Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 648, Thu May 25 21:28:21 2000 UTC	revision 1240, Tue Jun 11 15:35:34 2002 UTC
#	Line 8	Line 8
8	The form of an entry should be:	The form of an entry should be:
9
10	Name:	Name:
11	Date:	Date: yyyy/mm/dd
12	Tag: <post-commit CVS tag>	Tag: <post-commit CVS tag>
13	Description:	Description:
14
15		----------------------------------------------------------------------
16		Name: Matthias Blume
17		Date: 2002/06/11
18		Tag: blume-20020611-unixpath
19		Description:
20
21		Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
22
23		----------------------------------------------------------------------
24		Name: Matthias Blume
25		Date: 2002/06/10 16:35:00 EDT
26		Tag: blume-20020610-ieeereal
27		Description:
28
29		I back-ported my implementation of IEEEReal.fromString from the
30		idlbasis-devel branch so that we can test it.
31
32		Another small change is that ppDec tries to give more information
33		than just "<sig>" in the case of functors.  However, this code is
34		broken in some mysterious way if the functor's body's signature
35		has not been declared by ascription but gets inferred from the
36		implementation.  This needs fixing...
37
38		----------------------------------------------------------------------
39		Name: Matthias Blume
40		Date: 2002/05/31
41		Tag: blume-20020531-btrace-mode
42		Description:
43
44		Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
45		the wayside when I switched over to using Controls everywhere.)
46
47		----------------------------------------------------------------------
48		Name: Lal George
49		Date: 2002/05/23 12:21:40 EDT
50		Tag: george-20020523-visual-labels
51		Description:
52
53		Labels are now displayed in the graphical output to make
54		the fall-through and target blocks obvious.
55
56		----------------------------------------------------------------------
57		Name: Matthias Blume
58		Date: 2002/05/22 11:03:00 EDT
59		Tag: blume-20020522-shrink
60		Description:
61
62		John tweaked yesterday's fix for 1131 to handle an out-of-memory
63		situation that comes up when allocating huge arrays.
64
65		----------------------------------------------------------------------
66		Name: Matthias Blume
67		Date: 2002/05/21 16:00:00 EDT
68		Tag: Release_110_40
69		Description:
70
71		New working release (110.40).  New bootfiles.
72
73		[Also: John Reppy fixed GC bug 1131.]
74
75		----------------------------------------------------------------------
76		Name: Matthias Blume
77		Date: 2002/05/21 12:35:00 EDT
78		Tag: blume-20020521-cmdoc
79		Description:
80
81		CM documentation update.
82
83		----------------------------------------------------------------------
84		Name: Matthias Blume
85		Date: 2002/05/21 10:55:00 EDT
86		Tag: blume-20020521-misc
87		Description:
88
89		- John tweaked runtime to be silent on heap export (except when
90		GC messages are on).
91		- I added a few more things (cross-compiling versions of CMB) to
92		config/preloads (as suggestions).
93
94		----------------------------------------------------------------------
95		Name: Matthias Blume
96		Date: 2002/05/20 22:25:00 EDT
97		Tag: blume-20020520-controls
98		Description:
99
100		- Added ControlUtil structure to control-lib.cm.
101		- Use it throughout.
102		- Used Controls facility to define MLRISC controls (as opposed to
103		registering MLRISC control ref cells with Controls after the fact)
104		- Fixed messed-up controls priorities.
105
106		* Removed again all the stuff from config/preloads that one wouldn't
107		be able to preload at the time the initial heap image is built.
108		(Many libraries, e.g., CML, do not exist yet at this time.  The
109		only libraries that can be preloaded via config/preloads are those
110		that come bundled with the bootfiles.)
111
112		----------------------------------------------------------------------
113		Name: Matthias Blume
114		Date: 2002/05/20 10:59:00 EDT
115		Tag: blume-20020520-preloads
116		Description:
117
118		Added a lot of commented-out suggestions for things to be included
119		in config/preloads.
120
121		----------------------------------------------------------------------
122		Name: Allen Leung
123		Date: 2002/05/18 14:20:00 EDT
124		Tag: leunga-20020518-mdl
125		Description:
126
127		o Made the mdl tool stuff compile and run again.
128		o I've disabled all the stuff that depends on RTL specifications; they
129		are all badly broken anyway.
130
131		----------------------------------------------------------------------
132		Name: Matthias Blume
133		Date: 2002/05/17 16:49:00 EDT
134		Tag: blume-20020517-controls
135		Description:
136
137		0. John Reppy made several modifications to the SML/NJ library.
138		In particular, there is a shiny new controls-lib.cm.
139
140		1. Pushed new controls interface through compiler so that everything
141		compiles again.
142
143		2. Added FormatComb and FORMAT_COMB to the CML version of the
144		SML/NJ library (so that CML compiles again).
145
146		3. Modified init scripts because XXX_DEFAULT environment variables
147		are no longer with us.  (Boot-time initialization is now done
148		using the same environment variables that are also used for
149		startup-time initialization of controls.)
150
151		----------------------------------------------------------------------
152		Name: Lal George
153		Date: 2002/05/15 09:20:10 EDT
154		Tag: george-20020515-pseudo-op-decls
155		Description:
156
157		All pseudo-ops emitted before the first segment declaration
158		such as TEXT, DATA, and BSS directives are assumed to be global
159		declarations and are emitted first in the assembly file. This is
160		useful in a number of situations where one has pseudo-ops that are not
161		specific to any segment, and also works around the constraint that one
162		cannot have client pseudo-ops in the TEXT segment.
163
164		Because no segment is associated with these declarations it is
165		an error to allocate any space or objects before the first segment
166		directive and an exception will be raised. However, we cannot make
167		this check for client pseudo-ops.
168
169		These top level declarations are a field in the CFG graph_info.
170		In theory you can continue to add to this field after the CFG has been
171		built -- provided you know what you are doing;-)
172
173		----------------------------------------------------------------------
174		Name: Matthias Blume
175		Date: 2002/05/13 16:40:00 EDT
176		Tag: blume-20020513-pp-etc
177		Description:
178
179		A few minor bugfixes:
180
181		- Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
182		(Bogus printouts for redefined bindings still occur.  Compiler
183		bug should no longer occur now.  We need to redo the prettyprinter
184		from scratch.)
185
186		- CM pathname printer now also adds escape sequences for ( and )
187
188		- commend and docu fixes for ml-nlffi
189
190		----------------------------------------------------------------------
191		Name: Matthias Blume
192		Date: 2002/05/10 16:40:00 EDT
193		Tag: blume-20020510-erg-textio
194		Description:
195
196		Applied the following bugfix provided by Emden Gansner:
197
198		Output is corrupted when outputSubstr is used rather than output.
199
200		The problem occurs when a substring
201
202		ss = (s, dataStart, dataLen)
203
204		where dataStart > 0, fills a stream buffer with avail bytes left.
205		avail bytes of s, starting at index dataStart, are copied into the
206		buffer, the buffer is flushed, and then the remaining dataLen-avail
207		bytes of ss are copied into the beginning of the buffer. Instead of
208		starting this copy at index dataStart+avail in s, the current code
209		starts the copy at index avail.
210
211		Fix:
212		In text-io-fn.sml, change line 695 from
213		val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
214		to
215		val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
216
217		----------------------------------------------------------------------
218		Name: Matthias Blume
219		Date: 2002/04/12 13:55:00 EDT
220		Tag: blume-20020412-assyntax
221		Description:
222
223		1. Grabbed newer assyntax.h from the XFree86 project.
224		2. Fiddled with how to compile X86.prim.asm without warnings.
225		3. (Very) Minor cleanup in CM.
226
227		----------------------------------------------------------------------
228		Name: Matthias Blume
229		Date: 2002/04/01 (no joke!) 17:07:00 EST
230		Tag: blume-20020401-x86div
231		Description:
232
233		Added full support for div/mod/rem/quot on the x86, using the machine
234		instruction's two results (without clumsily recomputing the remainder)
235		directly where appropriate.
236
237		Some more extensive power-of-two support was added to the x86 instruction
238		selector (avoiding expensive divs, mods, and muls where they can be
239		replaced with cheaper shifts and masks).  However, this sort of thing
240		ought to be done earlier, e.g., within the CPS optimizer so that
241		all architectures benefit from it.
242
243		The compiler compiles to a fixed point, but changes might be somewhat
244		fragile nevertheless.  Please, report any strange things that you might
245		see wrt. div/mod/quot/rem...
246
247		----------------------------------------------------------------------
248		Name: Matthias Blume
249		Date: 2002/03/29 17:22:00
250		Tag: blume-20020329-div
251		Description:
252
253		Fixed my broken div/mod logic.  Unfortunately, this means that the
254		inline code for div/mod now has one more comparison than before.
255		Fast paths (quotient > 0 or remainder = 0) are not affected, though.
256		The problem was with quotient = 0, because that alone does not tell
257		us which way the rounding went.  One then has to look at whether
258		remainder and divisor have the same sign...  :(
259
260		Anyway, I replaced the bootfiles with fresh ones...
261
262		----------------------------------------------------------------------
263		Name: Matthias Blume
264		Date: 2002/03/29 14:10:00 EST
265		Tag: blume-20020329-inlprims
266		Description:
267
268		NEW BOOTFILES!!!    Version number bumped to 110.39.3.
269
270		Primops have changed. This means that the bin/boot-file formats have
271		changed as well.
272
273		To make sure that there is no confusion, I made a new version.
274
275
276		CHANGES:
277
278		* removed REMT from mltree (remainder should never overflow).
279
280		* added primops to deal with divisions of all flavors to the frontend
281
282		* handled these primops all the way through so they map to their respective
283		MLRISC support
284
285		* used these primops in the implementation of Int, Int32, Word, Word32
286
287		* removed INLDIV, INLMOD, and INLREM as they are no longer necessary
288
289		* parameterized INLMIN, INLMAX, and INLABS by a numkind
290
291		* translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
292		floating point
293
294		* used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
295		and Real (but Real.abs maps to a separate floating-point-only primop)
296
297
298		TODO items:
299
300		* Hacked Alpha32 instruction selection, disabling the selection of REMx
301		instructions because the machine instruction encoder cannot handle
302		them.  (Hppa, PPC, and Sparc instruction selection did not handle
303		REM in the first place, and REM is supported by the x86 machine coder.)
304
305		* Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
306		selection phase.  (The two can be streamlined because the hardware
307		delivers both quotient and remainder at the same time anyway.)
308
309		* Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
310		(Currently the behavior is inconsistent both across architectures and
311		wrt. the draft Basis spec.)
312
313		* Word8 should eventually be handled natively, too.
314
315		* There seems to be one serious bug in mltree-gen.sml.  It appears, though,
316		as if there currently is no execution path that could trigger it in
317		SML/NJ.  (The assumptions underlying functions arith and promotable do not
318		hold for things like multiplication and division.)
319
320		----------------------------------------------------------------------
321		Name: Matthias Blume
322		Date: 2002/03/27 16:27:00 EST
323		Tag: blume-20020327-mlrisc-divisions
324		Description:
325
326		Added support for all four division operations (ML's div, mod, quot,
327		and rem) to MLRISC.  In the course of doing so, I also rationalized
328		the naming (no more annoying switch-around of DIV and QUOT), by
329		parameterizing the operation by div_rounding_mode (which can be either
330		DIV_TO_ZERO or DIV_TO_NEGINF).
331
332		The generic MLTreeGen functor takes care of compiling all four
333		operations down to only round-to-zero div.
334
335		Missing pieces:
336
337		* Doing something smarter than relying on MLTreeGen on architectures
338		like, e.g., the x86 where hardware division delivers both quotient and
339		remainder at the same time.  With this, the implementation of the
340		round-to-neginf operations could be further streamlined.
341
342		* Remove inlining support for div/mod/rem from the frontend and replace it
343		with primops that get carried through to the backend.  Do this for all
344		int and word types.
345
346		----------------------------------------------------------------------
347		Name: Matthias Blume
348		Date: 2002/03/25 17:25:00 EST
349		Tag: blume-20020325-divmod
350		Description:
351
352		I improved (hopefully without breaking them) the implementation of Int.div,
353		Int.mod, and Int.rem.   For this, the code in translate.sml now takes
354		advantage of the following observations:
355
356		Let  q = x quot y      r = x rem y
357		d = x div  y      m = x mod y
358
359		where "quot" is the round-to-zero version of integer division that
360		hardware usually provides.  Then we have:
361
362		r = x - q * y        where neither the * nor the - will overflow
363		d = if q >= 0 orelse x = q * y then q else q - 1
364		where neither the * nor the - will overflow
365		m = if q >= 0 orelse r = 0 then r else r + y
366		where the + will not overflow
367
368		This results in substantial simplification of the generated code.
369		The following table shows the number of CFG nodes and edges generated
370		for
371		fun f (x, y) = x OPER y
372		(* with OPER \in div, mod, quot, rem *)
373
374
375		OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
376		--------------------------------------------------------
377		div |         24 |         39 |         12 |         16
378		mod |         41 |         71 |         12 |         16
379		quot |          8 |         10 |          8 |         10
380		rem |         10 |         14 |          8 |         10
381
382
383		----------------------------------------------------------------------
384		Name: Matthias Blume
385		Date: 2002/03/25 22:06:00 EST
386		Tag: blume-20020325-cprotobug
387		Description:
388
389		Fixed a bug in cproto (c prototype decoder).
390
391		----------------------------------------------------------------------
392		Name: Matthias Blume
393		Date: 2002/03/25 16:00:00 EST
394		Tag: blume-20020325-raw-primops
395		Description:
396
397		I did some cleanup to Allen's new primop code and
398		replaced yesterday's bootfiles with new ones.
399		(But they are stored in the same place.)
400
401		----------------------------------------------------------------------
402		Name: Matthias Blume
403		Date: 2002/03/24 22:40:00 EST
404		Tag: blume-20020324-bootfiles
405		Description:
406
407		Made the bootfiles that Allen asked for.
408
409		----------------------------------------------------------------------
410		Name: Allen Leung
411		Date: 2002/03/23 15:50:00 EST
412		Tag: leunga-20020323-flint-cps-rcc-primops
413		Description:
414
415		1. Changes to FLINT primops:
416
417		(* make a call to a C-function;
418		* The primop carries C function prototype information and specifies
419		* which of its (ML-) arguments are floating point. C prototype
420		* information is for use by the backend, ML information is for
421		* use by the CPS converter. *)
422		| RAW_CCALL of { c_proto: CTypes.c_proto,
423		ml_args: ccall_type list,
424		ml_res_opt: ccall_type option,
425		reentrant : bool
426		} option
427		(* Allocate uninitialized storage on the heap.
428		* The record is meant to hold short-lived C objects, i.e., they
429		* are not ML pointers.  With the tag, the representation is
430		* the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
431		*)
432		| RAW_RECORD of {tag:bool,sz:int}
433		and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
434
435		2.  These CPS primops are now overloaded:
436
437		rawload of {kind:numkind}
438		rawstore of {kind:numkind}
439
440		The one argument form is:
441
442		rawload {kind} address
443
444		The two argument form is:
445
446		rawload {kind} [ml object, byte-offset]
447
448		3. RAW_CCALL/RCC now takes two extra arguments:
449
450		a. The first is whether the C call is reentrant, i.e., whether
451		ML state should be saved and restored.
452		b. The second argument is a string argument specifying the name of
453		library and the C function.
454
455		These things are currently not handled in the code generator, yet.
456
457		4. In CProto,
458
459		An encoding type of "bool" means "ml object" and is mapped into
460		C prototype of PTR.  Note that "bool" is different than "string",
461		even though "string" is also mapped into PTR, because "bool"
462		is assigned an CPS type of BOGt, while "string" is assigned INT32t.
463
464		5. Pickler/unpicker
465
466		Changed to handle RAW_RECORD and newest RAW_CCALL
467
468		6. MLRiscGen,
469
470		1. Changed to handle the new rawload/rawstore/rawrecord operators.
471		2. Code for handling C Calls has been moved to a new module CPSCCalls,
472		in the file CodeGen/cpscompile/cps-c-calls.sml
473
474		7. Added the conditional move operator
475
476		condmove of branch
477
478		to cps.  Generation of this is still buggy so it is currently
479		disabled.
480
481		----------------------------------------------------------------------
482		Name: Lal George
483		Date: 2002/03/22 14:18:25 EST
484		Tag: george-20020322-cps-branch-prob
485		Description:
486
487		Implemented the Ball-Larus branch prediction-heuristics, and
488		incorporated graphical viewers for control flow graphs.
