaned/corrected implementation of Date.
1722		(Still need to check for correctness; implement better canonicalizeDate.)
1723		11. Added "scan" to signature IEEE_REAL.
1724		12. Some improvements to IntInf [in particular: efficiency-hack for
1725		mod and rem when second operand is 2 (for parity checks).]
1726		13. Changed representation of type Time.time, using a single IntInf.int
1727		value counting microseconds.  This considerably simplified the
1728		implementation of structure Time.  We now support negative time
1729		values; scan and fromString handle signs.
1730		14. Functor PrimIO now takes two additional arguments (VectorSlice and
1731		ArraySlice).
1732
1733		----------------------------------------------------------------------
1734		Name: Matthias Blume (blume (at) tti - c (dot) org)
1735		Date: 2003/08/28 17:00:00 CDT
1736		Tag: blume-20030828-intinf
1737		Description:
1738
1739		This is a major update which comes with a version number bump
1740		(110.42.99 -- yes, we are really close to 110.43 :-), NEW BOOTFILES,
1741		and an implementation of IntInf in the Basis.
1742
1743		There are a fairly large number of related changes and updates throughout
1744		the system:
1745
1746		Basis:
1747		- Implemented IntInf.
1748		- Made LargeInt a projection of IntInf (by filtering through INTEGER).
1749		- Added some missing Real64 operations, most notably Real.toLargeInt.
1750		- Added FixedInt as a synonym for Int32.
1751
1752		compiler:
1753		* Added support for a built-in intinf type.
1754		- literals
1755		- pattern matching
1756		- conversion shortcuts (Int32.fromLarge o Int.toLarge etc.)
1757		- overloading on literals and operations
1758
1759		This required adding a primitive type intinf, some additional
1760		primops, and implementations for several non-trivial intinf
1761		operations in Core.  (The intinf type is completely abstract
1762		to the compiler; all operations get delegated back to the Core.)
1763
1764		* Intinf equality is handled by polyequal.  However, the compiler
1765		does not print its usual warning in this case (since polyequal
1766		is the right thing to do there).
1767
1768		* Improved the organization of structure InlineT.
1769
1770		* A word about conversion primops:
1771		If conversions involving intinf do not cancel out during
1772		CPS contract, then the compiler must insert calls to Core functions.
1773		Since all core access must be resolved already during the FLINT
1774		translate phase, it would be too late a the time of CPS contract
1775		to add new Core calls.  For this reason, conversion primops
1776		for intinf carry two arguments:  1. the numeric argument that
1777		they are supposed to convert, and 2. the Core function that
1778		can help with this conversion if necessary.  If CPS contract
1779		eliminates a primop, then the associated Core function becomes
1780		dead and goes away.  Intinf conversion primops that do not get
1781		eliminated by CPS contract get rewritten into calls of their
1782		core functions by a separate, new phase.
1783
1784		interactive system:
1785		- Control.Print.intinfDepth controls max length of intinf constants
1786		being printed. (Analogous to Control.Print.stringDepth.)
1787		- Cleanup in printutil and pputil: got rid of unused stuff and
1788		duplicates; replaced some of the code with code that makes better
1789		use of library functionality.
1790
1791		CM:
1792		Bugfix: parse-errors in init group (system/smlnj/init/init.cmi)
1793		are no longer silent.
1794
1795		CKIT:
1796		Fixed mismatched uses of Int32 and LargeInt.  I always decided
1797		in favor of LargeInt -- which is now the same as IntInf.
1798		CKIT-knowledgable people should check whether this is what's
1799		intended and otherwise change things back to using Int32 or
1800		FixedInt.
1801
1802		Throughout the code:
1803		Started using IntInf.int literals and built-in operations
1804		(e.g., comparison with 0) where this seems appropriate.
1805
1806
1807		----------------------------------------------------------------------
1808		Name: Dave MacQueen (dbm@cs.uchicago.edu)
1809		Date: 2003/08/13 11:36:00 CDT
1810		Tag: dbm-20030813-mcz-merge1
1811		Description:
1812
1813		Merging changes from the mcz-branch development branch into trunk.
1814		These changes involve replacement of the emulated old prettyprinter
1815		interface with direct use of the SML/NJ Lib PP library, and fixing
1816		of a couple of bugs (895, 1186) relating to error messages.  A new
1817		prettyprinter for ast datatypes (Elaborator/print/ppast.{sig,sml})
1818		has been added.
1819
1820		----------------------------------------------------------------------
1821		Name: Matthias Blume (blume (at) tti - c (dot) org)
1822		Date: 2003/08/11 15:45:00 CDT
1823		Tag: blume-20030811-windows
1824		Description:
1825
1826		Version number bumped to 110.42.9.  NEW BOOTFILES!!!
1827
1828		http://smlnj.cs.uchicago.edu/dist/working/110.42.9/
1829
1830		This patch restores SML/NJ's ability to run under win32.  There are a
1831		number of changes, including fixes for several bugs that had gone
1832		unnoticed until now:
1833
1834		- uname "CYGWIN_NT*" is recognized as win32 (This is relevant only when
1835		trying to run the win32 version from within cygwin.)
1836
1837		- There are a number of simple .bat scripts that substitute for their
1838		corresponding Unix shell-scripts.  (See below.)
1839
1840		- The internals of ml-build have been modified slightly.  The main
1841		difference is that instead of calling ".link-sml" (or link-sml.bat)
1842		using OS.Process.system, the ML process delegates this task back
1843		to the script.  Otherwise problems arise in mixed environments such
1844		as Cygwin where scripts look and work like Unix scripts, but
1845		where OS.Process.system cannot run them.
1846
1847		- In CM, the srcpath pickler used native pathname syntax -- which
1848		is incorrect in the case of cross-compilation.  The new pickle format
1849		is independent of platform-specific naming conventions.
1850
1851		- Path configuration files (such as lib/pathconfig) can now choose
1852		between native and standard syntax.  Placing a line of the form
1853
1854		standard!
1855
1856		into the file causes all subsequent paths to be interpreted using
1857		CM standard pathname syntax (= Unix conventions); a line
1858
1859		native!
1860
1861		switches back to native style.  This was needed so that
1862		path config files can be written portably, see src/system/pathconfig.
1863
1864		- Runtime system:
1865
1866		- win32-filesys.c:  get_file_time and set_file_time now
1867		access modification time, not creation time.
1868
1869		- I/O code made aware of new array representation.
1870
1871		- Bug fixes in X86.prim.masm.
1872
1873		- src/system/makeml made aware of win32. (For use under cygwin
1874		and other Unix-environments for windows.)
1875
1876		- In Basis, fixed off-by-one error in win32-io.sml (function vecF)
1877		which caused BinIO.inputAll to fail consistently.
1878
1879		.bat scripts:
1880
1881		Windows .bat scripts assume that SMLNJ_HOME is defined.
1882
1883		- sml.bat, ml-yacc.bat, ml-lex.bat:  Driver scripts for standalone
1884		applications (sml, ml-yacc, ml-lex).
1885		- ml-build.bat: analogous to ml-build.
1886		- config\install.bat:  Analogous to config/install.sh.  This requires
1887		that SMLNJ_HOME is set and that Microsoft Visual C is ready to use.
1888		(nmake etc. must be on the path, and vcvars32 must have been run.)
1889		Moreover, sources for ml-lex and ml-yacc need to exist under src,
1890		and the bootfile hierarchy must have been unpacked under
1891		sml.boot.x86-win32.
1892		The script is very primitive and does a poor job at error checking.
1893		It only installs the base system, ml-lex, and ml-yacc.  No other
1894		libraries are being installed (i.e., you get only those that
1895		are part of the compiler.)
1896		- link-sml.bat:  analogous to .link-sml, but not currently used
1897
1898		Unrelated bug fixes:
1899
1900		- ml-nlffigen now exports structures ST_* corresponding to incomplete
1901		types.
1902		- Added getDevice to PP/src/pp-debug-fn.sml.  (Would not compile
1903		otherwise.)
