Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 1136, Tue Mar 12 19:44:02 2002 UTC	revision 1335, Tue May 27 21:54:23 2003 UTC
#	Line 11	Line 11
11	Date: yyyy/mm/dd	Date: yyyy/mm/dd
12	Tag: <post-commit CVS tag>	Tag: <post-commit CVS tag>
13	Description:	Description:
14
15		----------------------------------------------------------------------
16		Name: Matthias Blume (blume@tti-c.org)
17		Date: 2003/05/27 16:55:00 CDT
18		Tag: blume-20030527-polyeq
19		Description:
20
21		Tried to eliminated most cases of polymorphic equality.
22
23		----------------------------------------------------------------------
24		Name: Matthias Blume (blume@tti-c.org)
25		Date: 2003/05/21 17:45:00 CDT
26		Tag: blume-20030517-complete
27		Description:
28
29		Two changes:
30
31		1. Added a flag for controlling whether non-exhaustive bindings will
32		be treated as errors (default is false).
33		2. Cleaned up the *entire* source tree so that CMB.make goes through
34		without a single non-exhaustive match- or bind warning.
35
36		----------------------------------------------------------------------
37		Name: Matthias Blume (blume@tti-c.org)
38		Date: 2003/05/17 10:20:00 CDT
39		Tag: blume-20030517-absyn
40		Description:
41
42		1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
43
44		This mainly affects the quality of error messages.  However, some
45		of the code is now more straightforward than before. (Treatment of
46		the above four constructs in translate.sml is much simpler than
47		the "macro-expansion" that was going on before.  Plus, the mach-
48		compiler no longer gets invoked just to be able to compile an
49		if-expression.)
50
51		2. The ErrorMsg.Error exception is now caught and absorbed by the
52		interactive loop.
53
54		----------------------------------------------------------------------
55		Name: Allen Leung
56		Date: 2003/05/16 13:05:00 CDT
57		Tag: leunga-20030516-cygwin-runtime
58		Description:
59
60		Ported the runtime system to cygwin, which uses the unix
61		x86-unix bin files.  Missing/buggy features:
62
63		o getnetbyname, getnetbyaddr: these functions seem to be missing in
64		the Cygwin library.
65		o Ctrl-C handling may be flaky.
66		o Windows system calls and Windows I/O are not supported.
67
68		A new set of binfiles is located at:
69
70		http://www.dorsai.org/~leunga/boot.x86-unix.tgz
71
72		This is only needed for bootstrapping the cygwin version of smlnj.
73		Other x86 versions can use the existing binfiles.
74
75		----------------------------------------------------------------------
76		Name: Matthias Blume
77		Date: 2003/04/08 15:42:00 CDT
78		Tag: blume-20030408-listpair
79		Description:
80
81		1. Added a target 'mlrisc' to installer.
82
83		2. Added missing elements to structure ListPair.
84
85		----------------------------------------------------------------------
86		Name: Allen Leung
87		Date: 2003/01/07 10:40:00 EST
88		Tag: leunga-20030107-int-rem
89		Description:
90
91		Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
92		The arguments to the SUBL instruction were swapped.
93
94		----------------------------------------------------------------------
95		Name: Matthias Blume
96		Date: 2002/12/12 16:25:00 EST
97		Tag: blume-20021212-risc-ra
98		Description:
99
100		Fixed a serious bug in the rewrite code for FP spilling/reloading that
101		sent the RA into an infinite loop when floating point registers get
102		spilled.  (Because of this bug, e.g., nucleic stopped compiling between
103		110.37 and 110.38.)
104		There was another set of potential problems related to the handling of
105		MLRISC annotations (but those did not yet cause real problems, apparently).
106
107		----------------------------------------------------------------------
108		Name: Matthias Blume
109		Date: 2002/12/06 22:40:00 EST
110		Tag: blume-20021206-cm-fileid
111		Description:
112
113		Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
114		This fixes the problem of CM getting confused by files that suddenly
115		change their identity (e.g., by getting unlinked and recreated by some
116		text editor such as vi).  There might be a better/cheaper/cleaner way
117		of doing this, but for now this will have to do.