489
490		Ball-Larus Heuristics:
491		---------------------
492		See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
493
494		By design it uses the Dempster-Shafer theory for combining
495		probabilities.  For example, in the function:
496
497		fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
498
499		the ball-larus heuristics predicts that the n=0 is unlikely
500		(OH-heuristic), and the 'then' branch is unlikely because of the
501		RH-heuristic -- giving the 'then' branch an even lower combined
502		probability using the Dempster-Shafer theory.
503
504		Finally, John Reppy's loop analysis in MLRISC, further lowers the
505		probability of the 'then' branch because of the loop in the else
506		branch.
507
508
509		Graphical Viewing:
510		------------------
511		I merely plugged in Allen's graphical viewers into the compiler. The
512		additional code is not much. At the top level, saying:
513
514		Control.MLRISC.getFlag "cfg-graphical-view" := true;
515
516		will display the graphical view of the control flow graph just before
517		back-patching.  daVinci must be in your path for this to work. If
518		daVinci is not available, then the default viewer can be changed
519		using:
520
521		Control.MLRISC.getString "viewer"
522
523		which can be set to "dot" or "vcg" for the corresponding viewers. Of
524		course, these viewers must be in your path.
525
526		The above will display the compilation unit at the level of clusters,
527		many of which are small, boring, and un-interesting. Also setting:
528
529		Control.MLRISC.getInt "cfg-graphical-view_size"
530
531		will display clusters that are larger than the value set by the above.
532
533
534		----------------------------------------------------------------------
535		Name: Matthias Blume
536		Date: 2002/03/21 22:20:00 EST
537		Tag: blume-20020321-kmp-bugfix
538		Description:
539
540		Changed the interface to the KMP routine in PreString and fixed
541		a minor bug in one place where it was used.
542
543		----------------------------------------------------------------------
544		Name: Allen Leung
545		Date: 2002/03/21 20:30:00 EST
546		Tag: leunga-20020321-cfg
547		Description:
548
549		Fixed a potential problem in cfg edge splitting.
550
551		----------------------------------------------------------------------
552		Name: Allen Leung
553		Date: 2002/03/21 17:15:00 EST
554		Tag: leunga-20020321-x86-fp-cfg
555		Description:
556
557		1. Recoded the buggy parts of x86-fp.
558
559		a. All the block reordering code has been removed.
560		We now depend on the block placement phases to do this work.
561
562		b. Critical edge splitting code has been simplified and moved into the
563		CFG modules, as where they belong.
564
565		Both of these were quite buggy and complex.  The code is now much, much
566		simpler.
567
568		2. X86 backend.
569
570		a. Added instructions for 64-bit support.  Instruction selection for
571		64-bit has not been committed, however, since that
572		requires changes to MLTREE which haven't been approved by
573		Lal and John.
574
575		b. Added support for FUCOMI and FUCOMIP when generating code for
576		PentiumPro and above.  We only generate these instructions in
577		the fast-fp mode.
578
579		c. Added cases for JP and JNP in X86FreqProps.
580
581		3. CFG
582
583		CFG now has a bunch of methods for edge splitting and merging.
584
585		4. Machine description.
586
587		John's simplification of MLTREE_BASIS.fcond broke a few machine
588		description things:
589
590		rtl-build.{sig,sml} and hppa.mdl fixed.
591
592		NOTE: the machine description stuff in the repository is still broken.
593		Again, I can't put my fixes in because that involves
594		changes to MLTREE.
595
596		----------------------------------------------------------------------
597		Name: Matthias Blume
598		Date: 2002/03/20 15:55:00 EST
599		Tag: blume-20020320-kmp
600		Description:
601
602		Implemented Knuth-Morris-Pratt string matching in PreString and used
603		it for String.isSubstring, Substring.isSubstring, and
604		Substring.position.
605
606		(Might need some stress-testing.  Simple examples worked fine.)
607
608		----------------------------------------------------------------------
609		Name: Matthias Blume
610		Date: 2002/03/19 16:37:00 EST
611		Tag: blume-20020319-witnesses
612		Description:
613
614		Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
615
616		This implements a generic mechanism for changing constness qualifiers
617		anywhere within big C types without resorting to outright "casts".
618		(So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
619		modify the constness at the outermost level.)
620		The implementation of "convert" is based on the idea of "witness"
621		values -- values that are not used by the operation but whose types
622		"testify" to their applicability.  On the implementation side, "convert"
623		is simply a projection (returning its second curried argument).  With
624		cross-module inlining, it should not result in any machine code being
625		generated.
626
627		----------------------------------------------------------------------
628		Name: Matthias Blume
629		Date: 2002/03/15 16:40:00 EST
630		Tag: blume-20020315-basis
631		Description:
632
633		Provided (preliminary?) implementations for
634
635		{String,Substring}.{concatWith,isSuffix,isSubstring}
636
637		and
638
639		Substring.full
640
641		Those are in the Basis spec but they were missing in SML/NJ.
642
643		----------------------------------------------------------------------
644		Name: Matthias Blume
645		Date: 2002/03/14 21:30:00 EST
646		Tag: blume-20020314-controls
647		Description:
648
649		Controls:
650		---------
651
652		1. Factored out the recently-added Controls : CONTROLS stuff and put
653		it into its own library $/controls-lib.cm.  The source tree for
654		this is under src/smlnj-lib/Controls.
655
656		2. Changed the names of types and functions in this interface, so they
657		make a bit more "sense":
658
659		module -> registry
660		'a registry -> 'a group
661
662		3. The interface now deals in ref cells only.  The getter/setter interface
663		is (mostly) gone.
664
665		4. Added a function that lets one register an already-existing ref cell.
666
667		5. Made the corresponding modifications to the rest of the code so that
668		everything compiles again.
669
670		6. Changed the implementation of Controls.MLRISC back to something closer
671		to the original.  In particular, this module (and therefore MLRISC)
672		does not depend on Controls.  There now is some link-time code in
673		int-sys.sml that registers the MLRISC controls with the Controls
674		module.
675
676		CM:
677		---
678
679		* One can now specify the lambda-split aggressiveness in init.cmi.
680
681		----------------------------------------------------------------------
682		Name: Allen Leung
683		Date: 2002/03/13 17:30:00 EST
684		Tag: leunga-20020313-x86-fp-unary
685		Description:
686
687		Bug fix for:
688
689		> leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
690		> Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
691		> - fun f(x,(y,z)) = Real.~ y;
692		> [autoloading]
693		> [autoloading done]
694		>       fchsl   (%eax), 184(%esp)
695		> Error: MLRisc bug: X86MCEmitter.emitInstr
696		>
697		> uncaught exception Error
698		>   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
699
700		The problem was that the code generator did not generate any fp registers
701		in this case, and the ra didn't know that it needed to run the X86FP phase to
702		translate the pseudo fp instruction.   This only happened with unary fp
703		operators in certain situations.
704
705		----------------------------------------------------------------------
706		Name: Matthias Blume
707		Date: 2002/03/13 14:00:00 EST
708		Tag: blume-20020313-overload-etc
709		Description:
710
711		1. Added _overload as a synonym for overload for backward compatibility.
712		(Control.overloadKW must be true for either version to be accepted.)
713
714		2. Fixed bug in install script that caused more things to be installed
715		than what was requested in config/targets.
716
717		3. Made CM aware of the (_)overload construct so that autoloading
718		works.
719
720		----------------------------------------------------------------------
721		Name: Matthias Blume
722		Date: 2002/03/12 22:03:00 EST
723		Tag: blume-20020312-url
724		Description:
725
726		Forgot to update BOOT and srcarchiveurl.
727
728		----------------------------------------------------------------------
729		Name: Matthias Blume
730		Date: 2002/03/12 17:30:00 EST
731		Tag: blume-20020312-version110392
732		Description:
733
734		Yet another version number bump (because of small changes to the
735		binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
736
737		Changes:
738
739		The new pid generation scheme described a few weeks ago was overly
740		complicated.  I implemented a new mechanism that is simpler and
741		provides a bit more "stability":  Once CM has seen a compilation
742		unit, it keeps its identity constant (as long as you do not delete
743		those crucial CM/GUID/* files).  This means that when you change
744		an interface, compile, then go back to the old interface, and
745		compile again, you arrive at the original pid.
746
747		There now also is a mechanism that instructs CM to use the plain
748		environment hash as a module's pid (effectively making its GUID
749		the empty string).  For this, "noguid" must be specified as an
750		option to the .sml file in question within its .cm file.
751		This is most useful for code that is being generated by tools such
752		as ml-nlffigen (because during development programmers tend to
753		erase the tool's entire output directory tree including CM's cached
754		GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
755		revert to the old, broken behavior of SML/NJ, but in specific cases
756		where there is no danger of interface confusion, its use is ok
757		(I think).
758
759		ml-nlffigen by default generates "noguid" annotations.  They can be
760		turned off by specifying -guid in its command line.
761
762		----------------------------------------------------------------------
763		Name: Lal George
764		Date: 2002/03/12 12 14:42:36 EST
765		Tag: george-20020312-frequency-computation
766		Description:
767
768		Integrated jump chaining and static block frequency into the
769		compiler. More details and numbers later.
770
771		----------------------------------------------------------------------
772		Name: Lal George
773		Date: 2002/03/11 11 22:38:53 EST
774		Tag: george-20020311-jump-chain-elim
775		Description:
776
777		Tested the jump chain elimination on all architectures (except the
778		hppa).  This is on by default right now and is profitable for the
779		alpha and x86, however, it may not be profitable for the sparc and ppc
780		when compiling the compiler.
781
782		The gc test will typically jump to a label at the end of the cluster,
783		where there is another jump to an external cluster containing the actual
784		code to invoke gc. This is to allow factoring of common gc invocation
785		sequences. That is to say, we generate:
786
787		f:
788		testgc
789		ja   L1      % jump if above to L1
790
791		L1:
792		jmp L2
793
794
795		After jump chain elimination the 'ja L1' instructions is converted to
796		'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
797		up being implemented in their long form (if L2 is far away) using:
798
799		jbe     L3      % jump if below or equal to L3
800		jmp     L2
801		L3:
802		...
803
804
805		For large compilation units L2  may be far away.
806
807
808		----------------------------------------------------------------------
809		Name: Matthias Blume
810		Date: 2002/03/11 13:30:00 EST
811		Tag: blume-20020311-mltreeeval
812		Description:
813
814		A functor parameter was missing.
815
816		----------------------------------------------------------------------
817		Name: Allen Leung
818		Date: 2002/03/11 10:30:00 EST
819		Tag: leunga-20020311-runtime-string0
820		Description:
821
822		The representation of the empty string now points to a
823		legal null terminated C string instead of unit.  It is now possible
824		to convert an ML string into C string with InlineT.CharVector.getData.
825		This compiles into one single machine instruction.
826
827		----------------------------------------------------------------------
828		Name: Allen Leung
829		Date: 2002/03/10 23:55:00 EST
830		Tag: leunga-20020310-x86-call
831		Description:
832
833		Added machine generation for CALL instruction (relative displacement mode)
834
835		----------------------------------------------------------------------
836		Name: Matthias Blume
837		Date: 2002/03/08 16:05:00
838		Tag: blume-20020308-entrypoints
839		Description:
840
841		Version number bumped to 110.39.1.  NEW BOOTFILES!
842
843		Entrypoints: non-zero offset into a code object where execution should begin.
844
845		- Added the notion of an entrypoint to CodeObj.
846		- Added reading/writing of entrypoint info to Binfile.
847		- Made runtime system bootloader aware of entrypoints.
848		- Use the address of the label of the first function given to mlriscGen
849		as the entrypoint.  This address is currently always 0, but it will
850		not be 0 once we turn on block placement.
851		- Removed the linkage cluster code (which was The Other Way(tm) of dealing
852		with entry points) from mlriscGen.
853
854		----------------------------------------------------------------------
855		Name: Allen Leung
856		Date: 2002/03/07 20:45:00 EST
857		Tag: leunga-20020307-x86-cmov
858		Description:
859
860		Bug fixes for CMOVcc on x86.
861
862		1. Added machine code generation for CMOVcc
863		2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
864		3. CMOVcc cannot have an immediate operand as argument.
865
866		----------------------------------------------------------------------
867		Name: Matthias Blume
868		Date: 2002/03/07 16:15:00 EST
869		Tag: blume-20020307-controls
870		Description:
871
872		This is a very large but mostly boring patch which makes (almost)
873		every tuneable compiler knob (i.e., pretty much everything under
874		Control.* plus a few other things) configurable via both the command
875		line and environment variables in the style CM did its configuration
876		until now.
877
878		Try starting sml with '-h' (or, if you are brave, '-H')
879
880		To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
881		implements the underlying generic mechanism.
882
883		The interface to some of the existing such facilities has changed somewhat.
884		For example, the MLRiscControl module now provides mkFoo instead of getFoo.
885		(The getFoo interface is still there for backward-compatibility, but its
886		use is deprecated.)
887
888		The ml-build script passes -Cxxx=yyy command-line arguments through so
889		that one can now twiddle the compiler settings when using this "batch"
890		compiler.
891
892		TODO items:
893
894		We should go through and throw out all controls that are no longer
895		connected to anything.  Moreover, we should go through and provide
896		meaningful (and correct!) documentation strings for those controls
897		that still are connected.
898
899		Currently, multiple calls to Controls.new are accepted (only the first
900		has any effect).  Eventually we should make sure that every control
901		is being made (via Controls.new) exactly once.  Future access can then
902		be done using Controls.acc.
903
904		Finally, it would probably be a good idea to use the getter-setter
905		interface to controls rather than ref cells.  For the time being, both
906		styles are provided by the Controls module, but getter-setter pairs are
907		better if thread-safety is of any concern because they can be wrapped.
908
909		*****************************************
910
911		One bug fix: The function blockPlacement in three of the MLRISC
912		backpatch files used to be hard-wired to one of two possibilities at
913		link time (according to the value of the placementFlag).  But (I
914		think) it should rather sense the flag every time.
915
916		*****************************************
917
918		Other assorted changes (by other people who did not supply a HISTORY entry):
919
920		1. the cross-module inliner now works much better (Monnier)
921		2. representation of weights, frequencies, and probabilities in MLRISC
922		changed in preparation of using those for weighted block placement
923		(Reppy, George)
924
925		----------------------------------------------------------------------
926		Name: Lal George
927		Date: 2002/03/07 14:44:24 EST 2002
928		Tag: george-20020307-weighted-block-placement
929
930		Tested the weighted block placement optimization on all architectures
931		(except the hppa) using AMPL to generate the block and edge frequencies.
932		Changes were required in the machine properties to correctly
933		categorize trap instructions. There is an MLRISC flag
934		"weighted-block-placement" that can be used to enable weighted block
935		placement, but this will be ineffective without block/edge
936		frequencies (coming soon).
937
938
939		----------------------------------------------------------------------
940		Name: Lal George
941		Date: 2002/03/05 17:24:48 EST
942		Tag: george-20020305-linkage-cluster
943
944		In order to support the block placement optimization, a new cluster
945		is generated as the very first cluster (called the linkage cluster).
946		It contains a single jump to the 'real' entry point for the compilation
947		unit. Block placement has no effect on the linkage cluster itself, but
948		all the other clusters  have full freedom in the manner in which they
949		reorder blocks or functions.
950
951		On the x86 the typical linkage code that is generated is:
952		----------------------
953		.align 2
954		L0:
955		addl    $L1-L0, 72(%esp)
956		jmp     L1
957
958
959		.align  2
960		L1:
961		----------------------
962
963		72(%esp) is the memory location for the stdlink register. This
964		must contain the address of the CPS function being called. In the
965		above example, it contains the address of  L0; before
966		calling L1 (the real entry point for the compilation unit), it
967		must contain the address for L1, and hence
968
969		addl $L1-L0, 72(%esp)
970
971		I have tested this on all architectures except the hppa.The increase
972		in code size is of course negligible
973
974		----------------------------------------------------------------------
975		Name: Allen Leung
976		Date: 2002/03/03 13:20:00 EST
977		Tag: leunga-20020303-mlrisc-tools
978
979		Added #[ ... ] expressions to mlrisc tools
980
981		----------------------------------------------------------------------
982		Name: Matthias Blume
983		Date: 2002/02/27 12:29:00 EST
984		Tag: blume-20020227-cdebug
985		Description:
986
987		- made types in structure C and C_Debug to be equal
988		- got rid of code duplication (c-int.sml vs. c-int-debug.sml)
989		- there no longer is a C_Int_Debug (C_Debug is directly derived from C)
990
991		----------------------------------------------------------------------
992		Name: Matthias Blume
993		Date: 2002/02/26 12:00:00 EST
994		Tag: blume-20020226-ffi
995		Description:
996
997		1. Fixed a minor bug in CM's "noweb" tool:
998		If numbering is turned off, then truly don't number (i.e., do not
999		supply the -L option to noweb).  The previous behavior was to supply
1000		-L'' -- which caused noweb to use the "default" line numbering scheme.