1904
1905		----------------------------------------------------------------------
1906		Name: Dave MacQueen (macqueen@cs.uchicago.edu)
1907		Date: 2003/06/17
1908		Tag: macqueen-20030617-bug895
1909		Description:
1910
1911		Modified compiler/Elaborator/print/pptype.sml to fix bug 895.
1912		Tag will be used for new development branch (mcz-branch) for
1913		use by MacQueen, (Lucasz) Zairek, and (George) Cao at uchicago.
1914
1915		----------------------------------------------------------------------
1916		Name: Matthias Blume (blume (at) tti - c (dot) org)
1917		Date: 2003/05/27 16:55:00 CDT
1918		Tag: blume-20030527-polyeq
1919		Description:
1920
1921		Tried to eliminated most cases of polymorphic equality.
1922
1923		----------------------------------------------------------------------
1924		Name: Matthias Blume (blume (at) tti - c (dot) org)
1925		Date: 2003/05/21 17:45:00 CDT
1926		Tag: blume-20030517-complete
1927		Description:
1928
1929		Two changes:
1930
1931		1. Added a flag for controlling whether non-exhaustive bindings will
1932		be treated as errors (default is false).
1933		2. Cleaned up the *entire* source tree so that CMB.make goes through
1934		without a single non-exhaustive match- or bind warning.
1935
1936		----------------------------------------------------------------------
1937		Name: Matthias Blume (blume (at) tti - c (dot) org)
1938		Date: 2003/05/17 10:20:00 CDT
1939		Tag: blume-20030517-absyn
1940		Description:
1941
1942		1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
1943
1944		This mainly affects the quality of error messages.  However, some
1945		of the code is now more straightforward than before. (Treatment of
1946		the above four constructs in translate.sml is much simpler than
1947		the "macro-expansion" that was going on before.  Plus, the mach-
1948		compiler no longer gets invoked just to be able to compile an
1949		if-expression.)
1950
1951		2. The ErrorMsg.Error exception is now caught and absorbed by the
1952		interactive loop.
1953
1954		----------------------------------------------------------------------
1955		Name: Allen Leung
1956		Date: 2003/05/16 13:05:00 CDT
1957		Tag: leunga-20030516-cygwin-runtime
1958		Description:
1959
1960		Ported the runtime system to cygwin, which uses the unix
1961		x86-unix bin files.  Missing/buggy features:
1962
1963		o getnetbyname, getnetbyaddr: these functions seem to be missing in
1964		the Cygwin library.
1965		o Ctrl-C handling may be flaky.
1966		o Windows system calls and Windows I/O are not supported.
1967
1968		A new set of binfiles is located at:
1969
1970		http://www.dorsai.org/~leunga/boot.x86-unix.tgz
1971
1972		This is only needed for bootstrapping the cygwin version of smlnj.
1973		Other x86 versions can use the existing binfiles.
1974
1975		----------------------------------------------------------------------
1976		Name: Matthias Blume
1977		Date: 2003/04/08 15:42:00 CDT
1978		Tag: blume-20030408-listpair
1979		Description:
1980
1981		1. Added a target 'mlrisc' to installer.
1982
1983		2. Added missing elements to structure ListPair.
1984
1985		----------------------------------------------------------------------
1986		Name: Allen Leung
1987		Date: 2003/01/07 10:40:00 EST
1988		Tag: leunga-20030107-int-rem
1989		Description:
1990
1991		Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
1992		The arguments to the SUBL instruction were swapped.
1993
1994		----------------------------------------------------------------------
1995		Name: Matthias Blume
1996		Date: 2002/12/12 16:25:00 EST
1997		Tag: blume-20021212-risc-ra
1998		Description:
1999
2000		Fixed a serious bug in the rewrite code for FP spilling/reloading that
2001		sent the RA into an infinite loop when floating point registers get
2002		spilled.  (Because of this bug, e.g., nucleic stopped compiling between
2003		110.37 and 110.38.)
2004		There was another set of potential problems related to the handling of
2005		MLRISC annotations (but those did not yet cause real problems, apparently).
2006
2007		----------------------------------------------------------------------
2008		Name: Matthias Blume
2009		Date: 2002/12/06 22:40:00 EST
2010		Tag: blume-20021206-cm-fileid
2011		Description:
2012
2013		Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
2014		This fixes the problem of CM getting confused by files that suddenly
2015		change their identity (e.g., by getting unlinked and recreated by some
2016		text editor such as vi).  There might be a better/cheaper/cleaner way
2017		of doing this, but for now this will have to do.
2018
2019		----------------------------------------------------------------------
2020		Name: Matthias Blume
2021		Date: 2002/10/28 09:50:00 EST
2022		Tag: blume-20021028-typecheck
2023		Description:
2024
2025		Exported structure Typecheck from $smlnj/viscomp/core.cm.
2026
2027		----------------------------------------------------------------------
2028		Name: Matthias Blume
2029		Date: 2002/10/17 09:10:00 EDT
2030		Tag: Release_110_42
2031		Description:
2032
2033		In good old tradition, there has been a slight hiccup so that we have
2034		to patch 110.42 after the fact.   The old release tag has been replaced
2035		(see below).
2036
2037		The change solves a problem with two competing approaches the
2038		configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
2039		each other's way.
2040
2041		This change only affects the runtime system code and the installer script.
2042		(No new bootfiles.)
2043
2044		----------------------------------------------------------------------
2045		Name: Matthias Blume
2046		Date: 2002/10/16 12:00:00 EDT
2047		Tag: Release_110_42_removed
2048		Description:
2049
2050		New working release.  New bootfiles.
2051
2052		----------------------------------------------------------------------
2053		Name: Matthias Blume
2054		Date: 2002/10/10 13:10:00 EDT
2055		Tag: blume-20021010-ppc-divs
2056		Description:
2057
2058		The mltree operator DIVS must be implemented with an overflow check on
2059		the PPC because the hardware indicates divide-by-zero using "overflow" as
2060		well.
2061
2062		----------------------------------------------------------------------
2063		Name: Matthias Blume
2064		Date: 2002/07/23 11:45:00
2065		Tag: blume-20020723-smlnj-home
2066		Description:
2067
2068		Sml now senses the SMLNJ_HOME environment variable.  If this is set,
2069		then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
2070		CM_PATHCONFIG is also set), the path configuration file is assumed
2071		to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
2072		the entire tree to some other place and everything will "just work".
2073
2074		(Companion commands such as ml-build and ml-makedepend also sense this
2075		variable.)
2076
2077		----------------------------------------------------------------------
2078		Name: Matthias Blume
2079		Date: 2002/07/12 21:19:00 EDT
2080		Tag: blume-20020712-liveness
2081		Description:
2082
2083		Exported two useful "step" functions from liveness module (MLRISC).
2084
2085		----------------------------------------------------------------------
2086		Name: Matthias Blume
2087		Date: 2002/07/05 16:00 EDT
2088		Tag: Release_110_41
2089		Description:
2090
2091		New working release.  New bootfiles.
2092
2093		----------------------------------------------------------------------
2094		Name: Matthias Blume
2095		Date: 2002/07/05 10:25:00 EDT
2096		Tag: blume-20020705-btimp
2097		Description:
2098
2099		Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
2100		other clients can set up backtracing support.
2101
2102		----------------------------------------------------------------------
2103		Name: Matthias Blume
2104		Date: 2002/06/25 17:23:00 EDT
2105		Tag: blume-20020625-fpmax
2106		Description:
2107
2108		Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
2109		case.  (The sense of the isNaN test was reversed -- which made min and
2110		max always return their first argument.)
2111
2112		----------------------------------------------------------------------
2113		Name: Matthias Blume
2114		Date: 2002/06/11
2115		Tag: blume-20020611-unixpath
2116		Description:
2117
2118		Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
2119
2120		----------------------------------------------------------------------
2121		Name: Matthias Blume
2122		Date: 2002/06/10 16:35:00 EDT
2123		Tag: blume-20020610-ieeereal
2124		Description:
2125
2126		I back-ported my implementation of IEEEReal.fromString from the
2127		idlbasis-devel branch so that we can test it.
