Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 1045, Mon Jan 28 22:13:13 2002 UTC	revision 1186, Fri Apr 12 17:54:31 2002 UTC
#	Line 14	Line 14
14
15	----------------------------------------------------------------------	----------------------------------------------------------------------
16	Name: Matthias Blume	Name: Matthias Blume
17		Date: 2002/04/12 13:55:00 EDT
18		Tag: blume-20020412-assyntax
19		Description:
20
21		1. Grabbed newer assyntax.h from the XFree86 project.
22		2. Fiddled with how to compile X86.prim.asm without warnings.
23		3. (Very) Minor cleanup in CM.
24
25		----------------------------------------------------------------------
26		Name: Matthias Blume
27		Date: 2002/04/01 (no joke!) 17:07:00 EST
28		Tag: blume-20020401-x86div
29		Description:
30
31		Added full support for div/mod/rem/quot on the x86, using the machine
32		instruction's two results (without clumsily recomputing the remainder)
33		directly where appropriate.
34
35		Some more extensive power-of-two support was added to the x86 instruction
36		selector (avoiding expensive divs, mods, and muls where they can be
37		replaced with cheaper shifts and masks).  However, this sort of thing
38		ought to be done earlier, e.g., within the CPS optimizer so that
39		all architectures benefit from it.
40
41		The compiler compiles to a fixed point, but changes might be somewhat
42		fragile nevertheless.  Please, report any strange things that you might
43		see wrt. div/mod/quot/rem...
44
45		----------------------------------------------------------------------
46		Name: Matthias Blume
47		Date: 2002/03/29 17:22:00
48		Tag: blume-20020329-div
49		Description:
50
51		Fixed my broken div/mod logic.  Unfortunately, this means that the
52		inline code for div/mod now has one more comparison than before.
53		Fast paths (quotient > 0 or remainder = 0) are not affected, though.
54		The problem was with quotient = 0, because that alone does not tell
55		us which way the rounding went.  One then has to look at whether
56		remainder and divisor have the same sign...  :(
57
58		Anyway, I replaced the bootfiles with fresh ones...
59
60		----------------------------------------------------------------------
61		Name: Matthias Blume
62		Date: 2002/03/29 14:10:00 EST
63		Tag: blume-20020329-inlprims
64		Description:
65
66		NEW BOOTFILES!!!    Version number bumped to 110.39.3.
67
68		Primops have changed. This means that the bin/boot-file formats have
69		changed as well.
70
71		To make sure that there is no confusion, I made a new version.
72
73
74		CHANGES:
75
76		* removed REMT from mltree (remainder should never overflow).
77
78		* added primops to deal with divisions of all flavors to the frontend
79
80		* handled these primops all the way through so they map to their respective
81		MLRISC support
82
83		* used these primops in the implementation of Int, Int32, Word, Word32
84
85		* removed INLDIV, INLMOD, and INLREM as they are no longer necessary
86
87		* parameterized INLMIN, INLMAX, and INLABS by a numkind
88
89		* translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
90		floating point
91
92		* used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
93		and Real (but Real.abs maps to a separate floating-point-only primop)
94
95
96		TODO items:
97
98		* Hacked Alpha32 instruction selection, disabling the selection of REMx
99		instructions because the machine instruction encoder cannot handle
100		them.  (Hppa, PPC, and Sparc instruction selection did not handle
101		REM in the first place, and REM is supported by the x86 machine coder.)
102
103		* Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
104		selection phase.  (The two can be streamlined because the hardware
105		delivers both quotient and remainder at the same time anyway.)
106
107		* Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
108		(Currently the behavior is inconsistent both across architectures and
109		wrt. the draft Basis spec.)
110
111		* Word8 should eventually be handled natively, too.
112
113		* There seems to be one serious bug in mltree-gen.sml.  It appears, though,
114		as if there currently is no execution path that could trigger it in
115		SML/NJ.  (The assumptions underlying functions arith and promotable do not
116		hold for things like multiplication and division.)
117
118		----------------------------------------------------------------------
119		Name: Matthias Blume
120		Date: 2002/03/27 16:27:00 EST
121		Tag: blume-20020327-mlrisc-divisions
122		Description:
123
124		Added support for all four division operations (ML's div, mod, quot,
125		and rem) to MLRISC.  In the course of doing so, I also rationalized
126		the naming (no more annoying switch-around of DIV and QUOT), by
127		parameterizing the operation by div_rounding_mode (which can be either
128		DIV_TO_ZERO or DIV_TO_NEGINF).
129
130		The generic MLTreeGen functor takes care of compiling all four
131		operations down to only round-to-zero div.
132
133		Missing pieces:
134
135		* Doing something smarter than relying on MLTreeGen on architectures
136		like, e.g., the x86 where hardware division delivers both quotient and
137		remainder at the same time.  With this, the implementation of the
138		round-to-neginf operations could be further streamlined.
139
140		* Remove inlining support for div/mod/rem from the frontend and replace it
141		with primops that get carried through to the backend.  Do this for all
142		int and word types.
143
144		----------------------------------------------------------------------
145		Name: Matthias Blume
146		Date: 2002/03/25 17:25:00 EST
147		Tag: blume-20020325-divmod
148		Description:
149
150		I improved (hopefully without breaking them) the implementation of Int.div,
151		Int.mod, and Int.rem.   For this, the code in translate.sml now takes
152		advantage of the following observations:
153
154		Let  q = x quot y      r = x rem y
155		d = x div  y      m = x mod y
156
157		where "quot" is the round-to-zero version of integer division that
158		hardware usually provides.  Then we have:
159
160		r = x - q * y        where neither the * nor the - will overflow
161		d = if q >= 0 orelse x = q * y then q else q - 1
162		where neither the * nor the - will overflow
163		m = if q >= 0 orelse r = 0 then r else r + y
164		where the + will not overflow
165
166		This results in substantial simplification of the generated code.
167		The following table shows the number of CFG nodes and edges generated
168		for
169		fun f (x, y) = x OPER y
170		(* with OPER \in div, mod, quot, rem *)
171
172
173		OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
174		--------------------------------------------------------
175		div |         24 |         39 |         12 |         16
176		mod |         41 |         71 |         12 |         16
177		quot |          8 |         10 |          8 |         10
178		rem |         10 |         14 |          8 |         10
179
180
181		----------------------------------------------------------------------
182		Name: Matthias Blume
183		Date: 2002/03/25 22:06:00 EST
184		Tag: blume-20020325-cprotobug
185		Description:
186
187		Fixed a bug in cproto (c prototype decoder).
188
189		----------------------------------------------------------------------
190		Name: Matthias Blume
191		Date: 2002/03/25 16:00:00 EST
192		Tag: blume-20020325-raw-primops
193		Description:
194
195		I did some cleanup to Allen's new primop code and
196		replaced yesterday's bootfiles with new ones.