118
119		----------------------------------------------------------------------
120		Name: Matthias Blume
121		Date: 2002/10/28 09:50:00 EST
122		Tag: blume-20021028-typecheck
123		Description:
124
125		Exported structure Typecheck from $smlnj/viscomp/core.cm.
126
127		----------------------------------------------------------------------
128		Name: Matthias Blume
129		Date: 2002/10/17 09:10:00 EDT
130		Tag: Release_110_42
131		Description:
132
133		In good old tradition, there has been a slight hiccup so that we have
134		to patch 110.42 after the fact.   The old release tag has been replaced
135		(see below).
136
137		The change solves a problem with two competing approaches the
138		configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
139		each other's way.
140
141		This change only affects the runtime system code and the installer script.
142		(No new bootfiles.)
143
144		----------------------------------------------------------------------
145		Name: Matthias Blume
146		Date: 2002/10/16 12:00:00 EDT
147		Tag: Release_110_42_removed
148		Description:
149
150		New working release.  New bootfiles.
151
152		----------------------------------------------------------------------
153		Name: Matthias Blume
154		Date: 2002/10/10 13:10:00 EDT
155		Tag: blume-20021010-ppc-divs
156		Description:
157
158		The mltree operator DIVS must be implemented with an overflow check on
159		the PPC because the hardware indicates divide-by-zero using "overflow" as
160		well.
161
162		----------------------------------------------------------------------
163		Name: Matthias Blume
164		Date: 2002/07/23 11:45:00
165		Tag: blume-20020723-smlnj-home
166		Description:
167
168		Sml now senses the SMLNJ_HOME environment variable.  If this is set,
169		then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
170		CM_PATHCONFIG is also set), the path configuration file is assumed
171		to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
172		the entire tree to some other place and everything will "just work".
173
174		(Companion commands such as ml-build and ml-makedepend also sense this
175		variable.)
176
177		----------------------------------------------------------------------
178		Name: Matthias Blume
179		Date: 2002/07/12 21:19:00 EDT
180		Tag: blume-20020712-liveness
181		Description:
182
183		Exported two useful "step" functions from liveness module (MLRISC).
184
185		----------------------------------------------------------------------
186		Name: Matthias Blume
187		Date: 2002/07/05 16:00 EDT
188		Tag: Release_110_41
189		Description:
190
191		New working release.  New bootfiles.
192
193		----------------------------------------------------------------------
194		Name: Matthias Blume
195		Date: 2002/07/05 10:25:00 EDT
196		Tag: blume-20020705-btimp
197		Description:
198
199		Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
200		other clients can set up backtracing support.
201
202		----------------------------------------------------------------------
203		Name: Matthias Blume
204		Date: 2002/06/25 17:23:00 EDT
205		Tag: blume-20020625-fpmax
206		Description:
207
208		Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
209		case.  (The sense of the isNaN test was reversed -- which made min and
210		max always return their first argument.)
211
212		----------------------------------------------------------------------
213		Name: Matthias Blume
214		Date: 2002/06/11
215		Tag: blume-20020611-unixpath
216		Description:
217
218		Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
219
220		----------------------------------------------------------------------
221		Name: Matthias Blume
222		Date: 2002/06/10 16:35:00 EDT
223		Tag: blume-20020610-ieeereal
224		Description:
225
226		I back-ported my implementation of IEEEReal.fromString from the
227		idlbasis-devel branch so that we can test it.
228
229		Another small change is that ppDec tries to give more information
230		than just "<sig>" in the case of functors.  However, this code is
231		broken in some mysterious way if the functor's body's signature
232		has not been declared by ascription but gets inferred from the
233		implementation.  This needs fixing...
234
235		----------------------------------------------------------------------
236		Name: Matthias Blume
237		Date: 2002/05/31
238		Tag: blume-20020531-btrace-mode
239		Description:
240
241		Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
242		the wayside when I switched over to using Controls everywhere.)
243
244		----------------------------------------------------------------------
245		Name: Lal George
246		Date: 2002/05/23 12:21:40 EDT
247		Tag: george-20020523-visual-labels
248		Description:
249
250		Labels are now displayed in the graphical output to make
251		the fall-through and target blocks obvious.