2128
2129		Another small change is that ppDec tries to give more information
2130		than just "<sig>" in the case of functors.  However, this code is
2131		broken in some mysterious way if the functor's body's signature
2132		has not been declared by ascription but gets inferred from the
2133		implementation.  This needs fixing...
2134
2135		----------------------------------------------------------------------
2136		Name: Matthias Blume
2137		Date: 2002/05/31
2138		Tag: blume-20020531-btrace-mode
2139		Description:
2140
2141		Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
2142		the wayside when I switched over to using Controls everywhere.)
2143
2144		----------------------------------------------------------------------
2145		Name: Lal George
2146		Date: 2002/05/23 12:21:40 EDT
2147		Tag: george-20020523-visual-labels
2148		Description:
2149
2150		Labels are now displayed in the graphical output to make
2151		the fall-through and target blocks obvious.
2152
2153		----------------------------------------------------------------------
2154		Name: Matthias Blume
2155		Date: 2002/05/22 11:03:00 EDT
2156		Tag: blume-20020522-shrink
2157		Description:
2158
2159		John tweaked yesterday's fix for 1131 to handle an out-of-memory
2160		situation that comes up when allocating huge arrays.
2161
2162		----------------------------------------------------------------------
2163		Name: Matthias Blume
2164		Date: 2002/05/21 16:00:00 EDT
2165		Tag: Release_110_40
2166		Description:
2167
2168		New working release (110.40).  New bootfiles.
2169
2170		[Also: John Reppy fixed GC bug 1131.]
2171
2172		----------------------------------------------------------------------
2173		Name: Matthias Blume
2174		Date: 2002/05/21 12:35:00 EDT
2175		Tag: blume-20020521-cmdoc
2176		Description:
2177
2178		CM documentation update.
2179
2180		----------------------------------------------------------------------
2181		Name: Matthias Blume
2182		Date: 2002/05/21 10:55:00 EDT
2183		Tag: blume-20020521-misc
2184		Description:
2185
2186		- John tweaked runtime to be silent on heap export (except when
2187		GC messages are on).
2188		- I added a few more things (cross-compiling versions of CMB) to
2189		config/preloads (as suggestions).
2190
2191		----------------------------------------------------------------------
2192		Name: Matthias Blume
2193		Date: 2002/05/20 22:25:00 EDT
2194		Tag: blume-20020520-controls
2195		Description:
2196
2197		- Added ControlUtil structure to control-lib.cm.
2198		- Use it throughout.
2199		- Used Controls facility to define MLRISC controls (as opposed to
2200		registering MLRISC control ref cells with Controls after the fact)
2201		- Fixed messed-up controls priorities.
2202
2203		* Removed again all the stuff from config/preloads that one wouldn't
2204		be able to preload at the time the initial heap image is built.
2205		(Many libraries, e.g., CML, do not exist yet at this time.  The
2206		only libraries that can be preloaded via config/preloads are those
2207		that come bundled with the bootfiles.)
2208
2209		----------------------------------------------------------------------
2210		Name: Matthias Blume
2211		Date: 2002/05/20 10:59:00 EDT
2212		Tag: blume-20020520-preloads
2213		Description:
2214
2215		Added a lot of commented-out suggestions for things to be included
2216		in config/preloads.
2217
2218		----------------------------------------------------------------------
2219		Name: Allen Leung
2220		Date: 2002/05/18 14:20:00 EDT
2221		Tag: leunga-20020518-mdl
2222		Description:
2223
2224		o Made the mdl tool stuff compile and run again.
2225		o I've disabled all the stuff that depends on RTL specifications; they
2226		are all badly broken anyway.
2227
2228		----------------------------------------------------------------------
2229		Name: Matthias Blume
2230		Date: 2002/05/17 16:49:00 EDT
2231		Tag: blume-20020517-controls
2232		Description:
2233
2234		0. John Reppy made several modifications to the SML/NJ library.
2235		In particular, there is a shiny new controls-lib.cm.
2236
2237		1. Pushed new controls interface through compiler so that everything
2238		compiles again.
2239
2240		2. Added FormatComb and FORMAT_COMB to the CML version of the
2241		SML/NJ library (so that CML compiles again).
2242
2243		3. Modified init scripts because XXX_DEFAULT environment variables
2244		are no longer with us.  (Boot-time initialization is now done
2245		using the same environment variables that are also used for
2246		startup-time initialization of controls.)
2247
2248		----------------------------------------------------------------------
2249		Name: Lal George
2250		Date: 2002/05/15 09:20:10 EDT
2251		Tag: george-20020515-pseudo-op-decls
2252		Description:
2253
2254		All pseudo-ops emitted before the first segment declaration
2255		such as TEXT, DATA, and BSS directives are assumed to be global
2256		declarations and are emitted first in the assembly file. This is
2257		useful in a number of situations where one has pseudo-ops that are not
2258		specific to any segment, and also works around the constraint that one
2259		cannot have client pseudo-ops in the TEXT segment.
2260
2261		Because no segment is associated with these declarations it is
2262		an error to allocate any space or objects before the first segment
2263		directive and an exception will be raised. However, we cannot make
2264		this check for client pseudo-ops.
2265
2266		These top level declarations are a field in the CFG graph_info.
2267		In theory you can continue to add to this field after the CFG has been
2268		built -- provided you know what you are doing;-)
2269
2270		----------------------------------------------------------------------
2271		Name: Matthias Blume
2272		Date: 2002/05/13 16:40:00 EDT
2273		Tag: blume-20020513-pp-etc
2274		Description:
2275
2276		A few minor bugfixes:
2277
2278		- Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
2279		(Bogus printouts for redefined bindings still occur.  Compiler
2280		bug should no longer occur now.  We need to redo the prettyprinter
2281		from scratch.)
2282
2283		- CM pathname printer now also adds escape sequences for ( and )
2284
2285		- commend and docu fixes for ml-nlffi
2286
2287		----------------------------------------------------------------------
2288		Name: Matthias Blume
2289		Date: 2002/05/10 16:40:00 EDT
2290		Tag: blume-20020510-erg-textio
2291		Description:
2292
2293		Applied the following bugfix provided by Emden Gansner:
2294
2295		Output is corrupted when outputSubstr is used rather than output.
2296
2297		The problem occurs when a substring
2298
2299		ss = (s, dataStart, dataLen)
2300
2301		where dataStart > 0, fills a stream buffer with avail bytes left.
2302		avail bytes of s, starting at index dataStart, are copied into the
2303		buffer, the buffer is flushed, and then the remaining dataLen-avail
2304		bytes of ss are copied into the beginning of the buffer. Instead of
2305		starting this copy at index dataStart+avail in s, the current code
2306		starts the copy at index avail.
2307
2308		Fix:
2309		In text-io-fn.sml, change line 695 from
2310		val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
2311		to
2312		val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
2313
2314		----------------------------------------------------------------------
2315		Name: Matthias Blume
2316		Date: 2002/04/12 13:55:00 EDT
2317		Tag: blume-20020412-assyntax
2318		Description:
2319
2320		1. Grabbed newer assyntax.h from the XFree86 project.
2321		2. Fiddled with how to compile X86.prim.asm without warnings.
2322		3. (Very) Minor cleanup in CM.
2323
2324		----------------------------------------------------------------------
2325		Name: Matthias Blume
2326		Date: 2002/04/01 (no joke!) 17:07:00 EST
2327		Tag: blume-20020401-x86div
2328		Description:
2329
2330		Added full support for div/mod/rem/quot on the x86, using the machine
2331		instruction's two results (without clumsily recomputing the remainder)
2332		directly where appropriate.
2333
2334		Some more extensive power-of-two support was added to the x86 instruction
2335		selector (avoiding expensive divs, mods, and muls where they can be
2336		replaced with cheaper shifts and masks).  However, this sort of thing
2337		ought to be done earlier, e.g., within the CPS optimizer so that
2338		all architectures benefit from it.
2339
2340		The compiler compiles to a fixed point, but changes might be somewhat
2341		fragile nevertheless.  Please, report any strange things that you might
2342		see wrt. div/mod/quot/rem...