197		(But they are stored in the same place.)
198
199		----------------------------------------------------------------------
200		Name: Matthias Blume
201		Date: 2002/03/24 22:40:00 EST
202		Tag: blume-20020324-bootfiles
203		Description:
204
205		Made the bootfiles that Allen asked for.
206
207		----------------------------------------------------------------------
208		Name: Allen Leung
209		Date: 2002/03/23 15:50:00 EST
210		Tag: leunga-20020323-flint-cps-rcc-primops
211		Description:
212
213		1. Changes to FLINT primops:
214
215		(* make a call to a C-function;
216		* The primop carries C function prototype information and specifies
217		* which of its (ML-) arguments are floating point. C prototype
218		* information is for use by the backend, ML information is for
219		* use by the CPS converter. *)
220		| RAW_CCALL of { c_proto: CTypes.c_proto,
221		ml_args: ccall_type list,
222		ml_res_opt: ccall_type option,
223		reentrant : bool
224		} option
225		(* Allocate uninitialized storage on the heap.
226		* The record is meant to hold short-lived C objects, i.e., they
227		* are not ML pointers.  With the tag, the representation is
228		* the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
229		*)
230		| RAW_RECORD of {tag:bool,sz:int}
231		and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
232
233		2.  These CPS primops are now overloaded:
234
235		rawload of {kind:numkind}
236		rawstore of {kind:numkind}
237
238		The one argument form is:
239
240		rawload {kind} address
241
242		The two argument form is:
243
244		rawload {kind} [ml object, byte-offset]
245
246		3. RAW_CCALL/RCC now takes two extra arguments:
247
248		a. The first is whether the C call is reentrant, i.e., whether
249		ML state should be saved and restored.
250		b. The second argument is a string argument specifying the name of
251		library and the C function.
252
253		These things are currently not handled in the code generator, yet.
254
255		4. In CProto,
256
257		An encoding type of "bool" means "ml object" and is mapped into
258		C prototype of PTR.  Note that "bool" is different than "string",
259		even though "string" is also mapped into PTR, because "bool"
260		is assigned an CPS type of BOGt, while "string" is assigned INT32t.
261
262		5. Pickler/unpicker
263
264		Changed to handle RAW_RECORD and newest RAW_CCALL
265
266		6. MLRiscGen,
267
268		1. Changed to handle the new rawload/rawstore/rawrecord operators.
269		2. Code for handling C Calls has been moved to a new module CPSCCalls,
270		in the file CodeGen/cpscompile/cps-c-calls.sml
271
272		7. Added the conditional move operator
273
274		condmove of branch
275
276		to cps.  Generation of this is still buggy so it is currently
277		disabled.
278
279		----------------------------------------------------------------------
280		Name: Lal George
281		Date: 2002/03/22 14:18:25 EST
282		Tag: george-20020322-cps-branch-prob
283		Description:
284
285		Implemented the Ball-Larus branch prediction-heuristics, and
286		incorporated graphical viewers for control flow graphs.
287
288		Ball-Larus Heuristics:
289		---------------------
290		See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
291
292		By design it uses the Dempster-Shafer theory for combining
293		probabilities.  For example, in the function:
294
295		fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
296
297		the ball-larus heuristics predicts that the n=0 is unlikely
298		(OH-heuristic), and the 'then' branch is unlikely because of the
299		RH-heuristic -- giving the 'then' branch an even lower combined
300		probability using the Dempster-Shafer theory.
301
302		Finally, John Reppy's loop analysis in MLRISC, further lowers the
303		probability of the 'then' branch because of the loop in the else
304		branch.
305
306
307		Graphical Viewing:
308		------------------
309		I merely plugged in Allen's graphical viewers into the compiler. The
310		additional code is not much. At the top level, saying:
311
312		Control.MLRISC.getFlag "cfg-graphical-view" := true;
313
314		will display the graphical view of the control flow graph just before
315		back-patching.  daVinci must be in your path for this to work. If
316		daVinci is not available, then the default viewer can be changed
317		using:
318
319		Control.MLRISC.getString "viewer"
320
321		which can be set to "dot" or "vcg" for the corresponding viewers. Of
322		course, these viewers must be in your path.
323
324		The above will display the compilation unit at the level of clusters,
325		many of which are small, boring, and un-interesting. Also setting:
326
327		Control.MLRISC.getInt "cfg-graphical-view_size"
328
329		will display clusters that are larger than the value set by the above.
330
331
332		----------------------------------------------------------------------
333		Name: Matthias Blume
334		Date: 2002/03/21 22:20:00 EST
335		Tag: blume-20020321-kmp-bugfix
336		Description:
337
338		Changed the interface to the KMP routine in PreString and fixed
339		a minor bug in one place where it was used.
340
341		----------------------------------------------------------------------
342		Name: Allen Leung
343		Date: 2002/03/21 20:30:00 EST
344		Tag: leunga-20020321-cfg
345		Description:
346
347		Fixed a potential problem in cfg edge splitting.
348
349		----------------------------------------------------------------------
350		Name: Allen Leung
351		Date: 2002/03/21 17:15:00 EST
352		Tag: leunga-20020321-x86-fp-cfg
353		Description:
354
355		1. Recoded the buggy parts of x86-fp.
356
357		a. All the block reordering code has been removed.
358		We now depend on the block placement phases to do this work.
359
360		b. Critical edge splitting code has been simplified and moved into the
361		CFG modules, as where they belong.
362
363		Both of these were quite buggy and complex.  The code is now much, much
364		simpler.