252
253		----------------------------------------------------------------------
254		Name: Matthias Blume
255		Date: 2002/05/22 11:03:00 EDT
256		Tag: blume-20020522-shrink
257		Description:
258
259		John tweaked yesterday's fix for 1131 to handle an out-of-memory
260		situation that comes up when allocating huge arrays.
261
262		----------------------------------------------------------------------
263		Name: Matthias Blume
264		Date: 2002/05/21 16:00:00 EDT
265		Tag: Release_110_40
266		Description:
267
268		New working release (110.40).  New bootfiles.
269
270		[Also: John Reppy fixed GC bug 1131.]
271
272		----------------------------------------------------------------------
273		Name: Matthias Blume
274		Date: 2002/05/21 12:35:00 EDT
275		Tag: blume-20020521-cmdoc
276		Description:
277
278		CM documentation update.
279
280		----------------------------------------------------------------------
281		Name: Matthias Blume
282		Date: 2002/05/21 10:55:00 EDT
283		Tag: blume-20020521-misc
284		Description:
285
286		- John tweaked runtime to be silent on heap export (except when
287		GC messages are on).
288		- I added a few more things (cross-compiling versions of CMB) to
289		config/preloads (as suggestions).
290
291		----------------------------------------------------------------------
292		Name: Matthias Blume
293		Date: 2002/05/20 22:25:00 EDT
294		Tag: blume-20020520-controls
295		Description:
296
297		- Added ControlUtil structure to control-lib.cm.
298		- Use it throughout.
299		- Used Controls facility to define MLRISC controls (as opposed to
300		registering MLRISC control ref cells with Controls after the fact)
301		- Fixed messed-up controls priorities.
302
303		* Removed again all the stuff from config/preloads that one wouldn't
304		be able to preload at the time the initial heap image is built.
305		(Many libraries, e.g., CML, do not exist yet at this time.  The
306		only libraries that can be preloaded via config/preloads are those
307		that come bundled with the bootfiles.)
308
309		----------------------------------------------------------------------
310		Name: Matthias Blume
311		Date: 2002/05/20 10:59:00 EDT
312		Tag: blume-20020520-preloads
313		Description:
314
315		Added a lot of commented-out suggestions for things to be included
316		in config/preloads.
317
318		----------------------------------------------------------------------
319		Name: Allen Leung
320		Date: 2002/05/18 14:20:00 EDT
321		Tag: leunga-20020518-mdl
322		Description:
323
324		o Made the mdl tool stuff compile and run again.
325		o I've disabled all the stuff that depends on RTL specifications; they
326		are all badly broken anyway.
327
328		----------------------------------------------------------------------
329		Name: Matthias Blume
330		Date: 2002/05/17 16:49:00 EDT
331		Tag: blume-20020517-controls
332		Description:
333
334		0. John Reppy made several modifications to the SML/NJ library.
335		In particular, there is a shiny new controls-lib.cm.
336
337		1. Pushed new controls interface through compiler so that everything
338		compiles again.
339
340		2. Added FormatComb and FORMAT_COMB to the CML version of the
341		SML/NJ library (so that CML compiles again).
342
343		3. Modified init scripts because XXX_DEFAULT environment variables
344		are no longer with us.  (Boot-time initialization is now done
345		using the same environment variables that are also used for
346		startup-time initialization of controls.)
347
348		----------------------------------------------------------------------
349		Name: Lal George
350		Date: 2002/05/15 09:20:10 EDT
351		Tag: george-20020515-pseudo-op-decls
352		Description:
353
354		All pseudo-ops emitted before the first segment declaration
355		such as TEXT, DATA, and BSS directives are assumed to be global
356		declarations and are emitted first in the assembly file. This is
357		useful in a number of situations where one has pseudo-ops that are not
358		specific to any segment, and also works around the constraint that one
359		cannot have client pseudo-ops in the TEXT segment.
360
361		Because no segment is associated with these declarations it is
362		an error to allocate any space or objects before the first segment
363		directive and an exception will be raised. However, we cannot make
364		this check for client pseudo-ops.