2343
2344		----------------------------------------------------------------------
2345		Name: Matthias Blume
2346		Date: 2002/03/29 17:22:00
2347		Tag: blume-20020329-div
2348		Description:
2349
2350		Fixed my broken div/mod logic.  Unfortunately, this means that the
2351		inline code for div/mod now has one more comparison than before.
2352		Fast paths (quotient > 0 or remainder = 0) are not affected, though.
2353		The problem was with quotient = 0, because that alone does not tell
2354		us which way the rounding went.  One then has to look at whether
2355		remainder and divisor have the same sign...  :(
2356
2357		Anyway, I replaced the bootfiles with fresh ones...
2358
2359		----------------------------------------------------------------------
2360		Name: Matthias Blume
2361		Date: 2002/03/29 14:10:00 EST
2362		Tag: blume-20020329-inlprims
2363		Description:
2364
2365		NEW BOOTFILES!!!    Version number bumped to 110.39.3.
2366
2367		Primops have changed. This means that the bin/boot-file formats have
2368		changed as well.
2369
2370		To make sure that there is no confusion, I made a new version.
2371
2372
2373		CHANGES:
2374
2375		* removed REMT from mltree (remainder should never overflow).
2376
2377		* added primops to deal with divisions of all flavors to the frontend
2378
2379		* handled these primops all the way through so they map to their respective
2380		MLRISC support
2381
2382		* used these primops in the implementation of Int, Int32, Word, Word32
2383
2384		* removed INLDIV, INLMOD, and INLREM as they are no longer necessary
2385
2386		* parameterized INLMIN, INLMAX, and INLABS by a numkind
2387
2388		* translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
2389		floating point
2390
2391		* used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
2392		and Real (but Real.abs maps to a separate floating-point-only primop)
2393
2394
2395		TODO items:
2396
2397		* Hacked Alpha32 instruction selection, disabling the selection of REMx
2398		instructions because the machine instruction encoder cannot handle
2399		them.  (Hppa, PPC, and Sparc instruction selection did not handle
2400		REM in the first place, and REM is supported by the x86 machine coder.)
2401
2402		* Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
2403		selection phase.  (The two can be streamlined because the hardware
2404		delivers both quotient and remainder at the same time anyway.)
2405
2406		* Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
2407		(Currently the behavior is inconsistent both across architectures and
2408		wrt. the draft Basis spec.)
2409
2410		* Word8 should eventually be handled natively, too.
2411
2412		* There seems to be one serious bug in mltree-gen.sml.  It appears, though,
2413		as if there currently is no execution path that could trigger it in
2414		SML/NJ.  (The assumptions underlying functions arith and promotable do not
2415		hold for things like multiplication and division.)
2416
2417		----------------------------------------------------------------------
2418		Name: Matthias Blume
2419		Date: 2002/03/27 16:27:00 EST
2420		Tag: blume-20020327-mlrisc-divisions
2421		Description:
2422
2423		Added support for all four division operations (ML's div, mod, quot,
2424		and rem) to MLRISC.  In the course of doing so, I also rationalized
2425		the naming (no more annoying switch-around of DIV and QUOT), by
2426		parameterizing the operation by div_rounding_mode (which can be either
2427		DIV_TO_ZERO or DIV_TO_NEGINF).
2428
2429		The generic MLTreeGen functor takes care of compiling all four
2430		operations down to only round-to-zero div.
2431
2432		Missing pieces:
2433
2434		* Doing something smarter than relying on MLTreeGen on architectures
2435		like, e.g., the x86 where hardware division delivers both quotient and
2436		remainder at the same time.  With this, the implementation of the
2437		round-to-neginf operations could be further streamlined.
2438
2439		* Remove inlining support for div/mod/rem from the frontend and replace it
2440		with primops that get carried through to the backend.  Do this for all
2441		int and word types.
2442
2443		----------------------------------------------------------------------
2444		Name: Matthias Blume
2445		Date: 2002/03/25 17:25:00 EST
2446		Tag: blume-20020325-divmod
2447		Description:
2448
2449		I improved (hopefully without breaking them) the implementation of Int.div,
2450		Int.mod, and Int.rem.   For this, the code in translate.sml now takes
2451		advantage of the following observations:
2452
2453		Let  q = x quot y      r = x rem y
2454		d = x div  y      m = x mod y
2455
2456		where "quot" is the round-to-zero version of integer division that
2457		hardware usually provides.  Then we have:
2458
2459		r = x - q * y        where neither the * nor the - will overflow
2460		d = if q >= 0 orelse x = q * y then q else q - 1
2461		where neither the * nor the - will overflow
2462		m = if q >= 0 orelse r = 0 then r else r + y
2463		where the + will not overflow
2464
2465		This results in substantial simplification of the generated code.
2466		The following table shows the number of CFG nodes and edges generated
2467		for
2468		fun f (x, y) = x OPER y
2469		(* with OPER \in div, mod, quot, rem *)
2470
2471
2472		OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
2473		--------------------------------------------------------
2474		div |         24 |         39 |         12 |         16
2475		mod |         41 |         71 |         12 |         16
2476		quot |          8 |         10 |          8 |         10
2477		rem |         10 |         14 |          8 |         10
2478
2479
2480		----------------------------------------------------------------------
2481		Name: Matthias Blume
2482		Date: 2002/03/25 22:06:00 EST
2483		Tag: blume-20020325-cprotobug
2484		Description:
2485
2486		Fixed a bug in cproto (c prototype decoder).
2487
2488		----------------------------------------------------------------------
2489		Name: Matthias Blume
2490		Date: 2002/03/25 16:00:00 EST
2491		Tag: blume-20020325-raw-primops
2492		Description:
2493
2494		I did some cleanup to Allen's new primop code and
2495		replaced yesterday's bootfiles with new ones.
2496		(But they are stored in the same place.)
2497
2498		----------------------------------------------------------------------
2499		Name: Matthias Blume
2500		Date: 2002/03/24 22:40:00 EST
2501		Tag: blume-20020324-bootfiles
2502		Description:
2503
2504		Made the bootfiles that Allen asked for.
2505
2506		----------------------------------------------------------------------
2507		Name: Allen Leung
2508		Date: 2002/03/23 15:50:00 EST
2509		Tag: leunga-20020323-flint-cps-rcc-primops
2510		Description:
2511
2512		1. Changes to FLINT primops:
2513
2514		(* make a call to a C-function;
2515		* The primop carries C function prototype information and specifies
2516		* which of its (ML-) arguments are floating point. C prototype
2517		* information is for use by the backend, ML information is for
2518		* use by the CPS converter. *)
2519		| RAW_CCALL of { c_proto: CTypes.c_proto,
2520		ml_args: ccall_type list,
2521		ml_res_opt: ccall_type option,
2522		reentrant : bool
2523		} option
2524		(* Allocate uninitialized storage on the heap.
2525		* The record is meant to hold short-lived C objects, i.e., they
2526		* are not ML pointers.  With the tag, the representation is
2527		* the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
2528		*)
2529		| RAW_RECORD of {tag:bool,sz:int}
2530		and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
2531
2532		2.  These CPS primops are now overloaded:
2533
2534		rawload of {kind:numkind}
2535		rawstore of {kind:numkind}
2536
2537		The one argument form is:
2538
2539		rawload {kind} address
2540
2541		The two argument form is:
2542
2543		rawload {kind} [ml object, byte-offset]
2544
2545		3. RAW_CCALL/RCC now takes two extra arguments:
2546
2547		a. The first is whether the C call is reentrant, i.e., whether
2548		ML state should be saved and restored.
2549		b. The second argument is a string argument specifying the name of
2550		library and the C function.
2551
2552		These things are currently not handled in the code generator, yet.
2553
2554		4. In CProto,
2555
2556		An encoding type of "bool" means "ml object" and is mapped into
2557		C prototype of PTR.  Note that "bool" is different than "string",
2558		even though "string" is also mapped into PTR, because "bool"
2559		is assigned an CPS type of BOGt, while "string" is assigned INT32t.
2560
2561		5. Pickler/unpicker
2562
2563		Changed to handle RAW_RECORD and newest RAW_CCALL
2564
2565		6. MLRiscGen,
2566
2567		1. Changed to handle the new rawload/rawstore/rawrecord operators.