365
366		2. X86 backend.
367
368		a. Added instructions for 64-bit support.  Instruction selection for
369		64-bit has not been committed, however, since that
370		requires changes to MLTREE which haven't been approved by
371		Lal and John.
372
373		b. Added support for FUCOMI and FUCOMIP when generating code for
374		PentiumPro and above.  We only generate these instructions in
375		the fast-fp mode.
376
377		c. Added cases for JP and JNP in X86FreqProps.
378
379		3. CFG
380
381		CFG now has a bunch of methods for edge splitting and merging.
382
383		4. Machine description.
384
385		John's simplification of MLTREE_BASIS.fcond broke a few machine
386		description things:
387
388		rtl-build.{sig,sml} and hppa.mdl fixed.
389
390		NOTE: the machine description stuff in the repository is still broken.
391		Again, I can't put my fixes in because that involves
392		changes to MLTREE.
393
394		----------------------------------------------------------------------
395		Name: Matthias Blume
396		Date: 2002/03/20 15:55:00 EST
397		Tag: blume-20020320-kmp
398		Description:
399
400		Implemented Knuth-Morris-Pratt string matching in PreString and used
401		it for String.isSubstring, Substring.isSubstring, and
402		Substring.position.
403
404		(Might need some stress-testing.  Simple examples worked fine.)
405
406		----------------------------------------------------------------------
407		Name: Matthias Blume
408		Date: 2002/03/19 16:37:00 EST
409		Tag: blume-20020319-witnesses
410		Description:
411
412		Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
413
414		This implements a generic mechanism for changing constness qualifiers
415		anywhere within big C types without resorting to outright "casts".
416		(So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
417		modify the constness at the outermost level.)
418		The implementation of "convert" is based on the idea of "witness"
419		values -- values that are not used by the operation but whose types
420		"testify" to their applicability.  On the implementation side, "convert"
421		is simply a projection (returning its second curried argument).  With
422		cross-module inlining, it should not result in any machine code being
423		generated.
424
425		----------------------------------------------------------------------
426		Name: Matthias Blume
427		Date: 2002/03/15 16:40:00 EST
428		Tag: blume-20020315-basis
429		Description:
430
431		Provided (preliminary?) implementations for
432
433		{String,Substring}.{concatWith,isSuffix,isSubstring}
434
435		and
436
437		Substring.full
438
439		Those are in the Basis spec but they were missing in SML/NJ.
440
441		----------------------------------------------------------------------
442		Name: Matthias Blume
443		Date: 2002/03/14 21:30:00 EST
444		Tag: blume-20020314-controls
445		Description:
446
447		Controls:
448		---------
449
450		1. Factored out the recently-added Controls : CONTROLS stuff and put
451		it into its own library $/controls-lib.cm.  The source tree for
452		this is under src/smlnj-lib/Controls.
453
454		2. Changed the names of types and functions in this interface, so they
455		make a bit more "sense":
456
457		module -> registry
458		'a registry -> 'a group
459
460		3. The interface now deals in ref cells only.  The getter/setter interface
461		is (mostly) gone.
462
463		4. Added a function that lets one register an already-existing ref cell.
464
465		5. Made the corresponding modifications to the rest of the code so that
466		everything compiles again.
467
468		6. Changed the implementation of Controls.MLRISC back to something closer
469		to the original.  In particular, this module (and therefore MLRISC)
470		does not depend on Controls.  There now is some link-time code in
471		int-sys.sml that registers the MLRISC controls with the Controls
472		module.
473
474		CM:
475		---
476
477		* One can now specify the lambda-split aggressiveness in init.cmi.
478
479		----------------------------------------------------------------------
480		Name: Allen Leung
481		Date: 2002/03/13 17:30:00 EST
482		Tag: leunga-20020313-x86-fp-unary
483		Description:
484
485		Bug fix for:
486
487		> leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
488		> Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
489		> - fun f(x,(y,z)) = Real.~ y;
490		> [autoloading]
491		> [autoloading done]
492		>       fchsl   (%eax), 184(%esp)
493		> Error: MLRisc bug: X86MCEmitter.emitInstr
494		>
495		> uncaught exception Error
496		>   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
497
498		The problem was that the code generator did not generate any fp registers
499		in this case, and the ra didn't know that it needed to run the X86FP phase to
500		translate the pseudo fp instruction.   This only happened with unary fp
501		operators in certain situations.
502
503		----------------------------------------------------------------------
504		Name: Matthias Blume
505		Date: 2002/03/13 14:00:00 EST
506		Tag: blume-20020313-overload-etc
507		Description:
508
509		1. Added _overload as a synonym for overload for backward compatibility.
510		(Control.overloadKW must be true for either version to be accepted.)
511
512		2. Fixed bug in install script that caused more things to be installed
513		than what was requested in config/targets.
514
515		3. Made CM aware of the (_)overload construct so that autoloading
516		works.
517
518		----------------------------------------------------------------------
519		Name: Matthias Blume
520		Date: 2002/03/12 22:03:00 EST
521		Tag: blume-20020312-url
522		Description:
523
524		Forgot to update BOOT and srcarchiveurl.
525
526		----------------------------------------------------------------------
527		Name: Matthias Blume
528		Date: 2002/03/12 17:30:00 EST
529		Tag: blume-20020312-version110392
530		Description:
531
532		Yet another version number bump (because of small changes to the
533		binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
534
535		Changes:
536
537		The new pid generation scheme described a few weeks ago was overly
538		complicated.  I implemented a new mechanism that is simpler and
539		provides a bit more "stability":  Once CM has seen a compilation
540		unit, it keeps its identity constant (as long as you do not delete
541		those crucial CM/GUID/* files).  This means that when you change
542		an interface, compile, then go back to the old interface, and
543		compile again, you arrive at the original pid.
544
545		There now also is a mechanism that instructs CM to use the plain
546		environment hash as a module's pid (effectively making its GUID
547		the empty string).  For this, "noguid" must be specified as an
548		option to the .sml file in question within its .cm file.
