Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 583, Thu Mar 23 21:52:30 2000 UTC	revision 697, Tue Aug 8 03:33:13 2000 UTC
#	Line 12	Line 12
12	Tag: <post-commit CVS tag>	Tag: <post-commit CVS tag>
13	Description:	Description:
14	----------------------------------------------------------------------	----------------------------------------------------------------------
15		Name: Matthias Blume
16		Date: 2000/08/08 12:33:00 JST
17		Tag: blume-20000808-manual
18		Description:
19
20		Tiny update to CM manual.
21
22		----------------------------------------------------------------------
23		Name: Allen Leung
24		Date: 2000/08/7 19:31:00
25		Tag: leunga-20000807-a-whole-bunch-of-stuff
26
27		Moby, C--, SSA, x86, machine descriptions etc.  Should only affect C--
28		and Mobdy.
29
30		1.  x86
31
32		a.  Fixes to peephole module by John and Dan.
33		b.  Assembly fix to SETcc by Allen.
34		c.  Fix to c-call by John.
35		d.  Fix to spilling by John.  (This one deals with the missing FSTPT case)
36		e.  Instruction selection optimization to SETcc as suggested by John.
37
38		For example,
39
40		MV(32, x, COND(32, CMP(32, LT, a, b), LI 1, LI 0))
41
42		should generate:
43
44		MOVL a, x
45		SUBL b, x
46		SHRL 31, x
47
48		2.  IR stuff
49
50		A bunch of new DJ-graph related algorithms added.  These
51		speed up SSA construction.
52
53		3.  SSA + Scheduling
54
55		Added code for SSA and scheduling to the repository
56
57		----------------------------------------------------------------------
58		Name: Lal George
59		Date: 2000/07/27 11:53:14 EDT
60
61		Tag: lal-20000727-linux-ppc
62		Description:
63
64		Made changes to support Linux PPC.
65		p.s. I have confirmation that the 110.29 boot files work fine.
66
67		----------------------------------------------------------------------
68		Name: Matthias Blume
69		Date: 2000/07/27 17:40:00 JST
70		Tag: blume-20000727-scripts
71		Description:
72
73		!!!! WARNING !!!!
74		You must recompile the runtime system!
75		!!!! WARNING !!!!
76
77		This is basically another round of script-enhancements:
78
79		1. sml, ml-build, and ml-makedepend accept options -D and -U to define
80		and undefine CM preprocessor symbols.
81
82		2. ml-build avoids generating a new heap image if it finds that the
83		existing one is still ok.  (The condition is that no ML file had to
84		be recompiled and all ML files are found to be older that the heap
85		file.)
86
87		To make this work smoothly, I also hacked the runtime system as
88		well as SMLofNJ.SysInfo to get access to the heap image suffix
89		(.sparc-solaris, ...) that is currently being used.
90
91		Moreover, the signature of CM.mk_standalone has changed.  See the
92		CM manual.
93
94		3. ml-makedepend accepts additional options -n, -a, and -o.  (See the
95		CM manual for details.)
96
97		4. More CM manual updates:
98		- all of the above has been documented.
99		- there is now a section describing the (CM-related) command line
100		arguments that are accepted by the "sml" command
101
102		----------------------------------------------------------------------
103		Name: Matthias Blume
104		Date: 2000/07/25 16:20:00 JST
105		Tag: blume-20000725-makedepend
106		Description:
107
108		Added a script called ml-makedepend.  This can be used in makefiles
109		for Unix' make in a way very similar to the "makedepend" command for
110		C.
111
112		The script internally uses function CM.sources.
113
114		Synopsis:
115
116		ml-makedepend [-f makefile] cmfile targetname
117
118		The default for the makefile is "makefile" (or "Makefile" should
119		"makefile" not exist).
120
121		ml-makedepend adds a cmfile/targetname-specific section to this
122		makefile (after removing the previous version of this section).  The
123		section contains a single dependency specification with targetname on
124		the LHS (targetname is an arbitrary name), and a list of files derived
125		from the cmfile on the RHS.  Some of the files on the RHS are
126		ARCH/OPSYS-specific.  Therefore, ml-makedepend inserts references to
127		"make" variables $(ARCH) and $(OPSYS) in place of the corresponding
128		path names.  The makefile writer is responsible for making sure that
129		these variables have correct at the time "make" is invoked.
130
131		----------------------------------------------------------------------
132		Name: Matthias Blume
133		Date: 2000/07/22 23:30:00 JST
134		Tag: blume-20000722-urlupdate
135		Description:
136
137		Changed BOOT and config/srcarchiveurl to point to BL server:
138
139		ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.29/
140
141		----------------------------------------------------------------------
142		Name: Matthias Blume
143		Date: 2000/07/18 18:00:00 JST
144		Tag: blume-20000718-Version_110_29
145		Description:
146
147		1. Updated src/compiler/TopLevel/main/version.sml to version 110.29
148
149		2. Updated config/version to 110.29
150
151		3. Updated config/srcarchiveurl
152
153		3. New boot files!
154		ftp://ftp.cs.princeton.edu/pub/people/blume/sml/110.29-autofetch
155
156		----------------------------------------------------------------------
157		Name: Matthias Blume
158		Date: 2000/07/11 13:58:00 JST
159		Tag: blume-20000711-doctypo
160		Description:
161
162		Fixed a few typos in CM manual.
163
164		----------------------------------------------------------------------
165		Name: Allen Leung
166		Date: 2000/06/15 00:38:00
167		Tag: leunga-20000704-sparc-x86
168
169		1. x86 peephole improvement sp += k; sp -= k => nop  [from John]
170		2. fix to x86 RET bug [found by Dan Grossman]
171		3. sparc assembly bug fix for ticc instructions [found by Fermin]
172
173		Affects c-- and moby only
174
175		----------------------------------------------------------------------
176		Name: Matthias Blume
177		Date: 2000/07/04 15:26:00
178		Tag: blume-20000704-trigger
179		Description:
180
181		1. Improvements to CM manual.
182		2. SMLofNJ.Internals.BTrace.trigger reinstated as an alternative way
183		of getting a back-trace.  The function, when called, raises an
184		internal exception which explicitly carries the full back-trace history,
185		so it is unaffected by any intervening handle-raise pairs ("trivial"
186		or not).  The interactive loop will print that history once it arrives
187		at top level.
188		Short of having all exceptions implicitly carry the full history, the
189		recommended way of using this facility is:
190		- compile your program with instrumentation "on"
191		- run it, when it raises an exception, look at the history
192		- if the history is "cut off" because of some handler, go and modify
193		your program so that it explicitly calls BTrace.trigger
194		- recompile (still instrumented), and rerun; look at the full history
195
196		----------------------------------------------------------------------
197		Name: Matthias Blume
198		Date: 2000/07/03 15:36:00 JST
199		Tag: blume-20000702-manual
200		Description:
201
202		Small corrections and updates to CM manual.
203
204		----------------------------------------------------------------------
205		Name: Matthias Blume
206		Date: 2000/06/29 16:04:00 JST
207		Tag: blume-20000629-yacctool
208		Description:
209
210		Changes:
211
212		1. Class "mlyacc" now takes separate arguments to pass options to
213		generated .sml- and .sig-files independently.
214		2. Corresponding CM manual updates.
215		3. BTrace module now also reports call sites.  (However, for loop clusters
216		it only shows from where the cluster was entered.)  There are associated
217		modifications to core.sml, internals.{sig,sml}, btrace.sml, and btimp.sml.
218
219		----------------------------------------------------------------------
220		Name: Matthias Blume
221		Date: 2000/06/27 16:51:00 JST
222		Tag: blume-20000627-noweb
223		Description:
224
225		Changes:
226
227		1. Implemented "subdir" and "witness" options for noweb tool.
228		This caused some slight internal changes in CM's tool implementation.
229		2. Fixed bug in "tool plugin" mechanism.  This is essentially cleaning
230		some remaining issues from earlier path anchor changes.
231		3. Updated CM manual accordingly.
232
233		4. Changed implementation of back-tracing so that I now consider it
234		ready for prime-time.
235
236		In particular, you don't have to explicitly trigger the back-trace
237		anymore.  Instead, if you are running BTrace-instrumented code and
238		there is an uncaught exception (regardless of whether or not it was
239		raised in instrumented code), the top-level evalloop will print
240		the back-trace.