2568		2. Code for handling C Calls has been moved to a new module CPSCCalls,
2569		in the file CodeGen/cpscompile/cps-c-calls.sml
2570
2571		7. Added the conditional move operator
2572
2573		condmove of branch
2574
2575		to cps.  Generation of this is still buggy so it is currently
2576		disabled.
2577
2578		----------------------------------------------------------------------
2579		Name: Lal George
2580		Date: 2002/03/22 14:18:25 EST
2581		Tag: george-20020322-cps-branch-prob
2582		Description:
2583
2584		Implemented the Ball-Larus branch prediction-heuristics, and
2585		incorporated graphical viewers for control flow graphs.
2586
2587		Ball-Larus Heuristics:
2588		---------------------
2589		See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
2590
2591		By design it uses the Dempster-Shafer theory for combining
2592		probabilities.  For example, in the function:
2593
2594		fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
2595
2596		the ball-larus heuristics predicts that the n=0 is unlikely
2597		(OH-heuristic), and the 'then' branch is unlikely because of the
2598		RH-heuristic -- giving the 'then' branch an even lower combined
2599		probability using the Dempster-Shafer theory.
2600
2601		Finally, John Reppy's loop analysis in MLRISC, further lowers the
2602		probability of the 'then' branch because of the loop in the else
2603		branch.
2604
2605
2606		Graphical Viewing:
2607		------------------
2608		I merely plugged in Allen's graphical viewers into the compiler. The
2609		additional code is not much. At the top level, saying:
2610
2611		Control.MLRISC.getFlag "cfg-graphical-view" := true;
2612
2613		will display the graphical view of the control flow graph just before
2614		back-patching.  daVinci must be in your path for this to work. If
2615		daVinci is not available, then the default viewer can be changed
2616		using:
2617
2618		Control.MLRISC.getString "viewer"
2619
2620		which can be set to "dot" or "vcg" for the corresponding viewers. Of
2621		course, these viewers must be in your path.
2622
2623		The above will display the compilation unit at the level of clusters,
2624		many of which are small, boring, and un-interesting. Also setting:
2625
2626		Control.MLRISC.getInt "cfg-graphical-view_size"
2627
2628		will display clusters that are larger than the value set by the above.
2629
2630
2631		----------------------------------------------------------------------
2632		Name: Matthias Blume
2633		Date: 2002/03/21 22:20:00 EST
2634		Tag: blume-20020321-kmp-bugfix
2635		Description:
2636
2637		Changed the interface to the KMP routine in PreString and fixed
2638		a minor bug in one place where it was used.
2639
2640		----------------------------------------------------------------------
2641		Name: Allen Leung
2642		Date: 2002/03/21 20:30:00 EST
2643		Tag: leunga-20020321-cfg
2644		Description:
2645
2646		Fixed a potential problem in cfg edge splitting.
2647
2648		----------------------------------------------------------------------
2649		Name: Allen Leung
2650		Date: 2002/03/21 17:15:00 EST
2651		Tag: leunga-20020321-x86-fp-cfg
2652		Description:
2653
2654		1. Recoded the buggy parts of x86-fp.
2655
2656		a. All the block reordering code has been removed.
2657		We now depend on the block placement phases to do this work.
2658
2659		b. Critical edge splitting code has been simplified and moved into the
2660		CFG modules, as where they belong.
2661
2662		Both of these were quite buggy and complex.  The code is now much, much
2663		simpler.
2664
2665		2. X86 backend.
2666
2667		a. Added instructions for 64-bit support.  Instruction selection for
2668		64-bit has not been committed, however, since that
2669		requires changes to MLTREE which haven't been approved by
2670		Lal and John.
2671
2672		b. Added support for FUCOMI and FUCOMIP when generating code for
2673		PentiumPro and above.  We only generate these instructions in
2674		the fast-fp mode.
2675
2676		c. Added cases for JP and JNP in X86FreqProps.
2677
2678		3. CFG
2679
2680		CFG now has a bunch of methods for edge splitting and merging.
2681
2682		4. Machine description.
2683
2684		John's simplification of MLTREE_BASIS.fcond broke a few machine
2685		description things:
2686
2687		rtl-build.{sig,sml} and hppa.mdl fixed.
2688
2689		NOTE: the machine description stuff in the repository is still broken.
2690		Again, I can't put my fixes in because that involves
2691		changes to MLTREE.
2692
2693		----------------------------------------------------------------------
2694		Name: Matthias Blume
2695		Date: 2002/03/20 15:55:00 EST
2696		Tag: blume-20020320-kmp
2697		Description:
2698
2699		Implemented Knuth-Morris-Pratt string matching in PreString and used
2700		it for String.isSubstring, Substring.isSubstring, and
2701		Substring.position.
2702
2703		(Might need some stress-testing.  Simple examples worked fine.)
2704
2705		----------------------------------------------------------------------
2706		Name: Matthias Blume
2707		Date: 2002/03/19 16:37:00 EST
2708		Tag: blume-20020319-witnesses
2709		Description:
2710
2711		Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
2712
2713		This implements a generic mechanism for changing constness qualifiers
2714		anywhere within big C types without resorting to outright "casts".
2715		(So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
2716		modify the constness at the outermost level.)
2717		The implementation of "convert" is based on the idea of "witness"
2718		values -- values that are not used by the operation but whose types
2719		"testify" to their applicability.  On the implementation side, "convert"
2720		is simply a projection (returning its second curried argument).  With
2721		cross-module inlining, it should not result in any machine code being
2722		generated.
2723
2724		----------------------------------------------------------------------
2725		Name: Matthias Blume
2726		Date: 2002/03/15 16:40:00 EST
2727		Tag: blume-20020315-basis
2728		Description:
2729
2730		Provided (preliminary?) implementations for
2731
2732		{String,Substring}.{concatWith,isSuffix,isSubstring}
2733
2734		and
2735
2736		Substring.full
2737
2738		Those are in the Basis spec but they were missing in SML/NJ.
2739
2740		----------------------------------------------------------------------
2741		Name: Matthias Blume
2742		Date: 2002/03/14 21:30:00 EST
2743		Tag: blume-20020314-controls
2744		Description:
2745
2746		Controls:
2747		---------
2748
2749		1. Factored out the recently-added Controls : CONTROLS stuff and put
2750		it into its own library $/controls-lib.cm.  The source tree for
2751		this is under src/smlnj-lib/Controls.
2752
2753		2. Changed the names of types and functions in this interface, so they
2754		make a bit more "sense":
2755
2756		module -> registry
2757		'a registry -> 'a group
2758
2759		3. The interface now deals in ref cells only.  The getter/setter interface
2760		is (mostly) gone.
2761
2762		4. Added a function that lets one register an already-existing ref cell.
2763
2764		5. Made the corresponding modifications to the rest of the code so that
2765		everything compiles again.
2766
2767		6. Changed the implementation of Controls.MLRISC back to something closer
2768		to the original.  In particular, this module (and therefore MLRISC)
2769		does not depend on Controls.  There now is some link-time code in
2770		int-sys.sml that registers the MLRISC controls with the Controls
2771		module.
2772
2773		CM:
2774		---
2775
2776		* One can now specify the lambda-split aggressiveness in init.cmi.
2777
2778		----------------------------------------------------------------------
2779		Name: Allen Leung
2780		Date: 2002/03/13 17:30:00 EST
2781		Tag: leunga-20020313-x86-fp-unary
2782		Description:
2783
2784		Bug fix for:
2785
2786		> leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
2787		> Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
2788		> - fun f(x,(y,z)) = Real.~ y;
2789		> [autoloading]
2790		> [autoloading done]
2791		>       fchsl   (%eax), 184(%esp)
2792		> Error: MLRisc bug: X86MCEmitter.emitInstr
2793		>
2794		> uncaught exception Error
2795		>   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
2796
2797		The problem was that the code generator did not generate any fp registers
2798		in this case, and the ra didn't know that it needed to run the X86FP phase to
2799		translate the pseudo fp instruction.   This only happened with unary fp
2800		operators in certain situations.
