Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 1003, Fri Dec 7 02:45:32 2001 UTC	revision 1212, Tue May 21 16:34:29 2002 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
17		Date: 2002/05/21 12:35:00 EDT
18		Tag: blume-20020521-cmdoc
19		Description:
20
21		CM documentation update.
22
23		----------------------------------------------------------------------
24		Name: Matthias Blume
25		Date: 2002/05/21 10:55:00 EDT
26		Tag: blume-20020521-misc
27		Description:
28
29		- John tweaked runtime to be silent on heap export (except when
30		GC messages are on).
31		- I added a few more things (cross-compiling versions of CMB) to
32		config/preloads (as suggestions).
33
34		----------------------------------------------------------------------
35		Name: Matthias Blume
36		Date: 2002/05/20 22:25:00 EDT
37		Tag: blume-20020520-controls
38		Description:
39
40		- Added ControlUtil structure to control-lib.cm.
41		- Use it throughout.
42		- Used Controls facility to define MLRISC controls (as opposed to
43		registering MLRISC control ref cells with Controls after the fact)
44		- Fixed messed-up controls priorities.
45
46		* Removed again all the stuff from config/preloads that one wouldn't
47		be able to preload at the time the initial heap image is built.
48		(Many libraries, e.g., CML, do not exist yet at this time.  The
49		only libraries that can be preloaded via config/preloads are those
50		that come bundled with the bootfiles.)
51
52		----------------------------------------------------------------------
53		Name: Matthias Blume
54		Date: 2002/05/20 10:59:00 EDT
55		Tag: blume-20020520-preloads
56		Description:
57
58		Added a lot of commented-out suggestions for things to be included
59		in config/preloads.
60
61		----------------------------------------------------------------------
62		Name: Allen Leung
63		Date: 2002/05/18 14:20:00 EDT
64		Tag: leunga-20020518-mdl
65		Description:
66
67		o Made the mdl tool stuff compile and run again.
68		o I've disabled all the stuff that depends on RTL specifications; they
69		are all badly broken anyway.
70
71		----------------------------------------------------------------------
72		Name: Matthias Blume
73		Date: 2002/05/17 16:49:00 EDT
74		Tag: blume-20020517-controls
75		Description:
76
77		0. John Reppy made several modifications to the SML/NJ library.
78		In particular, there is a shiny new controls-lib.cm.
79
80		1. Pushed new controls interface through compiler so that everything
81		compiles again.
82
83		2. Added FormatComb and FORMAT_COMB to the CML version of the
84		SML/NJ library (so that CML compiles again).
85
86		3. Modified init scripts because XXX_DEFAULT environment variables
87		are no longer with us.  (Boot-time initialization is now done
88		using the same environment variables that are also used for
89		startup-time initialization of controls.)
90
91		----------------------------------------------------------------------
92		Name: Lal George
93		Date: 2002/05/15 09:20:10 EDT
94		Tag: george-20020515-pseudo-op-decls
95		Description:
96
97		All pseudo-ops emitted before the first segment declaration
98		such as TEXT, DATA, and BSS directives are assumed to be global
99		declarations and are emitted first in the assembly file. This is
100		useful in a number of situations where one has pseudo-ops that are not
101		specific to any segment, and also works around the constraint that one
102		cannot have client pseudo-ops in the TEXT segment.
103
104		Because no segment is associated with these declarations it is
105		an error to allocate any space or objects before the first segment
106		directive and an exception will be raised. However, we cannot make
107		this check for client pseudo-ops.
108
109		These top level declarations are a field in the CFG graph_info.
110		In theory you can continue to add to this field after the CFG has been
111		built -- provided you know what you are doing;-)
112
113		----------------------------------------------------------------------
114		Name: Matthias Blume
115		Date: 2002/05/13 16:40:00 EDT
116		Tag: blume-20020513-pp-etc
117		Description:
118
119		A few minor bugfixes:
120
121		- Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
122		(Bogus printouts for redefined bindings still occur.  Compiler
123		bug should no longer occur now.  We need to redo the prettyprinter
124		from scratch.)
125
126		- CM pathname printer now also adds escape sequences for ( and )
127
128		- commend and docu fixes for ml-nlffi
129
130		----------------------------------------------------------------------
131		Name: Matthias Blume
132		Date: 2002/05/10 16:40:00 EDT
133		Tag: blume-20020510-erg-textio
134		Description:
135
136		Applied the following bugfix provided by Emden Gansner:
137
138		Output is corrupted when outputSubstr is used rather than output.
139
140		The problem occurs when a substring
141
142		ss = (s, dataStart, dataLen)
143
144		where dataStart > 0, fills a stream buffer with avail bytes left.
145		avail bytes of s, starting at index dataStart, are copied into the
146		buffer, the buffer is flushed, and then the remaining dataLen-avail
147		bytes of ss are copied into the beginning of the buffer. Instead of
148		starting this copy at index dataStart+avail in s, the current code
149		starts the copy at index avail.
150
151		Fix:
152		In text-io-fn.sml, change line 695 from
153		val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
154		to
155		val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
156
157		----------------------------------------------------------------------
158		Name: Matthias Blume
159		Date: 2002/04/12 13:55:00 EDT
160		Tag: blume-20020412-assyntax
161		Description:
162
163		1. Grabbed newer assyntax.h from the XFree86 project.
164		2. Fiddled with how to compile X86.prim.asm without warnings.
165		3. (Very) Minor cleanup in CM.
166
167		----------------------------------------------------------------------
168		Name: Matthias Blume
169		Date: 2002/04/01 (no joke!) 17:07:00 EST
170		Tag: blume-20020401-x86div
171		Description:
172
173		Added full support for div/mod/rem/quot on the x86, using the machine
174		instruction's two results (without clumsily recomputing the remainder)
175		directly where appropriate.
176
177		Some more extensive power-of-two support was added to the x86 instruction
178		selector (avoiding expensive divs, mods, and muls where they can be
179		replaced with cheaper shifts and masks).  However, this sort of thing
180		ought to be done earlier, e.g., within the CPS optimizer so that
181		all architectures benefit from it.
182
183		The compiler compiles to a fixed point, but changes might be somewhat
184		fragile nevertheless.  Please, report any strange things that you might
185		see wrt. div/mod/quot/rem...
186
187		----------------------------------------------------------------------
188		Name: Matthias Blume
189		Date: 2002/03/29 17:22:00
190		Tag: blume-20020329-div
191		Description:
192
193		Fixed my broken div/mod logic.  Unfortunately, this means that the
194		inline code for div/mod now has one more comparison than before.
195		Fast paths (quotient > 0 or remainder = 0) are not affected, though.
196		The problem was with quotient = 0, because that alone does not tell
197		us which way the rounding went.  One then has to look at whether
198		remainder and divisor have the same sign...  :(
199
200		Anyway, I replaced the bootfiles with fresh ones...
201
202		----------------------------------------------------------------------
203		Name: Matthias Blume
204		Date: 2002/03/29 14:10:00 EST
205		Tag: blume-20020329-inlprims
206		Description:
207
208		NEW BOOTFILES!!!    Version number bumped to 110.39.3.
209
210		Primops have changed. This means that the bin/boot-file formats have
211		changed as well.
212
213		To make sure that there is no confusion, I made a new version.
214
215
216		CHANGES:
217
218		* removed REMT from mltree (remainder should never overflow).