1001		Thanks to Chris Richards for pointing this out (and supplying the fix).
1002
1003		2. Once again, I reworked some aspects of the FFI:
1004
1005		A. The incomplete/complete type business:
1006
1007		- Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1008		gone!
1009		- ML types representing an incomplete type are now *equal* to
1010		ML types representing their corresponding complete types (just like
1011		in C).  This is still safe because ml-nlffigen will not generate
1012		RTTI for incomplete types, nor will it generate functions that
1013		require access to such RTTI.   But when ML code generated from both
1014		incomplete and complete versions of the C type meet, the ML types
1015		are trivially interoperable.
1016
1017		NOTE:  These changes restore the full generality of the translation
1018		(which was previously lost when I eliminated functorization)!
1019
1020		B. Enum types:
1021
1022		- Structure C now has a type constructor "enum" that is similar to
1023		how the "su" constructor works.  However, "enum" is not a phantom
1024		type because each "T enum" has values (and is isomorphic to
1025		MLRep.Signed.int).
1026		- There are generic access operations for enum objects (using
1027		MLRep.Signed.int).
1028		- ml-nlffigen will generate a structure E_foo for each "enum foo".
1029		* The structure contains the definition of type "mlrep" (the ML-side
1030		representation type of the enum).  Normally, mlrep is the same
1031		as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1032		then mlrep will be defined as a datatype -- thus facilitating
1033		pattern matching on mlrep values.
1034		("-ec" will be suppressed if there are duplicate values in an
1035		enumeration.)
1036		* Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1037		will be generated for each C enum constant xxx.
1038		* Conversion functions m2i and i2m convert between mlrep and
1039		MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1040		* Coversion functions c and ml convert between mlrep and "tag enum".
1041		* Access functions (get/set) fetch and store mlrep values.
1042		- By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1043		enumerations are merged into one single enumeration represented by
1044		structure E_'.
1045
1046		----------------------------------------------------------------------
1047		Name: Allen Leung
1048		Date: 2002/02/25 04:45:00 EST
1049		Tag: leunga-20020225-cps-spill
1050
1051		This is a new implementation of the CPS spill phase.
1052		The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1053		In case of problems, replace it with the old file spill.sml
1054
1055		The current compiler runs into some serious performance problems when
1056		constructing a large record.  This can happen when we try to compile a
1057		structure with many items.  Even a very simple structure like the following
1058		makes the compiler slow down.
1059
1060		structure Foo = struct
1061		val x_1 = 0w1 : Word32.int
1062		val x_2 = 0w2 : Word32.int
1063		val x_3 = 0w3 : Word32.int
1064		...
1065		val x_N = 0wN : Word32.int
1066		end
1067
1068		The following table shows the compile time, from N=1000 to N=4000,
1069		with the old compiler:
1070
1071		N
1072		1000   CPS 100 spill                           0.04u  0.00s  0.00g
1073		MLRISC ra                               0.06u  0.00s  0.05g
1074		(spills = 0 reloads = 0)
1075		TOTAL                                   0.63u  0.07s  0.21g
1076
1077		1100   CPS 100 spill                           8.25u  0.32s  0.64g
1078		MLRISC ra                               5.68u  0.59s  3.93g
1079		(spills = 0 reloads = 0)
1080		TOTAL                                   14.71u  0.99s  4.81g
1081
1082		1500   CPS 100 spill                           58.55u  2.34s  1.74g
1083		MLRISC ra                               5.54u  0.65s  3.91g
1084		(spills = 543 reloads = 1082)
1085		TOTAL                                   65.40u  3.13s  6.00g
1086
1087		2000   CPS 100 spill                           126.69u  4.84s  3.08g
1088		MLRISC ra                               0.80u  0.10s  0.55g
1089		(spills = 42 reloads = 84)
1090		TOTAL                                   129.42u  5.10s  4.13g
1091
1092		3000   CPS 100 spill                           675.59u  19.03s  11.64g
1093		MLRISC ra                               2.69u  0.27s  1.38g
1094		(spills = 62 reloads = 124)
1095		TOTAL                                   682.48u  19.61s  13.99g
1096
1097		4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1098		MLRISC ra                               4.96u  0.27s  2.72g
1099		(spills = 85 reloads = 170)
1100		TOTAL                                   2375.26u  57.21s  48.00g
1101
1102		As you can see the old cps spill module suffers from some serious
1103		performance problem.  But since I cannot decipher the old code fully,
1104		instead of patching the problems up, I'm reimplementing it
1105		with a different algorithm.  The new code is more modular,
1106		smaller when compiled, and substantially faster
1107		(O(n log n) time and O(n) space).  Timing of the new spill module:
1108
1109		4000  CPS 100 spill                           0.02u  0.00s  0.00g
1110		MLRISC ra                               0.25u  0.02s  0.15g
1111		(spills=1 reloads=3)
1112		TOTAL                                   7.74u  0.34s  1.62g
1113
1114		Implementation details:
1115
1116		As far as I can tell, the purpose of the CPS spill module is to make sure the
1117		number of live variables at any program point (the bandwidth)
1118		does not exceed a certain limit, which is determined by the
1119		size of the spill area.
1120
1121		When the bandwidth is too large, we decrease the register pressure by
1122		packing live variables into spill records.  How we achieve this is
1123		completely different than what we did in the old code.
1124
1125		First, there is something about the MLRiscGen code generator
1126		that we should be aware of:
1127
1128		o MLRiscGen performs code motion!
1129
1130		In particular, it will move floating point computations and
1131		address computations involving only the heap pointer to
1132		their use sites (if there is only a single use).
1133		What this means is that if we have a CPS record construction
1134		statement
1135
1136		RECORD(k,vl,w,e)
1137
1138		we should never count the new record address w as live if w
1139		has only one use (which is often the case).
1140
1141		We should do something similar to floating point, but the transformation
1142		there is much more complex, so I won't deal with that.
1143
1144		Secondly, there are now two new cps primops at our disposal:
1145
1146		1. rawrecord of record_kind option
1147		This pure operator allocates some uninitialized storage from the heap.
1148		There are two forms:
1149
1150		rawrecord NONE [INT n]  allocates a tagless record of length n
1151		rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1152		and initializes the tag.
1153
1154		2. rawupdate of cty
1155		rawupdate cty (v,i,x)
1156		Assigns to x to the ith component of record v.
1157		The storelist is not updated.
1158
1159		We use these new primops for both spilling and increment record construction.
1160
1161		1. Spilling.
1162
1163		This is implemented with a linear scan algorithm (but generalized
1164		to trees).  The algorithm will create a single spill record at the
1165		beginning of the cps function and use rawupdate to spill to it,
1166		and SELECT or SELp to reload from it.  So both spills and reloads
1167		are fine-grain operations.  In contrast, in the old algorithm
1168		"spills" have to be bundled together in records.
1169
1170		Ideally, we should sink the spill record construction to where
1171		it is needed.  We can even split the spill record into multiple ones
1172		at the places where they are needed.  But CPS is not a good
1173		representation for global code motion, so I'll keep it simple and
1174		am not attempting this.
1175
1176		2. Incremental record construction (aka record splitting).
1177
1178		Long records with many component values which are simulatenously live
1179		(recall that single use record addresses are not considered to
1180		be live) are constructed with rawrecord and rawupdate.
1181		We allocate space on the heap with rawrecord first, then gradually
1182		fill it in with rawupdate.  This is the technique suggested to me
1183		by Matthias.
1184
1185		Some restrictions on when this is applicable:
1186		1. It is not a VECTOR record.  The code generator currently does not handle
1187		this case. VECTOR record uses double indirection like arrays.
1188		2. All the record component values are defined in the same "basic block"
1189		as the record constructor.  This is to prevent speculative
1190		record construction.
1191
1192		----------------------------------------------------------------------
1193		Name: Allen Leung
1194		Date: 2002/02/22 01:02:00 EST
1195		Tag: leunga-20020222-mlrisc-tools
1196
1197		Minor bug fixes in the parser and rewriter
1198
1199		----------------------------------------------------------------------
1200		Name: Allen Leung
1201		Date: 2002/02/21 20:20:00 EST
1202		Tag: leunga-20020221-peephole
1203
1204		Regenerated the peephole files.  Some contained typos in the specification
1205		and some didn't compile because of pretty printing bugs in the old version
1206		of 'nowhere'.
1207
1208		----------------------------------------------------------------------
1209		Name: Allen Leung
1210		Date: 2002/02/19 20:20:00 EST
1211		Tag: leunga-20020219-mlrisc-tools
1212		Description:
1213
1214		Minor bug fixes to the mlrisc-tools library:
1215
1216		1.  Fixed up parsing colon suffixed keywords
1217		2.  Added the ability to shut the error messages up
1218		3.  Reimplemented the pretty printer and fixed up/improved
1219		the pretty printing of handle and -> types.
1220		4.  Fixed up generation of literal symbols in the nowhere tool.
1221		5.  Added some SML keywords to to sml.sty
1222
1223		----------------------------------------------------------------------
1224		Name: Matthias Blume
1225		Date: 2002/02/19 16:20:00 EST
1226		Tag: blume-20020219-cmffi
1227		Description:
1228
1229		A wild mix of changes, some minor, some major:
1230
1231		* All C FFI-related libraries are now anchored under $c:
1232		$/c.cm      --> $c/c.cm
1233		$/c-int.cm  --> $c/internals/c-int.cm
1234		$/memory.cm --> $c/memory/memory.cm
1235
1236		* "make" tool (in CM) now treats its argument pathname slightly
1237		differently:
1238		1. If the native expansion is an absolute name, then before invoking
1239		the "make" command on it, CM will apply OS.Path.mkRelative
1240		(with relativeTo = OS.FileSys.getDir()) to it.
1241		2. The argument will be passed through to subsequent phases of CM
1242		processing without "going native".  In particular, if the argument
1243		was an anchored path, then "make" will not lose track of that anchor.
1244
1245		* Compiler backends now "know" their respective C calling conventions
1246		instead of having to be told about it by ml-nlffigen.  This relieves
1247		ml-nlffigen from one of its burdens.
1248
1249		* The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1250
1251		* Export C_DEBUG and C_Debug from $c/c.cm.
1252
1253		* C type encoding in ml-nlffi-lib has been improved to model the conceptual
1254		subtyping relationship between incomplete pointers and their complete
1255		counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1256		with the convention of instantiating 'o with ('t, 'c) obj whenever
1257		the pointer target type is complete.  In the incomplete case, 'o
1258		will be instantiated with some "'c iobj" -- a type obtained by
1259		using one of the functors PointerToIncompleteType or PointerToCompleteType.
1260
1261		Operations that work on both incomplete and complete pointer types are
1262		typed as taking an 'o ptr while operations that require the target to
1263		be known are typed as taking some ('t, 'c) obj ptr.
1264
1265		voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1266		where void is an eqtype without any values.  This makes it possible
1267		to work on voidptr values using functions meant to operate on light
1268		incomplete pointers.
1269
1270		* As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1271		been vastly simplified.
1272
1273		----------------------------------------------------------------------
1274		Name: Matthias Blume
1275		Date: 2002/02/19 10:48:00 EST
1276		Tag: blume-20020219-pqfix
1277		Description:
1278
1279		Applied Chris Okasaki's bug fix for priority queues.
1280
1281		----------------------------------------------------------------------
1282		Name: Matthias Blume
1283		Date: 2002/02/15 17:05:00
1284		Tag: Release_110_39
1285		Description:
1286
1287		Last-minute retagging is becoming a tradition... :-(
1288
1289		This is the working release 110.39.
1290
1291		----------------------------------------------------------------------
1292		Name: Matthias Blume
1293		Date: 2002/02/15 16:00:00 EST
1294		Tag: Release_110_39-orig
1295		Description:
1296
1297		Working release 110.39.  New bootfiles.
1298
1299		(Update: There was a small bug in the installer so it wouldn't work
1300		with all shells.  So I retagged. -Matthias)
1301
1302		----------------------------------------------------------------------
1303		Name: Matthias Blume
1304		Date: 2002/02/15 14:17:00 EST
1305		Tag: blume-20020215-showbindings
1306		Description:
1307
1308		Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1309		the latter can be useful for exploring what bindings are available at
1310		the interactive prompt.  (The first function returns only the list
1311		of symbols that are really bound, the second prints those but also the
1312		ones that CM's autoloading mechanism knows about.)
1313
1314		----------------------------------------------------------------------
1315		Name: Matthias Blume
1316		Date: 2002/02/15 12:08:00 EST
1317		Tag: blume-20020215-iptrs
1318		Description:
1319
1320		Two improvements to ml-nlffigen:
1321
1322		1. Write files only if they do not exist or if their current contents
1323		do not coincide with what's being written.  (That is, avoid messing
1324		with the time stamps unless absolutely necessary.)
1325
1326		2. Implement a "repository" mechanism for generated files related
1327		to "incomplete pointer types".   See the README file for details.
1328
1329		----------------------------------------------------------------------
1330		Name: Matthias Blume
1331		Date: 2002/02/14 11:50:00 EST
1332		Tag: blume-20020214-quote
1333		Description:
1334
1335		Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1336		because of the new and improved tag generation scheme.  (Thanks to Allen
1337		Leung for pointing it out.)
1338
1339		----------------------------------------------------------------------
1340		Name: Lal George
1341		Date: 2002/02/14 09:55:27 EST 2002
1342		Tag: george-20020214-isabelle-bug
1343		Description:
1344
1345		Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1346		of Isabelle on the x86.
1347
1348		From Allen:
1349		-----------
1350		I've found the problem:
1351
1352		in ra-core.sml, I use the counter "blocked" to keep track of the
1353		true number of elements in the freeze queue.  When the counter goes
1354		to zero, I skip examining the queue.  But I've messed up the
1355		bookkeeping in combine():
1356
1357		else ();
1358		case !ucol of
1359		PSEUDO => (if !cntv > 0 then
1360		(if !cntu > 0 then blocked := !blocked - 1 else ();
1361		^^^^^^^^^^^^^^^^^^^^^^^
1362		moveu := mergeMoveList(!movev, !moveu)
1363		)
1364		else ();
1365
1366		combine() is called to coalesce two nodes u and v.
1367		I think I was thinking that if the move counts of u and v are both
1368		greater than zero then after they are coalesced then one node is
1369		removed from the freeze queue.  Apparently I was thinking that
1370		both u and v are of low degree, but that's clearly not necessarily true.
1371
1372
1373		02/12/2002:
1374		Here's the patch.  HOL now compiles.
1375
1376		I don't know how this impact on performance (compile
1377		time or runtime).  This bug caused the RA (especially on the x86)
1378		to go thru the potential spill phase when there are still nodes on the
1379		freeze queue.
1380
1381
1382
1383
1384		----------------------------------------------------------------------
1385		Name: Matthias Blume
1386		Date: 2002/02/13 22:40:00 EST
1387		Tag: blume-20020213-fptr-rtti
1388		Description:
1389
1390		Fixed a bug in ml-nlffigen that was introduced with one of the previous
1391		updates.
1392
1393		----------------------------------------------------------------------
1394		Name: Matthias Blume
1395		Date: 2002/02/13 16:41:00 EST
1396		Tag: blume-20020213-cmlpq
1397		Description:
1398
1399		Added new priority queue export symbols (which have just been added to
1400		smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1401		not compile and the installer would choke.)