2801
2802		----------------------------------------------------------------------
2803		Name: Matthias Blume
2804		Date: 2002/03/13 14:00:00 EST
2805		Tag: blume-20020313-overload-etc
2806		Description:
2807
2808		1. Added _overload as a synonym for overload for backward compatibility.
2809		(Control.overloadKW must be true for either version to be accepted.)
2810
2811		2. Fixed bug in install script that caused more things to be installed
2812		than what was requested in config/targets.
2813
2814		3. Made CM aware of the (_)overload construct so that autoloading
2815		works.
2816
2817		----------------------------------------------------------------------
2818		Name: Matthias Blume
2819		Date: 2002/03/12 22:03:00 EST
2820		Tag: blume-20020312-url
2821		Description:
2822
2823		Forgot to update BOOT and srcarchiveurl.
2824
2825		----------------------------------------------------------------------
2826		Name: Matthias Blume
2827		Date: 2002/03/12 17:30:00 EST
2828		Tag: blume-20020312-version110392
2829		Description:
2830
2831		Yet another version number bump (because of small changes to the
2832		binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
2833
2834		Changes:
2835
2836		The new pid generation scheme described a few weeks ago was overly
2837		complicated.  I implemented a new mechanism that is simpler and
2838		provides a bit more "stability":  Once CM has seen a compilation
2839		unit, it keeps its identity constant (as long as you do not delete
2840		those crucial CM/GUID/* files).  This means that when you change
2841		an interface, compile, then go back to the old interface, and
2842		compile again, you arrive at the original pid.
2843
2844		There now also is a mechanism that instructs CM to use the plain
2845		environment hash as a module's pid (effectively making its GUID
2846		the empty string).  For this, "noguid" must be specified as an
2847		option to the .sml file in question within its .cm file.
2848		This is most useful for code that is being generated by tools such
2849		as ml-nlffigen (because during development programmers tend to
2850		erase the tool's entire output directory tree including CM's cached
2851		GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
2852		revert to the old, broken behavior of SML/NJ, but in specific cases
2853		where there is no danger of interface confusion, its use is ok
2854		(I think).
2855
2856		ml-nlffigen by default generates "noguid" annotations.  They can be
2857		turned off by specifying -guid in its command line.
2858
2859		----------------------------------------------------------------------
2860		Name: Lal George
2861		Date: 2002/03/12 12 14:42:36 EST
2862		Tag: george-20020312-frequency-computation
2863		Description:
2864
2865		Integrated jump chaining and static block frequency into the
2866		compiler. More details and numbers later.
2867
2868		----------------------------------------------------------------------
2869		Name: Lal George
2870		Date: 2002/03/11 11 22:38:53 EST
2871		Tag: george-20020311-jump-chain-elim
2872		Description:
2873
2874		Tested the jump chain elimination on all architectures (except the
2875		hppa).  This is on by default right now and is profitable for the
2876		alpha and x86, however, it may not be profitable for the sparc and ppc
2877		when compiling the compiler.
2878
2879		The gc test will typically jump to a label at the end of the cluster,
2880		where there is another jump to an external cluster containing the actual
2881		code to invoke gc. This is to allow factoring of common gc invocation
2882		sequences. That is to say, we generate:
2883
2884		f:
2885		testgc
2886		ja   L1      % jump if above to L1
2887
2888		L1:
2889		jmp L2
2890
2891
2892		After jump chain elimination the 'ja L1' instructions is converted to
2893		'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
2894		up being implemented in their long form (if L2 is far away) using:
2895
2896		jbe     L3      % jump if below or equal to L3
2897		jmp     L2
2898		L3:
2899		...
2900
2901
2902		For large compilation units L2  may be far away.
2903
2904
2905		----------------------------------------------------------------------
2906		Name: Matthias Blume
2907		Date: 2002/03/11 13:30:00 EST
2908		Tag: blume-20020311-mltreeeval
2909		Description:
2910
2911		A functor parameter was missing.
2912
2913		----------------------------------------------------------------------
2914		Name: Allen Leung
2915		Date: 2002/03/11 10:30:00 EST
2916		Tag: leunga-20020311-runtime-string0
2917		Description:
2918
2919		The representation of the empty string now points to a
2920		legal null terminated C string instead of unit.  It is now possible
2921		to convert an ML string into C string with InlineT.CharVector.getData.
2922		This compiles into one single machine instruction.
2923
2924		----------------------------------------------------------------------
2925		Name: Allen Leung
2926		Date: 2002/03/10 23:55:00 EST
2927		Tag: leunga-20020310-x86-call
2928		Description:
2929
2930		Added machine generation for CALL instruction (relative displacement mode)
2931
2932		----------------------------------------------------------------------
2933		Name: Matthias Blume
2934		Date: 2002/03/08 16:05:00
2935		Tag: blume-20020308-entrypoints
2936		Description:
2937
2938		Version number bumped to 110.39.1.  NEW BOOTFILES!
2939
2940		Entrypoints: non-zero offset into a code object where execution should begin.
2941
2942		- Added the notion of an entrypoint to CodeObj.
2943		- Added reading/writing of entrypoint info to Binfile.
2944		- Made runtime system bootloader aware of entrypoints.
2945		- Use the address of the label of the first function given to mlriscGen
2946		as the entrypoint.  This address is currently always 0, but it will
2947		not be 0 once we turn on block placement.
2948		- Removed the linkage cluster code (which was The Other Way(tm) of dealing
2949		with entry points) from mlriscGen.
2950
2951		----------------------------------------------------------------------
2952		Name: Allen Leung
2953		Date: 2002/03/07 20:45:00 EST
2954		Tag: leunga-20020307-x86-cmov
2955		Description:
2956
2957		Bug fixes for CMOVcc on x86.
2958
2959		1. Added machine code generation for CMOVcc
2960		2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
2961		3. CMOVcc cannot have an immediate operand as argument.
2962
2963		----------------------------------------------------------------------
2964		Name: Matthias Blume
2965		Date: 2002/03/07 16:15:00 EST
2966		Tag: blume-20020307-controls
2967		Description:
2968
2969		This is a very large but mostly boring patch which makes (almost)
2970		every tuneable compiler knob (i.e., pretty much everything under
2971		Control.* plus a few other things) configurable via both the command
2972		line and environment variables in the style CM did its configuration
2973		until now.
2974
2975		Try starting sml with '-h' (or, if you are brave, '-H')
2976
2977		To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
2978		implements the underlying generic mechanism.
2979
2980		The interface to some of the existing such facilities has changed somewhat.
2981		For example, the MLRiscControl module now provides mkFoo instead of getFoo.
2982		(The getFoo interface is still there for backward-compatibility, but its
2983		use is deprecated.)
2984
2985		The ml-build script passes -Cxxx=yyy command-line arguments through so
2986		that one can now twiddle the compiler settings when using this "batch"
2987		compiler.
2988
2989		TODO items:
2990
2991		We should go through and throw out all controls that are no longer
2992		connected to anything.  Moreover, we should go through and provide
2993		meaningful (and correct!) documentation strings for those controls
2994		that still are connected.
2995
2996		Currently, multiple calls to Controls.new are accepted (only the first
2997		has any effect).  Eventually we should make sure that every control
2998		is being made (via Controls.new) exactly once.  Future access can then
2999		be done using Controls.acc.
3000
3001		Finally, it would probably be a good idea to use the getter-setter
3002		interface to controls rather than ref cells.  For the time being, both
3003		styles are provided by the Controls module, but getter-setter pairs are
3004		better if thread-safety is of any concern because they can be wrapped.
3005
3006		*****************************************
3007
3008		One bug fix: The function blockPlacement in three of the MLRISC
3009		backpatch files used to be hard-wired to one of two possibilities at
3010		link time (according to the value of the placementFlag).  But (I
3011		think) it should rather sense the flag every time.