219
220		* added primops to deal with divisions of all flavors to the frontend
221
222		* handled these primops all the way through so they map to their respective
223		MLRISC support
224
225		* used these primops in the implementation of Int, Int32, Word, Word32
226
227		* removed INLDIV, INLMOD, and INLREM as they are no longer necessary
228
229		* parameterized INLMIN, INLMAX, and INLABS by a numkind
230
231		* translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
232		floating point
233
234		* used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
235		and Real (but Real.abs maps to a separate floating-point-only primop)
236
237
238		TODO items:
239
240		* Hacked Alpha32 instruction selection, disabling the selection of REMx
241		instructions because the machine instruction encoder cannot handle
242		them.  (Hppa, PPC, and Sparc instruction selection did not handle
243		REM in the first place, and REM is supported by the x86 machine coder.)
244
245		* Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
246		selection phase.  (The two can be streamlined because the hardware
247		delivers both quotient and remainder at the same time anyway.)
248
249		* Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
250		(Currently the behavior is inconsistent both across architectures and
251		wrt. the draft Basis spec.)
252
253		* Word8 should eventually be handled natively, too.
254
255		* There seems to be one serious bug in mltree-gen.sml.  It appears, though,
256		as if there currently is no execution path that could trigger it in
257		SML/NJ.  (The assumptions underlying functions arith and promotable do not
258		hold for things like multiplication and division.)
259
260		----------------------------------------------------------------------
261		Name: Matthias Blume
262		Date: 2002/03/27 16:27:00 EST
263		Tag: blume-20020327-mlrisc-divisions
264		Description:
265
266		Added support for all four division operations (ML's div, mod, quot,
267		and rem) to MLRISC.  In the course of doing so, I also rationalized
268		the naming (no more annoying switch-around of DIV and QUOT), by
269		parameterizing the operation by div_rounding_mode (which can be either
270		DIV_TO_ZERO or DIV_TO_NEGINF).
271
272		The generic MLTreeGen functor takes care of compiling all four
273		operations down to only round-to-zero div.
274
275		Missing pieces:
276
277		* Doing something smarter than relying on MLTreeGen on architectures
278		like, e.g., the x86 where hardware division delivers both quotient and
279		remainder at the same time.  With this, the implementation of the
280		round-to-neginf operations could be further streamlined.
281
282		* Remove inlining support for div/mod/rem from the frontend and replace it
283		with primops that get carried through to the backend.  Do this for all
284		int and word types.
285
286		----------------------------------------------------------------------
287		Name: Matthias Blume
288		Date: 2002/03/25 17:25:00 EST
289		Tag: blume-20020325-divmod
290		Description:
291
292		I improved (hopefully without breaking them) the implementation of Int.div,
293		Int.mod, and Int.rem.   For this, the code in translate.sml now takes
294		advantage of the following observations:
295
296		Let  q = x quot y      r = x rem y
297		d = x div  y      m = x mod y
298
299		where "quot" is the round-to-zero version of integer division that
300		hardware usually provides.  Then we have:
301
302		r = x - q * y        where neither the * nor the - will overflow
303		d = if q >= 0 orelse x = q * y then q else q - 1
304		where neither the * nor the - will overflow
305		m = if q >= 0 orelse r = 0 then r else r + y
306		where the + will not overflow
307
308		This results in substantial simplification of the generated code.
309		The following table shows the number of CFG nodes and edges generated
310		for
311		fun f (x, y) = x OPER y
312		(* with OPER \in div, mod, quot, rem *)
313
314
315		OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
316		--------------------------------------------------------
317		div |         24 |         39 |         12 |         16
318		mod |         41 |         71 |         12 |         16
319		quot |          8 |         10 |          8 |         10
320		rem |         10 |         14 |          8 |         10
321
322
323		----------------------------------------------------------------------
324		Name: Matthias Blume
325		Date: 2002/03/25 22:06:00 EST
326		Tag: blume-20020325-cprotobug
327		Description:
328
329		Fixed a bug in cproto (c prototype decoder).
330
331		----------------------------------------------------------------------
332		Name: Matthias Blume
333		Date: 2002/03/25 16:00:00 EST
334		Tag: blume-20020325-raw-primops
335		Description:
336
337		I did some cleanup to Allen's new primop code and
338		replaced yesterday's bootfiles with new ones.
339		(But they are stored in the same place.)
340
341		----------------------------------------------------------------------
342		Name: Matthias Blume
343		Date: 2002/03/24 22:40:00 EST
344		Tag: blume-20020324-bootfiles
345		Description:
346
347		Made the bootfiles that Allen asked for.
348
349		----------------------------------------------------------------------
350		Name: Allen Leung
351		Date: 2002/03/23 15:50:00 EST
352		Tag: leunga-20020323-flint-cps-rcc-primops
353		Description:
354
355		1. Changes to FLINT primops:
356
357		(* make a call to a C-function;
358		* The primop carries C function prototype information and specifies
359		* which of its (ML-) arguments are floating point. C prototype
360		* information is for use by the backend, ML information is for
361		* use by the CPS converter. *)
362		| RAW_CCALL of { c_proto: CTypes.c_proto,
363		ml_args: ccall_type list,
364		ml_res_opt: ccall_type option,
365		reentrant : bool
366		} option
367		(* Allocate uninitialized storage on the heap.
368		* The record is meant to hold short-lived C objects, i.e., they
369		* are not ML pointers.  With the tag, the representation is
370		* the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
371		*)
372		| RAW_RECORD of {tag:bool,sz:int}
373		and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
374
375		2.  These CPS primops are now overloaded:
376
377		rawload of {kind:numkind}
378		rawstore of {kind:numkind}
379
380		The one argument form is:
381
382		rawload {kind} address
383
384		The two argument form is:
385
386		rawload {kind} [ml object, byte-offset]
387
388		3. RAW_CCALL/RCC now takes two extra arguments:
389
390		a. The first is whether the C call is reentrant, i.e., whether
391		ML state should be saved and restored.
392		b. The second argument is a string argument specifying the name of
393		library and the C function.
394
395		These things are currently not handled in the code generator, yet.
396
397		4. In CProto,
398
399		An encoding type of "bool" means "ml object" and is mapped into
400		C prototype of PTR.  Note that "bool" is different than "string",
401		even though "string" is also mapped into PTR, because "bool"
402		is assigned an CPS type of BOGt, while "string" is assigned INT32t.
403
404		5. Pickler/unpicker
405
406		Changed to handle RAW_RECORD and newest RAW_CCALL
407
408		6. MLRiscGen,
409
410		1. Changed to handle the new rawload/rawstore/rawrecord operators.
411		2. Code for handling C Calls has been moved to a new module CPSCCalls,
412		in the file CodeGen/cpscompile/cps-c-calls.sml
413
414		7. Added the conditional move operator
415
416		condmove of branch
417
418		to cps.  Generation of this is still buggy so it is currently
419		disabled.
420
421		----------------------------------------------------------------------
422		Name: Lal George
423		Date: 2002/03/22 14:18:25 EST
424		Tag: george-20020322-cps-branch-prob
425		Description:
426
427		Implemented the Ball-Larus branch prediction-heuristics, and
428		incorporated graphical viewers for control flow graphs.
429
430		Ball-Larus Heuristics:
431		---------------------
432		See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
433
434		By design it uses the Dempster-Shafer theory for combining
435		probabilities.  For example, in the function:
436
437		fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
438
439		the ball-larus heuristics predicts that the n=0 is unlikely
440		(OH-heuristic), and the 'then' branch is unlikely because of the
441		RH-heuristic -- giving the 'then' branch an even lower combined
442		probability using the Dempster-Shafer theory.
443
444		Finally, John Reppy's loop analysis in MLRISC, further lowers the
445		probability of the 'then' branch because of the loop in the else
446		branch.
447
448
449		Graphical Viewing:
450		------------------
451		I merely plugged in Allen's graphical viewers into the compiler. The
452		additional code is not much. At the top level, saying:
453
454		Control.MLRISC.getFlag "cfg-graphical-view" := true;
455
456		will display the graphical view of the control flow graph just before
457		back-patching.  daVinci must be in your path for this to work. If
458		daVinci is not available, then the default viewer can be changed
459		using:
460
461		Control.MLRISC.getString "viewer"
462
463		which can be set to "dot" or "vcg" for the corresponding viewers. Of
464		course, these viewers must be in your path.