365
366		These top level declarations are a field in the CFG graph_info.
367		In theory you can continue to add to this field after the CFG has been
368		built -- provided you know what you are doing;-)
369
370		----------------------------------------------------------------------
371		Name: Matthias Blume
372		Date: 2002/05/13 16:40:00 EDT
373		Tag: blume-20020513-pp-etc
374		Description:
375
376		A few minor bugfixes:
377
378		- Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
379		(Bogus printouts for redefined bindings still occur.  Compiler
380		bug should no longer occur now.  We need to redo the prettyprinter
381		from scratch.)
382
383		- CM pathname printer now also adds escape sequences for ( and )
384
385		- commend and docu fixes for ml-nlffi
386
387		----------------------------------------------------------------------
388		Name: Matthias Blume
389		Date: 2002/05/10 16:40:00 EDT
390		Tag: blume-20020510-erg-textio
391		Description:
392
393		Applied the following bugfix provided by Emden Gansner:
394
395		Output is corrupted when outputSubstr is used rather than output.
396
397		The problem occurs when a substring
398
399		ss = (s, dataStart, dataLen)
400
401		where dataStart > 0, fills a stream buffer with avail bytes left.
402		avail bytes of s, starting at index dataStart, are copied into the
403		buffer, the buffer is flushed, and then the remaining dataLen-avail
404		bytes of ss are copied into the beginning of the buffer. Instead of
405		starting this copy at index dataStart+avail in s, the current code
406		starts the copy at index avail.
407
408		Fix:
409		In text-io-fn.sml, change line 695 from
410		val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
411		to
412		val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
413
414		----------------------------------------------------------------------
415		Name: Matthias Blume
416		Date: 2002/04/12 13:55:00 EDT
417		Tag: blume-20020412-assyntax
418		Description:
419
420		1. Grabbed newer assyntax.h from the XFree86 project.
421		2. Fiddled with how to compile X86.prim.asm without warnings.
422		3. (Very) Minor cleanup in CM.
423
424		----------------------------------------------------------------------
425		Name: Matthias Blume
426		Date: 2002/04/01 (no joke!) 17:07:00 EST
427		Tag: blume-20020401-x86div
428		Description:
429
430		Added full support for div/mod/rem/quot on the x86, using the machine
431		instruction's two results (without clumsily recomputing the remainder)
432		directly where appropriate.
433
434		Some more extensive power-of-two support was added to the x86 instruction
435		selector (avoiding expensive divs, mods, and muls where they can be
436		replaced with cheaper shifts and masks).  However, this sort of thing
437		ought to be done earlier, e.g., within the CPS optimizer so that
438		all architectures benefit from it.
439
440		The compiler compiles to a fixed point, but changes might be somewhat
441		fragile nevertheless.  Please, report any strange things that you might
442		see wrt. div/mod/quot/rem...
443
444		----------------------------------------------------------------------
445		Name: Matthias Blume
446		Date: 2002/03/29 17:22:00
447		Tag: blume-20020329-div
448		Description:
449
450		Fixed my broken div/mod logic.  Unfortunately, this means that the
451		inline code for div/mod now has one more comparison than before.
452		Fast paths (quotient > 0 or remainder = 0) are not affected, though.
453		The problem was with quotient = 0, because that alone does not tell
454		us which way the rounding went.  One then has to look at whether
455		remainder and divisor have the same sign...  :(
456
457		Anyway, I replaced the bootfiles with fresh ones...
458
459		----------------------------------------------------------------------
460		Name: Matthias Blume
461		Date: 2002/03/29 14:10:00 EST
462		Tag: blume-20020329-inlprims
463		Description:
464
465		NEW BOOTFILES!!!    Version number bumped to 110.39.3.
466
467		Primops have changed. This means that the bin/boot-file formats have
468		changed as well.
469
470		To make sure that there is no confusion, I made a new version.
471
472
473		CHANGES:
474
475		* removed REMT from mltree (remainder should never overflow).