3012
3013		*****************************************
3014
3015		Other assorted changes (by other people who did not supply a HISTORY entry):
3016
3017		1. the cross-module inliner now works much better (Monnier)
3018		2. representation of weights, frequencies, and probabilities in MLRISC
3019		changed in preparation of using those for weighted block placement
3020		(Reppy, George)
3021
3022		----------------------------------------------------------------------
3023		Name: Lal George
3024		Date: 2002/03/07 14:44:24 EST 2002
3025		Tag: george-20020307-weighted-block-placement
3026
3027		Tested the weighted block placement optimization on all architectures
3028		(except the hppa) using AMPL to generate the block and edge frequencies.
3029		Changes were required in the machine properties to correctly
3030		categorize trap instructions. There is an MLRISC flag
3031		"weighted-block-placement" that can be used to enable weighted block
3032		placement, but this will be ineffective without block/edge
3033		frequencies (coming soon).
3034
3035
3036		----------------------------------------------------------------------
3037		Name: Lal George
3038		Date: 2002/03/05 17:24:48 EST
3039		Tag: george-20020305-linkage-cluster
3040
3041		In order to support the block placement optimization, a new cluster
3042		is generated as the very first cluster (called the linkage cluster).
3043		It contains a single jump to the 'real' entry point for the compilation
3044		unit. Block placement has no effect on the linkage cluster itself, but
3045		all the other clusters  have full freedom in the manner in which they
3046		reorder blocks or functions.
3047
3048		On the x86 the typical linkage code that is generated is:
3049		----------------------
3050		.align 2
3051		L0:
3052		addl    $L1-L0, 72(%esp)
3053		jmp     L1
3054
3055
3056		.align  2
3057		L1:
3058		----------------------
3059
3060		72(%esp) is the memory location for the stdlink register. This
3061		must contain the address of the CPS function being called. In the
3062		above example, it contains the address of  L0; before
3063		calling L1 (the real entry point for the compilation unit), it
3064		must contain the address for L1, and hence
3065
3066		addl $L1-L0, 72(%esp)
3067
3068		I have tested this on all architectures except the hppa.The increase
3069		in code size is of course negligible
3070
3071		----------------------------------------------------------------------
3072		Name: Allen Leung
3073		Date: 2002/03/03 13:20:00 EST
3074		Tag: leunga-20020303-mlrisc-tools
3075
3076		Added #[ ... ] expressions to mlrisc tools
3077
3078		----------------------------------------------------------------------
3079		Name: Matthias Blume
3080		Date: 2002/02/27 12:29:00 EST
3081		Tag: blume-20020227-cdebug
3082		Description:
3083
3084		- made types in structure C and C_Debug to be equal
3085		- got rid of code duplication (c-int.sml vs. c-int-debug.sml)
3086		- there no longer is a C_Int_Debug (C_Debug is directly derived from C)
3087
3088		----------------------------------------------------------------------
3089		Name: Matthias Blume
3090		Date: 2002/02/26 12:00:00 EST
3091		Tag: blume-20020226-ffi
3092		Description:
3093
3094		1. Fixed a minor bug in CM's "noweb" tool:
3095		If numbering is turned off, then truly don't number (i.e., do not
3096		supply the -L option to noweb).  The previous behavior was to supply
3097		-L'' -- which caused noweb to use the "default" line numbering scheme.
3098		Thanks to Chris Richards for pointing this out (and supplying the fix).
3099
3100		2. Once again, I reworked some aspects of the FFI:
3101
3102		A. The incomplete/complete type business:
3103
3104		- Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
3105		gone!
3106		- ML types representing an incomplete type are now *equal* to
3107		ML types representing their corresponding complete types (just like
3108		in C).  This is still safe because ml-nlffigen will not generate
3109		RTTI for incomplete types, nor will it generate functions that
3110		require access to such RTTI.   But when ML code generated from both
3111		incomplete and complete versions of the C type meet, the ML types
3112		are trivially interoperable.
3113
3114		NOTE:  These changes restore the full generality of the translation
3115		(which was previously lost when I eliminated functorization)!
3116
3117		B. Enum types:
3118
3119		- Structure C now has a type constructor "enum" that is similar to
3120		how the "su" constructor works.  However, "enum" is not a phantom
3121		type because each "T enum" has values (and is isomorphic to
3122		MLRep.Signed.int).
3123		- There are generic access operations for enum objects (using
3124		MLRep.Signed.int).
3125		- ml-nlffigen will generate a structure E_foo for each "enum foo".
3126		* The structure contains the definition of type "mlrep" (the ML-side
3127		representation type of the enum).  Normally, mlrep is the same
3128		as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
3129		then mlrep will be defined as a datatype -- thus facilitating
3130		pattern matching on mlrep values.
3131		("-ec" will be suppressed if there are duplicate values in an
3132		enumeration.)
3133		* Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
3134		will be generated for each C enum constant xxx.
3135		* Conversion functions m2i and i2m convert between mlrep and
3136		MLRep.Signed.int.  (Without "-ec", these functions are identities.)
3137		* Coversion functions c and ml convert between mlrep and "tag enum".
3138		* Access functions (get/set) fetch and store mlrep values.
3139		- By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
3140		enumerations are merged into one single enumeration represented by
3141		structure E_'.
3142
3143		----------------------------------------------------------------------
3144		Name: Allen Leung
3145		Date: 2002/02/25 04:45:00 EST
3146		Tag: leunga-20020225-cps-spill
3147
3148		This is a new implementation of the CPS spill phase.
3149		The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
3150		In case of problems, replace it with the old file spill.sml
3151
3152		The current compiler runs into some serious performance problems when
3153		constructing a large record.  This can happen when we try to compile a
3154		structure with many items.  Even a very simple structure like the following
3155		makes the compiler slow down.
3156
3157		structure Foo = struct
3158		val x_1 = 0w1 : Word32.int
3159		val x_2 = 0w2 : Word32.int
3160		val x_3 = 0w3 : Word32.int
3161		...
3162		val x_N = 0wN : Word32.int
3163		end
3164
3165		The following table shows the compile time, from N=1000 to N=4000,
3166		with the old compiler:
3167
3168		N
3169		1000   CPS 100 spill                           0.04u  0.00s  0.00g
3170		MLRISC ra                               0.06u  0.00s  0.05g
3171		(spills = 0 reloads = 0)
3172		TOTAL                                   0.63u  0.07s  0.21g
3173
3174		1100   CPS 100 spill                           8.25u  0.32s  0.64g
3175		MLRISC ra                               5.68u  0.59s  3.93g
3176		(spills = 0 reloads = 0)
3177		TOTAL                                   14.71u  0.99s  4.81g
3178
3179		1500   CPS 100 spill                           58.55u  2.34s  1.74g
3180		MLRISC ra                               5.54u  0.65s  3.91g
3181		(spills = 543 reloads = 1082)
3182		TOTAL                                   65.40u  3.13s  6.00g
3183
3184		2000   CPS 100 spill                           126.69u  4.84s  3.08g
3185		MLRISC ra                               0.80u  0.10s  0.55g
3186		(spills = 42 reloads = 84)
3187		TOTAL                                   129.42u  5.10s  4.13g
3188
3189		3000   CPS 100 spill                           675.59u  19.03s  11.64g
3190		MLRISC ra                               2.69u  0.27s  1.38g
3191		(spills = 62 reloads = 124)
3192		TOTAL                                   682.48u  19.61s  13.99g
3193
3194		4000   CPS 100 spill                           2362.82u  56.28s  43.60g
3195		MLRISC ra                               4.96u  0.27s  2.72g
3196		(spills = 85 reloads = 170)
3197		TOTAL                                   2375.26u  57.21s  48.00g
3198
3199		As you can see the old cps spill module suffers from some serious
3200		performance problem.  But since I cannot decipher the old code fully,
3201		instead of patching the problems up, I'm reimplementing it
3202		with a different algorithm.  The new code is more modular,
3203		smaller when compiled, and substantially faster
3204		(O(n log n) time and O(n) space).  Timing of the new spill module:
3205
3206		4000  CPS 100 spill                           0.02u  0.00s  0.00g
3207		MLRISC ra                               0.25u  0.02s  0.15g
3208		(spills=1 reloads=3)
3209		TOTAL                                   7.74u  0.34s  1.62g
3210
3211		Implementation details:
3212
3213		As far as I can tell, the purpose of the CPS spill module is to make sure the
3214		number of live variables at any program point (the bandwidth)
3215		does not exceed a certain limit, which is determined by the
3216		size of the spill area.