465
466		The above will display the compilation unit at the level of clusters,
467		many of which are small, boring, and un-interesting. Also setting:
468
469		Control.MLRISC.getInt "cfg-graphical-view_size"
470
471		will display clusters that are larger than the value set by the above.
472
473
474		----------------------------------------------------------------------
475		Name: Matthias Blume
476		Date: 2002/03/21 22:20:00 EST
477		Tag: blume-20020321-kmp-bugfix
478		Description:
479
480		Changed the interface to the KMP routine in PreString and fixed
481		a minor bug in one place where it was used.
482
483		----------------------------------------------------------------------
484		Name: Allen Leung
485		Date: 2002/03/21 20:30:00 EST
486		Tag: leunga-20020321-cfg
487		Description:
488
489		Fixed a potential problem in cfg edge splitting.
490
491		----------------------------------------------------------------------
492		Name: Allen Leung
493		Date: 2002/03/21 17:15:00 EST
494		Tag: leunga-20020321-x86-fp-cfg
495		Description:
496
497		1. Recoded the buggy parts of x86-fp.
498
499		a. All the block reordering code has been removed.
500		We now depend on the block placement phases to do this work.
501
502		b. Critical edge splitting code has been simplified and moved into the
503		CFG modules, as where they belong.
504
505		Both of these were quite buggy and complex.  The code is now much, much
506		simpler.
507
508		2. X86 backend.
509
510		a. Added instructions for 64-bit support.  Instruction selection for
511		64-bit has not been committed, however, since that
512		requires changes to MLTREE which haven't been approved by
513		Lal and John.
514
515		b. Added support for FUCOMI and FUCOMIP when generating code for
516		PentiumPro and above.  We only generate these instructions in
517		the fast-fp mode.
518
519		c. Added cases for JP and JNP in X86FreqProps.
520
521		3. CFG
522
523		CFG now has a bunch of methods for edge splitting and merging.
524
525		4. Machine description.
526
527		John's simplification of MLTREE_BASIS.fcond broke a few machine
528		description things:
529
530		rtl-build.{sig,sml} and hppa.mdl fixed.
531
532		NOTE: the machine description stuff in the repository is still broken.
533		Again, I can't put my fixes in because that involves
534		changes to MLTREE.
535
536		----------------------------------------------------------------------
537		Name: Matthias Blume
538		Date: 2002/03/20 15:55:00 EST
539		Tag: blume-20020320-kmp
540		Description:
541
542		Implemented Knuth-Morris-Pratt string matching in PreString and used
543		it for String.isSubstring, Substring.isSubstring, and
544		Substring.position.
545
546		(Might need some stress-testing.  Simple examples worked fine.)
547
548		----------------------------------------------------------------------
549		Name: Matthias Blume
550		Date: 2002/03/19 16:37:00 EST
551		Tag: blume-20020319-witnesses
552		Description:
553
554		Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
555
556		This implements a generic mechanism for changing constness qualifiers
557		anywhere within big C types without resorting to outright "casts".
558		(So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
559		modify the constness at the outermost level.)
560		The implementation of "convert" is based on the idea of "witness"
561		values -- values that are not used by the operation but whose types
562		"testify" to their applicability.  On the implementation side, "convert"
563		is simply a projection (returning its second curried argument).  With
564		cross-module inlining, it should not result in any machine code being
565		generated.
566
567		----------------------------------------------------------------------
568		Name: Matthias Blume
569		Date: 2002/03/15 16:40:00 EST
570		Tag: blume-20020315-basis
571		Description:
572
573		Provided (preliminary?) implementations for
574
575		{String,Substring}.{concatWith,isSuffix,isSubstring}
576
577		and
578
579		Substring.full
580
581		Those are in the Basis spec but they were missing in SML/NJ.
582
583		----------------------------------------------------------------------
584		Name: Matthias Blume
585		Date: 2002/03/14 21:30:00 EST
586		Tag: blume-20020314-controls
587		Description:
588
589		Controls:
590		---------
591
592		1. Factored out the recently-added Controls : CONTROLS stuff and put
593		it into its own library $/controls-lib.cm.  The source tree for
594		this is under src/smlnj-lib/Controls.
595
596		2. Changed the names of types and functions in this interface, so they
597		make a bit more "sense":
598
599		module -> registry
600		'a registry -> 'a group
601
602		3. The interface now deals in ref cells only.  The getter/setter interface
603		is (mostly) gone.
604
605		4. Added a function that lets one register an already-existing ref cell.
606
607		5. Made the corresponding modifications to the rest of the code so that
608		everything compiles again.
609
610		6. Changed the implementation of Controls.MLRISC back to something closer
611		to the original.  In particular, this module (and therefore MLRISC)
612		does not depend on Controls.  There now is some link-time code in
613		int-sys.sml that registers the MLRISC controls with the Controls
614		module.
615
616		CM:
617		---
618
619		* One can now specify the lambda-split aggressiveness in init.cmi.
620
621		----------------------------------------------------------------------
622		Name: Allen Leung
623		Date: 2002/03/13 17:30:00 EST
624		Tag: leunga-20020313-x86-fp-unary
625		Description:
626
627		Bug fix for:
628
629		> leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
630		> Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
631		> - fun f(x,(y,z)) = Real.~ y;
632		> [autoloading]
633		> [autoloading done]
634		>       fchsl   (%eax), 184(%esp)
635		> Error: MLRisc bug: X86MCEmitter.emitInstr
636		>
637		> uncaught exception Error
638		>   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
639
640		The problem was that the code generator did not generate any fp registers
641		in this case, and the ra didn't know that it needed to run the X86FP phase to
642		translate the pseudo fp instruction.   This only happened with unary fp
643		operators in certain situations.
644
645		----------------------------------------------------------------------
646		Name: Matthias Blume
647		Date: 2002/03/13 14:00:00 EST
648		Tag: blume-20020313-overload-etc
649		Description:
650
651		1. Added _overload as a synonym for overload for backward compatibility.
652		(Control.overloadKW must be true for either version to be accepted.)
653
654		2. Fixed bug in install script that caused more things to be installed
655		than what was requested in config/targets.
656
657		3. Made CM aware of the (_)overload construct so that autoloading
658		works.
659
660		----------------------------------------------------------------------
661		Name: Matthias Blume
662		Date: 2002/03/12 22:03:00 EST
663		Tag: blume-20020312-url
664		Description:
665
666		Forgot to update BOOT and srcarchiveurl.
667
668		----------------------------------------------------------------------
669		Name: Matthias Blume
670		Date: 2002/03/12 17:30:00 EST
671		Tag: blume-20020312-version110392
672		Description:
673
674		Yet another version number bump (because of small changes to the
675		binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
676
677		Changes:
678
679		The new pid generation scheme described a few weeks ago was overly
680		complicated.  I implemented a new mechanism that is simpler and
681		provides a bit more "stability":  Once CM has seen a compilation
682		unit, it keeps its identity constant (as long as you do not delete
683		those crucial CM/GUID/* files).  This means that when you change
684		an interface, compile, then go back to the old interface, and
685		compile again, you arrive at the original pid.
686
687		There now also is a mechanism that instructs CM to use the plain
688		environment hash as a module's pid (effectively making its GUID
689		the empty string).  For this, "noguid" must be specified as an
690		option to the .sml file in question within its .cm file.
691		This is most useful for code that is being generated by tools such
692		as ml-nlffigen (because during development programmers tend to
693		erase the tool's entire output directory tree including CM's cached
694		GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
695		revert to the old, broken behavior of SML/NJ, but in specific cases
696		where there is no danger of interface confusion, its use is ok
697		(I think).