1402
1403		----------------------------------------------------------------------
1404		Name: Matthias Blume
1405		Date: 2002/02/13 16:15:00 EST
1406		Tag: blume-20020213-various
1407		Description:
1408
1409		1. More tweaks to ml-nlffigen:
1410
1411		- better internal datastructures (resulting in slight speedup)
1412		- "-match" option requires exact match
1413		- "localized" gensym counters (untagged structs/unions nested within
1414		other structs/unions or within typedefs get a fresh counter; their
1415		tag will be prefixed by a concatenation of their parents' tags)
1416		- bug fixes (related to calculation of transitive closure of types
1417		to be included in the output)
1418
1419		2. Minor Basis updates:
1420
1421		- added implementations for List.collate and Option.app
1422
1423		----------------------------------------------------------------------
1424		Name: Matthias Blume
1425		Date: 2002/02/11 15:55:00 EST
1426		Tag: blume-20020211-gensym
1427		Description:
1428
1429		Added a "-gensym" option to command line of ml-nlffigen.  This can be
1430		used to specify a "stem" -- a string that is inserted in all "gensym'd"
1431		names (ML structure names that correspond to unnamed C structs, unions,
1432		and enums), so that separate runs of ml-nlffigen do not clash.
1433
1434		----------------------------------------------------------------------
1435		Name: Matthias Blume
1436		Date: 2002/02/11 12:05:00 EST
1437		Tag: blume-20020211-gensml
1438		Description:
1439
1440		A quick fix for a problem with GenSML (in the pgraph-util library):
1441		Make generation of toplevel "local" optional.  (Strictly speaking,
1442		signature definitions within "local" are not legal SML.)
1443
1444		Other than that: updates to INSTALL and cm/TODO.
1445
1446		----------------------------------------------------------------------
1447		Name: Matthias Blume
1448		Date: 2002/02/08 15:00:00 EST
1449		Tag: blume-20020208-uniquepid
1450		Description:
1451
1452		0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1453
1454		1. The installer (config/install.sh) has gotten smarter:
1455
1456		- Configuration options are a bit easier to specify now
1457		(in config/targets).
1458		- Bug in recognizing .tar.bz2 files fixed.
1459		- Installer automatically resolves dependencies between
1460		configuration options (e.g., if you ask for eXene, you will
1461		also get cml -- regardless whether you asked for it or not).
1462		- Installer can run in "quieter mode" by setting the environment
1463		variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1464		"completely silent", though.
1465		- Build HashCons library as part of smlnj-lib.
1466
1467		2. A new scheme for assigning persistent identifiers to compilation
1468		units (and, by extension, to types etc.) has been put into place.
1469		This fixes a long-standing bug where types and even dynamic values
1470		can get internally confused, thereby compromising type safety
1471		(abstraction) and dynamic correctness.  See
1472
1473		http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1474
1475		for an example of how things could go wrong until now.
1476
1477		The downside of the new scheme is that pids are not quite as
1478		persistent as they used to be: CM will generate a fresh pid
1479		for every compilation unit that it thinks it sees for the first
1480		time.  That means that if you compile starting from a clean, fresh
1481		source tree at two different times, you end up with different
1482		binaries.
1483
1484		Cutoff recompilation, however, has not been compromised because
1485		CM keeps pid information in special caches between runs.
1486
1487		----------------------------------------------------------------------
1488		Name: Lal George
1489		Date: 2002/02/07 15:34:13 EST 2002
1490		Tag: <none>
1491		Description:
1492
1493		Compilers that generate assembly code may produce  global labels
1494		whose value is resolved at link time. The various peephole optimization
1495		modules did not take this in account.
1496
1497		TODO. The Labels.addrOf function should really return an option
1498		type so that clients are forced to deal with this issue, rather
1499		than an exception being raised.
1500
1501		----------------------------------------------------------------------
1502		Name: Lal George
1503		Date: 2002/02/06 13:55:02 EST
1504		Tag: george-20020206-ra-breakup
1505		Description:
1506
1507		1. A bug fix from Allen.
1508
1509		A typo causes extra fstp %st(0)'s to be generated at compensation
1510		edges, which might cause stack underflow traps at runtime.  This
1511		occurs in fft where there are extraneous fstps right before the 'into'
1512		trap instruction (in this case they are harmless since none of the
1513		integers overflow.)
1514
1515		2. Pulled out various utility modules that were embedded in the modules
1516		of the register allocator. I need these modules for other purposes, but
1517		they are not complete enough to put into a library (just yet).
1518		----------------------------------------------------------------------
1519		Name: Matthias Blume
1520		Date: 2002/01/31 16:05:00 EST
1521		Tag: blume-20020131-sparc-ccalls
1522		Description:
1523
1524		1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1525		of extra stack space by mistake.  Fixed.
1526
1527		2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1528
1529		----------------------------------------------------------------------
1530		Name: Allen Leung
1531		Date: 2002/01/30
1532		Tag: leunga-20020130-nowhere-bug-fix
1533		Description:
1534
1535		MLRISC bug fixes:
1536		1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1537		2. MachineInt.fromString was negating its value.
1538
1539		----------------------------------------------------------------------
1540		Name: Matthias Blume
1541		Date: 2002/01/29
1542		Tag: blume-20020129-INSTALL
1543		Description:
1544
1545		- Added somewhat detailed installation instructions (file INSTALL).
1546		- Fixed curl-detection bug in config/install.sh.
1547		- It is now possible to select the URL getter using the URLGETTER
1548		environment variable:
1549
1550		not set / "unknown"      --> automatic detection (script tries wget,
1551		curl, and lynx)
1552		"wget" / "curl" / "lynx" --> use the specified program (script "knows"
1553		how to properly invoke them)
1554		other                    --> use $URLGETTER directly, it must take
1555		precisely two command-line arguments
1556		(source URL and destination file name)
1557
1558		----------------------------------------------------------------------
1559		Name: Matthias Blume
1560		Date: 2002/01/28
1561		Tag: blume-20020128-sparc-ccalls
1562		Description:
1563
1564		- Fixed problem with calculation of "used" registers in sparc-c-calls.
1565		- Make use of the allocParam argument in sparc-c-calls.
1566
1567		----------------------------------------------------------------------
1568		Name: Matthias Blume
1569		Date: 2002/01/28
1570		Tag: blume-20020128-allocParam
1571		Description:
1572
1573		John Reppy:  Changes c-calls API to accept client-callback for
1574		allocating extra stack space.
1575		me: Corresponding changes to mlriscGen (using a dummy argument that
1576		does not change the current behavior).
1577
1578		----------------------------------------------------------------------
1579		Name: Matthias Blume
1580		Date: 2002/01/28 12:00:00
1581		Tag: Release_110_38
1582		Description:
1583
1584		This time for real!!!
1585
1586		----------------------------------------------------------------------
1587		Name: Matthias Blume
1588		Date: 2002/01/28 10:56:00 EST
1589		Tag: blume-20020128-retraction
1590		Description:
1591
1592		0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1593		with blume-Release_110_38-retracted.)
1594
1595		1. Fixed a problem with incorrect rounding modes in real64.sml.
1596		(Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1597
1598		2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1599		and enums fixed.  The naming of corresponding ML identifiers should
1600		now be consistent again.
1601
1602		----------------------------------------------------------------------
1603		Name: Allen Leung
1604		Date: 2002/01/27
1605		Tag: leunga-20020127-nowhere
1606		Description:
1607
1608		Added a target called nowhere in the configuration scripts.
1609		Enabling this will build the MLRISC 'nowhere' tool (for translating
1610		programs with where-clauses into legal SML code) during installation.
1611
1612		----------------------------------------------------------------------
1613		Name: Matthias Blume
1614		Date: 2002/01/25 21:27:00 EST
1615		Tag: blume-Release_110_38-retracted
1616		Description:
1617
1618		Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1619
1620		README will be added later.
1621
1622		!!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1623		(Release_110_38) removed.  Reason: Last-minute bug fixes.
1624
1625		----------------------------------------------------------------------
1626		Name: Matthias Blume
1627		Date: 2002/01/25
1628		Tag: blume-20020125-ffi
1629		Description:
1630
1631		A large number of tweaks and improvements to ml-nlffi-lib and
1632		ml-nlffigen:
1633
1634		- ML represenation types have been streamlined
1635		- getter and setter functions work with concrete values, not abstract
1636		ones where possible
1637		- ml-nlffigen command line more flexible (see README file there)
1638		- some bugs have been fixed (hopefully)
1639
1640		----------------------------------------------------------------------
1641		Name: Lal George
1642		Date: 2002/01/24
1643		Tag: george-20020124-risc-ra-interface
1644		Description:
1645
1646		There is a dramatic simplification in the interface to the
1647		register allocator for RISC architectures as a result of making
1648		parallel copy instructions explicit.
1649
1650		----------------------------------------------------------------------
1651		Name: Matthias Blume
1652		Date: 2002/01/22
1653		Tag: blume-20020122-x86-ccalls
1654		Description:
1655
1656		Bug fix for c-calls on x86 (having to do with how char- and
1657		short-arguments are being handled).
1658
1659		----------------------------------------------------------------------
1660		Name: Matthias Blume
1661		Date: 2002/01/21
1662		Tag: blume-20020121-ff
1663		Description:
1664
1665		Another day of fiddling with the FFI...
1666
1667		1. Bug fix/workaround:  CKIT does not complain about negative array
1668		dimensions, so ml-nlffigen has to guard itself against this possibility.
1669		(Otherwise a negative dimension would send it into an infinite loop.)
1670
1671		2. Some of the abstract types (light objects, light pointers, most "base"
1672		types) in structure C are now eqtypes.
1673
1674		3. Added constructors and test functions for NULL function pointers.
1675
1676		----------------------------------------------------------------------
1677		Name: Matthias Blume
1678		Date: 2002/01/18
1679		Tag: blume-20020118-ready-for-new-release
1680		Description:
1681
1682		Made config/srcarchiveurl point to a new place.  (Will provide boot
1683		files shortly.)
1684
1685		Maybe we christen this to be 110.38?
1686
1687		----------------------------------------------------------------------
1688		Name: Matthias Blume
1689		Date: 2002/01/18
1690		Tag: blume-20020118-more-ffifiddle
1691		Description:
1692
1693		Today's FFI fiddling:
1694
1695		- Provided a structure CGetSet with "convenient" versions of C.Get.* and
1696		C.Set.* that use concrete (MLRep.*) arguments and results instead
1697		of abstract ones.
1698
1699		- Provided word-style bit operations etc. for "int" representation
1700		types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1701		and Long.
1702
1703		----------------------------------------------------------------------
1704		Name: Matthias Blume
1705		Date: 2002/01/18
1706		Tag: blume-20020118-use-x86-fp
1707		Description:
1708
1709		Now that x86-fast-fp seems to be working, I turned it back on again
1710		by default.  (Seems to work fine now, even with the FFI.)
1711
1712		Other than that, I added some documentation about the FFI to
1713		src/ml-nlffigen/README and updated the FFI test examples in
1714		src/ml-nlffi-lib/Tests/*.
1715
1716		----------------------------------------------------------------------
1717		Name: Allen Leung
1718		Date: 2002/01/17
1719		Tag: leunga-20020117-x86-fast-fp-call
1720		Description:
1721
1722		1. Fixed a problem with handling return fp values when x86's fast fp
1723		mode is turned on.
1724
1725		2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
1726		of %f32.
1727
1728		3. Added a constructor INT32lit to the ast of MLRISC tools.
1729
1730		----------------------------------------------------------------------
1731		Name: Matthias Blume
1732		Date: 2002/01/16
1733		Tag: blume-20020116-ffifiddle
1734		Description:
1735
1736		More fiddling with the FFI interface:
1737
1738		- Make constness 'c instead of rw wherever possible.  This eliminates
1739		the need for certain explicit coercions.  (However, due to ML's
1740		value polymorphism, there will still be many cases where explicit
1741		coercions are necessary.  Phantom types are not the whole answer
1742		to modeling a subtyping relationship in ML.)
1743
1744		- ro/rw coersions for pointers added.  (Avoids the detour through */&.)
1745
1746		- "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
1747		clumsy workaround for varargs problem.)
1748
1749		----------------------------------------------------------------------
1750		Name: Lal George
1751		Date: 2002/01/15
1752		Tag: <none>
1753		Description:
1754
1755		1. Since COPY instructions are no longer native to the architecture,
1756		a generic functor can be used to implement the expandCopies function.
1757
1758		2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1759		TEXT segment.
1760
1761		----------------------------------------------------------------------
1762		Name: Matthias Blume
1763		Date: 2002/01/15
1764		Tag: blume-20020115-ffiupdates
1765		Description:
1766
1767		1. Fix for bug resulting in single-precision float values being returned
1768		incorrectly from FFI calls.
1769
1770		2. Small modifications to C FFI API:
1771
1772		- memory-allocation routines return straight objects (no options)
1773		and raise an exception in out-of-memory situations
1774		- unsafe extensions to cast between function pointers and pointers
1775		from/to ints
1776		- added structure C_Debug as an alternative to structure C where
1777		pointer-dereferencing (|*| and |*!) always check for null-pointers
1778		- added open_lib' to DynLinkage;  open_lib' works like open_lib
1779		but also takes a (possibly empty) list of existing library handles
1780		that the current library depends on
1781
1782		----------------------------------------------------------------------
1783		Name: Matthias Blume
1784		Date: 2002/01/10
1785		Tag: blume-20020110-newffigen
1786		Description:
1787
1788		1. Updates to portable graph code.
1789
1790		2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
1791		more scalable now so that even huge interfaces such as the one
1792		for GTK compile in finite time and space. :-)
1793		See src/ml-nlffigen/README for details on what's new.
1794
1795		----------------------------------------------------------------------
1796		Name: Lal George
1797		Date: 2001/01/09 14:31:35 EST 2002
1798		Tag: george-20011206-rm-native-copy
1799		Description:
1800
1801		Removed the native COPY and FCOPY instructions
1802		from all the architectures and replaced it with the
1803		explicit COPY instruction from the previous commit.
1804
1805		It is now possible to simplify many of the optimizations
1806		modules that manipulate copies. This has not been
1807		done in this change.
1808
1809		----------------------------------------------------------------------
1810		Name: Lal George
1811		Date: 2001/12/06 16:50:13 EST 2001
1812		Tag: george-20011206-mlrisc-instruction
1813		Description:
1814
1815		Changed the representation of instructions from being fully abstract
1816		to being partially concrete. That is to say:
1817
1818		from
1819		type instruction
1820
1821		to
1822		type instr                              (* machine instruction *)
1823
1824		datatype instruction =
1825		LIVE of {regs: C.cellset, spilled: C.cellset}
1826		| KILL of {regs: C.cellset, spilled: C.cellset}
1827		| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
1828		| ANNOTATION of {i: instruction, a: Annotations.annotation}
1829		| INSTR of instr
1830
1831		This makes the handling of certain special instructions that appear on
1832		all architectures easier and uniform.
1833
1834		LIVE and KILL say that a list of registers are live or killed at the
1835		program point where they appear. No spill code is generated when an
1836		element of the 'regs' field is spilled, but the register is moved to
1837		the 'spilled' (which is present, more for debugging than anything else).
1838
1839		LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
1840		We used to generate:
1841
1842		DEFFREG f1
1843		f1 := f2 + f3
1844		trapb
1845
1846		but now generate:
1847
1848		f1 := f2 + f3
1849		trapb
1850		LIVE {regs=[f1,f2,f3], spilled=[]}
1851
1852		Furthermore, the DEFFREG (hack) required that all floating point instruction
1853		use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
1854		defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
1855		in a cleaner alpha implementation. (Hopefully, intel will not get rid of
1856		this architecture).
1857
1858		COPYXXX is intended to replace the parallel COPY and FCOPY  available on
1859		all the architectures. This will result in further simplification of the
1860		register allocator that must be aware of them for coalescing purposes, and
1861		will also simplify certain aspects of the machine description that provides
1862		callbacks related to parallel copies.
1863
1864		ANNOTATION should be obvious, and now INSTR represents the honest to God
1865		machine instruction set!
1866
1867		The <arch>/instructions/<arch>Instr.sml files define certain utility
1868		functions for making porting easier -- essentially converting upper case
1869		to lower case. All machine instructions (of type instr) are in upper case,
1870		and the lower case form generates an MLRISC instruction. For example on
1871		the alpha we have:
1872
1873		datatype instr =
1874		LDA of {r:cell, b:cell, d:operand}
1875		| ...
1876
1877		val lda : {r:cell, b:cell, d:operand} -> instruction
1878		...
1879
1880		where lda is just (INSTR o LDA), etc.