241
242		Features:
243
244		- Instrumented and uninstrumented code work together seemlessly.
245		(Of course, uninstrumented code is never mentioned in actual
246		back-traces.)
247
248		- Asymptotic time- and space-complexity of instrumented code is
249		equal to that of uninstrumented code.  (This means that
250		tail-recursion is preserved by the instrumentation phase.)
251
252		- Modules whose code has been instrumented in different sessions
253		work together without problem.
254
255		- There is no penalty whatsoever on uninstrumented code.
256
257		- There is no penalty on "raise" expressions, even in
258		instrumented code.
259
260		A potential bug (or perhaps it is a feature, too):
261
262		A back-trace reaches no further than the outermost instrumented
263		non-trivial "raise".  Here, a "trivial" raise is one that is the
264		sole RHS of a "handle" rule.  Thus, back-traces reach trough
265
266		<exp> handle e => raise e
267
268		and even
269
270		<exp> handle Foo => raise Bar
271
272		and, of course, through
273
274		<exp> handle Foo => ...
275
276		if the exception was not Foo.
277
278		Back-traces always reach right through any un-instrumented code
279		including any of its "handle" expressions, trivial or not.
280
281		To try this out, do the following:
282
283		- Erase all existing binfiles for your program.
284		(You may keep binfiles for those modules where you think you
285		definitely don't need back-tracing.)
286		- Turn on back-trace instrumentation:
287		SMLofNJ.Internals.BTrace.mode (SOME true);
288		- Recompile your program.  (I.e., run "CM.make" or "use".)
289		- You may now turn instrumentation off again (if you want):
290		SMLofNJ.Internals.BTrace.mode (SOME false);
291		- Run your program as usual.  If it raises an exception that
292		reaches the interactive toplevel, then a back-trace will
293		automatically be printed.  After that, the toplevel loop
294		will print the exception history as usual.
295
296		----------------------------------------------------------------------
297		Name: Matthias Blume
298		Date: 2000/06/26 09:56:46 JST
299		Tag: blume-20000626-setup
300		Description:
301
302		CM: - setup-parameter to "sml" added; this can be used to run arbitrary
303		ML code before and after compiling a file (e.g., to set compiler
304		flags)
305
306		Compiler: - improved btrace API (in core.sml, internals.{sig,sml})
307		- associated changes to btrace.sml (BTrace instrumentation pass)
308		- cleaner implementation of btimp.sml (BTrace tracing and report
309		module)
310
311		CM manual: * new path encoding documented
312		* description of setup-parameter to "sml" added
313
314		The biggest user-visible change to back-tracing is that it is no
315		longer necessary to compile all traced modules within the same
316		session.  (This was a real limitation.)
317
318		----------------------------------------------------------------------
319		Name: Matthias Blume
320		Date: 2000/06/24 12:40:00 JST
321		Tag: blume-20000624-startup
322		Description:
323
324		Fixes startup slowdown problem.  (I was calling SrcPath.sync a _tad_
325		bit too often -- to put it mildly. :)
326
327		----------------------------------------------------------------------
328		Name: Matthias Blume
329		Date: 2000/06/23 18:20:00 JST
330		Tag: blume-20000623-btrace
331		Description:
332
333		This updates adds a backtrace facility to aid programmers in debugging
334		their programs.  This involves the following changes:
335
336		1. Module system/smlnj/init/core.sml (structure _Core) now has hooks for
337		keeping track of the current call stack.  When programs are compiled
338		in a special mode, the compiler will insert calls to these hooks
339		into the user program.
340		"Hook" means that it is possible for different implementations of
341		back-tracing to register themselves (at different times).
342
343		2. compiler/MiscUtil/profile/btrace.sml implements the annotation phase
344		as an Absyn.dec->Absyn.dec rewrite.  Normally this phase is turned off.
345		It can be turned on using this call:
346		SMLofNJ.Internals.BTrace.mode (SOME true);
347		Turning it off again:
348		SMLofNJ.Internals.BTrace.mode (SOME false);
349		Querying the current status:
350		SMLofNJ.Internals.BTrace.mode NONE;
351		Annotated programs are about twice as big as normal ones, and they
352		run a factor of 2 to 4 slower with a dummy back-trace plugin (one
353		where all hooks do nothing).  The slowdown with a plugin that is
354		actually useful (such as the one supplied by default) is even greater,
355		but in the case of the default plugin it is still only an constant
356		factor (amortized).
357
358		3. system/Basis/Implementation/NJ/internals.{sig,sml} have been augmented
359		with a sub-structure BTrace for controlling back-tracing.  In particular,
360		the above-mentioned function "mode" controls whether the annotation
361		phase is invoked by the compiler.  Another important function is
362		"trigger": when called it aborts the current execution and causes
363		the top-level loop to print a full back-trace.
364
365		4. compiler/MiscUtil/profile/btimp.sml is the current default plugin
366		for back-tracing.  It keeps track of the dynamic call stack and in
367		addition to that it keeps a partial history at each "level" of that
368		stack.  For example, if a tail-calls b, b tail-calls c, and c tail-calls
369		d and b (at separate times, dynamically), then the report will show:
370
371		GOTO   d
372		/c
373		GOTO  \b
374		CALL   a
375
376		This shows that there was an initial non-tail call of a, then a
377		tail-call to b or c, looping behavior in a cluster of functions that
378		consist of b and c, and then a goto from that cluster (i.e., either from
379		b or from c) to d.
380
381		Note that (depending on the user program) the amount of information
382		that the back-trace module has to keep track of at each level is bounded
383		by a constant.  Thus, the whole implementation has the same asymptotical
384		complexity as the original program (both in space and in time).
385
386		5. compiler/TopLevel/interact/evalloop.sml has been modified to
387		handle the special exception SMLofNJ.Internals.BTrace.BTrace
388		which is raised by the "trigger" function mentioned above.
389
390		Notes on usage:
391
392		- Annotated code works well together with unannotated code:
393		Unannotated calls simply do not show up at all in the backtrace.
394
395		- It is not a good idea to let modules that were annotated during
396		different sessions run at the same time.  This is because the compiler
397		chooses small integers to identify individual functions, and there
398		will be clashes if different modules were compiled in separate sessions.
399		(Nothing will crash, and you will even be told about the clashes, but
400		back-trace information will in general not be useful.)
401
402		- Back-tracing can be confused by callcc and capture.
403
404		- The only way of getting a back-trace right now is to explicitly
405		invoke the "trigger" function from your user program.  Eventually, we
406		should make every exception carry back-trace information (if
407		available).  But since this creates more overhead at "raise"-time
408		(similar to the current exnHistory overhead), I have not yet
409		implemented this.  (The implementation will be rather easy.)  With
410		exceptions carrying back-trace information, this facility will be even
411		more useful because users don't need to modify their programs...
412
413		- While it is possible to compile the compiler with back-trace
414		annotations turned on (I did it to get some confidence in
415		correctness), you must make absolutely sure that core.sml and
416		btimp.sml are compiled WITHOUT annotation!  (core.sml cannot actually
417		be compiled with annotation because there is no core access yet, but
418		if you compile btimp.sml with annotation, then the system will go into
419		an infinite recursion and crash.)
420		Since CM currently does not know about BTrace, the only way to turn
421		annotations on and off for different modules of the compiler is to
422		interrupt CMB.make, change the settings, and re-invoke it.  Of course,
423		this is awkward and clumsy.