698
699		ml-nlffigen by default generates "noguid" annotations.  They can be
700		turned off by specifying -guid in its command line.
701
702		----------------------------------------------------------------------
703		Name: Lal George
704		Date: 2002/03/12 12 14:42:36 EST
705		Tag: george-20020312-frequency-computation
706		Description:
707
708		Integrated jump chaining and static block frequency into the
709		compiler. More details and numbers later.
710
711		----------------------------------------------------------------------
712		Name: Lal George
713		Date: 2002/03/11 11 22:38:53 EST
714		Tag: george-20020311-jump-chain-elim
715		Description:
716
717		Tested the jump chain elimination on all architectures (except the
718		hppa).  This is on by default right now and is profitable for the
719		alpha and x86, however, it may not be profitable for the sparc and ppc
720		when compiling the compiler.
721
722		The gc test will typically jump to a label at the end of the cluster,
723		where there is another jump to an external cluster containing the actual
724		code to invoke gc. This is to allow factoring of common gc invocation
725		sequences. That is to say, we generate:
726
727		f:
728		testgc
729		ja   L1      % jump if above to L1
730
731		L1:
732		jmp L2
733
734
735		After jump chain elimination the 'ja L1' instructions is converted to
736		'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
737		up being implemented in their long form (if L2 is far away) using:
738
739		jbe     L3      % jump if below or equal to L3
740		jmp     L2
741		L3:
742		...
743
744
745		For large compilation units L2  may be far away.
746
747
748		----------------------------------------------------------------------
749		Name: Matthias Blume
750		Date: 2002/03/11 13:30:00 EST
751		Tag: blume-20020311-mltreeeval
752		Description:
753
754		A functor parameter was missing.
755
756		----------------------------------------------------------------------
757		Name: Allen Leung
758		Date: 2002/03/11 10:30:00 EST
759		Tag: leunga-20020311-runtime-string0
760		Description:
761
762		The representation of the empty string now points to a
763		legal null terminated C string instead of unit.  It is now possible
764		to convert an ML string into C string with InlineT.CharVector.getData.
765		This compiles into one single machine instruction.
766
767		----------------------------------------------------------------------
768		Name: Allen Leung
769		Date: 2002/03/10 23:55:00 EST
770		Tag: leunga-20020310-x86-call
771		Description:
772
773		Added machine generation for CALL instruction (relative displacement mode)
774
775		----------------------------------------------------------------------
776		Name: Matthias Blume
777		Date: 2002/03/08 16:05:00
778		Tag: blume-20020308-entrypoints
779		Description:
780
781		Version number bumped to 110.39.1.  NEW BOOTFILES!
782
783		Entrypoints: non-zero offset into a code object where execution should begin.
784
785		- Added the notion of an entrypoint to CodeObj.
786		- Added reading/writing of entrypoint info to Binfile.
787		- Made runtime system bootloader aware of entrypoints.
788		- Use the address of the label of the first function given to mlriscGen
789		as the entrypoint.  This address is currently always 0, but it will
790		not be 0 once we turn on block placement.
791		- Removed the linkage cluster code (which was The Other Way(tm) of dealing
792		with entry points) from mlriscGen.
793
794		----------------------------------------------------------------------
795		Name: Allen Leung
796		Date: 2002/03/07 20:45:00 EST
797		Tag: leunga-20020307-x86-cmov
798		Description:
799
800		Bug fixes for CMOVcc on x86.
801
802		1. Added machine code generation for CMOVcc
803		2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
804		3. CMOVcc cannot have an immediate operand as argument.
805
806		----------------------------------------------------------------------
807		Name: Matthias Blume
808		Date: 2002/03/07 16:15:00 EST
809		Tag: blume-20020307-controls
810		Description:
811
812		This is a very large but mostly boring patch which makes (almost)
813		every tuneable compiler knob (i.e., pretty much everything under
814		Control.* plus a few other things) configurable via both the command
815		line and environment variables in the style CM did its configuration
816		until now.
817
818		Try starting sml with '-h' (or, if you are brave, '-H')
819
820		To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
821		implements the underlying generic mechanism.
822
823		The interface to some of the existing such facilities has changed somewhat.
824		For example, the MLRiscControl module now provides mkFoo instead of getFoo.
825		(The getFoo interface is still there for backward-compatibility, but its
826		use is deprecated.)
827
828		The ml-build script passes -Cxxx=yyy command-line arguments through so
829		that one can now twiddle the compiler settings when using this "batch"
830		compiler.
831
832		TODO items:
833
834		We should go through and throw out all controls that are no longer
835		connected to anything.  Moreover, we should go through and provide
836		meaningful (and correct!) documentation strings for those controls
837		that still are connected.
838
839		Currently, multiple calls to Controls.new are accepted (only the first
840		has any effect).  Eventually we should make sure that every control
841		is being made (via Controls.new) exactly once.  Future access can then
842		be done using Controls.acc.
843
844		Finally, it would probably be a good idea to use the getter-setter
845		interface to controls rather than ref cells.  For the time being, both
846		styles are provided by the Controls module, but getter-setter pairs are
847		better if thread-safety is of any concern because they can be wrapped.
848
849		*****************************************
850
851		One bug fix: The function blockPlacement in three of the MLRISC
852		backpatch files used to be hard-wired to one of two possibilities at
853		link time (according to the value of the placementFlag).  But (I
854		think) it should rather sense the flag every time.
855
856		*****************************************
857
858		Other assorted changes (by other people who did not supply a HISTORY entry):
859
860		1. the cross-module inliner now works much better (Monnier)
861		2. representation of weights, frequencies, and probabilities in MLRISC
862		changed in preparation of using those for weighted block placement
863		(Reppy, George)
864
865		----------------------------------------------------------------------
866		Name: Lal George
867		Date: 2002/03/07 14:44:24 EST 2002
868		Tag: george-20020307-weighted-block-placement
869
870		Tested the weighted block placement optimization on all architectures
871		(except the hppa) using AMPL to generate the block and edge frequencies.
872		Changes were required in the machine properties to correctly
873		categorize trap instructions. There is an MLRISC flag
874		"weighted-block-placement" that can be used to enable weighted block
875		placement, but this will be ineffective without block/edge
876		frequencies (coming soon).
877
878
879		----------------------------------------------------------------------
880		Name: Lal George
881		Date: 2002/03/05 17:24:48 EST
882		Tag: george-20020305-linkage-cluster
883
884		In order to support the block placement optimization, a new cluster
885		is generated as the very first cluster (called the linkage cluster).
886		It contains a single jump to the 'real' entry point for the compilation
887		unit. Block placement has no effect on the linkage cluster itself, but
888		all the other clusters  have full freedom in the manner in which they
889		reorder blocks or functions.
890
891		On the x86 the typical linkage code that is generated is:
892		----------------------
893		.align 2
894		L0:
895		addl    $L1-L0, 72(%esp)
896		jmp     L1
897
898
899		.align  2
900		L1:
901		----------------------
902
903		72(%esp) is the memory location for the stdlink register. This
904		must contain the address of the CPS function being called. In the
905		above example, it contains the address of  L0; before
906		calling L1 (the real entry point for the compilation unit), it
907		must contain the address for L1, and hence
908
909		addl $L1-L0, 72(%esp)
910
911		I have tested this on all architectures except the hppa.The increase
912		in code size is of course negligible
913
914		----------------------------------------------------------------------
915		Name: Allen Leung
916		Date: 2002/03/03 13:20:00 EST
917		Tag: leunga-20020303-mlrisc-tools
918
919		Added #[ ... ] expressions to mlrisc tools
920
921		----------------------------------------------------------------------
922		Name: Matthias Blume
923		Date: 2002/02/27 12:29:00 EST
924		Tag: blume-20020227-cdebug
925		Description:
926
927		- made types in structure C and C_Debug to be equal
928		- got rid of code duplication (c-int.sml vs. c-int-debug.sml)
929		- there no longer is a C_Int_Debug (C_Debug is directly derived from C)
930
931		----------------------------------------------------------------------
932		Name: Matthias Blume
933		Date: 2002/02/26 12:00:00 EST
934		Tag: blume-20020226-ffi
935		Description:
936
937		1. Fixed a minor bug in CM's "noweb" tool:
938		If numbering is turned off, then truly don't number (i.e., do not
939		supply the -L option to noweb).  The previous behavior was to supply
940		-L'' -- which caused noweb to use the "default" line numbering scheme.