549		This is most useful for code that is being generated by tools such
550		as ml-nlffigen (because during development programmers tend to
551		erase the tool's entire output directory tree including CM's cached
552		GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
553		revert to the old, broken behavior of SML/NJ, but in specific cases
554		where there is no danger of interface confusion, its use is ok
555		(I think).
556
557		ml-nlffigen by default generates "noguid" annotations.  They can be
558		turned off by specifying -guid in its command line.
559
560		----------------------------------------------------------------------
561		Name: Lal George
562		Date: 2002/03/12 12 14:42:36 EST
563		Tag: george-20020312-frequency-computation
564		Description:
565
566		Integrated jump chaining and static block frequency into the
567		compiler. More details and numbers later.
568
569		----------------------------------------------------------------------
570		Name: Lal George
571		Date: 2002/03/11 11 22:38:53 EST
572		Tag: george-20020311-jump-chain-elim
573		Description:
574
575		Tested the jump chain elimination on all architectures (except the
576		hppa).  This is on by default right now and is profitable for the
577		alpha and x86, however, it may not be profitable for the sparc and ppc
578		when compiling the compiler.
579
580		The gc test will typically jump to a label at the end of the cluster,
581		where there is another jump to an external cluster containing the actual
582		code to invoke gc. This is to allow factoring of common gc invocation
583		sequences. That is to say, we generate:
584
585		f:
586		testgc
587		ja   L1      % jump if above to L1
588
589		L1:
590		jmp L2
591
592
593		After jump chain elimination the 'ja L1' instructions is converted to
594		'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
595		up being implemented in their long form (if L2 is far away) using:
596
597		jbe     L3      % jump if below or equal to L3
598		jmp     L2
599		L3:
600		...
601
602
603		For large compilation units L2  may be far away.
604
605
606		----------------------------------------------------------------------
607		Name: Matthias Blume
608		Date: 2002/03/11 13:30:00 EST
609		Tag: blume-20020311-mltreeeval
610		Description:
611
612		A functor parameter was missing.
613
614		----------------------------------------------------------------------
615		Name: Allen Leung
616		Date: 2002/03/11 10:30:00 EST
617		Tag: leunga-20020311-runtime-string0
618		Description:
619
620		The representation of the empty string now points to a
621		legal null terminated C string instead of unit.  It is now possible
622		to convert an ML string into C string with InlineT.CharVector.getData.
623		This compiles into one single machine instruction.
624
625		----------------------------------------------------------------------
626		Name: Allen Leung
627		Date: 2002/03/10 23:55:00 EST
628		Tag: leunga-20020310-x86-call
629		Description:
630
631		Added machine generation for CALL instruction (relative displacement mode)
632
633		----------------------------------------------------------------------
634		Name: Matthias Blume
635		Date: 2002/03/08 16:05:00
636		Tag: blume-20020308-entrypoints
637		Description:
638
639		Version number bumped to 110.39.1.  NEW BOOTFILES!
640
641		Entrypoints: non-zero offset into a code object where execution should begin.
642
643		- Added the notion of an entrypoint to CodeObj.
644		- Added reading/writing of entrypoint info to Binfile.
645		- Made runtime system bootloader aware of entrypoints.
646		- Use the address of the label of the first function given to mlriscGen
647		as the entrypoint.  This address is currently always 0, but it will
648		not be 0 once we turn on block placement.
649		- Removed the linkage cluster code (which was The Other Way(tm) of dealing
650		with entry points) from mlriscGen.
651
652		----------------------------------------------------------------------
653		Name: Allen Leung
654		Date: 2002/03/07 20:45:00 EST
655		Tag: leunga-20020307-x86-cmov
656		Description:
657
658		Bug fixes for CMOVcc on x86.
659
660		1. Added machine code generation for CMOVcc
661		2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
662		3. CMOVcc cannot have an immediate operand as argument.
663
664		----------------------------------------------------------------------
665		Name: Matthias Blume
666		Date: 2002/03/07 16:15:00 EST
667		Tag: blume-20020307-controls
668		Description:
669
670		This is a very large but mostly boring patch which makes (almost)
671		every tuneable compiler knob (i.e., pretty much everything under
672		Control.* plus a few other things) configurable via both the command
673		line and environment variables in the style CM did its configuration
674		until now.
675
676		Try starting sml with '-h' (or, if you are brave, '-H')
677
678		To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
679		implements the underlying generic mechanism.
680
681		The interface to some of the existing such facilities has changed somewhat.
682		For example, the MLRiscControl module now provides mkFoo instead of getFoo.
683		(The getFoo interface is still there for backward-compatibility, but its
684		use is deprecated.)
685
686		The ml-build script passes -Cxxx=yyy command-line arguments through so
687		that one can now twiddle the compiler settings when using this "batch"
688		compiler.
689
690		TODO items:
691
692		We should go through and throw out all controls that are no longer
693		connected to anything.  Moreover, we should go through and provide
694		meaningful (and correct!) documentation strings for those controls
695		that still are connected.
696
697		Currently, multiple calls to Controls.new are accepted (only the first
698		has any effect).  Eventually we should make sure that every control
699		is being made (via Controls.new) exactly once.  Future access can then
700		be done using Controls.acc.
701
702		Finally, it would probably be a good idea to use the getter-setter
703		interface to controls rather than ref cells.  For the time being, both
704		styles are provided by the Controls module, but getter-setter pairs are
705		better if thread-safety is of any concern because they can be wrapped.
706
707		*****************************************
708
709		One bug fix: The function blockPlacement in three of the MLRISC
710		backpatch files used to be hard-wired to one of two possibilities at
711		link time (according to the value of the placementFlag).  But (I
712		think) it should rather sense the flag every time.
713
714		*****************************************
715
716		Other assorted changes (by other people who did not supply a HISTORY entry):
717
718		1. the cross-module inliner now works much better (Monnier)
719		2. representation of weights, frequencies, and probabilities in MLRISC
720		changed in preparation of using those for weighted block placement
721		(Reppy, George)
722
723		----------------------------------------------------------------------
724		Name: Lal George
725		Date: 2002/03/07 14:44:24 EST 2002
726		Tag: george-20020307-weighted-block-placement
727
728		Tested the weighted block placement optimization on all architectures
729		(except the hppa) using AMPL to generate the block and edge frequencies.