424
425		Sample sessions:
426
427		Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
428		- SMLofNJ.Internals.BTrace.mode (SOME true);
429		[autoloading]
430		[autoloading done]
431		val it = false : bool
432		- structure X = struct
433		-     fun main n = let
434		-         fun a (x, 0) = d x
435		-           | a (x, n) = b (x, n - 1)
436		-         and b (x, n) = c (x, n)
437		-         and c (x, n) = a (x, n)
438		-         and d x = e (x, 3)
439		-         and e (x, 0) = f x
440		-           | e (x, n) = e (x, n - 1)
441		-         and f 0 = SMLofNJ.Internals.BTrace.trigger ()
442		-           | f n = n * g (n - 1)
443		-         and g n = a (n, 3)
444		-     in
445		-         f n
446		-     end
447		- end;
448		structure X : sig val main : int -> int end
449		- X.main 3;
450		*** BACK-TRACE ***
451		GOTO   stdIn:4.2-13.20: X.main[2].f
452		GOTO-( stdIn:4.2-13.20: X.main[2].e
453		GOTO   stdIn:4.2-13.20: X.main[2].d
454		/ stdIn:4.2-13.20: X.main[2].a
455		| stdIn:4.2-13.20: X.main[2].b
456		GOTO-\ stdIn:4.2-13.20: X.main[2].c
457		CALL   stdIn:4.2-13.20: X.main[2].g
458		GOTO   stdIn:4.2-13.20: X.main[2].f
459		GOTO-( stdIn:4.2-13.20: X.main[2].e
460		GOTO   stdIn:4.2-13.20: X.main[2].d
461		/ stdIn:4.2-13.20: X.main[2].a
462		| stdIn:4.2-13.20: X.main[2].b
463		GOTO-\ stdIn:4.2-13.20: X.main[2].c
464		CALL   stdIn:4.2-13.20: X.main[2].g
465		GOTO   stdIn:4.2-13.20: X.main[2].f
466		GOTO-( stdIn:4.2-13.20: X.main[2].e
467		GOTO   stdIn:4.2-13.20: X.main[2].d
468		/ stdIn:4.2-13.20: X.main[2].a
469		| stdIn:4.2-13.20: X.main[2].b
470		GOTO-\ stdIn:4.2-13.20: X.main[2].c
471		CALL   stdIn:4.2-13.20: X.main[2].g
472		GOTO   stdIn:4.2-13.20: X.main[2].f
473		CALL   stdIn:2.15-17.4: X.main[2]
474		-
475
476		(Note that because of a FLINt bug the above code currently does not
477		compile without BTrace turned on.)
478
479		Here is another example, using my modified Tiger compiler:
480
481		Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
482		- SMLofNJ.Internals.BTrace.mode (SOME true);
483		[autoloading]
484		[autoloading done]
485		val it = false : bool
486		- CM.make "sources.cm";
487		[autoloading]
488		...
489		[autoloading done]
490		[scanning sources.cm]
491		[parsing (sources.cm):parse.sml]
492		[creating directory CM/SKEL ...]
493		[parsing (sources.cm):tiger.lex.sml]
494		...
495		[wrote CM/sparc-unix/semant.sml]
496		[compiling (sources.cm):main.sml]
497		[wrote CM/sparc-unix/main.sml]
498		[New bindings added.]
499		val it = true : bool
500		- Main.compile ("../testcases/merge.tig", "foo.out");
501		*** BACK-TRACE ***
502		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trvar
503		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
504		CALL   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
505		GOTO   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
506		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
507		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
508		CALL   lib/semant.sml:488.3-505.6: SemantFun[2].transDec.trdec[2].transBody[2]
509		/ lib/semant.sml:411.65-543.8: SemantFun[2].transDec
510		CALL-\ lib/semant.sml:413.2-540.9: SemantFun[2].transDec.trdec[2]
511		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
512		CALL   lib/semant.sml:8.52-558.4: SemantFun[2].transProg[2]
513		CALL   main.sml:1.18-118.4: Main.compile[2]
514		-
515
516		----------------------------------------------------------------------
517		Name: Matthias Blumen
518		Date: 2000/06/21 18:00:00 JST
519		Tag: blume-20000621-manual
520		Description:
521
522		CM manual update: Path environments documented.
523
524		----------------------------------------------------------------------
525		Name: Matthias Blume
526		Date: 2000/06/19 13:40:00
527		Tag: blume-20000619-manual
528		Description:
529
530		CM manual and system/README update.  This only covers the fact that
531		there are no more implicit anchors.  (Path environments and the "bind"
532		option to "cm" have yet to be documented.)
533
534		----------------------------------------------------------------------
535		Name: Matthias Blume
536		Date: 2000/06/19 11:05:00 JST
537		Tag: blume-20000619-chdir-bugfix
538		Description:
539
540		Fixed a bug in new SrcPath module that sometimes led to a bad chDir call.
541
542		----------------------------------------------------------------------
543		Name: Matthias Blume
544		Date: 2000/06/18 22:00:10 JST
545		Tag: blume-20000618-implicit-anchors-really-gone
546		Description:
547
548		I updates the previous HISTORY entry where I forgot to mention that
549		implicit anchors are no longer with us.
550
551		The current update also gets rid of the (now useless) controller
552		CM.Control.implicit_anchors.
553
554		----------------------------------------------------------------------
555		Name: Matthias Blume
556		Date: 2000/06/16 17:30:00 JST
557		Tag: blume-20000616-anchorenv
558		Description:
559
560		This patch implements the long anticipated (just kidding :) "anchor
561		environment" mechanism.  In the course of doing this, I also
562		re-implemented CM's internal "SrcPath" module from scratch.  The new
563		one should be more robust in certain boundary cases.  In any case, it
564		is a lot cleaner than its predecessor (IMHO).
565
566		This time, although there is yet another boot file format change, I
567		kept the unpickler backward-compatible.  As a result, no new bootfiles
568		are necessary and bootstrapping is straightforward.  (You cannot read
569		new bootfiles into an old system, but the other way around is no
570		problem.)
571
572		Visible changes:
573
574		** 0. Implicit path anchors (without the leading $-symbol) are no
575		longer recognized at all. This means that such path names are not
576		illegal either.  For example, the name basis.cm simply refers to a
577		local file called "basis.cm" (i.e, the name is an ordinary path
578		relative to .cm-files directory).  Or, to put it differently, only
579		names that start with $ are anchored paths.
580
581		** 1. The $<singlearc> abbreviation for $/<singlearc> has finally
582		vanished.
583
584		John (Reppy) had critizised this as soon as I originally proposed and
585		implemented it, but at that time I did not really deeply believe
586		him. :) Now I came full-circle because I need the $<singlearc> syntax
587		in another place where it cannot be seen as an abbreviation for
588		$/<singlearc>.  To avoid the confusion, $<singlearc> now means what it
589		seems to mean (i.e., it "expands" into the corresponding anchor
590		value).
591
592		However, when paths are used as members in CM description files, it
593		continues to be true that there must be at least another arc after the
594		anchor.  This is now enforced separately during semantic analysis
595		(i.e., from a lexical/syntactical point of view, the notation is ok.)
596
597		** 2. The "cm" class now accepts an option "bind".  The option's value
598		is a sub-option list of precisely two items -- one labeled "anchor"
599		and the other one labeled "value".  As you might expect, "anchor" is
600		used to specify an anchor name to be bound, and "value" specifies what
601		the anchor is being bound to.
602
603		The value must be a directory name and can be given in either standard
604		syntax (including the possibility that it is itself an anchored path)
605		or native syntax.
606
607		Examples:
608
609		foo.cm (bind:(anchor:bar value:$mystuff/bar))
610		lib.cm (bind:(anchor:a value:"H:\\x\\y\\z"))  (* only works under windows *)
611
612		and so on.
613
614		The meaning of this is that the .cm-file will be processed with an
615		augmented anchor environment where the given anchor(s) is/are bound to
616		the given values(s).
617
618		The rationale for having this feature is this: Suppose you are trying
619		to use two different (already stable) libraries a.cm and b.cm (that
620		you perhaps didn't write yourself).  Further, suppose each of these
621		two libraries internally uses its own auxiliary library $aux/lib.cm.
622		Normally you would now have a problem because the anchor "lib" can not
623		be bound to more than one value globally.  Therefore, the project that
624		uses both a.cm and b.cm must locally redirect the anchor to some other
625		place:
626
627		a.cm (bind:(anchor:lib value:/usr/lib/smlnj/a-stuff))
628		b.cm (bind:(anchor:lib value:/usr/lib/smlnj/b-stuff))
629
630		This hard-wires $lib/aux.cm to /usr/lib/smlnj/a-stuff/aux.cm or
631		/usr/lib/smlnj/b-stuff/aux.cm, respectively.
632
633		Hard-wiring path names is a bit inflexible (and CM will verbosely warn
634		you when you do so at the time of CM.stabilize).  Therefore, you can
635		also use an anchored path as the value:
636
637		a.cm (bind:(anchor:lib value:$a-lib))
638		b.cm (bind:(anchor:lib value:$b-lib))
639
640		Now you can globally configure (using the usual CM.Anchor.anchor or
641		pathconfig machinery) bindings for "a-lib" and "b-lib".  Since "lib"
642		itself is always locally bound, setting it globally is no longer
643		meaningful or necessary (but it does not hurt either).  In fact, "lib"
644		can still be used as a global anchor for separate purposes.  As a
645		matter of fact, one can locally define "lib" in terms of a global
646		"lib":
647
648		a.cm (bind:(anchor:lib value:$lib/a))
649		b.cm (bind:(anchor:lib value:$lib/b))
650
651		** 3: The encoding of path names has changed.  This affects the way
652		path names are shown in CM's progress report and also the internal
653		protocol encoding used for parallel make.