476
477		* added primops to deal with divisions of all flavors to the frontend
478
479		* handled these primops all the way through so they map to their respective
480		MLRISC support
481
482		* used these primops in the implementation of Int, Int32, Word, Word32
483
484		* removed INLDIV, INLMOD, and INLREM as they are no longer necessary
485
486		* parameterized INLMIN, INLMAX, and INLABS by a numkind
487
488		* translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
489		floating point
490
491		* used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
492		and Real (but Real.abs maps to a separate floating-point-only primop)
493
494
495		TODO items:
496
497		* Hacked Alpha32 instruction selection, disabling the selection of REMx
498		instructions because the machine instruction encoder cannot handle
499		them.  (Hppa, PPC, and Sparc instruction selection did not handle
500		REM in the first place, and REM is supported by the x86 machine coder.)
501
502		* Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
503		selection phase.  (The two can be streamlined because the hardware
504		delivers both quotient and remainder at the same time anyway.)
505
506		* Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
507		(Currently the behavior is inconsistent both across architectures and
508		wrt. the draft Basis spec.)
509
510		* Word8 should eventually be handled natively, too.
511
512		* There seems to be one serious bug in mltree-gen.sml.  It appears, though,
513		as if there currently is no execution path that could trigger it in
514		SML/NJ.  (The assumptions underlying functions arith and promotable do not
515		hold for things like multiplication and division.)
516
517		----------------------------------------------------------------------
518		Name: Matthias Blume
519		Date: 2002/03/27 16:27:00 EST
520		Tag: blume-20020327-mlrisc-divisions
521		Description:
522
523		Added support for all four division operations (ML's div, mod, quot,
524		and rem) to MLRISC.  In the course of doing so, I also rationalized
525		the naming (no more annoying switch-around of DIV and QUOT), by
526		parameterizing the operation by div_rounding_mode (which can be either
527		DIV_TO_ZERO or DIV_TO_NEGINF).
528
529		The generic MLTreeGen functor takes care of compiling all four
530		operations down to only round-to-zero div.
531
532		Missing pieces:
533
534		* Doing something smarter than relying on MLTreeGen on architectures
535		like, e.g., the x86 where hardware division delivers both quotient and
536		remainder at the same time.  With this, the implementation of the
537		round-to-neginf operations could be further streamlined.
538
539		* Remove inlining support for div/mod/rem from the frontend and replace it
540		with primops that get carried through to the backend.  Do this for all
541		int and word types.
542
543		----------------------------------------------------------------------
544		Name: Matthias Blume
545		Date: 2002/03/25 17:25:00 EST
546		Tag: blume-20020325-divmod
547		Description:
548
549		I improved (hopefully without breaking them) the implementation of Int.div,
550		Int.mod, and Int.rem.   For this, the code in translate.sml now takes
551		advantage of the following observations:
552
553		Let  q = x quot y      r = x rem y
554		d = x div  y      m = x mod y
555
556		where "quot" is the round-to-zero version of integer division that
557		hardware usually provides.  Then we have:
558
559		r = x - q * y        where neither the * nor the - will overflow
560		d = if q >= 0 orelse x = q * y then q else q - 1
561		where neither the * nor the - will overflow
562		m = if q >= 0 orelse r = 0 then r else r + y
563		where the + will not overflow
564
565		This results in substantial simplification of the generated code.
566		The following table shows the number of CFG nodes and edges generated
567		for
568		fun f (x, y) = x OPER y
569		(* with OPER \in div, mod, quot, rem *)
570
571
572		OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
573		--------------------------------------------------------
574		div |         24 |         39 |         12 |         16
575		mod |         41 |         71 |         12 |         16
576		quot |          8 |         10 |          8 |         10
577		rem |         10 |         14 |          8 |         10
578
579
580		----------------------------------------------------------------------
581		Name: Matthias Blume
582		Date: 2002/03/25 22:06:00 EST
583		Tag: blume-20020325-cprotobug
584		Description:
585
586		Fixed a bug in cproto (c prototype decoder).
587
588		----------------------------------------------------------------------
589		Name: Matthias Blume
590		Date: 2002/03/25 16:00:00 EST
591		Tag: blume-20020325-raw-primops
592		Description:
593
594		I did some cleanup to Allen's new primop code and
595		replaced yesterday's bootfiles with new ones.
596		(But they are stored in the same place.)