941		Thanks to Chris Richards for pointing this out (and supplying the fix).
942
943		2. Once again, I reworked some aspects of the FFI:
944
945		A. The incomplete/complete type business:
946
947		- Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
948		gone!
949		- ML types representing an incomplete type are now *equal* to
950		ML types representing their corresponding complete types (just like
951		in C).  This is still safe because ml-nlffigen will not generate
952		RTTI for incomplete types, nor will it generate functions that
953		require access to such RTTI.   But when ML code generated from both
954		incomplete and complete versions of the C type meet, the ML types
955		are trivially interoperable.
956
957		NOTE:  These changes restore the full generality of the translation
958		(which was previously lost when I eliminated functorization)!
959
960		B. Enum types:
961
962		- Structure C now has a type constructor "enum" that is similar to
963		how the "su" constructor works.  However, "enum" is not a phantom
964		type because each "T enum" has values (and is isomorphic to
965		MLRep.Signed.int).
966		- There are generic access operations for enum objects (using
967		MLRep.Signed.int).
968		- ml-nlffigen will generate a structure E_foo for each "enum foo".
969		* The structure contains the definition of type "mlrep" (the ML-side
970		representation type of the enum).  Normally, mlrep is the same
971		as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
972		then mlrep will be defined as a datatype -- thus facilitating
973		pattern matching on mlrep values.
974		("-ec" will be suppressed if there are duplicate values in an
975		enumeration.)
976		* Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
977		will be generated for each C enum constant xxx.
978		* Conversion functions m2i and i2m convert between mlrep and
979		MLRep.Signed.int.  (Without "-ec", these functions are identities.)
980		* Coversion functions c and ml convert between mlrep and "tag enum".
981		* Access functions (get/set) fetch and store mlrep values.
982		- By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
983		enumerations are merged into one single enumeration represented by
984		structure E_'.
985
986		----------------------------------------------------------------------
987		Name: Allen Leung
988		Date: 2002/02/25 04:45:00 EST
989		Tag: leunga-20020225-cps-spill
990
991		This is a new implementation of the CPS spill phase.
992		The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
993		In case of problems, replace it with the old file spill.sml
994
995		The current compiler runs into some serious performance problems when
996		constructing a large record.  This can happen when we try to compile a
997		structure with many items.  Even a very simple structure like the following
998		makes the compiler slow down.
999
1000		structure Foo = struct
1001		val x_1 = 0w1 : Word32.int
1002		val x_2 = 0w2 : Word32.int
1003		val x_3 = 0w3 : Word32.int
1004		...
1005		val x_N = 0wN : Word32.int
1006		end
1007
1008		The following table shows the compile time, from N=1000 to N=4000,
1009		with the old compiler:
1010
1011		N
1012		1000   CPS 100 spill                           0.04u  0.00s  0.00g
1013		MLRISC ra                               0.06u  0.00s  0.05g
1014		(spills = 0 reloads = 0)
1015		TOTAL                                   0.63u  0.07s  0.21g
1016
1017		1100   CPS 100 spill                           8.25u  0.32s  0.64g
1018		MLRISC ra                               5.68u  0.59s  3.93g
1019		(spills = 0 reloads = 0)
1020		TOTAL                                   14.71u  0.99s  4.81g
1021
1022		1500   CPS 100 spill                           58.55u  2.34s  1.74g
1023		MLRISC ra                               5.54u  0.65s  3.91g
1024		(spills = 543 reloads = 1082)
1025		TOTAL                                   65.40u  3.13s  6.00g
1026
1027		2000   CPS 100 spill                           126.69u  4.84s  3.08g
1028		MLRISC ra                               0.80u  0.10s  0.55g
1029		(spills = 42 reloads = 84)
1030		TOTAL                                   129.42u  5.10s  4.13g
1031
1032		3000   CPS 100 spill                           675.59u  19.03s  11.64g
1033		MLRISC ra                               2.69u  0.27s  1.38g
1034		(spills = 62 reloads = 124)
1035		TOTAL                                   682.48u  19.61s  13.99g
1036
1037		4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1038		MLRISC ra                               4.96u  0.27s  2.72g
1039		(spills = 85 reloads = 170)
1040		TOTAL                                   2375.26u  57.21s  48.00g
1041
1042		As you can see the old cps spill module suffers from some serious
1043		performance problem.  But since I cannot decipher the old code fully,
1044		instead of patching the problems up, I'm reimplementing it
1045		with a different algorithm.  The new code is more modular,
1046		smaller when compiled, and substantially faster
1047		(O(n log n) time and O(n) space).  Timing of the new spill module:
1048
1049		4000  CPS 100 spill                           0.02u  0.00s  0.00g
1050		MLRISC ra                               0.25u  0.02s  0.15g
1051		(spills=1 reloads=3)
1052		TOTAL                                   7.74u  0.34s  1.62g
1053
1054		Implementation details:
1055
1056		As far as I can tell, the purpose of the CPS spill module is to make sure the
1057		number of live variables at any program point (the bandwidth)
1058		does not exceed a certain limit, which is determined by the
1059		size of the spill area.
1060
1061		When the bandwidth is too large, we decrease the register pressure by
1062		packing live variables into spill records.  How we achieve this is
1063		completely different than what we did in the old code.
1064
1065		First, there is something about the MLRiscGen code generator
1066		that we should be aware of:
1067
1068		o MLRiscGen performs code motion!
1069
1070		In particular, it will move floating point computations and
1071		address computations involving only the heap pointer to
1072		their use sites (if there is only a single use).
1073		What this means is that if we have a CPS record construction
1074		statement
1075
1076		RECORD(k,vl,w,e)
1077
1078		we should never count the new record address w as live if w
1079		has only one use (which is often the case).
1080
1081		We should do something similar to floating point, but the transformation
1082		there is much more complex, so I won't deal with that.
1083
1084		Secondly, there are now two new cps primops at our disposal:
1085
1086		1. rawrecord of record_kind option
1087		This pure operator allocates some uninitialized storage from the heap.
1088		There are two forms:
1089
1090		rawrecord NONE [INT n]  allocates a tagless record of length n
1091		rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1092		and initializes the tag.
1093
1094		2. rawupdate of cty
1095		rawupdate cty (v,i,x)
1096		Assigns to x to the ith component of record v.
1097		The storelist is not updated.
1098
1099		We use these new primops for both spilling and increment record construction.
1100
1101		1. Spilling.
1102
1103		This is implemented with a linear scan algorithm (but generalized
1104		to trees).  The algorithm will create a single spill record at the
1105		beginning of the cps function and use rawupdate to spill to it,
1106		and SELECT or SELp to reload from it.  So both spills and reloads
1107		are fine-grain operations.  In contrast, in the old algorithm
1108		"spills" have to be bundled together in records.
1109
1110		Ideally, we should sink the spill record construction to where
1111		it is needed.  We can even split the spill record into multiple ones
1112		at the places where they are needed.  But CPS is not a good
1113		representation for global code motion, so I'll keep it simple and
1114		am not attempting this.
1115
1116		2. Incremental record construction (aka record splitting).
1117
1118		Long records with many component values which are simulatenously live
1119		(recall that single use record addresses are not considered to
1120		be live) are constructed with rawrecord and rawupdate.