654
655		The encoding now uses one or more ':'-separated segments.  Each
656		segments corresponds to a file that has been specified relative to the
657		file given by its preceding segment.  The first segment is either
658		relative to the CWD, absolute, or anchored.  Each segment itself is
659		basically a Unix pathname; all segments but the first are relative.
660
661		Example:
662
663		$foo/bar/baz.cm:a/b/c.sml
664
665		This path denotes the file bar/a/b/c.sml relative to the directory
666		denoted by anchor "foo".  Notice that the encoding also includes
667		baz.cm which is the .cm-file that listed a/b/c.sml.  As usual, such
668		paths are resolved relative to the .cm-files directory, so baz.cm must
669		be ignored to get the "real" pathname.
670
671		To make this fact more obvious, CM puts the names of such "virtual
672		arcs" into parentheses when they appear in progress reports. (No
673		parentheses will appear in the internal protocol encoding.)  Thus,
674		what you really see is:
675
676		$foo/bar/(baz.cm):a/b/c.sml
677
678		I find this notation to be much more informative than before.
679
680		Another new feature of the encoding is that special characters
681		including parentheses, colons, (back)slashes, and white space are
682		written as \ddd (where ddd is the decimal encoding of the character).
683
684		*** The CM manual still needs to be updated.
685
686		----------------------------------------------------------------------
687		Name: Allen Leung
688		Date: 2000/06/15 00:38:00
689		Tag: leunga-20000615-x86-peephole
690
691		x86 Peephole fix by Fermin.  Affects c-- and moby only.
692
693		----------------------------------------------------------------------
694		Name: Matthias Blume
695		Date: 2000/06/12 11:40:00
696		Tag: blume-20000612-parmakefix
697		Description:
698
699		More cleanup after changing the file naming scheme: This time I
700		repaired the parallel make mechanism for CMB.make which I broke earlier.
701
702		----------------------------------------------------------------------
703		Name: Allen Leung
704		Date: 2000/06/09 01:25:00
705		Tag: leunga-20000609-various
706
707		None of these things should affect normal SML/NJ operations
708
709		1. Peephole improvements provided by Fermin (c--)
710		2. New annotation DEFUSE for adding extra dependence (moby)
711		3. New X86 LOCK instructions (moby)
712		4. New machine description language for reservation tables (scheduling)
713		5. Fixes to various optimization/analysis modules (branch chaining, dominator
714		trees etc.)
715		6. I've changed the CM files so that they can work with versions
716		110.0.6, 110.25 and 110.28
717
718		----------------------------------------------------------------------
719		Name: Matthias Blume
720		Date: 2000/06/09 12:40:00
721		Tag: blume-20000609-log
722		Description:
723
724		- Removed all(?) remaining RCS Log entries from sources.
725
726		- Fixed bug in ml-yacc and ml-lex sources (use explicit anchors for
727		anchored paths).
728
729		----------------------------------------------------------------------
730		Name: Matthias Blume
731		Date: 2000/06/07 17:00:00 JST
732		Tag: blume-20000607-no-implicit-anchors
733		Description:
734
735		1. This update changes the default setting for
736		CM.Control.implicit_anchors from true to false.  This means that
737		implicit anchors are no longer permitted by default.  I also tried to
738		make sure that nothing else still relies on implicit anchors.
739		(This is the next step on the schedule towards a CM that does not even
740		have the notion of implicit anchors anymore.)
741
742		2. More CM manual updates.
743
744		3. I managed to track down and fix the pickling bug I mentioned last
745		time.  Because of the previously existing workaround, this entails no
746		immediate practical changes.
747
748		----------------------------------------------------------------------
749		Name: Matthias Blume
750		Date: 2000/06/06 11:15:00 JST
751		Tag: blume-20000606-lazierpickle
752		Description:
753
754		!!!! NEW BOOT FILES !!!!
755
756		* The main purpose of this update is to make library pickles lazier in
757		order to reduce the initial space penalty for autoloading a library.
758		As a result, it is now possible to have $smlnj/compiler.cm
759		pre-registered.  This should take care of the many complaints or
760		inquiries about missing structure Compiler.  This required changes to
761		CM's internal data structures and small tweaks to some algorithms.
762
763		As a neat additional effect, it is no longer necessary (for the sake
764		of lean heap image files) to distinguish between a "minimal" CM and a
765		"full" CM.  Now, there is only one CM (i.e., the "full" version:
766		$smlnj/cm.cm aka $smlnj/cm/full.cm), and it is always available at the
767		interactive top level. ($smlnj/cm/minimal.cm is gone.)
768
769		To make the life of compiler-hackers easier, "makeml" now also
770		pre-registers $smlnj/cmb.cm (aka $smlnj/cmb/current.cm).  In other
771		words, after you bootstrap a new sml for the first time, you will not
772		have to autoload $smlnj/cmb.cm again afterwards.  (The first time
773		around you will still have to do it, though.)
774
775		* A second change consists of major updates to the CM manual.  There
776		are now several appendices with summary information and also a full
777		specification of the CM description file syntax.
778
779		* In directory src/system I added the script "allcross".  This script
780		invokes sml and cross-compiles the compiler for all supported
781		architectures.  (Useful when providing a new set of boot files.)
782
783		* There seems to be a latent bug in my "lazy pickles" mechanism.  I
784		added a small tweak to pickle-util.sml to work around this problem,
785		but it is not a proper fix yet.  I will investigate further.  (The
786		effect of the bug was an inflation of library pickle size.)
787
788		* Version number increased to 110.28.1 (to avoid compatibility problems).
789
790		----------------------------------------------------------------------
791		Name: Allen Leung
792		Date: 2000/05/25 17:28 EDT
793		Tag: leunga-20000525-ra
794		Description:
795
796		Fixed a bug in freezing phase of the register allocator.
797
798		----------------------------------------------------------------------
799		Name: Allen Leung
800		Date: 2000/05/15 22:53 EDT
801		Tag: leunga-20000515-alpha-x86-ra
802		Description:
803
804		1. Alpha
805
806		Slight cleanup.  Removed the instruction SGNXL
807
808		2. X86
809
810		Added the following instructions to the instruction set:
811
812		ROLx, RORx,
813		BTx, BTSx, BTLx, BTRx,
814		XCHGx, and variants with the LOCK prefix
815
816		3. Register Allocation
817
818		The module ra-rewrite-with-renaming has been improved.
819
820		These have no effect on SML/NJ.
821
822		----------------------------------------------------------------------
823		Name: Matthias Blume
824		Date: 2000/05/15 16:20:00 JST
825		Tag: blume-20000515-lightrebuild
826		Description:
827
828		1. I added an alternative to "-rebuild" to "makeml".  The difference is
829		that prior to calling CMB.make' the CM-variable "LIGHT" will be
830		defined.  In effect, the command will not build any cross-compiler
831		backends and therefore finish more quickly.
832
833		The "fixpt" script also takes a "-light" switch to be able to use
834		this new facility while compiling for a fixpoint.
835
836		2. I replaced all mentions of anchored paths in group owner specifications
837		with simple relative paths (usually starting with "..").
838		The rationale is that a library's internal workings should not be
839		compromised by the lack of some anchor.  (An anchor is necessary
840		for someone who wants to refer to the library by an anchored path,
841		but it should not be necessary to build the same library in the first
842		place.)
843
844		3. I changed the way CM's tool mechanism determines the shell command
845		string used for things like ml-yacc etc. so that it does not break
846		when CM.Control.implicit_anchors is turned off.
847
848		----------------------------------------------------------------------
849		Name: Matthias Blume
850		Date: 2000/05/12 18:20:00 JST
851		Tag: blume-20000512-ml-build
852		Description:
853
854		Fixed a bug in config/_ml-build that prevented ml-yacc and ml-lex from
855		getting installed properly (by config/install.sh).