597
598		----------------------------------------------------------------------
599		Name: Matthias Blume
600		Date: 2002/03/24 22:40:00 EST
601		Tag: blume-20020324-bootfiles
602		Description:
603
604		Made the bootfiles that Allen asked for.
605
606		----------------------------------------------------------------------
607		Name: Allen Leung
608		Date: 2002/03/23 15:50:00 EST
609		Tag: leunga-20020323-flint-cps-rcc-primops
610		Description:
611
612		1. Changes to FLINT primops:
613
614		(* make a call to a C-function;
615		* The primop carries C function prototype information and specifies
616		* which of its (ML-) arguments are floating point. C prototype
617		* information is for use by the backend, ML information is for
618		* use by the CPS converter. *)
619		| RAW_CCALL of { c_proto: CTypes.c_proto,
620		ml_args: ccall_type list,
621		ml_res_opt: ccall_type option,
622		reentrant : bool
623		} option
624		(* Allocate uninitialized storage on the heap.
625		* The record is meant to hold short-lived C objects, i.e., they
626		* are not ML pointers.  With the tag, the representation is
627		* the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
628		*)
629		| RAW_RECORD of {tag:bool,sz:int}
630		and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
631
632		2.  These CPS primops are now overloaded:
633
634		rawload of {kind:numkind}
635		rawstore of {kind:numkind}
636
637		The one argument form is:
638
639		rawload {kind} address
640
641		The two argument form is:
642
643		rawload {kind} [ml object, byte-offset]
644
645		3. RAW_CCALL/RCC now takes two extra arguments:
646
647		a. The first is whether the C call is reentrant, i.e., whether
648		ML state should be saved and restored.
649		b. The second argument is a string argument specifying the name of
650		library and the C function.
651
652		These things are currently not handled in the code generator, yet.
653
654		4. In CProto,
655
656		An encoding type of "bool" means "ml object" and is mapped into
657		C prototype of PTR.  Note that "bool" is different than "string",
658		even though "string" is also mapped into PTR, because "bool"
659		is assigned an CPS type of BOGt, while "string" is assigned INT32t.
660
661		5. Pickler/unpicker
662
663		Changed to handle RAW_RECORD and newest RAW_CCALL
664
665		6. MLRiscGen,
666
667		1. Changed to handle the new rawload/rawstore/rawrecord operators.
668		2. Code for handling C Calls has been moved to a new module CPSCCalls,
669		in the file CodeGen/cpscompile/cps-c-calls.sml
670
671		7. Added the conditional move operator
672
673		condmove of branch
674
675		to cps.  Generation of this is still buggy so it is currently
676		disabled.
677
678		----------------------------------------------------------------------
679		Name: Lal George
680		Date: 2002/03/22 14:18:25 EST
681		Tag: george-20020322-cps-branch-prob
682		Description:
683
684		Implemented the Ball-Larus branch prediction-heuristics, and
685		incorporated graphical viewers for control flow graphs.
686
687		Ball-Larus Heuristics:
688		---------------------
689		See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
690
691		By design it uses the Dempster-Shafer theory for combining
692		probabilities.  For example, in the function:
693
694		fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
695
696		the ball-larus heuristics predicts that the n=0 is unlikely
697		(OH-heuristic), and the 'then' branch is unlikely because of the
698		RH-heuristic -- giving the 'then' branch an even lower combined
699		probability using the Dempster-Shafer theory.
700
701		Finally, John Reppy's loop analysis in MLRISC, further lowers the
702		probability of the 'then' branch because of the loop in the else
703		branch.
704
705
706		Graphical Viewing:
707		------------------
708		I merely plugged in Allen's graphical viewers into the compiler. The
709		additional code is not much. At the top level, saying:
710
711		Control.MLRISC.getFlag "cfg-graphical-view" := true;
712
713		will display the graphical view of the control flow graph just before
714		back-patching.  daVinci must be in your path for this to work. If
715		daVinci is not available, then the default viewer can be changed
716		using:
717
718		Control.MLRISC.getString "viewer"
719
720		which can be set to "dot" or "vcg" for the corresponding viewers. Of
721		course, these viewers must be in your path.
722
723		The above will display the compilation unit at the level of clusters,
724		many of which are small, boring, and un-interesting. Also setting:
725
726		Control.MLRISC.getInt "cfg-graphical-view_size"
727
728		will display clusters that are larger than the value set by the above.