730		Changes were required in the machine properties to correctly
731		categorize trap instructions. There is an MLRISC flag
732		"weighted-block-placement" that can be used to enable weighted block
733		placement, but this will be ineffective without block/edge
734		frequencies (coming soon).
735
736
737		----------------------------------------------------------------------
738		Name: Lal George
739		Date: 2002/03/05 17:24:48 EST
740		Tag: george-20020305-linkage-cluster
741
742		In order to support the block placement optimization, a new cluster
743		is generated as the very first cluster (called the linkage cluster).
744		It contains a single jump to the 'real' entry point for the compilation
745		unit. Block placement has no effect on the linkage cluster itself, but
746		all the other clusters  have full freedom in the manner in which they
747		reorder blocks or functions.
748
749		On the x86 the typical linkage code that is generated is:
750		----------------------
751		.align 2
752		L0:
753		addl    $L1-L0, 72(%esp)
754		jmp     L1
755
756
757		.align  2
758		L1:
759		----------------------
760
761		72(%esp) is the memory location for the stdlink register. This
762		must contain the address of the CPS function being called. In the
763		above example, it contains the address of  L0; before
764		calling L1 (the real entry point for the compilation unit), it
765		must contain the address for L1, and hence
766
767		addl $L1-L0, 72(%esp)
768
769		I have tested this on all architectures except the hppa.The increase
770		in code size is of course negligible
771
772		----------------------------------------------------------------------
773		Name: Allen Leung
774		Date: 2002/03/03 13:20:00 EST
775		Tag: leunga-20020303-mlrisc-tools
776
777		Added #[ ... ] expressions to mlrisc tools
778
779		----------------------------------------------------------------------
780		Name: Matthias Blume
781		Date: 2002/02/27 12:29:00 EST
782		Tag: blume-20020227-cdebug
783		Description:
784
785		- made types in structure C and C_Debug to be equal
786		- got rid of code duplication (c-int.sml vs. c-int-debug.sml)
787		- there no longer is a C_Int_Debug (C_Debug is directly derived from C)
788
789		----------------------------------------------------------------------
790		Name: Matthias Blume
791		Date: 2002/02/26 12:00:00 EST
792		Tag: blume-20020226-ffi
793		Description:
794
795		1. Fixed a minor bug in CM's "noweb" tool:
796		If numbering is turned off, then truly don't number (i.e., do not
797		supply the -L option to noweb).  The previous behavior was to supply
798		-L'' -- which caused noweb to use the "default" line numbering scheme.
799		Thanks to Chris Richards for pointing this out (and supplying the fix).
800
801		2. Once again, I reworked some aspects of the FFI:
802
803		A. The incomplete/complete type business:
804
805		- Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
806		gone!
807		- ML types representing an incomplete type are now *equal* to
808		ML types representing their corresponding complete types (just like
809		in C).  This is still safe because ml-nlffigen will not generate
810		RTTI for incomplete types, nor will it generate functions that
811		require access to such RTTI.   But when ML code generated from both
812		incomplete and complete versions of the C type meet, the ML types
813		are trivially interoperable.
814
815		NOTE:  These changes restore the full generality of the translation
816		(which was previously lost when I eliminated functorization)!
817
818		B. Enum types:
819
820		- Structure C now has a type constructor "enum" that is similar to
821		how the "su" constructor works.  However, "enum" is not a phantom
822		type because each "T enum" has values (and is isomorphic to
823		MLRep.Signed.int).
824		- There are generic access operations for enum objects (using
825		MLRep.Signed.int).
826		- ml-nlffigen will generate a structure E_foo for each "enum foo".
827		* The structure contains the definition of type "mlrep" (the ML-side
828		representation type of the enum).  Normally, mlrep is the same
829		as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
830		then mlrep will be defined as a datatype -- thus facilitating
831		pattern matching on mlrep values.
832		("-ec" will be suppressed if there are duplicate values in an
833		enumeration.)
834		* Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
835		will be generated for each C enum constant xxx.
836		* Conversion functions m2i and i2m convert between mlrep and
837		MLRep.Signed.int.  (Without "-ec", these functions are identities.)
838		* Coversion functions c and ml convert between mlrep and "tag enum".
839		* Access functions (get/set) fetch and store mlrep values.
840		- By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
841		enumerations are merged into one single enumeration represented by
842		structure E_'.
843
844		----------------------------------------------------------------------
845		Name: Allen Leung
846		Date: 2002/02/25 04:45:00 EST
847		Tag: leunga-20020225-cps-spill
848
849		This is a new implementation of the CPS spill phase.
850		The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
851		In case of problems, replace it with the old file spill.sml
852
853		The current compiler runs into some serious performance problems when
854		constructing a large record.  This can happen when we try to compile a
855		structure with many items.  Even a very simple structure like the following
856		makes the compiler slow down.
857
858		structure Foo = struct
859		val x_1 = 0w1 : Word32.int
860		val x_2 = 0w2 : Word32.int
861		val x_3 = 0w3 : Word32.int
862		...