856
857		----------------------------------------------------------------------
858		Name: Matthias Blume
859		Date: 2000/05/12 17:30:00 JST
860		Tag: blume-20000512-anchors
861		Description:
862
863		!!! NEW BOOT FILES !!!
864
865		This change is in preparation of fading out support for "implicitly
866		anchored path names".  I went through all sources and used the
867		explicit (and relatively new) $-notation.  See system/README and the
868		CM manual for more info on this.
869
870		I also modified the anchoring scheme for some things such as "smlnj",
871		"MLRISC", "cm", etc. to take advantage of the fact that explicit
872		anchors are more expressive: anchor name and first arc do not have to
873		coincide.  This entails the following user-visible change:
874
875		You have to write $smlnj/foo/bar instead of smlnj/foo/bar.  In
876		particular, when you fire up sml with a command-line argument, say,
877		e.g.:
878
879		sml '$smlnj/cmb.cm'
880
881		At the ML toplevel prompt:
882
883		CM.autoload "$smlnj/cmb.cm";
884
885		There is also a new controller in CM.Control that can be used to turn
886		off all remaining support for implicit anchors by saying:
887
888		CM.autoload "$smlnj/
889		#set CM.Control.implicit_anchors false;
890
891		This causes CM to reject implicitly anchored paths.  This is (for the
892		time being) less permissive than the "final" version where there will
893		be no more such implicit anchors and relative paths will be just that:
894		relative.
895
896		The next step (version after next version?) will be to make the
897		default for CM.Control.implicit_anchors false.  After the dust has
898		settled, I can then produce the "final" version of this...
899
900		Note: Since bootstrapping is a bit tricky, I provided new boot files.
901
902		----------------------------------------------------------------------
903		Name: Matthias Blume
904		Date: 2000/05/11 16:30:00 JST
905		Tag: blume-20000511-sources
906		Description:
907
908		The main change is that I added function CM.sources as a generalized
909		version of the earlier CM.makedepend.  This entails the following
910		additional changes:
911
912		- CM.makedepend has been dropped.
913
914		- CM manual has been updated.
915
916		- TOOLS signature and API have been changed.
917
918		----------------------------------------------------------------------
919		Name: Allen Leung
920		Date: 2000/05/10 21:17 EDT
921		Tag: leunga-20000510-moby-c--ssa
922		Description:
923
924		Various bug fixes and new features for C--, Moby and MLRISC optimizations.
925		None of these affect SML/NJ.
926
927		1. Register Allocation
928
929		a. A new ra spilling module (ra/ra-spill-with-renaming) is implemented.
930		This module tries to remove local (i.e. basic block level) redundancies
931		during spilling.
932
933		b. A new framework for performing region based register allocation.
934		Not yet entirely functional.
935
936		2. X86
937
938		a. DefUse for POP was missing the stack pointer [found by Lal]
939		b. Reload for CALL was incorrect in X86Spill [found by John]
940		c. Various fixes in X86Spill so that it can be used correctly for
941		the new spilling module.
942
943		3. SSA/IR
944
945		a. New module ir/dj-dataflow.sml implements elimination based
946		data flow analysis.
947
948		4. MLRiscGen
949
950		a. Fix for gc type annotation
951
952		5. MDGen
953
954		Various fixes for machine description -> ml code translation.  For ssa
955		only.
956
957		----------------------------------------------------------------------
958		Name: Allen Leung
959		Date: 2000/05/08 22:17 EDT
960		Tag: leunga-20000508-labexp
961		Description:
962
963		Fermin has found a few assembly problems with constant expressions
964		generated in LabelExp.  Mostly, the problems involve extra parentheses,
965		which choke on dumb assemblers.  This is his fix.
966
967		----------------------------------------------------------------------
968		Name: Dave MacQueen
969		Date: 2000/04/09 14:00 EDT
970		Tag: dbm-20000502-Version_110_28
971		Description:
972
973		1. Updated src/compiler/TopLevel/main/version.sml to version 110.28
974
975		2. Updated config/version to 110.28
976
977		3. Updated config/srcarchiveurl
978
979		3. New boot files!
980		ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.28/
981
982		----------------------------------------------------------------------
983		Name: Matthias Blume
984		Date: 2000/05/01 19:05:00 JST
985		Tag: blume-20000501-noweb
986		Description:
987
988		A new noweb tool has been added.  The existing system is entirely
989		unaffected by this, but some CM users have asked for renewed noweb
990		support.  Everything is documented in the CM manual.
991
992		New (plugin) libraries:
993
994		noweb-tool.cm
995		nw-ext.cm
996
997		----------------------------------------------------------------------
998		Name: Dave MacQueen
999		Date: 2000/04/30 12:40PM EDT
1000		Tag: dbm-20000430-bug_fixes
1001		Description:
1002
1003		1. Fix for bug 1498
1004		smlnj/src/system/Basis/Implementation/Unsafe/object.sig
1005		smlnj/src/system/Basis/Implementation/Unsafe/object.sml
1006		added toRealArray function
1007		smlnj/src/compiler/MiscUtil/print/ppobj.sml
1008		added check for tag Obj.RealArray to array printing case in ppObj
1009
1010		2. Fix for bug 1510
1011		smlnj/src/compiler/Semant/types/typesutil.sml
1012		fixed definition of dummyargs (used by equalTycon) so that
1013		dummy args are distinct types
1014
1015		----------------------------------------------------------------------
1016		Name: Matthias Blume
1017		Date: 2000/04/30 01:00:00 JST
1018		Tag: blume-20000430-versions
1019		Description:
1020
1021		1. CM version numbering added.  This is an implementation of Lal's
1022		proposal for adding version numbers and version checking to .cm
1023		files.  Lal said that his proposal was just that -- a proposal.
1024		For the time being I went ahead and implemented it so that people
1025		can comment on it.  Everything is completely backward-compatible
1026		(except for the stable library format, i.e., new bootfiles!).
1027
1028		As usual, see the CM manual for details.
1029
1030		2. An alternative syntax for anchored paths has been implemented.
1031		Dave has recently voiced the same concerns that I had when I did
1032		this, so there should be some support.  My take is that eventually
1033		I will let support for the current syntax (where anchors are
1034		"implicit") fade out in favor of the new, explicit syntax.
1035		In order to be backward-compatible, both old and new syntax are
1036		currently supported.
1037
1038		Again, see the CM manual for details.
1039
1040		3. Parallel make is trying to be slightly smarter:  When the master
1041		process finds a "bottleneck", i.e., when there is only one
1042		compilation unit that can be compiled and everybody else is
1043		waiting on it, then it will simply compile it directly instead
1044		of clumsily telling one of the slaves to do it.
1045
1046		4. Support for "unsharing" added.  This is necessary in order to be
1047		able to have two different versions of the same library running
1048		at the same time (e.g., for trying out a new MLRISC while still
1049		having the old MLRISC linked into the current compiler, etc.)
1050		See the CM manual.
1051
1052		5. Simple "makedepend" functionality added for generating Makefile
1053		dependency information.  (This is rather crude at the moment.
1054		Expect some changes here in the future.)
1055
1056		6. ".fun" added as a recognized suffix for ML files. Also documented
1057		explicitly in the manual that the fallback behavior (unknown suffix
1058		-> ML file) is not an official feature!
1059
1060		7. Small changes to the pickler for stable libraries.
1061
1062		8. Several internal changes to CM (for cleanup/improvement).
1063
1064
1065		!!!! NEW BINFILES !!!!
1066
1067		----------------------------------------------------------------------
1068		Name: Matthias Blume
1069		Date: 2000/04/28 17:30:00 JST
1070		Tag: blume-20000428-pathconfig
1071		Description:
1072
1073		1. I changed config/install.sh to remove duplicate entries from the
1074		lib/pathconfig file at the end.  Moreover, the final version of
1075		lib/pathconfig is sorted alphabetically.  The same (sorting) is done
1076		in src/system/installml.
1077
1078		2. The config/install.sh script now consistently uses relative
1079		pathnames in lib/pathconfig whenever the anchor is in the lib
1080		directory.  (So far this was true for the libraries that come
1081		pre-compiled and bundled as part of the bootfiles but not for
1082		libraries that are compiled by the script itself.)