729
730
731		----------------------------------------------------------------------
732		Name: Matthias Blume
733		Date: 2002/03/21 22:20:00 EST
734		Tag: blume-20020321-kmp-bugfix
735		Description:
736
737		Changed the interface to the KMP routine in PreString and fixed
738		a minor bug in one place where it was used.
739
740		----------------------------------------------------------------------
741		Name: Allen Leung
742		Date: 2002/03/21 20:30:00 EST
743		Tag: leunga-20020321-cfg
744		Description:
745
746		Fixed a potential problem in cfg edge splitting.
747
748		----------------------------------------------------------------------
749		Name: Allen Leung
750		Date: 2002/03/21 17:15:00 EST
751		Tag: leunga-20020321-x86-fp-cfg
752		Description:
753
754		1. Recoded the buggy parts of x86-fp.
755
756		a. All the block reordering code has been removed.
757		We now depend on the block placement phases to do this work.
758
759		b. Critical edge splitting code has been simplified and moved into the
760		CFG modules, as where they belong.
761
762		Both of these were quite buggy and complex.  The code is now much, much
763		simpler.
764
765		2. X86 backend.
766
767		a. Added instructions for 64-bit support.  Instruction selection for
768		64-bit has not been committed, however, since that
769		requires changes to MLTREE which haven't been approved by
770		Lal and John.
771
772		b. Added support for FUCOMI and FUCOMIP when generating code for
773		PentiumPro and above.  We only generate these instructions in
774		the fast-fp mode.
775
776		c. Added cases for JP and JNP in X86FreqProps.
777
778		3. CFG
779
780		CFG now has a bunch of methods for edge splitting and merging.
781
782		4. Machine description.
783
784		John's simplification of MLTREE_BASIS.fcond broke a few machine
785		description things:
786
787		rtl-build.{sig,sml} and hppa.mdl fixed.
788
789		NOTE: the machine description stuff in the repository is still broken.
790		Again, I can't put my fixes in because that involves
791		changes to MLTREE.
792
793		----------------------------------------------------------------------
794		Name: Matthias Blume
795		Date: 2002/03/20 15:55:00 EST
796		Tag: blume-20020320-kmp
797		Description:
798
799		Implemented Knuth-Morris-Pratt string matching in PreString and used
800		it for String.isSubstring, Substring.isSubstring, and
801		Substring.position.
802
803		(Might need some stress-testing.  Simple examples worked fine.)
804
805		----------------------------------------------------------------------
806		Name: Matthias Blume
807		Date: 2002/03/19 16:37:00 EST
808		Tag: blume-20020319-witnesses
809		Description:
810
811		Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
812
813		This implements a generic mechanism for changing constness qualifiers
814		anywhere within big C types without resorting to outright "casts".
815		(So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
816		modify the constness at the outermost level.)
817		The implementation of "convert" is based on the idea of "witness"
818		values -- values that are not used by the operation but whose types
819		"testify" to their applicability.  On the implementation side, "convert"
820		is simply a projection (returning its second curried argument).  With
821		cross-module inlining, it should not result in any machine code being
822		generated.
823
824		----------------------------------------------------------------------
825		Name: Matthias Blume
826		Date: 2002/03/15 16:40:00 EST
827		Tag: blume-20020315-basis
828		Description:
829
830		Provided (preliminary?) implementations for
831
832		{String,Substring}.{concatWith,isSuffix,isSubstring}
833
834		and
835
836		Substring.full
837
838		Those are in the Basis spec but they were missing in SML/NJ.
839
840		----------------------------------------------------------------------
841		Name: Matthias Blume
842		Date: 2002/03/14 21:30:00 EST
843		Tag: blume-20020314-controls
844		Description:
845
846		Controls:
847		---------
848
849		1. Factored out the recently-added Controls : CONTROLS stuff and put
850		it into its own library $/controls-lib.cm.  The source tree for
851		this is under src/smlnj-lib/Controls.
852
853		2. Changed the names of types and functions in this interface, so they
854		make a bit more "sense":
855
856		module -> registry
857		'a registry -> 'a group
858
859		3. The interface now deals in ref cells only.  The getter/setter interface
860		is (mostly) gone.