863		val x_N = 0wN : Word32.int
864		end
865
866		The following table shows the compile time, from N=1000 to N=4000,
867		with the old compiler:
868
869		N
870		1000   CPS 100 spill                           0.04u  0.00s  0.00g
871		MLRISC ra                               0.06u  0.00s  0.05g
872		(spills = 0 reloads = 0)
873		TOTAL                                   0.63u  0.07s  0.21g
874
875		1100   CPS 100 spill                           8.25u  0.32s  0.64g
876		MLRISC ra                               5.68u  0.59s  3.93g
877		(spills = 0 reloads = 0)
878		TOTAL                                   14.71u  0.99s  4.81g
879
880		1500   CPS 100 spill                           58.55u  2.34s  1.74g
881		MLRISC ra                               5.54u  0.65s  3.91g
882		(spills = 543 reloads = 1082)
883		TOTAL                                   65.40u  3.13s  6.00g
884
885		2000   CPS 100 spill                           126.69u  4.84s  3.08g
886		MLRISC ra                               0.80u  0.10s  0.55g
887		(spills = 42 reloads = 84)
888		TOTAL                                   129.42u  5.10s  4.13g
889
890		3000   CPS 100 spill                           675.59u  19.03s  11.64g
891		MLRISC ra                               2.69u  0.27s  1.38g
892		(spills = 62 reloads = 124)
893		TOTAL                                   682.48u  19.61s  13.99g
894
895		4000   CPS 100 spill                           2362.82u  56.28s  43.60g
896		MLRISC ra                               4.96u  0.27s  2.72g
897		(spills = 85 reloads = 170)
898		TOTAL                                   2375.26u  57.21s  48.00g
899
900		As you can see the old cps spill module suffers from some serious
901		performance problem.  But since I cannot decipher the old code fully,
902		instead of patching the problems up, I'm reimplementing it
903		with a different algorithm.  The new code is more modular,
904		smaller when compiled, and substantially faster
905		(O(n log n) time and O(n) space).  Timing of the new spill module:
906
907		4000  CPS 100 spill                           0.02u  0.00s  0.00g
908		MLRISC ra                               0.25u  0.02s  0.15g
909		(spills=1 reloads=3)
910		TOTAL                                   7.74u  0.34s  1.62g
911
912		Implementation details:
913
914		As far as I can tell, the purpose of the CPS spill module is to make sure the
915		number of live variables at any program point (the bandwidth)
916		does not exceed a certain limit, which is determined by the
917		size of the spill area.
918
919		When the bandwidth is too large, we decrease the register pressure by
920		packing live variables into spill records.  How we achieve this is
921		completely different than what we did in the old code.
922
923		First, there is something about the MLRiscGen code generator
924		that we should be aware of:
925
926		o MLRiscGen performs code motion!
927
928		In particular, it will move floating point computations and
929		address computations involving only the heap pointer to
930		their use sites (if there is only a single use).
931		What this means is that if we have a CPS record construction
932		statement
933
934		RECORD(k,vl,w,e)
935
936		we should never count the new record address w as live if w
937		has only one use (which is often the case).
938
939		We should do something similar to floating point, but the transformation
940		there is much more complex, so I won't deal with that.
941
942		Secondly, there are now two new cps primops at our disposal:
943
944		1. rawrecord of record_kind option
945		This pure operator allocates some uninitialized storage from the heap.
946		There are two forms:
947
948		rawrecord NONE [INT n]  allocates a tagless record of length n
949		rawrecord (SOME rk) [INT n] allocates a tagged record of length n
950		and initializes the tag.
951
952		2. rawupdate of cty
953		rawupdate cty (v,i,x)
954		Assigns to x to the ith component of record v.
955		The storelist is not updated.
956
957		We use these new primops for both spilling and increment record construction.
958
959		1. Spilling.
960
961		This is implemented with a linear scan algorithm (but generalized
962		to trees).  The algorithm will create a single spill record at the
963		beginning of the cps function and use rawupdate to spill to it,
964		and SELECT or SELp to reload from it.  So both spills and reloads
965		are fine-grain operations.  In contrast, in the old algorithm
966		"spills" have to be bundled together in records.
967
968		Ideally, we should sink the spill record construction to where
969		it is needed.  We can even split the spill record into multiple ones
970		at the places where they are needed.  But CPS is not a good
971		representation for global code motion, so I'll keep it simple and
972		am not attempting this.
973
974		2. Incremental record construction (aka record splitting).
975
976		Long records with many component values which are simulatenously live
977		(recall that single use record addresses are not considered to
978		be live) are constructed with rawrecord and rawupdate.
979		We allocate space on the heap with rawrecord first, then gradually
980		fill it in with rawupdate.  This is the technique suggested to me
981		by Matthias.
982
983		Some restrictions on when this is applicable:
984		1. It is not a VECTOR record.  The code generator currently does not handle
985		this case. VECTOR record uses double indirection like arrays.
986		2. All the record component values are defined in the same "basic block"
987		as the record constructor.  This is to prevent speculative
988		record construction.
989
990		----------------------------------------------------------------------
991		Name: Allen Leung
992		Date: 2002/02/22 01:02:00 EST
993		Tag: leunga-20020222-mlrisc-tools
994
995		Minor bug fixes in the parser and rewriter
996
997		----------------------------------------------------------------------
998		Name: Allen Leung
999		Date: 2002/02/21 20:20:00 EST
1000		Tag: leunga-20020221-peephole
1001
1002		Regenerated the peephole files.  Some contained typos in the specification
1003		and some didn't compile because of pretty printing bugs in the old version
1004		of 'nowhere'.
1005
1006		----------------------------------------------------------------------
1007		Name: Allen Leung
1008		Date: 2002/02/19 20:20:00 EST
1009		Tag: leunga-20020219-mlrisc-tools
1010		Description:
1011
1012		Minor bug fixes to the mlrisc-tools library:
1013
1014		1.  Fixed up parsing colon suffixed keywords
1015		2.  Added the ability to shut the error messages up
1016		3.  Reimplemented the pretty printer and fixed up/improved
1017		the pretty printing of handle and -> types.
1018		4.  Fixed up generation of literal symbols in the nowhere tool.
1019		5.  Added some SML keywords to to sml.sty
1020
1021		----------------------------------------------------------------------
1022		Name: Matthias Blume
1023		Date: 2002/02/19 16:20:00 EST
1024		Tag: blume-20020219-cmffi
1025		Description:
1026
1027		A wild mix of changes, some minor, some major:
1028
1029		* All C FFI-related libraries are now anchored under $c:
1030		$/c.cm      --> $c/c.cm
1031		$/c-int.cm  --> $c/internals/c-int.cm
1032		$/memory.cm --> $c/memory/memory.cm
1033
1034		* "make" tool (in CM) now treats its argument pathname slightly
1035		differently:
1036		1. If the native expansion is an absolute name, then before invoking
1037		the "make" command on it, CM will apply OS.Path.mkRelative
1038		(with relativeTo = OS.FileSys.getDir()) to it.
1039		2. The argument will be passed through to subsequent phases of CM
1040		processing without "going native".  In particular, if the argument
1041		was an anchored path, then "make" will not lose track of that anchor.