1083
1084		----------------------------------------------------------------------
1085		Name: Matthias Blume
1086		Date: 2000/04/26 13:10:00 JST
1087		Tag: blume-20000426-fun_suffix
1088		Description:
1089
1090		Added ".fun" as a recognized file name suffix (for ML code).
1091
1092		----------------------------------------------------------------------
1093		Name: Allen Leung
1094		Date: 2000/04/25 17:00:00 EST
1095		Tag: leunga-20000425-alpha-ra
1096		Description:
1097
1098		1. Alpha
1099
1100		PSEUDOARITH was missing in AlphaRewrite.  This causes an endless loop
1101		in C--.
1102
1103		2. RA
1104
1105		Added a flag "ra-dump-size" to print out the size of the flowgraph
1106		and the interference graph.
1107
1108		----------------------------------------------------------------------
1109		Name: Dave MacQueen
1110		Date: 2000/04/25/
1111		Tag: dbm-20000425-mlyacc_doc_examples
1112		Description:
1113		Updated mlyacc.tex sections 5 and 7 for SML '97 and CM.
1114		Updated all three examples in src/ml-yacc/examples to run
1115		under 110.* using CM.make.
1116
1117		----------------------------------------------------------------------
1118		Name: Allen Leung
1119		Date: 2000/04/20 23:04:00 EST
1120		Tag: leunga-20000420-ssa-c---stuff
1121		Description:
1122
1123		This update synchronizes my repository with Yale's.  Most of these
1124		changes, however, do not affect SML/NJ at all (the RA is an exception).
1125
1126		1. Register Allocator
1127
1128		a. An improvement in the interference graph construction:
1129		Given a copy
1130
1131		s <- t
1132
1133		no interference edge between s and t is added for this definition of s.
1134
1135		b. I've added two new spill heuristic modules that Fermin and I developed
1136		(in the new library RA.cm). These are unused in SML/NJ but maybe
1137		useful for others (Moby?)
1138
1139		2. X86
1140
1141		a. Various fixes in the backend provided by Fermin [C--] and Lal.
1142
1143		3. Alpha
1144
1145		a. Added the BSR instruction and code generation that goes with it [C--]
1146		b. Other fixes too numerous to recount provided by Fermin [C--]
1147
1148		4. Regmaps
1149
1150		a. The regmaps are not initialized with the identity physical bindings
1151		at creation time.  This is unneeded.
1152
1153		5. MLRISC Optimizations
1154
1155		a. The DJ-Graph module can now compute the iterated dominance frontiers
1156		intersects with liveness incrementally in linear time! Woohoo!
1157		This is now used in my new SSA construction algorithm.
1158
1159		b. THe branch reorganization module is now smarter about linear chains of
1160		basic blocks.
1161
1162
1163		----------------------------------------------------------------------
1164		Name: Matthias Blume
1165		Date: 2000/04/12 13:52:00 JST
1166		Tag: blume_main_v110p27_1
1167		Description:
1168
1169		Changed install.sh script to handle archive files without version number
1170		and to use "boot.<arch>-<os>" instead of "sml.boot.<arch>-<os>" for the
1171		name of the boot file archive.
1172
1173		----------------------------------------------------------------------
1174		Name: Dave MacQueen
1175		Date: 2000/04/09 14:00 EDT
1176		Tag: dbm-20000410-Version_110_27
1177		Description:
1178
1179		1. Updated src/compiler/TopLevel/main/version.sml to version 110.27
1180
1181		2. Updated src/config/version to 110.27
1182
1183		3. New boot files!
1184
1185		----------------------------------------------------------------------
1186		Name: Allen Leung
1187		Date: 2000/04/09 19:09:00 EST
1188		Tag: leunga-20000409-misc
1189		Description:
1190
1191		1.  Yet another fix for x86 assembly for idivl, imull, mull and friends.
1192
1193		2.  Miscellaneous improvements to MLRISC (unused in sml/nj)
1194
1195		----------------------------------------------------------------------
1196		Name: Stefan
1197		Date: 2000/04/07 10:00:00 EDT
1198		Tag: monnier-20000406-branch-handling
1199		Description:
1200
1201		Improved handling of branches (mostly those generated from
1202		polymorphic equality), removed switchoff and changed the
1203		default optimization settings (more cpsopt and less flintopt).
1204
1205		----------------------------------------------------------------------
1206		Name: Allen Leung
1207		Date: 2000/04/06 01:30:00 EST
1208		Tag: leunga-20000406-peephole-x86-SSA-2
1209		Description:
1210
1211		Forgot a few files.
1212
1213		----------------------------------------------------------------------
1214		Name: Allen Leung
1215		Date: 2000/04/06 00:36:00 EST
1216		Tag: leunga-20000406-peephole-x86-SSA
1217		Description:
1218
1219		1.  New Peephole code
1220
1221		2.  Minor improvement to X86 instruction selection
1222
1223		3.  Various fixes to SSA and machine description -> code translator
1224
1225		----------------------------------------------------------------------
1226		Name: Matthias Blume
1227		Date: 2000/04/05 12:30:00 JST
1228		Tag: blume_main_v110p26p2_3
1229		Description:
1230
1231		This update just merges three minor cosmetic updates to CM's sources
1232		to get ready for the 110.27 code freeze on Friday.  No functionality
1233		has changed.
1234
1235		----------------------------------------------------------------------
1236		Name: Allen Leung
1237		Date: 2000/04/04 19:39:00 EST
1238		Tag: leunga-20000404-x86-asm
1239		Description:
1240
1241		1.  Fixed a problem in X86 assembly.
1242
1243		Things like
1244
1245		jmp %eax
1246		jmp (%eax)
1247
1248		should be output as
1249
1250		jmp *%eax
1251		jmp *(%eax)
1252
1253		2.  Assembly output
1254
1255		Added a new flag
1256
1257		"asm-indent-copies" (default to false)
1258
1259		When this flag is on, parallel copies will be indented an extra level.
1260
1261		----------------------------------------------------------------------
1262		Name: Allen Leung
1263		Date: 2000/04/04 03:18:00 EST
1264		Tag: leunga-20000404-C--Moby
1265		Description:
1266
1267		All of these fixes are related to C--, Moby, and my own optimization
1268		stuff; so they shouldn't affect SML/NJ.
1269
1270		1.  X86
1271
1272		Various fixes related floating point, and extensions.
1273
1274		2.  Alpha
1275
1276		Some extra patterns related to loads with signed/zero extension
1277		provided by Fermin.
1278
1279		3.  Assembly
1280
1281		When generating assembly, resolve the value of client defined constants,
1282		instead of generating symbolic values.  This is controlled by the
1283		new flag "asm-resolve-constants", which is default to true.
1284
1285		4.  Machine Descriptions
1286
1287		a. The precedence parser was slightly broken when parsing infixr symbols.
1288		b. The type generalizing code had the bound variables reversed, resulting
1289		in a problem during arity raising.
1290		c. Various fixes in machine descriptions.
1291
1292		----------------------------------------------------------------------
1293		Name: Matthias Blume
1294		Date: 2000/04/03 16:05:00 JST
1295		Tag: blume_main_v110p26p2_2
1296		Description:
1297
1298		I eliminated coreEnv from compInfo.  Access to the "Core" structure is
1299		now done via the ordinary static environment that is context to each
1300		compilation unit.
1301
1302		To this end, I arranged that instead of "structure Core" as "structure
1303		_Core" is bound in the pervasive environment.  Core access is done via
1304		_Core (which can never be accidentally rebound because _Core is not a
1305		legal surface-syntax symbol).
1306
1307		The current solution is much cleaner because the core environment is
1308		now simply part of the pervasive environment which is part of every
1309		compilation unit's context anyway.  In particular, this eliminates all
1310		special-case handling that was necessary until now in order to deal
1311		with dynamic and symbolic parts of the core environment.
1312
1313		Remaining hackery (to bind the "magic" symbol _Core) is localized in the
1314		compilation manager's bootstrap compiler (actually: in the "init group"
1315		handling).  See the comments in src/system/smlnj/init/init.cmi for
1316		more details.
1317
1318		I also tried to track down all mentions of "Core" (as string argument
1319		to Symbol.strSymbol) in the compiler and replaced them with a
1320		reference to the new CoreSym.coreSym.  Seems cleaner since the actual
1321		name appears in one place only.
1322
1323		Binfile and bootfile format have not changed, but the switchover from
1324		the old "init.cmi" to the new one is a bit tricky, so I supplied new
1325		bootfiles anyway.