1881
1882		----------------------------------------------------------------------
1883		Name: Matthias Blume
1884		Date: 2001/11/22 21:40:00 EST
1885		Tag: Release_110_37
1886		Description:
1887
1888		Release 110.37.  This time for real.
1889
1890		----------------------------------------------------------------------
1891		Name: Matthias Blume
1892		Date: 2001/11/21 16:35:00 EST
1893		Tag: blume-20011121-foot-in-mouth
1894		Description:
1895
1896		Removed the "Release_110_37" tag because of a serious bug.
1897		This will be re-tagged once the bug is fixed.
1898
1899		----------------------------------------------------------------------
1900		Name: Matthias Blume
1901		Date: 2001/11/21 16:14:00 EST
1902		Tag: blume-20011121-forgottenfile
1903		Description:
1904
1905		Forgot to add a file.  (Just a .tex-file -- part of
1906		the CM manual source.)
1907
1908		----------------------------------------------------------------------
1909		Name: Matthias Blume
1910		Date: 2001/11/21 16:10:00 EST
1911		Tag: blume-20011121-invalid_110_37
1912		Description:
1913
1914		Note: I removed the original tag "Release_110_37" from this commit
1915		because we found a serious bug in all non-x86 backends.
1916		- Matthias
1917
1918		1. Modifications to the SML/NJ code generator and to the runtime system
1919		so that code object name strings are directly inserted into code
1920		objects at code generation time.  The only business the runtime system
1921		has with this is now to read the name strings on occasions.
1922		(The encoding of the name string has also changed somewhat.)
1923
1924		2. CM now implements a simple "set calculus" for specifying export lists.
1925		In particular, it is now possible to refer to the export lists of
1926		other libraries/groups/sources and form unions as well as differences.
1927		See the latest CM manual for details.
1928
1929		3. An separate notion of "proxy" libraries has again be eliminated from
1930		CM's model.  (Proxy libraries are now simply a special case of using
1931		the export list calculus.)
1932
1933		4. Some of the existing libraries now take advantage of the new set
1934		calculus.
1935		(Notice that not all libraries have been converted because some
1936		of the existing .cm-files are supposed to be backward compatible
1937		with 110.0.x.)
1938
1939		5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
1940		"General.exnMessage"!  The former relies on a certain hook to be
1941		initialized, and that often does not happen in the stand-alone case.)
1942
1943		----------------------------------------------------------------------
1944		Name: Lal George
1945		Date: 2001/11/21  13:56:18 EST
1946		Tag: george-2001121-pseudo-ops
1947		Description:
1948
1949		Implemented a complete redesign of MLRISC pseudo-ops. Now there
1950		ought to never be any question of incompatabilities with
1951		pseudo-op syntax expected by host assemblers.
1952
1953		For now, only modules supporting GAS syntax are implemented
1954		but more should follow, such as MASM, and vendor assembler
1955		syntax, e.g. IBM as, Sun as, etc.
1956
1957		----------------------------------------------------------------------
1958		Name: Matthias Blume
1959		Date: 2001/11/14 11:52:00 EST
1960		Tag: blume-20011114-srcname
1961		Description:
1962
1963		1. Routed the name of the current source file to mlriscgen where it
1964		should be directly emitted into the code object.  (This last part
1965		is yet to be done.)
1966
1967		2. Some cleanup of the pgraph code to make it match the proposal that
1968		I put out the other day.  (The proposal notwithstanding, things are
1969		still in flux here.)
1970
1971		----------------------------------------------------------------------
1972		Name: Lal George
1973		Date: 2001/11/14 09:44:04 EST
1974		Tag:
1975		Description:
1976
1977		Fix for a backpatching bug reported by Allen.
1978
1979		Because the boundary between short and long span-dependent
1980		instructions is +/- 128, there are an astounding number of
1981		span-dependent instructions whose size is over estimated.
1982
1983		Allen came up with the idea of letting the size of span
1984		dependent instructions be non-monotonic, for a maxIter
1985		number of times, after which the size must be monotonically
1986		increasing.
1987
1988		This table shows the number of span-dependent instructions
1989		whose size was over-estimated as a function of maxIter, for the
1990		file Parse/parse/ml.grm.sml:
1991
1992		maxIter            # of instructions:
1993		10                      687
1994		20                      438
1995		30                      198
1996		40                        0
1997
1998		In compiling the compiler, there is no significant difference in
1999		compilation speed between maxIter=10 and maxIter=40. Actually,
2000		my measurements showed that maxIter=40 was a tad faster than
2001		maxIter=10! Also 96% of the  files in the compiler reach a fix
2002		point within 13 iterations, so fixing maxIter at 40, while high,
2003		is okay.
2004
2005		----------------------------------------------------------------------
2006		Name: Matthias Blume
2007		Date: 2001/10/31 15:25:00 EST
2008		Tag: blume-20011031-pgraph
2009		Description:
2010
2011		CKIT:
2012		* Changed the "Function" constructor of type Ast.ctype to carry optional
2013		argument identifiers.
2014		* Changed the return type of TypeUtil.getFunction accordingly.
2015		* Type equality ignores the argument names.
2016		* TypeUtil.composite tries to preserve argument names but gives up quickly
2017		if there is a mismatch.
2018
2019		installation script:
2020		* attempts to use "curl" if available (unless "wget" is available as well)
2021
2022		CM:
2023		* has an experimental implementation of "portable graphs" which I will
2024		soon propose as an implementation-independent library format
2025		* there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2026
2027		NLFFI-LIB:
2028		* some cleanup (all cosmetic)
2029
2030		NLFFIGEN:
2031		* temporarily disabled the mechanism that suppresses ML output for
2032		C definitions whose identifiers start with an underscore character
2033		* generate val bindings for enum constants
2034		* user can request that only one style (light or heavy) is being used;
2035		default is to use both (command-line arguments: -heavy and -light)
2036		* fixed bug in handling of function types involving incomplete pointers
2037		* generate ML entry points that take record arguments (i.e., using
2038		named arguments) for C functions that have a prototype with named
2039		arguments
2040		(see changes to CKIT)
2041
2042		----------------------------------------------------------------------
2043		Name: Allen Leung
2044		Date: 2001/10/27 20:34:00 EDT
2045		Tag: leunga-20011027-x86-fast-fp-call
2046		Description:
2047
2048		Fixed the bug described in blume-20010920-slowfp.
2049
2050		The fix involves
2051		1. generating FCOPYs in FSTP in ia32-svid
2052		2. marking a CALL with the appropriate annotation
2053
2054		----------------------------------------------------------------------
2055		Name: Matthias Blume
2056		Date: 2001/10/16 11:32:00 EDT
2057		Tag: blume-20011016-netbsd
2058		Description:
2059
2060		Underscore patch from Chris Richards (fixing problem with compiling
2061		runtime system under recent NetBSD).
2062
2063		----------------------------------------------------------------------
2064		Name: Allen Leung
2065		Date: 2001/10/12 17:18:32 EDT 2001
2066		Tag: leung-20011012-x86-printflowgraph
2067		Description:
2068
2069		X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2070
2071		----------------------------------------------------------------------
2072		Name: Lal George
2073		Date: 2001/10/11 23:51:34 EDT
2074		Tag: george-20011011-too-many-instrs
2075		Description:
2076
2077		The representation of a program point never expected to see more
2078		than 65536 instructions in a basic block!
2079
2080		----------------------------------------------------------------------
2081		Name: Lal George
2082		Date: 2001/10/09 09:41:37 EDT
2083		Tag: george-20011008-mlrisc-labels
2084		Description:
2085
2086		Changed the machine description files to support printing of
2087		local and global labels in assembly code, based on host assembler
2088		conventions.
2089
2090		----------------------------------------------------------------------
2091		Name: Matthias Blume
2092		Date: 2001/09/25 15:25:00 EDT
2093		Tag: blume-20010925-exninfo
2094		Description:
2095
2096		I provided a non-hook implementation of exnName (at the toplevel) and
2097		made the "dummy" implementation of exnMessage (at the toplevel) more
2098		useful: if nothing gets "hooked in", then at least you are going to
2099		see the exception name and a message indicating why you don't see more.
2100
2101		[For the time being, programs that need exnMessage and want to use
2102		ml-build should either use General.exnMessage (strongly recommended) or
2103		refer to structure General at some other point so that CM sees a
2104		static dependency.]
2105
2106		[Similar remarks go for "print" and "use":  If you want to use their
2107		functionality in stand-alone programs generated by ml-build, then use
2108		TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2109
2110		----------------------------------------------------------------------
2111		Name: Matthias Blume
2112		Date: 2001/09/20 17:28:00 EDT
2113		Tag: blume-20010920-slowfp
2114		Description:
2115
2116		Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2117
2118		----------------------------------------------------------------------
2119		Name: Matthias Blume
2120		Date: 2001/09/20 17:20:00 EDT
2121		Tag: blume-20010920-canonicalpaths
2122		Description:
2123
2124		0. Updated the BOOT file (something that I forgot to do earlier).
2125
2126		1. Small internal change to CM so that it avoids "/../" in filenames
2127		as much as possible (but only where it is safe).
2128
2129		2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2130		that arguments that contain delimiters are passed through correctly.
2131		This change also means that all "special" arguments of the form
2132		@SMLxxx... must come first.
2133
2134		3. Changed install script to put relative anchor names for tool commands
2135		into pathconfig.
2136
2137		----------------------------------------------------------------------
2138		Name: Matthias Blume
2139		Date: 2001/09/18 15:35:00 EDT
2140		Tag: blume-20010918-readme11036
2141		Description:
2142
2143		Added README files.
2144
2145		----------------------------------------------------------------------
2146		Name: Matthias Blume
2147		Date: 2001/09/18 11:45:00 EDT
2148		Tag: Release_110_36 (retag)
2149		Description:
2150
2151		Fixed mistake in config/preloads. Retagged as 110.36.
2152
2153		----------------------------------------------------------------------
2154		Name: Matthias Blume
2155		Date: 2001/09/18 09:40:00 EDT
2156		Tag: Release_110_36_orig (tag changed)
2157		Description:
2158
2159		New version (110.36).  New bootfiles.
2160
2161		----------------------------------------------------------------------
2162		Name: Matthias Blume
2163		Date: 2001/09/14 16:15:00 EDT
2164		Tag: blume-20010914-x86fastfp
2165		Description:
2166
2167		John committed some changes that Allen made, in particular a (hopefully)
2168		correctly working version of the x86-fp module.
2169
2170		I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
2171		flag to "true".  Everything seems to compile to a fixpoint ok, and
2172		"mandelbrot" speeds up by about 15%.
2173
2174		----------------------------------------------------------------------
2175		Name: Matthias Blume
2176		Date: 2001/09/13 11:20:00 EDT
2177		Tag: blume-20010913-minimal
2178		Description:
2179
2180		1. Stefan Monnier's patch to fix a miscompilation problem that
2181		was brought to light by John Reppy's work on Moby.
2182
2183		2. Implemented a minimal "structure Compiler" that contains just
2184		"version" and "architecture".  The minimal version will be
2185		available when the full version is not.  This is for backward-
2186		compatibility with code that wants to test Compiler.version.
2187
2188		----------------------------------------------------------------------
2189		Name: Matthias Blume
2190		Date: 2001/08/28 14:03:00 EDT
2191		Tag: blume-20010828-ml-lex
2192		Description:
2193
2194		Fix for bug 1581, received from Neophytos Michael.
2195
2196		----------------------------------------------------------------------
2197		Name: Matthias Blume
2198		Date: 2001/08/27 11:20:00 EDT
2199		Tag: blume-20010827-readme11035
2200		Description:
2201
2202		Fleshed out the README file for 110.35.
2203
2204		----------------------------------------------------------------------
2205		Name: Matthias Blume
2206		Date: 2001/08/24 17:10:00 EDT
2207		Tag: Release_110_35
2208		Description:
2209
2210		New version number (110.35).  New bootfiles.
2211
2212		----------------------------------------------------------------------
2213		Name: Lal George
2214		Date: 2001/08/24 13:47:18 EDT 2001
2215		Tag: george-20010824-MLRISC-graphs
2216		Description:
2217
2218		removed clusters from MLRISC completely and replaced with graphs.
2219
2220		----------------------------------------------------------------------
2221		Name: Matthias Blume
2222		Date: 2001/08/23 17:50:00 EDT
2223		Tag: blume-20010823-toplevel
2224		Description:
2225
2226		- some reorganization of the code that implements various kinds of
2227		environments in the compiler (static, dynamic, symbolic, combined)
2228		- re-implemented the EnvRef module so that evalStream works properly
2229		(if the stream contains references to "use", "CM.make", etc.)
2230		- cleaned up evalloop.sml and interact.sml (but they need more cleaning)
2231
2232		----------------------------------------------------------------------
2233		Name: Matthias Blume
2234		Date: 2001/08/20 15:50 EDT
2235		Tag: blume20010820-slipup
2236		Description:
2237
2238		I forgot to commit a few files.  Here they are...
2239
2240		----------------------------------------------------------------------
2241		Name: Matthias Blume
2242		Date: 2001/08/20 15:35:00 EDT
2243		Tag: blume-20010820-debugprof
2244		Description:
2245
2246		!!!! NEW BOOTFILES !!!!
2247
2248		This is another round of reorganizing the compiler sources.  This
2249		time the main goal was to factor out all the "instrumentation"
2250		passes (for profiling and backtracing) into their own library.
2251		The difficulty was to do it in such a way that it does not depend
2252		on elaborate.cm but only on elabdata.cm.
2253
2254		Therefore there have been further changes to both elaborate.cm and
2255		elabdata.cm -- more "generic" things have been moved from the former
2256		to the latter.  As a result, I was forced to split the assignment
2257		of numbers indicating "primtyc"s into two portions: SML-generic and
2258		SML/NJ-specific.  Since it would have been awkward to maintain,
2259		I bit the bullet and actually _changed_ the mapping between these
2260		numbers and primtycs.  The bottom line of this is that you need
2261		a new set of bin- and bootfiles.
2262
2263		I have built new bootfiles for all architectures, so doing a fresh
2264		checkout and config/install.sh should be all you need.
2265
2266		The newly created library's name is
2267
2268		$smlnj/viscomp/debugprof.cm
2269
2270		and its sources live under
2271
2272		src/compiler/DebugProf
2273
2274		----------------------------------------------------------------------
2275		Name: Matthias Blume
2276		Date: 2001/08/15 17:15:00 EDT
2277		Tag: blume-20010815-compreorg
2278		Description:
2279
2280		This is a first cut at reorganizing the CM libraries that make up the
2281		core of the compiler.  The idea is to separate out pieces that could
2282		be used independently by tools, e.g., the parser, the typechecker, etc.
2283
2284		The current status is a step in this direction, but it is not quite
2285		satisfactory yet.  Expect more changes in the future.
2286
2287		Here is the current (new) organization...
2288
2289		What used to be $smlnj/viscomp/core.cm is now divided into
2290		six CM libraries:
2291
2292		$smlnj/viscomp/basics.cm
2293		/parser.cm
2294		/elabdata.cm
2295		/elaborate.cm
2296		/execute.cm
2297		/core.cm
2298
2299		The CM files for these libraries live under src/system/smlnj/viscomp.
2300		All these libraries are proxy libraries that contain precisely
2301		one CM library component.  Here are the locations of the components
2302		(all within the src/compiler tree):
2303
2304		Basics/basics.cm
2305		Parse/parser.cm
2306		ElabData/elabdata.cm
2307		Elaborator/elaborate.cm
2308		Execution/execute.cm
2309		core.cm
2310
2311		[This organization is the same that has been used already
2312		for a while for the architecture-specific parts of the visible
2313		compiler and for the old version of core.cm.]
2314
2315		As you will notice, many source files have been moved from their
2316		respective original locations to a new home in one of the above
2317		subtrees.
2318
2319		The division of labor between the new libraries is the following:
2320
2321		basics.cm:
2322		- Simple, basic definitions that pertain to many (or all) of
2323		the other libraries.
2324		parser.cm:
2325		- The SML parser, producing output of type Ast.dec.
2326		- The type family for Ast is also defined and exported here.
2327		elabdata.cm:
2328		- The datatypes that describe input and output of the elaborator.
2329		This includes types, absyn, and static environments.
2330		elaborator.cm:
2331		- The SML/NJ type checker and elaborator.
2332		This maps an Ast.dec (with a given static environment) to
2333		an Absyn.dec (with a new static environment).