1121		We allocate space on the heap with rawrecord first, then gradually
1122		fill it in with rawupdate.  This is the technique suggested to me
1123		by Matthias.
1124
1125		Some restrictions on when this is applicable:
1126		1. It is not a VECTOR record.  The code generator currently does not handle
1127		this case. VECTOR record uses double indirection like arrays.
1128		2. All the record component values are defined in the same "basic block"
1129		as the record constructor.  This is to prevent speculative
1130		record construction.
1131
1132		----------------------------------------------------------------------
1133		Name: Allen Leung
1134		Date: 2002/02/22 01:02:00 EST
1135		Tag: leunga-20020222-mlrisc-tools
1136
1137		Minor bug fixes in the parser and rewriter
1138
1139		----------------------------------------------------------------------
1140		Name: Allen Leung
1141		Date: 2002/02/21 20:20:00 EST
1142		Tag: leunga-20020221-peephole
1143
1144		Regenerated the peephole files.  Some contained typos in the specification
1145		and some didn't compile because of pretty printing bugs in the old version
1146		of 'nowhere'.
1147
1148		----------------------------------------------------------------------
1149		Name: Allen Leung
1150		Date: 2002/02/19 20:20:00 EST
1151		Tag: leunga-20020219-mlrisc-tools
1152		Description:
1153
1154		Minor bug fixes to the mlrisc-tools library:
1155
1156		1.  Fixed up parsing colon suffixed keywords
1157		2.  Added the ability to shut the error messages up
1158		3.  Reimplemented the pretty printer and fixed up/improved
1159		the pretty printing of handle and -> types.
1160		4.  Fixed up generation of literal symbols in the nowhere tool.
1161		5.  Added some SML keywords to to sml.sty
1162
1163		----------------------------------------------------------------------
1164		Name: Matthias Blume
1165		Date: 2002/02/19 16:20:00 EST
1166		Tag: blume-20020219-cmffi
1167		Description:
1168
1169		A wild mix of changes, some minor, some major:
1170
1171		* All C FFI-related libraries are now anchored under $c:
1172		$/c.cm      --> $c/c.cm
1173		$/c-int.cm  --> $c/internals/c-int.cm
1174		$/memory.cm --> $c/memory/memory.cm
1175
1176		* "make" tool (in CM) now treats its argument pathname slightly
1177		differently:
1178		1. If the native expansion is an absolute name, then before invoking
1179		the "make" command on it, CM will apply OS.Path.mkRelative
1180		(with relativeTo = OS.FileSys.getDir()) to it.
1181		2. The argument will be passed through to subsequent phases of CM
1182		processing without "going native".  In particular, if the argument
1183		was an anchored path, then "make" will not lose track of that anchor.
1184
1185		* Compiler backends now "know" their respective C calling conventions
1186		instead of having to be told about it by ml-nlffigen.  This relieves
1187		ml-nlffigen from one of its burdens.
1188
1189		* The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1190
1191		* Export C_DEBUG and C_Debug from $c/c.cm.
1192
1193		* C type encoding in ml-nlffi-lib has been improved to model the conceptual
1194		subtyping relationship between incomplete pointers and their complete
1195		counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1196		with the convention of instantiating 'o with ('t, 'c) obj whenever
1197		the pointer target type is complete.  In the incomplete case, 'o
1198		will be instantiated with some "'c iobj" -- a type obtained by
1199		using one of the functors PointerToIncompleteType or PointerToCompleteType.
1200
1201		Operations that work on both incomplete and complete pointer types are
1202		typed as taking an 'o ptr while operations that require the target to
1203		be known are typed as taking some ('t, 'c) obj ptr.
1204
1205		voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1206		where void is an eqtype without any values.  This makes it possible
1207		to work on voidptr values using functions meant to operate on light
1208		incomplete pointers.
1209
1210		* As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1211		been vastly simplified.
1212
1213		----------------------------------------------------------------------
1214		Name: Matthias Blume
1215		Date: 2002/02/19 10:48:00 EST
1216		Tag: blume-20020219-pqfix
1217		Description:
1218
1219		Applied Chris Okasaki's bug fix for priority queues.
1220
1221		----------------------------------------------------------------------
1222		Name: Matthias Blume
1223		Date: 2002/02/15 17:05:00
1224		Tag: Release_110_39
1225		Description:
1226
1227		Last-minute retagging is becoming a tradition... :-(
1228
1229		This is the working release 110.39.
1230
1231		----------------------------------------------------------------------
1232		Name: Matthias Blume
1233		Date: 2002/02/15 16:00:00 EST
1234		Tag: Release_110_39-orig
1235		Description:
1236
1237		Working release 110.39.  New bootfiles.
1238
1239		(Update: There was a small bug in the installer so it wouldn't work
1240		with all shells.  So I retagged. -Matthias)
1241
1242		----------------------------------------------------------------------
1243		Name: Matthias Blume
1244		Date: 2002/02/15 14:17:00 EST
1245		Tag: blume-20020215-showbindings
1246		Description:
1247
1248		Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1249		the latter can be useful for exploring what bindings are available at
1250		the interactive prompt.  (The first function returns only the list
1251		of symbols that are really bound, the second prints those but also the
1252		ones that CM's autoloading mechanism knows about.)
1253
1254		----------------------------------------------------------------------
1255		Name: Matthias Blume
1256		Date: 2002/02/15 12:08:00 EST
1257		Tag: blume-20020215-iptrs
1258		Description:
1259
1260		Two improvements to ml-nlffigen:
1261
1262		1. Write files only if they do not exist or if their current contents
1263		do not coincide with what's being written.  (That is, avoid messing
1264		with the time stamps unless absolutely necessary.)
1265
1266		2. Implement a "repository" mechanism for generated files related
1267		to "incomplete pointer types".   See the README file for details.
1268
1269		----------------------------------------------------------------------
1270		Name: Matthias Blume
1271		Date: 2002/02/14 11:50:00 EST
1272		Tag: blume-20020214-quote
1273		Description:
1274
1275		Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1276		because of the new and improved tag generation scheme.  (Thanks to Allen
1277		Leung for pointing it out.)
1278
1279		----------------------------------------------------------------------
1280		Name: Lal George
1281		Date: 2002/02/14 09:55:27 EST 2002
1282		Tag: george-20020214-isabelle-bug
1283		Description:
1284
1285		Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1286		of Isabelle on the x86.
1287
1288		From Allen:
1289		-----------
1290		I've found the problem:
1291
1292		in ra-core.sml, I use the counter "blocked" to keep track of the
1293		true number of elements in the freeze queue.  When the counter goes
1294		to zero, I skip examining the queue.  But I've messed up the
1295		bookkeeping in combine():
1296
1297		else ();
1298		case !ucol of
1299		PSEUDO => (if !cntv > 0 then
1300		(if !cntu > 0 then blocked := !blocked - 1 else ();
1301		^^^^^^^^^^^^^^^^^^^^^^^
1302		moveu := mergeMoveList(!movev, !moveu)
1303		)
1304		else ();
1305
1306		combine() is called to coalesce two nodes u and v.
1307		I think I was thinking that if the move counts of u and v are both
1308		greater than zero then after they are coalesced then one node is
1309		removed from the freeze queue.  Apparently I was thinking that
1310		both u and v are of low degree, but that's clearly not necessarily true.
1311
1312
1313		02/12/2002:
1314		Here's the patch.  HOL now compiles.
1315
1316		I don't know how this impact on performance (compile
1317		time or runtime).  This bug caused the RA (especially on the x86)
1318		to go thru the potential spill phase when there are still nodes on the
1319		freeze queue.