1326
1327		----------------------------------------------------------------------
1328		Name: Allen Leung
1329		Date: 2000/04/02 21:17:00 EST
1330		Tag: leunga-20000402-mltree
1331		Description:
1332
1333		1. Renamed the constructor CALL in MLTREE by popular demand.
1334		2. Added a bunch of files from my repository.  These are currently
1335		used by other non-SMLNJ backends.
1336
1337		----------------------------------------------------------------------
1338		Name: Allen Leung
1339		Date: 2000/03/31 21:15:00 EST
1340		Tag: leunga-20000331-aliasing
1341		Description:
1342
1343		This update contains a rewritten (and hopefully more correct) module
1344		for extracting aliasing information from CPS.
1345
1346		To turn on this feature:
1347
1348		Compiler.Control.CG.memDisambiguate := true
1349
1350		To pretty print the region information with assembly
1351
1352		Compiler.Control.MLRISC.getFlag "asm-show-region" := true;
1353
1354		To control how many levels of aliasing information are printed, use:
1355
1356		Compiler.Control.MLRISC.getInt "points-to-show-level" := n
1357
1358		The default of n is 3.
1359
1360		----------------------------------------------------------------------
1361		Name: David MacQueen
1362		Date: 2000/03/31 11:15:00 EST
1363		Tag: dbm-20000331-runtime_fix
1364		Description:
1365
1366		This update contains:
1367
1368		1. runtime/c-lib/c-libraries.c
1369		includes added in revision 1.2 caused compilation errors on hppa-hpux
1370
1371		2. fix for bug 1556
1372		system/Basis/Implementation/NJ/internal-signals.sml
1373
1374		----------------------------------------------------------------------
1375		Name: Matthias Blume
1376		Date: 2000/03/31 18:00:00 JST
1377		Tag: blume_main_v110p26p2_1
1378		Description:
1379
1380		This update contains:
1381
1382		1. A small change to CM's handling of stable libraries:
1383		CM now maintains one "global" modmap that is used for all stable
1384		libraries.  The use of such a global modmap maximizes sharing and
1385		minimizes the need for re-traversing parts of environments during
1386		modmap construction.  (However, this has minor impact since modmap
1387		construction seems to account for just one percent or less of total
1388		compile time.)
1389
1390		2. I added a "genmap" phase to the statistics.  This is where I got the
1391		"one percent" number (see above).
1392
1393		3. CM's new tool parameter mechanism just became _even_ better. :)
1394		- The parser understands named parameters and recursive options.
1395		- The "make" and "shell" tools use these new features.
1396		(This makes it a lot easier to cascade these tools.)
1397		- There is a small syntax change: named parameters use a
1398
1399		<name> : ( <option> ... )            or
1400		<name> : <string>
1401
1402		syntax.  Previously, named parameters were implemented in an
1403		ad-hoc fashion by each tool individually (by parsing strings)
1404		and had the form
1405
1406		<name>=<string>
1407
1408		See the CM manual for a full description of these issues.
1409
1410		----------------------------------------------------------------------
1411		Name: Matthias Blume
1412		Date: 2000/03/30 18:00:00 JST
1413		Tag: blume_main_v110p26p2_0
1414		Description:
1415
1416		!!!!! WARNING !!!!!!
1417		!!  New binfiles  !!
1418		!!!!!!!!!!!!!!!!!!!!
1419
1420		This update contains:
1421
1422		1. Moderate changes to CM:
1423
1424		- Changes to CM's tools mechanism.  In particular, it is now possible
1425		to have tools that accept additional "command line" parameters
1426		(specified in the .cm file at each instance where the tool's class is
1427		used).
1428
1429		This was done to accommodate the new "make" and "shell" tools which
1430		facilitate fairly seamless hookup to portions of code managed using
1431		Makefiles or Shell scripts.
1432
1433		There are no classes "shared" or "private" anymore.  Instead, the
1434		sharing annotation is now a parameter to the "sml" class.
1435
1436		There is a bit of generic machinery for implementing one's own
1437		tools that accept command-line parameters.  However, I am not yet fully
1438		satisfied with that part, so expect changes here in the future.
1439
1440		All existing tools are described in the CM manual.
1441
1442		- Slightly better error handling.  (CM now suppresses many followup
1443		error messages that tended to be more annoying than helpful.)
1444
1445		2. Major changes to the compiler's static environment data structures.
1446
1447		- no CMStaticEnv anymore.
1448		- no CMEnv, no "BareEnvironment" (actually, _only_ BareEnvironment,
1449		but it is called Environment), no conversions between different
1450		kinds of static environments
1451
1452		- There is still a notion of a "modmap", but such modmaps are generated
1453		on demand at the time when they are needed.  This sounds slow, but I
1454		sped up the code that generates modmaps enough for this not to lead to
1455		a slowdown of the compiler (at least I didn't detect any).
1456
1457		- To facilitate rapid modmap generation, static environments now
1458		contain an (optional) "modtree" structure.  Modtree annotations are
1459		constructed by the unpickler during unpickling.  (This means that
1460		the elaborator does not have to worry about modtrees at all.)
1461		Modtrees have the advantage that they are compositional in the same
1462		way as the environment data structure itself is compositional.
1463		As a result, modtrees never hang on to parts of an environment that
1464		has already been rendered "stale" by filtering or rebinding.
1465
1466		- I went through many, many trials and errors before arriving at the
1467		current solution.  (The initial idea of "linkpaths" did not work.)
1468		But the result of all this is that I have touched a lot of files that
1469		depend on the "modules" and "types" data structures (most of the
1470		elaborator). There were a lot of changes during my "linkpath" trials
1471		that could have been reverted to their original state but weren't.
1472		Please, don't be too harsh on me for messing with this code a bit more
1473		than what was strictly necessary...  (I _did_ resist the tempation
1474		of doing any "global reformatting" to avoid an untimely death at
1475		Dave's hands. :)
1476
1477		- One positive aspect of the previous point:  At least I made sure that
1478		all files that I touched now compile without warnings (other than
1479		"polyEqual").
1480
1481		- compiler now tends to run "leaner" (i.e., ties up less memory in
1482		redundant modmaps)
1483
1484		----------------------------------------------------------------------
1485		Name: Allen Leung
1486		Date: 2000/03/29 18:00:00
1487		Tag: leunga-20000327-mlriscGen_hppa_alpha_x86
1488		Boot files (optional): ftp://react-ilp.cs.nyu.edu/leunga/110.26.1-sml.boot.x86-unix-20000330.tar.gz
1489		Description:
1490
1491		This update contains *MAJOR* changes to the way code is generated from CPS
1492		in the module mlriscGen, and in various backend modules.
1493
1494		CHANGES
1495		=======
1496
1497		1. MLRiscGen: forward propagation fix.
1498
1499		There was a bug in forward propagation introduced at about the same time
1500		as the MLRISC x86 backend, which prohibits coalescing to be
1501		performed effectively in loops.
1502
1503		Effect: speed up of loops in RISC architectures.
1504		By itself, this actually slowed down certain benchmarks on the x86.
1505
1506		2. MLRiscGen:  forward propagating addresses from consing.
1507
1508		I've changed the way consing code is generated.  Basically I separated
1509		out the initialization part:
1510
1511		store tag,   offset(allocptr)
1512		store elem1, offset+4(allocptr)
1513		store elem2, offset+8(allocptr)
1514		...
1515		store elemn, offset+4n(allocptr)
1516
1517		and the address computation part:
1518
1519		celladdr <- offset+4+alloctpr
1520
1521		and move the address computation part
1522
1523		Effect:  register pressure is generally lower as a result.  This
1524		makes compilation of certain expressions much faster, such as
1525		long lists with non-trivial elements.
1526
1527		[(0,0), (0,0), .... (0,0)]
1528
1529		3. MLRiscGen: base pointer elimination.
1530
1531		As part of the linkage mechanism, we generate the sequence:
1532
1533		L:  ...  <- start of the code fragment
1534
1535		L1:
1536		base pointer <- linkreg - L1 + L
1537
1538		The base pointer was then used for computing relocatable addresses
1539		in the code fragment.  Frequently (such as in lots of continuations)
1540		this is not needed.  We now eliminate this sequence whenever possible.
1541
1542		For compile time efficiency, I'm using a very stupid local heuristic.
1543		But in general, this should be done as a control flow analysis.
1544
1545		Effect:  Smaller code size.  Speed up of most programs.
1546
1547		4. Hppa back end
1548
1549		Long jumps in span dependence resolution used to depend on the existence
1550		of the base pointer.