1042
1043		* Compiler backends now "know" their respective C calling conventions
1044		instead of having to be told about it by ml-nlffigen.  This relieves
1045		ml-nlffigen from one of its burdens.
1046
1047		* The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1048
1049		* Export C_DEBUG and C_Debug from $c/c.cm.
1050
1051		* C type encoding in ml-nlffi-lib has been improved to model the conceptual
1052		subtyping relationship between incomplete pointers and their complete
1053		counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1054		with the convention of instantiating 'o with ('t, 'c) obj whenever
1055		the pointer target type is complete.  In the incomplete case, 'o
1056		will be instantiated with some "'c iobj" -- a type obtained by
1057		using one of the functors PointerToIncompleteType or PointerToCompleteType.
1058
1059		Operations that work on both incomplete and complete pointer types are
1060		typed as taking an 'o ptr while operations that require the target to
1061		be known are typed as taking some ('t, 'c) obj ptr.
1062
1063		voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1064		where void is an eqtype without any values.  This makes it possible
1065		to work on voidptr values using functions meant to operate on light
1066		incomplete pointers.
1067
1068		* As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1069		been vastly simplified.
1070
1071		----------------------------------------------------------------------
1072		Name: Matthias Blume
1073		Date: 2002/02/19 10:48:00 EST
1074		Tag: blume-20020219-pqfix
1075		Description:
1076
1077		Applied Chris Okasaki's bug fix for priority queues.
1078
1079		----------------------------------------------------------------------
1080		Name: Matthias Blume
1081		Date: 2002/02/15 17:05:00
1082		Tag: Release_110_39
1083		Description:
1084
1085		Last-minute retagging is becoming a tradition... :-(
1086
1087		This is the working release 110.39.
1088
1089		----------------------------------------------------------------------
1090		Name: Matthias Blume
1091		Date: 2002/02/15 16:00:00 EST
1092		Tag: Release_110_39-orig
1093		Description:
1094
1095		Working release 110.39.  New bootfiles.
1096
1097		(Update: There was a small bug in the installer so it wouldn't work
1098		with all shells.  So I retagged. -Matthias)
1099
1100		----------------------------------------------------------------------
1101		Name: Matthias Blume
1102		Date: 2002/02/15 14:17:00 EST
1103		Tag: blume-20020215-showbindings
1104		Description:
1105
1106		Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1107		the latter can be useful for exploring what bindings are available at
1108		the interactive prompt.  (The first function returns only the list
1109		of symbols that are really bound, the second prints those but also the
1110		ones that CM's autoloading mechanism knows about.)
1111
1112		----------------------------------------------------------------------
1113		Name: Matthias Blume
1114		Date: 2002/02/15 12:08:00 EST
1115		Tag: blume-20020215-iptrs
1116		Description:
1117
1118		Two improvements to ml-nlffigen:
1119
1120		1. Write files only if they do not exist or if their current contents
1121		do not coincide with what's being written.  (That is, avoid messing
1122		with the time stamps unless absolutely necessary.)
1123
1124		2. Implement a "repository" mechanism for generated files related
1125		to "incomplete pointer types".   See the README file for details.
1126
1127		----------------------------------------------------------------------
1128		Name: Matthias Blume
1129		Date: 2002/02/14 11:50:00 EST
1130		Tag: blume-20020214-quote
1131		Description:
1132
1133		Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1134		because of the new and improved tag generation scheme.  (Thanks to Allen
1135		Leung for pointing it out.)
1136
1137		----------------------------------------------------------------------
1138		Name: Lal George
1139		Date: 2002/02/14 09:55:27 EST 2002
1140		Tag: george-20020214-isabelle-bug
1141		Description:
1142
1143		Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1144		of Isabelle on the x86.
1145
1146		From Allen:
1147		-----------
1148		I've found the problem:
1149
1150		in ra-core.sml, I use the counter "blocked" to keep track of the
1151		true number of elements in the freeze queue.  When the counter goes
1152		to zero, I skip examining the queue.  But I've messed up the
1153		bookkeeping in combine():
1154
1155		else ();
1156		case !ucol of
1157		PSEUDO => (if !cntv > 0 then
1158		(if !cntu > 0 then blocked := !blocked - 1 else ();
1159		^^^^^^^^^^^^^^^^^^^^^^^
1160		moveu := mergeMoveList(!movev, !moveu)
1161		)
1162		else ();
1163
1164		combine() is called to coalesce two nodes u and v.
1165		I think I was thinking that if the move counts of u and v are both
1166		greater than zero then after they are coalesced then one node is
1167		removed from the freeze queue.  Apparently I was thinking that
1168		both u and v are of low degree, but that's clearly not necessarily true.
1169
1170
1171		02/12/2002:
1172		Here's the patch.  HOL now compiles.
1173
1174		I don't know how this impact on performance (compile
1175		time or runtime).  This bug caused the RA (especially on the x86)
1176		to go thru the potential spill phase when there are still nodes on the
1177		freeze queue.
1178
1179
1180
1181
1182		----------------------------------------------------------------------
1183		Name: Matthias Blume
1184		Date: 2002/02/13 22:40:00 EST
1185		Tag: blume-20020213-fptr-rtti
1186		Description:
1187
1188		Fixed a bug in ml-nlffigen that was introduced with one of the previous
1189		updates.
1190
1191		----------------------------------------------------------------------
1192		Name: Matthias Blume
1193		Date: 2002/02/13 16:41:00 EST
1194		Tag: blume-20020213-cmlpq
1195		Description:
1196
1197		Added new priority queue export symbols (which have just been added to
1198		smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1199		not compile and the installer would choke.)