861
862		4. Added a function that lets one register an already-existing ref cell.
863
864		5. Made the corresponding modifications to the rest of the code so that
865		everything compiles again.
866
867		6. Changed the implementation of Controls.MLRISC back to something closer
868		to the original.  In particular, this module (and therefore MLRISC)
869		does not depend on Controls.  There now is some link-time code in
870		int-sys.sml that registers the MLRISC controls with the Controls
871		module.
872
873		CM:
874		---
875
876		* One can now specify the lambda-split aggressiveness in init.cmi.
877
878		----------------------------------------------------------------------
879		Name: Allen Leung
880		Date: 2002/03/13 17:30:00 EST
881		Tag: leunga-20020313-x86-fp-unary
882		Description:
883
884		Bug fix for:
885
886		> leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
887		> Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
888		> - fun f(x,(y,z)) = Real.~ y;
889		> [autoloading]
890		> [autoloading done]
891		>       fchsl   (%eax), 184(%esp)
892		> Error: MLRisc bug: X86MCEmitter.emitInstr
893		>
894		> uncaught exception Error
895		>   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
896
897		The problem was that the code generator did not generate any fp registers
898		in this case, and the ra didn't know that it needed to run the X86FP phase to
899		translate the pseudo fp instruction.   This only happened with unary fp
900		operators in certain situations.
901
902		----------------------------------------------------------------------
903		Name: Matthias Blume
904		Date: 2002/03/13 14:00:00 EST
905		Tag: blume-20020313-overload-etc
906		Description:
907
908		1. Added _overload as a synonym for overload for backward compatibility.
909		(Control.overloadKW must be true for either version to be accepted.)
910
911		2. Fixed bug in install script that caused more things to be installed
912		than what was requested in config/targets.
913
914		3. Made CM aware of the (_)overload construct so that autoloading
915		works.
916
917		----------------------------------------------------------------------
918		Name: Matthias Blume
919		Date: 2002/03/12 22:03:00 EST
920		Tag: blume-20020312-url
921		Description:
922
923		Forgot to update BOOT and srcarchiveurl.
924
925		----------------------------------------------------------------------
926		Name: Matthias Blume
927		Date: 2002/03/12 17:30:00 EST
928		Tag: blume-20020312-version110392
929		Description:
930
931		Yet another version number bump (because of small changes to the
932		binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
933
934		Changes:
935
936		The new pid generation scheme described a few weeks ago was overly
937		complicated.  I implemented a new mechanism that is simpler and
938		provides a bit more "stability":  Once CM has seen a compilation
939		unit, it keeps its identity constant (as long as you do not delete
940		those crucial CM/GUID/* files).  This means that when you change
941		an interface, compile, then go back to the old interface, and
942		compile again, you arrive at the original pid.
943
944		There now also is a mechanism that instructs CM to use the plain
945		environment hash as a module's pid (effectively making its GUID
946		the empty string).  For this, "noguid" must be specified as an
947		option to the .sml file in question within its .cm file.
948		This is most useful for code that is being generated by tools such
949		as ml-nlffigen (because during development programmers tend to
950		erase the tool's entire output directory tree including CM's cached
951		GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
952		revert to the old, broken behavior of SML/NJ, but in specific cases
953		where there is no danger of interface confusion, its use is ok
954		(I think).
955
956		ml-nlffigen by default generates "noguid" annotations.  They can be
957		turned off by specifying -guid in its command line.
958
959	----------------------------------------------------------------------	----------------------------------------------------------------------
960	Name: Lal George	Name: Lal George
961	Date: 2002/03/12 12 14:42:36 EST	Date: 2002/03/12 12 14:42:36 EST
#	Line 68	Line 1013
1013	----------------------------------------------------------------------	----------------------------------------------------------------------
1014	Name: Allen Leung	Name: Allen Leung
1015	Date: 2002/03/11 10:30:00 EST	Date: 2002/03/11 10:30:00 EST
1016	Tag: leunga-20020310-runtime-string0	Tag: leunga-20020311-runtime-string0
1017	Description:	Description:
1018
1019	The representation of the empty string now points to a	The representation of the empty string now points to a

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.1136   changed lines   Added in v.1335

---

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