2334		- This libraries implements certain modules that used to be
2335		structures as functors (to remove dependencies on FLINT).
2336		execute.cm:
2337		- Everything having to do with executing binary code objects.
2338		- Dynamic environments.
2339		core.cm:
2340		- SML/NJ-specific instantiations of the elaborator and MLRISC.
2341		- Top-level modules.
2342		- FLINT (this should eventually become its own library)
2343
2344		Notes:
2345
2346		I am not 100% happy with the way I separated the elaborator (and its
2347		data structures) from FLINT.  Two instances of the same problem:
2348
2349		1. Data structures contain certain fields that carry FLINT-specific
2350		information.  I hacked around this using exn and the property list
2351		module from smlnj-lib.  But the fact that there are middle-end
2352		specific fields around at all is a bit annoying.
2353
2354		2. The elaborator calculates certain FLINT-related information.  I tried
2355		to make this as abstract as I could using functorization, but, again,
2356		the fact that the elaborator has to perform calculations on behalf
2357		of the middle-end at all is not nice.
2358
2359		3. Having to used exn and property lists is unfortunate because it
2360		weakens type checking.  The other alternative (parameterizing
2361		nearly *everything*) is not appealing, though.
2362
2363		I removed the "rebinding =" warning hack because due to the new organization
2364		it was awkward to maintain it.  As a result, the compiler now issues some of
2365		these warnings when compiling init.cmi during bootstrap compilation. On
2366		the plus side, you also get a warning when you do, for example:
2367		val op = = Int32.+
2368		which was not the case up to now.
2369
2370		I placed "assign" and "deref" into the _Core structure so that the
2371		code that deals with the "lazy" keyword can find them there.  This
2372		removes the need for having access to the primitive environment
2373		during elaboration.
2374
2375		----------------------------------------------------------------------
2376		Name: Matthias Blume
2377		Date: 2001/08/13
2378		Tag: blume-20010813-closures
2379		Description:
2380
2381		This fix was sent to us by Zhong Shao.  It is supposed to improve the
2382		performance of certain loops by avoiding needless closure allocation.
2383
2384		----------------------------------------------------------------------
2385		Name: Lal George
2386		Date: 2001/07/31 10:03:23 EDT 2001
2387		Tag: george-20010731-x86-fmalloc
2388		Description: Fixed bug in x86 calls
2389
2390		There was a bug where call instructions would mysteriously
2391		vanish. The call instruction had to be one that returned
2392		a floating point value.
2393
2394		----------------------------------------------------------------------
2395		Name: Lal George
2396		Date: 2001/07/19 16:36:29 EDT 2001
2397		Tag: george-20010719-simple-cells
2398		Description:
2399
2400		I have dramatically simplified the interface for CELLS in MLRISC.
2401
2402		In summary, the cells interface is broken up into three parts:
2403
2404		1. CellsBasis : CELLS_BASIS
2405
2406		CellsBasis is a top level structure and common for all
2407		architectures.  it contains the definitions of basic datatypes
2408		and utility  functions over these types.
2409
2410		2. functor Cells() : CELLS
2411
2412		Cells generates an interface for CELLS that incorporates the
2413		specific resources on the target architecture, such as the
2414		presence of special register classes, their number and size,
2415		and various useful substructures.
2416
2417		3. <ARCH>CELLS
2418
2419		e.g. SparcCells: SPARCCELLS
2420
2421		<ARCH>CELLS usually contains additional bindings for special
2422		registers  on the architecture, such as:
2423
2424		val r0 : cell           (* register zero *)
2425		val y : cell            (* Y register *)
2426		val psr : cell          (* processor status register *)
2427		...
2428
2429		The structure returned by applying the Cells functor is opened
2430		in this interface.
2431
2432		The main implication of all this is that the datatypes for cells is
2433		split between CellsBasis and CELLS -- a fairly simple change for user
2434		code.
2435
2436		In the old scheme the CELLS interface had a definitional binding of
2437		the form:
2438
2439		signature CELLS = sig
2440
2441		structure CellsBasis = CellsBasis
2442
2443		...
2444
2445		end
2446
2447		With all the sharing constraints that goes on in MLRISC, this old
2448		design  quickly leads to errors such as:
2449
2450		"structure definition spec inside of sharing ... "
2451
2452
2453		and appears to require an unacceptable amount of sharing and where
2454		constraint hackery.
2455
2456		I think this error message (the interaction of definitional specs and
2457		sharing) requires more explanation on our web page.
2458
2459		----------------------------------------------------------------------
2460		Name: Matthias Blume
2461		Date: 2001/07/19 15:00:00 EDT
2462		Tag: blume-20010719-libreorg
2463		Description:
2464
2465		This update puts together a fairly extensive but straightforward change
2466		to the way the libraries that implement the interactive system are
2467		organized:
2468
2469		The biggest change is the elimination of structure Compiler.  As a
2470		replacement for this structure, there is now a CM library
2471		(known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
2472		that exports all the substructures of the original structure Compiler
2473		directly.  So instead of saying Compiler.Foo.bar one now simply
2474		says Foo.bar.  (The CM libraries actually export a collection of
2475		structures that is richer than the collection of substructures of
2476		structure Compiler.)
2477
2478		To make the transition smooth, there is a separate library called
2479		$smlnj/compiler/compiler.cm which puts together and exports the
2480		original structure Compiler (or at least something very close to it).
2481
2482		There are five members of the original structure Compiler
2483		that are not exported directly but which instead became members
2484		of a new structure Backend (described by signature BACKEND).  These are:
2485		structure Profile (: PROFILE), structure Compile (: COMPILE), structure
2486		Interact (: INTERACT), structure Machine (: MACHINE), and val
2487		architecture (: string).
2488
2489		Structure Compiler.Version has become structure CompilerVersion.
2490
2491		Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
2492		by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
2493		or x86, respectively.
2494		Each of these exports the same frontend structures that
2495		$smlnj/compiler.cm exports.  But they do not have a structure Backend
2496		and instead export some structure <Arch>Backend where <Arch> is Alpha32,
2497		Hppa, PPC, Sparc, or X86, respectively.
2498
2499		Library $smlnj/compiler/all.cm exports the union of the exports of
2500		$smlnj/compiler/<arch>.cm
2501
2502		There are no structures <Arch>Compiler anymore, use
2503		$smlnj/compiler/<arch>.cm instead.
2504
2505		Library host-compiler-0.cm is gone.  Instead, the internal library
2506		that instantiates CM is now called cm0.cm.  Selection of the host
2507		compiler (backend) is no longer done here but. (Responsibility for it
2508		now lies with $smlnj/compiler/current.cm.  This seems to be more
2509		logical.)
2510
2511		Many individual files have been moved or renamed.  Some files have
2512		been split into multiple files, and some "dead" files have been deleted.
2513
2514		Aside from these changes to library organization, there are also changes
2515		to the way the code itself is organized:
2516
2517		Structure Binfile has been re-implemented in such a way that it no
2518		longer needs any knowledge of the compiler.  It exclusively deals
2519		with the details of binfile layout.  It no longer invokes the
2520		compiler (for the purpose of creating new prospective binfile
2521		content), and it no longer has any knowledge of how to interpret
2522		pickles.
2523
2524		Structure Compile (: COMPILE) has been stripped down to the bare
2525		essentials of compilation.  It no longer deals with linking/execution.
2526		The interface has been cleaned up considerably.
2527
2528		Utility routines for dealing with linking and execution have been
2529		moved into their own substructures.
2530
2531		(The ultimate goal of these changes is to provide a light-weight
2532		binfile loader/linker (at least for, e.g., stable libraries) that
2533		does not require CM or the compiler to be present.)
2534
2535		CM documentation has been updated to reflect the changes to library
2536		organization.
2537
2538		----------------------------------------------------------------------
2539		Name: Matthias Blume
2540		Date: 2001/07/10 17:30:00 EDT
2541		Tag: Release_110_34
2542		Description:
2543
2544		Minor tweak to 110.34 (re-tagged):
2545
2546		- README.html file added to CVS repository
2547		- runtime compiles properly under FreeBSD 3.X and 4.X
2548
2549		----------------------------------------------------------------------
2550		Name: Matthias Blume
2551		Date: 2001/07/10 17:30:00 EDT
2552		Tag: Release_110_34
2553		Description:
2554
2555		New version number (110.34). New bootfiles.
2556
2557		----------------------------------------------------------------------
2558		Name: Matthias Blume
2559		Date: 2001/07/09 16:00:00 EDT
2560		Tag: blume-20010709-more-varargs
2561		Description:
2562
2563		I changed the handling of varargs in ml-nlffigen again:
2564		The ellipsis ... will now simply be ignored (with an accompanying warning).
2565
2566		The immediate effect is that you can actually call a varargs function
2567		from ML -- but you can't actually supply any arguments beyond the ones
2568		specified explicitly.  (For example, you can call printf with its format
2569		string, but you cannot pass additional arguments.)
2570
2571		This behavior is only marginally more useful than the one before, but
2572		it has the advantage that a function or, more importantly, a function
2573		type never gets dropped on the floor, thus avoiding follow-up problems with
2574		other types that refer to the offending one.
2575
2576		----------------------------------------------------------------------
2577		Name: Matthias Blume
2578		Date: 2001/07/09 11:25:00 EDT
2579		Tag: blume-20010709-varargs
2580		Description:
2581
2582		1. ckit-lib.cm now exports structure Error
2583		2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
2584		with a warning accompanied by a source location.  Moreover, it
2585		merely skips the offending function or type and proceeds with the
2586		rest of its work.u  As a result, one can safely feed C code containing
2587		"..." to ml-nlffigen.
2588		3. There are some internal improvements to CM, providing slightly
2589		more general string substitutions in the tools subsystem.
2590
2591		----------------------------------------------------------------------
2592		Name: Matthias Blume
2593		Date: 2001/06/27 15:10:00 EDT
2594		Tag: blume-20010627-concur
2595		Description:
2596
2597		Fixed a small bug in CM's handling of parallel compilation.
2598		(You could observe the bug by Control-C-interrupting an ordinary
2599		CMB.make or CM.stabilize and then attaching some compile servers.
2600		The result was that all of a sudden the previously interrupted
2601		compilation would continue on its own.  This was because of
2602		an over-optimization: CM did not bother to clean out certain queues
2603		when no servers were attached "anyway", resulting in the contents
2604		of these queues to grab control when new servers did get attached.)
2605
2606		There is also another minor update to the CM manual.
2607
2608		----------------------------------------------------------------------
2609		Name: Matthias Blume
2610		Date: 2001/06/26 16:15:00 EDT
2611		Tag: blume-20010626-cmdoc
2612		Description:
2613
2614		Minor typo fixed in CM manual (syntax diagram for libraries).
2615
2616		----------------------------------------------------------------------
2617		Name: Matthias Blume
2618		Date: 2001/06/25 22:55:00 EDT
2619		Tag: blume-20010625-x86pc
2620		Description:
2621
2622		Fixed a nasty bug in the X86 assembly code that caused signal
2623		handlers to fail (crash) randomly.
2624
2625		----------------------------------------------------------------------
2626		Name: Matthias Blume
2627		Date: 2001/06/25 12:05:00 EDT
2628		Tag: blume-20010625-nlffigen
2629		Description:
2630
2631		This update fixes a number of minor bugs in ml-nlffigen as reported by
2632		Nick Carter <nbc@andrew.cmu.edu>.
2633
2634		1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
2635		2. Default names for generated files are now derived from the name of
2636		the C file *without its directory*.  In particular, this causes generated
2637		files to be placed locally even if the C file is in some system directory.
2638		3. Default names for generated signatures and structures are also derived
2639		from the C file name without its directory.  This avoids silly things
2640		like "structure GL/GL".
2641		(Other silly names are still possible because ml-nlffigen does not do
2642		a thorough check of whether generated names are legal ML identifiers.
2643		When in doubt, use command line arguments to force particular names.)
2644
2645		----------------------------------------------------------------------
2646		Name: Matthias Blume
2647		Date: 2001/06/21 12:25:00 EDT
2648		Tag: blume-20010621-eXene
2649		Description:
2650
2651		eXene now compiles and (sort of) works again.
2652
2653		The library name (for version > 110.33) is $/eXene.cm.
2654
2655		I also added an new example in src/eXene/examples/nbody.  See the
2656		README file there for details.
2657
2658		----------------------------------------------------------------------
2659		Name: Matthias Blume
2660		Date: 2001/06/20 16:40:00 EDT
2661		Tag: blume-20010620-cml
2662		Description:
2663
2664		CML now compiles and works again.
2665
2666		Libraries (for version > 110.33):
2667
2668		$cml/cml.cm            Main CML library.
2669		$cml/basis.cm          CML's version of $/basis.cm.
2670		$cml/cml-internal.cm   Internal helper library.
2671		$cml/core-cml.cm       Internal helper library.
2672		$cml-lib/trace-cml.cm  Tracing facility.
2673		$cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
2674
2675		The installer (config/install.sh) has been taught how to properly
2676		install this stuff.
2677
2678		----------------------------------------------------------------------
2679		Name: Matthias Blume
2680		Date: 2001/06/19 17:55:00 EDT
2681		Tag: blume-20010619-instantiate
2682		Description:
2683
2684		This un-breaks the fix for bug 1432.
2685		(The bug was originally fixed in 110.9 but I broke it again some
2686		time after that.)
2687
2688		----------------------------------------------------------------------
2689		Name: Matthias Blume
2690		Date: 2001/06/19 17:25:00 EDT
2691		Tag: blume-20010619-signals
2692		Description:
2693
2694		This should (hopefully) fix the long-standing signal handling bug.
2695		(The runtime system was constructing a continuation record with an
2696		incorrect descriptor which would cause the GC to drop data on the floor...)
2697
2698		----------------------------------------------------------------------
2699		Name: Matthias Blume
2700		Date: 2001/06/15 15:05:00 EDT
2701		Tag: blume-20010615-moresparc
2702		Description:
2703
2704		Here is a short late-hour update related to Sparc c-calls:
2705
2706		-- made handling of double-word arguments a bit smarter
2707
2708		-- instruction selection phase tries to collapse certain clumsily
2709		constructed ML-Trees; typical example:
2710
2711		ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
2712
2713		This currently has no further impact on SML/NJ since mlriscGen does
2714		not seem to generate such patterns in the first place, and c-calls
2715		(which did generate them in the beginning) has meanwhile been fixed
2716		so as to avoid them as well.
2717
2718		----------------------------------------------------------------------
2719		Name: Matthias Blume
2720		Date: 2001/06/15 15:05:00 EDT
2721		Tag: blume-20010615-sparc
2722		Description:
2723
2724		The purpose of this update is to provide an implementation of NLFFI
2725		on Sparc machines.
2726
2727		Here are the changes in detail:
2728
2729		* src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
2730		the Sparc implementation of the c-calls API.
2731		* The Sparc backend of SML/NJ has been modified to uniformely use %fp
2732		for accessing the ML frame.  Thus, we have a real frame pointer and
2733		can freely modify %sp without need for an omit-frame-ptr phase.
2734		The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
2735		this case.
2736		* ml-nlffigen has been taught to produce code for different architectures
2737		and calling conventions.
2738		* In a way similar to what was done in the x86 case, the Sparc
2739		backend uses its own specific extension to mltree.  (For example,
2740		it needs to be able to generate UNIMP instructions which are part
2741		of the calling convention.)
2742		* ml-nlffi-lib was reorganized to make it more modular (in particular,
2743		to make it easier to plug in new machine- and os-dependent parts).
2744
2745		There are some other fairly unrelated bug fixes and cleanups as well:
2746
2747		* I further hacked the .cm files for MLRISC tools (like MDLGen) so
2748		that they properly share their libraries with existing SML/NJ libraries.
2749		* I fixed a minor cosmetic bug in CM, supressing certain spurious
2750		follow-up error messages.
2751		* Updates to CM/CMB documentation.
2752
2753		TODO items:
2754
2755		* MLRISC should use a different register as its asmTemp on the Sparc.
2756		(The current %o2 is a really bad choice because it is part of the
2757		calling conventions, so things might interfere in unexpected ways.)
2758
2759		----------------------------------------------------------------------
2760		Name: Matthias Blume
2761		Date: 2001/06/07
2762		Tag: blume-20010607-calls
2763		Description:
2764
2765		A number of internal changes related to C calls and calling conventions:
2766
2767		1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
2768		number of bytes popped implicitly (by the callee).  In most cases
2769		this field is 0 but on x86/win32 it is some non-zero value.  This
2770		is information provided for the benefit of the "omit-frameptr" pass.