1200
1201		----------------------------------------------------------------------
1202		Name: Matthias Blume
1203		Date: 2002/02/13 16:15:00 EST
1204		Tag: blume-20020213-various
1205		Description:
1206
1207		1. More tweaks to ml-nlffigen:
1208
1209		- better internal datastructures (resulting in slight speedup)
1210		- "-match" option requires exact match
1211		- "localized" gensym counters (untagged structs/unions nested within
1212		other structs/unions or within typedefs get a fresh counter; their
1213		tag will be prefixed by a concatenation of their parents' tags)
1214		- bug fixes (related to calculation of transitive closure of types
1215		to be included in the output)
1216
1217		2. Minor Basis updates:
1218
1219		- added implementations for List.collate and Option.app
1220
1221		----------------------------------------------------------------------
1222		Name: Matthias Blume
1223		Date: 2002/02/11 15:55:00 EST
1224		Tag: blume-20020211-gensym
1225		Description:
1226
1227		Added a "-gensym" option to command line of ml-nlffigen.  This can be
1228		used to specify a "stem" -- a string that is inserted in all "gensym'd"
1229		names (ML structure names that correspond to unnamed C structs, unions,
1230		and enums), so that separate runs of ml-nlffigen do not clash.
1231
1232		----------------------------------------------------------------------
1233		Name: Matthias Blume
1234		Date: 2002/02/11 12:05:00 EST
1235		Tag: blume-20020211-gensml
1236		Description:
1237
1238		A quick fix for a problem with GenSML (in the pgraph-util library):
1239		Make generation of toplevel "local" optional.  (Strictly speaking,
1240		signature definitions within "local" are not legal SML.)
1241
1242		Other than that: updates to INSTALL and cm/TODO.
1243
1244		----------------------------------------------------------------------
1245		Name: Matthias Blume
1246		Date: 2002/02/08 15:00:00 EST
1247		Tag: blume-20020208-uniquepid
1248		Description:
1249
1250		0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1251
1252		1. The installer (config/install.sh) has gotten smarter:
1253
1254		- Configuration options are a bit easier to specify now
1255		(in config/targets).
1256		- Bug in recognizing .tar.bz2 files fixed.
1257		- Installer automatically resolves dependencies between
1258		configuration options (e.g., if you ask for eXene, you will
1259		also get cml -- regardless whether you asked for it or not).
1260		- Installer can run in "quieter mode" by setting the environment
1261		variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1262		"completely silent", though.
1263		- Build HashCons library as part of smlnj-lib.
1264
1265		2. A new scheme for assigning persistent identifiers to compilation
1266		units (and, by extension, to types etc.) has been put into place.
1267		This fixes a long-standing bug where types and even dynamic values
1268		can get internally confused, thereby compromising type safety
1269		(abstraction) and dynamic correctness.  See
1270
1271		http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1272
1273		for an example of how things could go wrong until now.
1274
1275		The downside of the new scheme is that pids are not quite as
1276		persistent as they used to be: CM will generate a fresh pid
1277		for every compilation unit that it thinks it sees for the first
1278		time.  That means that if you compile starting from a clean, fresh
1279		source tree at two different times, you end up with different
1280		binaries.
1281
1282		Cutoff recompilation, however, has not been compromised because
1283		CM keeps pid information in special caches between runs.
1284
1285		----------------------------------------------------------------------
1286		Name: Lal George
1287		Date: 2002/02/07 15:34:13 EST 2002
1288		Tag: <none>
1289		Description:
1290
1291		Compilers that generate assembly code may produce  global labels
1292		whose value is resolved at link time. The various peephole optimization
1293		modules did not take this in account.
1294
1295		TODO. The Labels.addrOf function should really return an option
1296		type so that clients are forced to deal with this issue, rather
1297		than an exception being raised.
1298
1299		----------------------------------------------------------------------
1300		Name: Lal George
1301		Date: 2002/02/06 13:55:02 EST
1302		Tag: george-20020206-ra-breakup
1303		Description:
1304
1305		1. A bug fix from Allen.
1306
1307		A typo causes extra fstp %st(0)'s to be generated at compensation
1308		edges, which might cause stack underflow traps at runtime.  This
1309		occurs in fft where there are extraneous fstps right before the 'into'
1310		trap instruction (in this case they are harmless since none of the
1311		integers overflow.)
1312
1313		2. Pulled out various utility modules that were embedded in the modules
1314		of the register allocator. I need these modules for other purposes, but
1315		they are not complete enough to put into a library (just yet).
1316		----------------------------------------------------------------------
1317		Name: Matthias Blume
1318		Date: 2002/01/31 16:05:00 EST
1319		Tag: blume-20020131-sparc-ccalls
1320		Description:
1321
1322		1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1323		of extra stack space by mistake.  Fixed.
1324
1325		2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1326
1327		----------------------------------------------------------------------
1328		Name: Allen Leung
1329		Date: 2002/01/30
1330		Tag: leunga-20020130-nowhere-bug-fix
1331		Description:
1332
1333		MLRISC bug fixes:
1334		1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1335		2. MachineInt.fromString was negating its value.
1336
1337		----------------------------------------------------------------------
1338		Name: Matthias Blume
1339		Date: 2002/01/29
1340		Tag: blume-20020129-INSTALL
1341		Description:
1342
1343		- Added somewhat detailed installation instructions (file INSTALL).
1344		- Fixed curl-detection bug in config/install.sh.
1345		- It is now possible to select the URL getter using the URLGETTER
1346		environment variable:
1347
1348		not set / "unknown"      --> automatic detection (script tries wget,
1349		curl, and lynx)
1350		"wget" / "curl" / "lynx" --> use the specified program (script "knows"
1351		how to properly invoke them)
1352		other                    --> use $URLGETTER directly, it must take
1353		precisely two command-line arguments
1354		(source URL and destination file name)
1355
1356		----------------------------------------------------------------------
1357		Name: Matthias Blume
1358	Date: 2002/01/28	Date: 2002/01/28
1359	Tag: blume-20020128-sparc-ccalls	Tag: blume-20020128-sparc-ccalls
1360	Description:	Description:
#	Line 593	Line 1934
1934
1935	----------------------------------------------------------------------	----------------------------------------------------------------------
1936	Name: Matthias Blume	Name: Matthias Blume
	>>>>>>> 1.169
1937	Date: 2001/09/18 15:35:00 EDT	Date: 2001/09/18 15:35:00 EDT
1938	Tag: blume-20010918-readme11036	Tag: blume-20010918-readme11036
1939	Description:	Description:

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.1045   changed lines   Added in v.1186

---

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