3217
3218		When the bandwidth is too large, we decrease the register pressure by
3219		packing live variables into spill records.  How we achieve this is
3220		completely different than what we did in the old code.
3221
3222		First, there is something about the MLRiscGen code generator
3223		that we should be aware of:
3224
3225		o MLRiscGen performs code motion!
3226
3227		In particular, it will move floating point computations and
3228		address computations involving only the heap pointer to
3229		their use sites (if there is only a single use).
3230		What this means is that if we have a CPS record construction
3231		statement
3232
3233		RECORD(k,vl,w,e)
3234
3235		we should never count the new record address w as live if w
3236		has only one use (which is often the case).
3237
3238		We should do something similar to floating point, but the transformation
3239		there is much more complex, so I won't deal with that.
3240
3241		Secondly, there are now two new cps primops at our disposal:
3242
3243		1. rawrecord of record_kind option
3244		This pure operator allocates some uninitialized storage from the heap.
3245		There are two forms:
3246
3247		rawrecord NONE [INT n]  allocates a tagless record of length n
3248		rawrecord (SOME rk) [INT n] allocates a tagged record of length n
3249		and initializes the tag.
3250
3251		2. rawupdate of cty
3252		rawupdate cty (v,i,x)
3253		Assigns to x to the ith component of record v.
3254		The storelist is not updated.
3255
3256		We use these new primops for both spilling and increment record construction.
3257
3258		1. Spilling.
3259
3260		This is implemented with a linear scan algorithm (but generalized
3261		to trees).  The algorithm will create a single spill record at the
3262		beginning of the cps function and use rawupdate to spill to it,
3263		and SELECT or SELp to reload from it.  So both spills and reloads
3264		are fine-grain operations.  In contrast, in the old algorithm
3265		"spills" have to be bundled together in records.
3266
3267		Ideally, we should sink the spill record construction to where
3268		it is needed.  We can even split the spill record into multiple ones
3269		at the places where they are needed.  But CPS is not a good
3270		representation for global code motion, so I'll keep it simple and
3271		am not attempting this.
3272
3273		2. Incremental record construction (aka record splitting).
3274
3275		Long records with many component values which are simulatenously live
3276		(recall that single use record addresses are not considered to
3277		be live) are constructed with rawrecord and rawupdate.
3278		We allocate space on the heap with rawrecord first, then gradually
3279		fill it in with rawupdate.  This is the technique suggested to me
3280		by Matthias.
3281
3282		Some restrictions on when this is applicable:
3283		1. It is not a VECTOR record.  The code generator currently does not handle
3284		this case. VECTOR record uses double indirection like arrays.
3285		2. All the record component values are defined in the same "basic block"
3286		as the record constructor.  This is to prevent speculative
3287		record construction.
3288
3289		----------------------------------------------------------------------
3290		Name: Allen Leung
3291		Date: 2002/02/22 01:02:00 EST
3292		Tag: leunga-20020222-mlrisc-tools
3293
3294		Minor bug fixes in the parser and rewriter
3295
3296		----------------------------------------------------------------------
3297		Name: Allen Leung
3298		Date: 2002/02/21 20:20:00 EST
3299		Tag: leunga-20020221-peephole
3300
3301		Regenerated the peephole files.  Some contained typos in the specification
3302		and some didn't compile because of pretty printing bugs in the old version
3303		of 'nowhere'.
3304
3305		----------------------------------------------------------------------
3306		Name: Allen Leung
3307		Date: 2002/02/19 20:20:00 EST
3308		Tag: leunga-20020219-mlrisc-tools
3309		Description:
3310
3311		Minor bug fixes to the mlrisc-tools library:
3312
3313		1.  Fixed up parsing colon suffixed keywords
3314		2.  Added the ability to shut the error messages up
3315		3.  Reimplemented the pretty printer and fixed up/improved
3316		the pretty printing of handle and -> types.
3317		4.  Fixed up generation of literal symbols in the nowhere tool.
3318		5.  Added some SML keywords to to sml.sty
3319
3320		----------------------------------------------------------------------
3321		Name: Matthias Blume
3322		Date: 2002/02/19 16:20:00 EST
3323		Tag: blume-20020219-cmffi
3324		Description:
3325
3326		A wild mix of changes, some minor, some major:
3327
3328		* All C FFI-related libraries are now anchored under $c:
3329		$/c.cm      --> $c/c.cm
3330		$/c-int.cm  --> $c/internals/c-int.cm
3331		$/memory.cm --> $c/memory/memory.cm
3332
3333		* "make" tool (in CM) now treats its argument pathname slightly
3334		differently:
3335		1. If the native expansion is an absolute name, then before invoking
3336		the "make" command on it, CM will apply OS.Path.mkRelative
3337		(with relativeTo = OS.FileSys.getDir()) to it.
3338		2. The argument will be passed through to subsequent phases of CM
3339		processing without "going native".  In particular, if the argument
3340		was an anchored path, then "make" will not lose track of that anchor.
3341
3342		* Compiler backends now "know" their respective C calling conventions
3343		instead of having to be told about it by ml-nlffigen.  This relieves
3344		ml-nlffigen from one of its burdens.
3345
3346		* The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
3347
3348		* Export C_DEBUG and C_Debug from $c/c.cm.
3349
3350		* C type encoding in ml-nlffi-lib has been improved to model the conceptual
3351		subtyping relationship between incomplete pointers and their complete
3352		counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
3353		with the convention of instantiating 'o with ('t, 'c) obj whenever
3354		the pointer target type is complete.  In the incomplete case, 'o
3355		will be instantiated with some "'c iobj" -- a type obtained by
3356		using one of the functors PointerToIncompleteType or PointerToCompleteType.
3357
3358		Operations that work on both incomplete and complete pointer types are
3359		typed as taking an 'o ptr while operations that require the target to
3360		be known are typed as taking some ('t, 'c) obj ptr.
3361
3362		voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
3363		where void is an eqtype without any values.  This makes it possible
3364		to work on voidptr values using functions meant to operate on light
3365		incomplete pointers.
3366
3367		* As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
3368		been vastly simplified.
3369
3370		----------------------------------------------------------------------
3371		Name: Matthias Blume
3372		Date: 2002/02/19 10:48:00 EST
3373		Tag: blume-20020219-pqfix
3374		Description:
3375
3376		Applied Chris Okasaki's bug fix for priority queues.
3377
3378		----------------------------------------------------------------------
3379		Name: Matthias Blume
3380		Date: 2002/02/15 17:05:00
3381		Tag: Release_110_39
3382		Description:
3383
3384		Last-minute retagging is becoming a tradition... :-(
3385
3386		This is the working release 110.39.
3387
3388		----------------------------------------------------------------------
3389		Name: Matthias Blume
3390		Date: 2002/02/15 16:00:00 EST
3391		Tag: Release_110_39-orig
3392		Description:
3393
3394		Working release 110.39.  New bootfiles.
3395
3396		(Update: There was a small bug in the installer so it wouldn't work
3397		with all shells.  So I retagged. -Matthias)
3398
3399		----------------------------------------------------------------------
3400		Name: Matthias Blume
3401		Date: 2002/02/15 14:17:00 EST
3402		Tag: blume-20020215-showbindings
3403		Description:
3404
3405		Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
3406		the latter can be useful for exploring what bindings are available at
3407		the interactive prompt.  (The first function returns only the list
3408		of symbols that are really bound, the second prints those but also the
3409		ones that CM's autoloading mechanism knows about.)
3410
3411		----------------------------------------------------------------------
3412	Name: Matthias Blume	Name: Matthias Blume
3413	Date: 2002/02/15 12:08:00 EST	Date: 2002/02/15 12:08:00 EST
3414	Tag: blume-20020215-iptrs	Tag: blume-20020215-iptrs

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.1067   changed lines   Added in v.1759

---

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