2771		2. The CALL instruction on the x86 carries a similar "pops" field.
2772		The instruction selection phase copies its value from the ML-Tree
2773		CALL statement.
2774		3. On all other architectures, the instruction selection phase checks
2775		whether "pops=0" and complains if not.
2776		4. The c-calls implementation for x86 now accepts two calling conventions:
2777		"ccall" and "stdcall".  When "ccall" is selected, the caller cleans
2778		up after the call and pops is set to 0.  For "stdcall", the caller
2779		does nothing, leaving the cleanup to the callee; pops is set to
2780		the number of bytes that were pushed onto the stack.
2781		5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
2782		distinguish between "ccall" and "stdcall".
2783		6. The UNIMP instruction has been added to the supported Sparc instruction
2784		set. (This is needed for implementing the official C calling convention
2785		on this architecture.)
2786		7. I fixed some of the .cm files under src/MLRISC/Tools to make them
2787		work with the latest CM.
2788
2789		----------------------------------------------------------------------
2790		Name: Matthias Blume
2791		Date: 2001/06/05 15:10:00 EDT
2792		Tag: blume-20010605-cm-index
2793		Description:
2794
2795		0. The "lambdasplit" parameter for class "sml" in CM has been documented.
2796
2797		1. CM can now generate "index files".  These are human-readable files
2798		that list on a per-.cm-file basis each toplevel symbol defined or
2799		imported.  The location of the index file for
2800		<p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
2801		To enable index-file generation, set CM.Control.generate_index to true
2802		or export an environment-symbol: export CM_GENERATE_INDEX=true.
2803
2804		The CM manual has been updated accordingly.
2805
2806		2. I made some slight modifications to the c-calls API in MLRISC.
2807
2808		a) There is now a callback to support saving/restoring of
2809		dedicated but caller-save registers around the actual call
2810		instruction.
2811		b) One can optionally specify a comment-annotation for the
2812		call instruction.
2813
2814		3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
2815		(For example, the comment annotation shows the C prototype of
2816		the function being called.)
2817
2818		----------------------------------------------------------------------
2819		Name: Matthias Blume
2820		Date: 2001/06/01 13:30:00 EDT
2821		Tag: blume-20010601-nlffi-cleanup
2822		Description:
2823
2824		This is mostly a cleanup of MLFFI stuff:
2825
2826		- some signature files have been put into a more exposed place
2827		- the ugly 'f type parameter is gone (simplifies types tremendously!)
2828		- ml-nlffigen changed accordingly
2829		- tutorial updated
2830
2831		Other changes:
2832
2833		- author's affiliation in CM manual(s) updated
2834		- some more recognized keywords added to Allen's sml.sty
2835
2836		----------------------------------------------------------------------
2837		Name: Matthias Blume
2838		Date: 2001/05/25 15:30:00 EDT
2839		Tag: blume-20010525-iptr
2840		Description:
2841
2842		- put the official 110.33-README (as it appears on the ftp server) under
2843		CVS
2844		- fixed a small bug related to incomplete pointer types in
2845		ml-nlffigen
2846		- small cosmetic change to the ml-nlffi-lib's "arr" type constructor
2847		(it does not need the 'f type parameter)
2848
2849		----------------------------------------------------------------------
2850		Name: Matthias Blume
2851		Date: 2001/05/23 14:30:00 EDT
2852		Tag: Release_110_33
2853		Description:
2854
2855		New version number (110.33).  New bootfiles.
2856
2857		----------------------------------------------------------------------
2858		Name: Matthias Blume
2859		Date: 2001/05/22 18:06:00 EDT
2860		Tag: blume-20010522-targets
2861		Description:
2862
2863		Made install.sh use file config/targets.customized if it exists, falling
2864		back to config/targets if it doesn't.  This way one can have a customized
2865		version of the targets file without touching the "real thing", thus
2866		eliminating the constant fear of accidentally checking something bogus
2867		back into the CVS repository...  (File config/targets.customized must
2868		not be added to the repository!)
2869
2870		----------------------------------------------------------------------
2871		Name: Matthias Blume
2872		Date: 2001/05/22 16:30:00 EDT
2873		Tag: blume-20010522-minitut
2874		Description:
2875
2876		1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
2877		struct returns.
2878		2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
2879		incomplete, preliminary documentation for NLFFI.
2880
2881		----------------------------------------------------------------------
2882		Name: Matthias Blume
2883		Date: 2001/05/14 11:30:00 EDT
2884		Tag: blume-20010514-script
2885		Description:
2886
2887		Some bugs in install script fixed.
2888
2889		In addition to that I also made a slight change to the NLFFI API:
2890		Functors generated by ml-nlffigen now take the dynamic library as a
2891		straight functor argument, not as a suspended one.  (The original
2892		functor code used to force the suspension right away anyway, so there
2893		was nothing gained by this complication of the interface.)
2894
2895		----------------------------------------------------------------------
2896		Name: Matthias Blume
2897		Date: 2001/05/11 14:35:00 EDT
2898		Tag: blume-20010511-ml-nlffi
2899		Description:
2900
2901		I finally took the plunge and added my new FFI code to the main
2902		repository.  For x86-linux it is now ready for prime-time.
2903
2904		There are two new subdirectories of "src":
2905
2906		- ml-nlffi-lib:
2907		The utility library for programs using the FFI interface.
2908		Here is the implementation of $/c.cm and its associated low-level
2909		partners $/c-int.cm and $/memory.cm.
2910		- ml-nlffigen:
2911		A stand-alone program for generating ML glue code from C source
2912		code.
2913
2914		Building ml-nlffigen requires $/ckit-lib.cm.
2915
2916		The config/install.sh script has been updates to do the Right Thing
2917		(hopefully).
2918
2919		Notice that the source tree for the C-Kit will not be put under "src"
2920		but directly under the installation root directory.  (This is the
2921		structure that currently exists on the CVS server when you check out
2922		module "sml".)  Fortunately, config/install.sh knows about this oddity.
2923
2924		Bugs: No documentation yet.
2925
2926		----------------------------------------------------------------------
2927		Name: Matthias Blume
2928		Date: 2001/05/09 16:35:00 EDT
2929		Tag: blume-20010509-cpscontract
2930		Description:
2931
2932		Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
2933		wrapper/unwrapper elimination did not decrement usage counts and some
2934		dead variables got overlooked by the dead-up logic.)
2935
2936		----------------------------------------------------------------------
2937		Name: Lal George
2938		Date: 2001/05/08  17:26:09 EDT
2939		Tag: george-20010508-omit-frameptr
2940		Description:
2941
2942		Changes to implement the omit-frame-pointer optimization to support
2943		raw C calls. For now, there is only support on the Intel x86, but
2944		other architectures will follow as more experience is gained with this.
2945
2946
2947		----------------------------------------------------------------------
2948		Name: Matthias Blume
2949		Date: 2001/05/07 14:40:00 EDT
2950		Tag: blume-20010507-proxies
2951		Description:
2952
2953		I made into "proxy libraries" all libraries that qualify for such a
2954		change.  (A qualifying library is a library that has another library or
2955		groups as its sole member and repeats that member's export list
2956		verbatim.  A proxy library avoids this repetition by omitting its export
2957		list, effectively inheriting the list that its (only) member exports.
2958		See the CM manual for more explanation.)
2959		The main effect is that explicit export lists for these libraries
2960		do not have to be kepts in sync, making maintenance a bit easier.
2961
2962		I also added copyright notices to many .cm-files.
2963
2964		Last but not least, I made a new set of bootfiles.
2965
2966		----------------------------------------------------------------------
2967		Name: Matthias Blume
2968		Date: 2001/05/04 17:00:00 EDT
2969		Tag: blume-20010504-cm-lsplit
2970		Description:
2971
2972		0. John merged pending changes to $/smlnj-lib.cm
2973
2974		1. Allen's previous change accidentally backed out of one of Lal's
2975		earlier changes.  I undid this mistake (re-introducing Lal's change).
2976
2977		2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
2978		within the compiler where applicable.  There is some code simplification
2979		because of that.
2980
2981		3. The "split" phase (in FLINT) is now part of the default list of phases.
2982		Compiler.Control.LambdaSplitting.* can be used to globally control the
2983		lambda-splitting (cross-module-inlining) engine.  In addition to that,
2984		it can now also be controlled on a per-source basis: CM has been taught
2985		a new tool parameter applicable to ML source files.
2986
2987		- To turn lambda-splitting off completely:
2988		local open Compiler.Control.LambdaSplitting in
2989		val _ = set Off
2990		end
2991		- To make "no lambda-splitting" the global default (but allow per-source
2992		overriding); this is the initial setting:
2993		local open Compiler.Control.LambdaSplitting in
2994		val _ = set (Default NONE)
2995		end
2996		- To make "lambda-splitting with aggressiveness a" the global default
2997		(and allow per-source overriding):
2998		local open Compiler.Control.LambdaSplitting in
2999		val _ = set (Default (SOME a))
3000		end
3001
3002		- To turn lambda-splitting off for a given ML souce file (say: a.sml)
3003		write (in the respective .cm-file):
3004		a.sml (lambdasplitting:off)
3005		- To turn lambda-splitting for a.sml on with minimal aggressiveness:
3006		a.sml (lambdasplitting:on)
3007		- To turn lambda-splitting for a.sml on with aggressiveness <a> (where
3008		<a> is a decimal non-negative integer):
3009		a.sml (lambdasplitting:<a>)
3010		- To turn lambda-splitting for a.sml on with maximal aggressiveness:
3011		a.sml (lambdasplitting:infinity)
3012		- To use the global default for a.sml:
3013		a.sml (lambdasplitting:default)
3014		or simply
3015		a.sml
3016
3017		----------------------------------------------------------------------
3018		Name: Allen Leung
3019		Date: 2001/05/04 01:57:00 EDT
3020		Tag: leunga-20010504-sync
3021		Description:
3022
3023		MLRISC features.
3024
3025		1. Fix to CMPXCHG instructions.
3026		2. Changed RA interface to allow annotations in callbacks.
3027		3. Added a new method to the stream interface to allow annotations updates.
3028
3029		----------------------------------------------------------------------
3030		Name: Matthias Blume
3031		Date: 2001/05/01 11:45:00 EDT
3032		Tag: blume-20010501-pcedittmp
3033		Description:
3034
3035		Changed install.sh to use the current working directory instead of
3036		/usr/tmp for a temporary file (pcedittmp).  The previous choice
3037		of /usr/tmp caused trouble with MacOS X because of file premission
3038		problems.
3039
3040		----------------------------------------------------------------------
3041		Name: Matthias Blume
3042		Date: 2001/04/20 11:10:00 EDT
3043		Tag: blume-20010420-inMLflag
3044		Description:
3045
3046		- added vp_limitPtrMask to vproc-state.h
3047		(for use by the raw-C-calls mechanism to implement proper interrupt
3048		handling)
3049		- made the ML compiler aware of various data-structure offsets so it
3050		can generate code for accessing the vp_inML flag and vp_limitPtrMask
3051		- tweaked mlriscGen.sml to have it emit interrupt-handling code for
3052		raw C-calls
3053
3054		----------------------------------------------------------------------
3055		Name: Lal George
3056		Date: 2001/04/20 09:15:28 EDT
3057		Tag: george-20010420-macosX
3058		Description:
3059
3060		- Changes to port to Mac OS X; Darwin.
3061
3062		- In the process I found that sqrt was broken on the PPC, because the
3063		fsqrt instruction is not implemented.
3064
3065		----------------------------------------------------------------------
3066		Name: Matthias Blume
3067		Date: 2001/04/18 12:45:00 EDT
3068		Tag: blume-20010418-ccalls
3069		Description:
3070
3071		- fixed two off-by-4 errors in the x86-specific c-calls implementation
3072		(this bug prevented structure arguments containing pointers from being
3073		passed correctly)
3074		- changed the raw-C-call code in mlriscGen.sml in such a way that
3075		structure arguments are represented as a pointer to the beginning
3076		of the structure (instead of having a series of synthesized arguments,
3077		one for each structure member)
3078
3079		- made makeml script's verbosity level configurable via environment
3080		variable (MAKEML_VERBOSITY)
3081
3082		- eliminated placeholder implementations for f32l, w16s, i16s, and f32s
3083		in rawmem-x86.sml; we are now using the real thing
3084
3085		----------------------------------------------------------------------
3086		Name: Matthias Blume
3087		Date: 2001/03/22 16:25:00 EST
3088		Tag: blume-20010322-bootfiles
3089		Description:
3090
3091		Created a new set of bootfiles (for your automatic installation convenience).
3092
3093		----------------------------------------------------------------------
3094		Name: Matthias Blume
3095		Date: 2001/03/22 15:10:00 EST
3096		Tag: blume-20010322-rawmem-parcm
3097		Description:
3098
3099		1. All "raw memory access" primitives for the new FFI are implemented now
3100		(at least on the x86).
3101		2. Some further cleanup of CM's parallel make mechanism.
3102
3103		----------------------------------------------------------------------
3104		Name: Matthias Blume
3105		Date: 2001/03/19 17:53:00 EST
3106		Tag: blume-20010319-parallel
3107		Description:
3108
3109		Parallel make (using compile servers) now works again.
3110
3111		To this end, CM.stabilize and CMB.make have been modified to work in
3112		two passes when compile servers are attached:
3113		1. Compile everything, do not perform stabilization; this pass
3114		uses compile servers
3115		2. Stabilize everything; this pass does not use compile servers
3116		If there are no compile servers, the two passes are combined into one
3117		(as before).  Splitting the passes increases the inherent parallelism
3118		in the dependency graph because the entire graph including all
3119		libraries is available at the same time.  This, in turn, improves
3120		server utilization.  The downside is that the master process will
3121		have to do some extra work after compilation is done (because for
3122		technical reasons it must re-read all the binfiles during stabilization).
3123
3124		----------------------------------------------------------------------
3125		Name: Matthias Blume
3126		Date: 2001/03/16 12:22:00 EST
3127		Tag: blume-20010316-bootfiles
3128		Description:
3129
3130		Created a new set of bootfiles (for your automatic installation convenience).
3131
3132		----------------------------------------------------------------------
3133		Name: Matthias Blume
3134		Date: 2001/03/16 11:00:00 EST
3135		Tag: blume-20010316-MLTREE-fixup
3136		Description:
3137
3138		This is a minor fixup for an (untagged) earlier commit by Allen.
3139		(A file was missing).
3140
3141		----------------------------------------------------------------------
3142		Name: Allen Leung
3143		Date: Mon Mar  5 18:54:57 EST 2001
3144		Tag: leunga-20010305-cut-support
3145
3146		1. New support for alternative control-flow in MLTREE.
3147		Currently we support
3148
3149		FLOW_TO(CALL ...., [k1,...,kn])
3150
3151		This is needed for 'cuts to' in C-- and try/handle-like constructs
3152		in Moby
3153
3154		New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
3155
3156		2. Register Allocator
3157
3158		Changes in interface [from Fermin, John]
3159
3160		3. Alpha 8-bit SLL support [Fermin]
3161
3162		4. All architectures
3163
3164		A new module (ClusterExpandCopies) for expanding parallel copies.
3165
3166		----------------------------------------------------------------------
3167		Name: Allen Leung
3168		Date: 2001/02/27 23:07:00 EST
3169		Tag: leunga-20010227-minor-stuff
3170
3171		1. Alpha bug fix for CMOVNE
3172		2. Handle mltree COND(..,FCMP ...,...)
3173		3. Bug fix in simplifier
3174
3175		----------------------------------------------------------------------
3176		Name: Matthias Blume
3177		Date: 2001/01/30 17:50:00 EST
3178		Tag: blume-20010130-sync
3179		Description:
3180
3181		This is just a minor update to sync my devel branch with the main brach.
3182		The only visible change is the addition of some README files.
3183
3184		----------------------------------------------------------------------
3185		Name: Matthias Blume
3186		Date: 2001/01/12 23:30:00 JST
3187		Tag: blume-20010112-bootfiles
3188		Description:
3189
3190		Made a new set of bootfiles that goes with the current state of the
3191		repository.