1320
1321
1322
1323
1324		----------------------------------------------------------------------
1325		Name: Matthias Blume
1326		Date: 2002/02/13 22:40:00 EST
1327		Tag: blume-20020213-fptr-rtti
1328		Description:
1329
1330		Fixed a bug in ml-nlffigen that was introduced with one of the previous
1331		updates.
1332
1333		----------------------------------------------------------------------
1334		Name: Matthias Blume
1335		Date: 2002/02/13 16:41:00 EST
1336		Tag: blume-20020213-cmlpq
1337		Description:
1338
1339		Added new priority queue export symbols (which have just been added to
1340		smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1341		not compile and the installer would choke.)
1342
1343		----------------------------------------------------------------------
1344		Name: Matthias Blume
1345		Date: 2002/02/13 16:15:00 EST
1346		Tag: blume-20020213-various
1347		Description:
1348
1349		1. More tweaks to ml-nlffigen:
1350
1351		- better internal datastructures (resulting in slight speedup)
1352		- "-match" option requires exact match
1353		- "localized" gensym counters (untagged structs/unions nested within
1354		other structs/unions or within typedefs get a fresh counter; their
1355		tag will be prefixed by a concatenation of their parents' tags)
1356		- bug fixes (related to calculation of transitive closure of types
1357		to be included in the output)
1358
1359		2. Minor Basis updates:
1360
1361		- added implementations for List.collate and Option.app
1362
1363		----------------------------------------------------------------------
1364		Name: Matthias Blume
1365		Date: 2002/02/11 15:55:00 EST
1366		Tag: blume-20020211-gensym
1367		Description:
1368
1369		Added a "-gensym" option to command line of ml-nlffigen.  This can be
1370		used to specify a "stem" -- a string that is inserted in all "gensym'd"
1371		names (ML structure names that correspond to unnamed C structs, unions,
1372		and enums), so that separate runs of ml-nlffigen do not clash.
1373
1374		----------------------------------------------------------------------
1375		Name: Matthias Blume
1376		Date: 2002/02/11 12:05:00 EST
1377		Tag: blume-20020211-gensml
1378		Description:
1379
1380		A quick fix for a problem with GenSML (in the pgraph-util library):
1381		Make generation of toplevel "local" optional.  (Strictly speaking,
1382		signature definitions within "local" are not legal SML.)
1383
1384		Other than that: updates to INSTALL and cm/TODO.
1385
1386		----------------------------------------------------------------------
1387		Name: Matthias Blume
1388		Date: 2002/02/08 15:00:00 EST
1389		Tag: blume-20020208-uniquepid
1390		Description:
1391
1392		0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1393
1394		1. The installer (config/install.sh) has gotten smarter:
1395
1396		- Configuration options are a bit easier to specify now
1397		(in config/targets).
1398		- Bug in recognizing .tar.bz2 files fixed.
1399		- Installer automatically resolves dependencies between
1400		configuration options (e.g., if you ask for eXene, you will
1401		also get cml -- regardless whether you asked for it or not).
1402		- Installer can run in "quieter mode" by setting the environment
1403		variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1404		"completely silent", though.
1405		- Build HashCons library as part of smlnj-lib.
1406
1407		2. A new scheme for assigning persistent identifiers to compilation
1408		units (and, by extension, to types etc.) has been put into place.
1409		This fixes a long-standing bug where types and even dynamic values
1410		can get internally confused, thereby compromising type safety
1411		(abstraction) and dynamic correctness.  See
1412
1413		http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1414
1415		for an example of how things could go wrong until now.
1416
1417		The downside of the new scheme is that pids are not quite as
1418		persistent as they used to be: CM will generate a fresh pid
1419		for every compilation unit that it thinks it sees for the first
1420		time.  That means that if you compile starting from a clean, fresh
1421		source tree at two different times, you end up with different
1422		binaries.
1423
1424		Cutoff recompilation, however, has not been compromised because
1425		CM keeps pid information in special caches between runs.
1426
1427		----------------------------------------------------------------------
1428		Name: Lal George
1429		Date: 2002/02/07 15:34:13 EST 2002
1430		Tag: <none>
1431		Description:
1432
1433		Compilers that generate assembly code may produce  global labels
1434		whose value is resolved at link time. The various peephole optimization
1435		modules did not take this in account.
1436
1437		TODO. The Labels.addrOf function should really return an option
1438		type so that clients are forced to deal with this issue, rather
1439		than an exception being raised.
1440
1441		----------------------------------------------------------------------
1442		Name: Lal George
1443		Date: 2002/02/06 13:55:02 EST
1444		Tag: george-20020206-ra-breakup
1445		Description:
1446
1447		1. A bug fix from Allen.
1448
1449		A typo causes extra fstp %st(0)'s to be generated at compensation
1450		edges, which might cause stack underflow traps at runtime.  This
1451		occurs in fft where there are extraneous fstps right before the 'into'
1452		trap instruction (in this case they are harmless since none of the
1453		integers overflow.)
1454
1455		2. Pulled out various utility modules that were embedded in the modules
1456		of the register allocator. I need these modules for other purposes, but
1457		they are not complete enough to put into a library (just yet).
1458		----------------------------------------------------------------------
1459		Name: Matthias Blume
1460		Date: 2002/01/31 16:05:00 EST
1461		Tag: blume-20020131-sparc-ccalls
1462		Description:
1463
1464		1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1465		of extra stack space by mistake.  Fixed.
1466
1467		2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1468
1469		----------------------------------------------------------------------
1470		Name: Allen Leung
1471		Date: 2002/01/30
1472		Tag: leunga-20020130-nowhere-bug-fix
1473		Description:
1474
1475		MLRISC bug fixes:
1476		1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1477		2. MachineInt.fromString was negating its value.
1478
1479		----------------------------------------------------------------------
1480		Name: Matthias Blume
1481		Date: 2002/01/29
1482		Tag: blume-20020129-INSTALL
1483		Description:
1484
1485		- Added somewhat detailed installation instructions (file INSTALL).
1486		- Fixed curl-detection bug in config/install.sh.
1487		- It is now possible to select the URL getter using the URLGETTER
1488		environment variable:
1489
1490		not set / "unknown"      --> automatic detection (script tries wget,
1491		curl, and lynx)
1492		"wget" / "curl" / "lynx" --> use the specified program (script "knows"
1493		how to properly invoke them)
1494		other                    --> use $URLGETTER directly, it must take
1495		precisely two command-line arguments
1496		(source URL and destination file name)
1497
1498		----------------------------------------------------------------------
1499		Name: Matthias Blume
1500		Date: 2002/01/28
1501		Tag: blume-20020128-sparc-ccalls
1502		Description:
1503
1504		- Fixed problem with calculation of "used" registers in sparc-c-calls.
1505		- Make use of the allocParam argument in sparc-c-calls.
1506
1507		----------------------------------------------------------------------
1508		Name: Matthias Blume
1509		Date: 2002/01/28
1510		Tag: blume-20020128-allocParam
1511		Description:
1512
1513		John Reppy:  Changes c-calls API to accept client-callback for
1514		allocating extra stack space.
1515		me: Corresponding changes to mlriscGen (using a dummy argument that
1516		does not change the current behavior).
1517
1518		----------------------------------------------------------------------
1519		Name: Matthias Blume
1520		Date: 2002/01/28 12:00:00
1521		Tag: Release_110_38
1522		Description:
1523
1524		This time for real!!!
1525
1526		----------------------------------------------------------------------
1527		Name: Matthias Blume
1528		Date: 2002/01/28 10:56:00 EST
1529		Tag: blume-20020128-retraction
1530		Description:
1531
1532		0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1533		with blume-Release_110_38-retracted.)
1534
1535		1. Fixed a problem with incorrect rounding modes in real64.sml.
1536		(Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1537
1538		2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1539		and enums fixed.  The naming of corresponding ML identifiers should
1540		now be consistent again.