3192
3193		----------------------------------------------------------------------
3194		Name: Matthias Blume
3195		Date: 2001/01/12 21:20:00 JST
3196		Tag: blume-20010112-sync
3197		Description:
3198
3199		I am just flushing out some minor changes that had accumulated in
3200		my private branch in order to sync with the main tree.  (This is
3201		mainly because I had CVS trouble when trying to merge _into_ my
3202		private branch.)
3203
3204		Most people should be completely unaffected by this.
3205
3206		----------------------------------------------------------------------
3207		Name: Allen Leung
3208		Date: Thu Jan 11 21:03:00 EST 2001
3209		Tag: leunga-20010111-labexp=mltree
3210		Description:
3211
3212		1.  Removed the type LabelExp and replace it by MLTree.
3213		2.  Rewritten mltree-simplify with the pattern matcher tool.
3214		3.  There were some bugs in alpha code generator which would break
3215		64-bit code generation.
3216		4.  Redo the tools to generate code with the
3217		5.  The CM files in MLRISC (and in src/system/smlnj/MLRISC)
3218		are now generated by perl scripts.
3219
3220		----------------------------------------------------------------------
3221		Name: Matthias Blume
3222		Date: 2001/01/10 21:55:00 JST
3223		Tag: blume-20010110-rcc
3224		Description:
3225
3226		The RCC stuff now seems to work (but only on the x86).
3227		This required hacking of the c-calls interface (and -implementation) in
3228		MLRISC.
3229
3230		Normal compiler users should be unaffected.
3231
3232		----------------------------------------------------------------------
3233		Name: Matthias Blume
3234		Date: 2001/01/09 01:20:00 JST
3235		Tag: blume-20010109-rcc
3236		Description:
3237
3238		This is a fairly big patch, flushing out a large number of pending
3239		changes that I made to my development copy over the last couple of days.
3240
3241		Of practical relevance at this moment is a workaround for a pickling
3242		bug that Allen ran into the other day.  The cause of the bug itself is
3243		still unknown and it might be hard to fix it properly, but the
3244		workaround has some merits of its own (namely somewhat reducing pickling
3245		overhead for certain libraries).  Therefore, I think this solution should
3246		be satisfactory at this time.
3247
3248		The rest of the changes (i.e., the vast majority) has to do with my
3249		ongoing efforts of providing direct support for C function calls from
3250		ML.  At the moment there is a new primop "RAW_CCALL", typing magic
3251		in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new
3252		case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to
3253		translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to
3254		handle RCC.
3255
3256		The last part (the changes to mlriscGen.sml) are still known to be
3257		wrong on the x86 and not implemented on all other architectures.  But
3258		the infrastructure is in place. I had to change a few functor
3259		signatures in the backend to be able to route the CCalls interface
3260		from MLRISC there, and I had to specialize the mltree type (on the
3261		x86) to include the necessary extensions. (The extensions themselves
3262		were already there and redy to go in MLRISC/x86).
3263
3264		Everything should be very happy as soon as someone helps me with
3265		mlriscGen.sml...
3266
3267		In any case, nothing of this should matter to anyone as long as the
3268		new primop is not being used (which is going to be the case unless you
3269		find it where I hid it :). The rest of the compiler is completely
3270		unaffected.
3271
3272		----------------------------------------------------------------------
3273		Name: Matthias Blume
3274		Date: 2001/01/05 00:30:00 JST
3275		Tag: blume-20010105-primops
3276		Description:
3277
3278		Added some experimental support for work that I am doing right now.
3279		These changes mostly concern added primops, but there is also a new
3280		experimental C library in the runtime system (but currently not enabled
3281		anywhere except on Linux/X86).
3282
3283		In the course of adding primops (and playing with them), I discovered that
3284		Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly
3285		broken.  (Zhong was very right he labeled this stuff as "major gross hack".)
3286		To recover, I made type information in INL_PRIM mandatory and changed
3287		prim.sml as well as built-in.sml accordingly.  The InLine structure now
3288		has complete, correct type information (i.e., no bottom types).
3289
3290		Since all these changes mean that we need new binfiles, I also bumped the
3291		version number to 110.32.1.
3292
3293		----------------------------------------------------------------------
3294		Name: Matthias Blume
3295		Date: 2000/12/30 22:10:00 JST
3296		Tag: blume-20001230-various
3297		Description:
3298
3299		Added proxy libraries for MLRISC and let MLRISC libraries refer
3300		to each other using path anchors.  (See CM manual for explanation.)
3301
3302		Updated CM documentation.
3303
3304		Fixed some bugs in CM.
3305
3306		Implemented "proxy" libraries (= syntactic sugar for CM).
3307
3308		Added "-quiet" option to makeml and changed runtime system accordingly.
3309
3310		Added cleanup handler for exportML to reset timers and compiler stats.
3311
3312		----------------------------------------------------------------------
3313		Name: Lal George
3314		Date: 2000/12/22 22:22:58 EST 2000
3315		Tag: Release_110_32
3316		Description:
3317
3318		Infinite precision used throughout MLRISC.
3319		see MLRISC/mltree/machine-int.sig
3320
3321		----------------------------------------------------------------------
3322		Name: Matthias Blume
3323		Date: 2000/12/22 23:16:00 JST
3324		Tag: blume-20001222-warn
3325		Description:
3326
3327		Corrected wording and formatting of some CM warning message which I
3328		broke in my previous patch.
3329
3330		----------------------------------------------------------------------
3331		Name: Matthias Blume
3332		Date: 2000/12/22 21:20:00 JST
3333		Tag: blume-20001222-anchorenv
3334		Description:
3335
3336		Fixed CM's handling of anchor environments in connection with CMB.make.
3337
3338		----------------------------------------------------------------------
3339		Name: Matthias Blume
3340		Date: 2000/12/22 13:15:00 JST
3341		Tag: blume-20001222-cleanup
3342		Description:
3343
3344		Removed src/cm/ffi which does not (and did not) belong here.
3345
3346		----------------------------------------------------------------------
3347		Name: Matthias Blume
3348		Date: 2000/12/21 23:55:00 JST
3349		Tag: blume-20001221-exn
3350		Description:
3351
3352		Probably most important: CM no longer silently swallows all exceptions
3353		in the compiler.
3354		Plus: some other minor CM changes.  For example, CM now reports some
3355		sizes for generated binfiles (code, data, envpickle, lambdapickle).
3356
3357		----------------------------------------------------------------------
3358		Name: Matthias Blume
3359		Date: 2000/12/15 00:01:05 JST
3360		Tag: blume-20001215-dirtool
3361		Description:
3362
3363		- "dir" tool added.
3364		- improvements and cleanup to Tools structure
3365		- documentation updates
3366
3367		----------------------------------------------------------------------
3368		Name: Allen Leung
3369		Date: Thu Dec 14 03:45:24 EST 2000
3370		Description:
3371		Tag:  leunga-20001214-int-inf
3372		Description:
3373
3374		In IntInf, added these standard functions, which are missing from our
3375		implementation:
3376
3377		andb : int * int -> int
3378		xorb : int * int -> int
3379		orb  : int * int -> int
3380		notb : int -> int
3381		<<   : int * word -> int
3382		~>>  : int * word -> int
3383
3384		Not tested, I hope they are correct.
3385
3386		----------------------------------------------------------------------
3387		Name: Allen Leung
3388		Date: Fri Dec  8 19:23:26 EST 2000
3389		Description:
3390		Tag:  leunga-20001208-nowhere
3391		Description:
3392
3393		Slight improvements to the 'nowhere' tool to handle OR-patterns,
3394		to generate better error messages etc.  Plus a brief manual.
3395
3396		----------------------------------------------------------------------
3397		Name: Lal George
3398		Date: 2000/12/08 09:54:02 EST 2000
3399		Tag: Release_110_31
3400		Description:
3401
3402		- Version 110.31
3403		----------------------------------------------------------------------
3404		Name: Allen Leung
3405		Date: Thu Dec  7 22:01:04 EST 2000
3406		Tag:  leunga-20001207-cell-monster-hack
3407		Description:
3408
3409		Major MLRISC internal changes.  Affect all clients.
3410		Summary:
3411
3412		1.  Type CELLS.cell = int is now replaced by a datatype.
3413		As a result, the old regmap is now gone.  Almost all interfaces
3414		in MLRISC change as a consequence.
3415
3416		2.  A new brand version of machine description tool (v3.0) that generates
3417		modules expecting the new interface.  The old version is removed.
3418
3419		3.  The RA interface has been further abstracted into two new functors.
3420		RISC_RA and X86RA.  These functors have much simpler interfaces.
3421		[See also directory MLRISC/demo.]
3422
3423		4.  Some other new source->source code generation tools are available:
3424
3425		a. MLRISC/Tools/RewriteGen -- generate rewriters from rules.
3426		b. MLRISC/Tools/WhereGen -- expands conditional pattern matching rules.
3427		I use this tool to generate the peephole optimizers---with the new
3428		cell type changes, peephole rules are becoming difficult to write
3429		without conditional pattern matching.
3430
3431		5.  More Intmap -> IntHashTable change.  Previous changes by Matthias didn't
3432		cover the entire MLRISC source tree so many things broke.
3433
3434		6.  CM files have been moved to the subdirectory MLRISC/cm.
3435		They are moved because there are a lot of them and they clutter up the
3436		root dir.
3437
3438		7.  More detailed documentation to come...
3439
3440		NOTE: To rebuild from 110.30 (ftp distribution), you'll have to do
3441		a makeml -rebuild first.  This is because of other other
3442		changes that Matthias has made (see below).
3443
3444
3445		----------------------------------------------------------------------
3446		Name: Matthias Blume
3447		Date: 2000/11/30 23:12:00 JST
3448		Tag: blume-20001130-filereorg
3449		Description:
3450
3451		Some manual updates and some file reorganizations in CM.
3452
3453		----------------------------------------------------------------------
3454		Name: Matthias Blume
3455		Date: 2000/11/24 17:45:00 JST
3456		Tag: blume-20001124-link
3457		Description:
3458
3459		Drastically improved link traversal code for the case that the dynamic
3460		value was already loaded at bootstrap time.  As a result, CM and CMB
3461		now both load blazingly fast -- even on a very slow machine.  Also,
3462		memory consumption has been further reduced by this.
3463
3464		Warning: The format of the PIDMAP file has changed.  THerefore, to
3465		bootstrap you have to do this:
3466
3467		1. Run CMB.make
3468		2. Make a symbolic link for the boot directory:
3469		ln -s sml.boot.ARCH-OS xxx
3470		3. "Rebuild" the boot directory:
3471		./makeml -boot xxx -rebuild sml ; rm xxx
3472		4. Boot normally:
3473		./makeml
3474
3475		----------------------------------------------------------------------
3476		Name: Matthias Blume
3477		Date: 2000/11/21 21:20:00 JST
3478		Tag: blume-20001121-tools
3479		Description:
3480
3481		Continued hacking on autoloading problem -- with success this time.
3482		Also changed tool-plugin mechanism.  See new CM manual.
3483
3484		----------------------------------------------------------------------
3485		Name: Matthias Blume
3486		Date: 2000/11/19 14:30:00 JST
3487		Tag:  blume-20001119-autoload
3488		Description:
3489
3490		Some hacking to make autoloading faster.  Success for CMB, no success
3491		so far for CM.  There is a reduced structure CM' that autoloads faster.
3492		(This is a temporary, non-documented hack to be eliminated again when
3493		the general problem is solved.)
3494
3495		----------------------------------------------------------------------
3496		Name: Matthias Blume
3497		Date: 2000/11/17 14:10:00 JST
3498		Tag: blume-20001117-pickle-lib
3499		Description:
3500
3501		1. Eliminated comp-lib.cm
3502		2. Made pickle-lib.cm
3503		3. Eliminated all uses of intset.sml (from comp-lib.cm)
3504		4. Replaced all uses of intmap.{sig,sml} (from comp-lib.cm) with
3505		equivalent constructs from smlnj-lib.cm (INtHashTable).
3506		5. Point 4. also goes for those uses of intmap.* in MLRISC.
3507		Duplicated intmap modules thrown out.
3508		6. Hunted down all duplicated SCC code and replaced it with
3509		equivalent stuff (GraphSCCFn from smlnj-lib.cm).
3510		7. Rewrote Feedback module.
3511		8. Moved sortedlist.sml into viscomp-lib.cm.  Eventually it
3512		should be thrown out and equivalent modules from smlnj-lib.cm
3513		should be used (IntRedBlackSet, IntListSet, ...).
3514
3515		Confirmed that compiler compiles to fixpoint.
3516
3517		----------------------------------------------------------------------
3518		Name: Allen Leung
3519		Date: 2000/11/10 18:00:00
3520		Tag: leunga-20001110-new-x86-fp
3521
3522		A new x86 floating point code generator has been added.
3523		By default this is turned off.  To turn this on, do:
3524
3525		CM.autoload "$smlnj/compiler.cm";
3526		Compiler.Control.MLRISC.getFlag "x86-fast-fp" := true;
3527
3528		Changes:
3529
3530		1.  Changed FTAN to FPTAN so that the assembly output is correct.
3531		2.  Changed the extension callback for FTANGENT to generate:
3532
3533		fptan
3534		fstp  %st(0)
3535		instead of
3536		fptan
3537		fstpl ftempmem
3538
3539		3.  Numerous assembly fixes for x86.
3540
3541		5.  Cleaned up the machine code output module x86/x86MC.sml and added
3542		support for a whole bunch of instructions and addressing modes:
3543
3544		fadd/fsub/fsubr/fmul/fdiv/fdivr  %st, %st(n)
3545		faddp/fsubp/fsubrp/fmulp/fdivp/fdivrp  %st, %st(n)
3546		fadd/fsub/fsubr/fmul/fdiv/fdivr  %st(n), %st
3547		fiadd/fisub/fisubr/fimul/fidiv/fidivr mem
3548		fxch %st(n)
3549		fld %st(n)
3550		fst %st(n)
3551		fst mem
3552		fstp %st(n)
3553		fucom %st(n)
3554		fucomp %st(n)
3555
3556		All these are now generated when the fast fp mode is turned on.
3557
3558		6.  Removed the dedicated registers %st(0), ..., %st(7) from X86CpsRegs
3559
3560		----------------------------------------------------------------------
3561		Name: Matthias Blume
3562		Date: 2000/11/09 11:20:00 JST
3563		Tag: blume-20001109-scc
3564		Description:
3565
3566		Eliminated some code duplication:
3567
3568		1. Added "where" clause to GraphSCCFn in SML/NJ Library.
3569		(Otherwise the functor is useless.)
3570		2. Used GraphSCCFn where SCCUtilFun was used previously.
3571		3. Got rid of SCCUtilFun (in comp-lib.cm).
3572
3573		----------------------------------------------------------------------
3574		Name: Lal George
3575		Date: 2000/11/06 09:02:21 EST 2000
3576		Tag: Release_110_30
3577		Description:
3578
3579		- Version 110.30
3580		----------------------------------------------------------------------
3581		Name: Matthias Blume
3582		Date: 2000/11/04 14:45:00
3583		Tag: blume-20001104-mlbuild
3584		Description:
3585
3586		- Made ml-build faster on startup.
3587		- Documentation fixes.
3588
3589		----------------------------------------------------------------------
3590		Name: Matthias Blume
3591		Date: 2000/11/02 17:00:00 JST
3592		Tag: blume-20001102-condcomp
3593		Description:
3594
3595		- Small tweaks to pickler -- new BOOTFILES!
3596		- Version bumped to 110.29.2.
3597		- Added conditional compilation facility to init.cmi (see comment there).
3598		----------------------------------------------------------------------
3599		Name: Allen Leung
3600		Date: 2000/10/23 19:31:00
3601		Tag: leunga-20001023-demo-ra
3602
3603		1. Minor RA changes that improves spilling on x86 (affects Moby and C-- only)
3604		2. Test programs for the graph library updated
3605		3. Some new MLRISC demo programs added
3606
3607		----------------------------------------------------------------------
3608		Name: Matthias Blume
3609		Date: 2000/08/31 22:15:00 JST
3610		Tag: blume-20001017-errmsg
3611		Description:
3612
3613		More error message grief: Where there used to be no messages, there
3614		now were some that had bogus error regions.  Fixed.
3615
3616		----------------------------------------------------------------------
3617		Name: Matthias Blume
3618		Date: 2000/08/31 17:30:00 JST
3619		Tag: blume-20001017-v110p29p1
3620		Description:
3621
3622		I made a version 110.29.1 with new bootfiles.