1541
1542		----------------------------------------------------------------------
1543		Name: Allen Leung
1544		Date: 2002/01/27
1545		Tag: leunga-20020127-nowhere
1546		Description:
1547
1548		Added a target called nowhere in the configuration scripts.
1549		Enabling this will build the MLRISC 'nowhere' tool (for translating
1550		programs with where-clauses into legal SML code) during installation.
1551
1552		----------------------------------------------------------------------
1553		Name: Matthias Blume
1554		Date: 2002/01/25 21:27:00 EST
1555		Tag: blume-Release_110_38-retracted
1556		Description:
1557
1558		Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1559
1560		README will be added later.
1561
1562		!!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1563		(Release_110_38) removed.  Reason: Last-minute bug fixes.
1564
1565		----------------------------------------------------------------------
1566		Name: Matthias Blume
1567		Date: 2002/01/25
1568		Tag: blume-20020125-ffi
1569		Description:
1570
1571		A large number of tweaks and improvements to ml-nlffi-lib and
1572		ml-nlffigen:
1573
1574		- ML represenation types have been streamlined
1575		- getter and setter functions work with concrete values, not abstract
1576		ones where possible
1577		- ml-nlffigen command line more flexible (see README file there)
1578		- some bugs have been fixed (hopefully)
1579
1580		----------------------------------------------------------------------
1581		Name: Lal George
1582		Date: 2002/01/24
1583		Tag: george-20020124-risc-ra-interface
1584		Description:
1585
1586		There is a dramatic simplification in the interface to the
1587		register allocator for RISC architectures as a result of making
1588		parallel copy instructions explicit.
1589
1590		----------------------------------------------------------------------
1591		Name: Matthias Blume
1592		Date: 2002/01/22
1593		Tag: blume-20020122-x86-ccalls
1594		Description:
1595
1596		Bug fix for c-calls on x86 (having to do with how char- and
1597		short-arguments are being handled).
1598
1599		----------------------------------------------------------------------
1600		Name: Matthias Blume
1601		Date: 2002/01/21
1602		Tag: blume-20020121-ff
1603		Description:
1604
1605		Another day of fiddling with the FFI...
1606
1607		1. Bug fix/workaround:  CKIT does not complain about negative array
1608		dimensions, so ml-nlffigen has to guard itself against this possibility.
1609		(Otherwise a negative dimension would send it into an infinite loop.)
1610
1611		2. Some of the abstract types (light objects, light pointers, most "base"
1612		types) in structure C are now eqtypes.
1613
1614		3. Added constructors and test functions for NULL function pointers.
1615
1616		----------------------------------------------------------------------
1617		Name: Matthias Blume
1618		Date: 2002/01/18
1619		Tag: blume-20020118-ready-for-new-release
1620		Description:
1621
1622		Made config/srcarchiveurl point to a new place.  (Will provide boot
1623		files shortly.)
1624
1625		Maybe we christen this to be 110.38?
1626
1627		----------------------------------------------------------------------
1628		Name: Matthias Blume
1629		Date: 2002/01/18
1630		Tag: blume-20020118-more-ffifiddle
1631		Description:
1632
1633		Today's FFI fiddling:
1634
1635		- Provided a structure CGetSet with "convenient" versions of C.Get.* and
1636		C.Set.* that use concrete (MLRep.*) arguments and results instead
1637		of abstract ones.
1638
1639		- Provided word-style bit operations etc. for "int" representation
1640		types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1641		and Long.
1642
1643		----------------------------------------------------------------------
1644		Name: Matthias Blume
1645		Date: 2002/01/18
1646		Tag: blume-20020118-use-x86-fp
1647		Description:
1648
1649		Now that x86-fast-fp seems to be working, I turned it back on again
1650		by default.  (Seems to work fine now, even with the FFI.)
1651
1652		Other than that, I added some documentation about the FFI to
1653		src/ml-nlffigen/README and updated the FFI test examples in
1654		src/ml-nlffi-lib/Tests/*.
1655
1656		----------------------------------------------------------------------
1657		Name: Allen Leung
1658		Date: 2002/01/17
1659		Tag: leunga-20020117-x86-fast-fp-call
1660		Description:
1661
1662		1. Fixed a problem with handling return fp values when x86's fast fp
1663		mode is turned on.
1664
1665		2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
1666		of %f32.
1667
1668		3. Added a constructor INT32lit to the ast of MLRISC tools.
1669
1670		----------------------------------------------------------------------
1671		Name: Matthias Blume
1672		Date: 2002/01/16
1673		Tag: blume-20020116-ffifiddle
1674		Description:
1675
1676		More fiddling with the FFI interface:
1677
1678		- Make constness 'c instead of rw wherever possible.  This eliminates
1679		the need for certain explicit coercions.  (However, due to ML's
1680		value polymorphism, there will still be many cases where explicit
1681		coercions are necessary.  Phantom types are not the whole answer
1682		to modeling a subtyping relationship in ML.)
1683
1684		- ro/rw coersions for pointers added.  (Avoids the detour through */&.)
1685
1686		- "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
1687		clumsy workaround for varargs problem.)
1688
1689		----------------------------------------------------------------------
1690		Name: Lal George
1691		Date: 2002/01/15
1692		Tag: <none>
1693		Description:
1694
1695		1. Since COPY instructions are no longer native to the architecture,
1696		a generic functor can be used to implement the expandCopies function.
1697
1698		2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1699		TEXT segment.
1700
1701		----------------------------------------------------------------------
1702		Name: Matthias Blume
1703		Date: 2002/01/15
1704		Tag: blume-20020115-ffiupdates
1705		Description:
1706
1707		1. Fix for bug resulting in single-precision float values being returned
1708		incorrectly from FFI calls.
1709
1710		2. Small modifications to C FFI API:
1711
1712		- memory-allocation routines return straight objects (no options)
1713		and raise an exception in out-of-memory situations
1714		- unsafe extensions to cast between function pointers and pointers
1715		from/to ints
1716		- added structure C_Debug as an alternative to structure C where
1717		pointer-dereferencing (|*| and |*!) always check for null-pointers
1718		- added open_lib' to DynLinkage;  open_lib' works like open_lib
1719		but also takes a (possibly empty) list of existing library handles
1720		that the current library depends on
1721
1722		----------------------------------------------------------------------
1723		Name: Matthias Blume
1724		Date: 2002/01/10
1725		Tag: blume-20020110-newffigen
1726		Description:
1727
1728		1. Updates to portable graph code.
1729
1730		2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
1731		more scalable now so that even huge interfaces such as the one
1732		for GTK compile in finite time and space. :-)
1733		See src/ml-nlffigen/README for details on what's new.
1734
1735		----------------------------------------------------------------------
1736		Name: Lal George
1737		Date: 2001/01/09 14:31:35 EST 2002
1738		Tag: george-20011206-rm-native-copy
1739		Description:
1740
1741		Removed the native COPY and FCOPY instructions
1742		from all the architectures and replaced it with the
1743		explicit COPY instruction from the previous commit.
1744
1745		It is now possible to simplify many of the optimizations
1746		modules that manipulate copies. This has not been
1747		done in this change.
1748
1749	----------------------------------------------------------------------	----------------------------------------------------------------------
1750	Name: Lal George	Name: Lal George
1751	Date: 2001/12/06 16:50:13 EST 2001	Date: 2001/12/06 16:50:13 EST 2001
#	Line 341	Line 2076
2076
2077	----------------------------------------------------------------------	----------------------------------------------------------------------
2078	Name: Matthias Blume	Name: Matthias Blume
	>>>>>>> 1.169
2079	Date: 2001/09/18 15:35:00 EDT	Date: 2001/09/18 15:35:00 EDT
2080	Tag: blume-20010918-readme11036	Tag: blume-20010918-readme11036
2081	Description:	Description:

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.1003   changed lines   Added in v.1212

---

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