1551
1552		A jump to a long label L was expanded into the following sequence:
1553
1554		LDIL %hi(L-8192), %r29
1555		LDO  %lo(L-8192)(%r29), %r29
1556		ADD  %r29, baseptr, %r29
1557		BV,n %r0(%r29)
1558
1559		In the presence of change (3) above, this will not work.  I've changed
1560		it so that the following sequence of instructions are generated, which
1561		doesn't mention the base pointer at all:
1562
1563		BL,n  L', %r29           /* branch and link, L' + 4 -> %r29 */
1564		L':  ADDIL L-(L'+4), %r29     /* Compute address of L */
1565		BV,n  %r0(%r29)          /* Jump */
1566
1567		5. Alpha back end
1568
1569		New alpha instructions LDB/LDW have been added, as per Fermin's
1570		suggestions.   This is unrelated to all other changes.
1571
1572		6. X86 back end
1573
1574		I've changed andl to testl in the floating point test sequence
1575		whenever appropriate.  The Intel optimization guide states that
1576		testl is preferable to andl.
1577
1578		7. RA (x86 only)
1579
1580		I've improved the spill propagation algorithm, using an approximation
1581		of maximal weighted independent sets.   This seems to be necessary to
1582		alleviate the negative effect in light of the slow down in (1).
1583
1584		I'll write down the algorithm one of these days.
1585
1586		8. MLRiscGen: frequencies
1587
1588		I've added an annotation that states that all call gc blocks have zero
1589		execution frequencies.  This improves register allocation on the x86.
1590
1591		BENCHMARKS
1592		==========
1593
1594		I've only perform the comparison on 110.25.
1595
1596		The platforms are:
1597
1598		HPPA  A four processor HP machine (E9000) with 5G of memory.
1599		X86   A 300Hhz Pentium II with 128M of memory, and
1600		SPARC An Ultra sparc 2 with 512M of memory.
1601
1602		I used the following parameters for the SML benchmarks:
1603
1604		@SMLalloc
1605		HPPA    256k
1606		SPARC   512k
1607		X86     256k
1608
1609		COMPILATION TIME
1610		----------------
1611		Here are the numbers comparing the compilation times of the compilers.
1612		I've only compared 110.25 compiling the new sources versus
1613		a fixpoint version of the new compiler compiling the same.
1614
1615		110.25                                  New
1616		Total  Time in RA  Spill+Reload   Total  Time In RA Spill+Reload
1617		HPPA   627s    116s        2684+3584     599s    95s       1003+1879
1618		SPARC  892s    173s        2891+3870     708s    116s      1004+1880
1619		X86    999s    315s       94006+130691   987s    296s    108877+141957
1620
1621		110.25         New
1622		Code Size      Code Size
1623		HPPA   8596736         8561421
1624		SPARC  8974299         8785143
1625		X86    9029180         8716783
1626
1627		So in summary, things are at least as good as before.   Dramatic
1628		reduction in compilation is obtained on the Sparc; I can't explain it,
1629		but it is reproducible.  Perhaps someone should try to reproduce this
1630		on their own machines.
1631
1632		SML BENCHMARKS
1633		--------------
1634
1635		On the average, all benchmarks perform at least as well as before.
1636
1637		HPPA         Compilation Time     Spill+Reload      Run Time
1638		110.25  New            110.25    New   110.25  New
1639
1640		barnesHut  3.158  3.015  4.75%    1+1       0+0   2.980  2.922   2.00%
1641		boyer  6.152  5.708  7.77%    0+0       0+0   0.218  0.213   2.34%
1642		count-graphs  1.168  1.120  4.32%    0+0       0+0  22.705 23.073  -1.60%
1643		fft  0.877  0.792 10.74%    1+3       1+3   0.602  0.587   2.56%
1644		knuthBendix  3.180  2.857 11.32%    0+0       0+0   0.675  0.662   2.02%
1645		lexgen  6.190  5.290 17.01%    0+0       0+0   0.913  0.788  15.86%
1646		life  0.803  0.703 14.22%   25+25      0+0   0.153  0.140   9.52%
1647		logic  2.048  2.007  2.08%    6+6       1+1   4.133  4.008   3.12%
1648		mandelbrot  0.077  0.080 -4.17%    0+0       0+0   0.765  0.712   7.49%
1649		mlyacc 22.932 20.937  9.53%  154+181    32+57  0.468  0.430   8.91%
1650		nucleic  5.183  5.060  2.44%    2+2       0+0   0.125  0.120   4.17%
1651		ratio-regions  3.357  3.142  6.84%    0+0       0+0  116.225 113.173 2.70%
1652		ray  1.283  1.290 -0.52%    0+0       0+0   2.887  2.855   1.11%
1653		simple  6.307  6.032  4.56%   28+30      5+7   3.705  3.658   1.28%
1654		tsp  0.888  0.862  3.09%    0+0       0+0   7.040  6.893   2.13%
1655		vliw 24.378 23.455  3.94%  106+127    25+45  2.758  2.707   1.91%
1656		--------------------------------------------------------------------------
1657		Average                     6.12%                                   4.09%
1658
1659		SPARC        Compilation Time     Spill+Reload      Run Time
1660		110.25  New            110.25    New   110.25  New
1661
1662		barnesHut  3.778  3.592  5.20%    2+2       0+0   3.648  3.453    5.65%
1663		boyer  6.632  6.110  8.54%    0+0       0+0   0.258  0.242    6.90%
1664		count-graphs  1.435  1.325  8.30%    0+0       0+0  33.672 34.737   -3.07%
1665		fft  0.980  0.940  4.26%    3+9       2+6   0.838  0.827    1.41%
1666		knuthBendix  3.590  3.138 14.39%    0+0       0+0   0.962  0.967   -0.52%
1667		lexgen  6.593  6.072  8.59%    1+1       0+0   1.077  1.078   -0.15%
1668		life  0.972  0.868 11.90%   26+26      0+0   0.143  0.140    2.38%
1669		logic  2.525  2.387  5.80%    7+7       1+1   5.625  5.158    9.05%
1670		mandelbrot  0.090  0.093 -3.57%    0+0       0+0   0.855  0.728   17.39%
1671		mlyacc 26.732 23.827 12.19%  162+189    32+57  0.550  0.560   -1.79%
1672		nucleic  6.233  6.197  0.59%    3+3       0+0   0.163  0.173   -5.77%
1673		ratio-regions  3.780  3.507  7.79%    0+0       0+0 133.993 131.035   2.26%
1674		ray  1.595  1.550  2.90%    1+1       0+0   3.440  3.418    0.63%
1675		simple  6.972  6.487  7.48%   29+32      5+7   3.523  3.525   -0.05%
1676		tsp  1.115  1.063  4.86%    0+0       0+0   7.393  7.265    1.77%
1677		vliw 27.765 24.818 11.87%  110+135    25+45  2.265  2.135    6.09%
1678		----------------------------------------------------------------------------
1679		Average                     6.94%                                    2.64%
1680
1681		X86          Compilation Time     Spill+Reload      Run Time
1682		110.25  New            110.25    New   110.25  New
1683
1684		barnesHut  5.530  5.420  2.03%  593+893   597+915   3.532  3.440   2.66%
1685		boyer  8.768  7.747 13.19%  493+199   301+289   0.327  0.297  10.11%
1686		count-graphs  2.040  2.010  1.49%  298+394   315+457  26.578 28.660  -7.26%
1687		fft  1.327  1.302  1.92%  112+209   115+210   1.055  0.962   9.71%
1688		knuthBendix  5.218  5.475 -4.69%  451+598   510+650   0.928  0.932  -0.36%
1689		lexgen  9.970  9.623  3.60% 1014+841  1157+885   0.947  0.928   1.97%
1690		life  1.183  1.183  0.00%  162+182   145+148   0.127  0.103  22.58%
1691		logic  3.285  3.512 -6.45%  514+684   591+836   5.682  5.577   1.88%
1692		mandelbrot  0.147  0.143  2.33%   38+41     33+54    0.703  0.690   1.93%
1693		mlyacc 35.457 32.763  8.22% 3496+4564 3611+4860  0.552  0.550   0.30%
1694		nucleic  7.100  6.888  3.07%  239+168   201+158   0.175  0.173   0.96%
1695		ratio-regions  6.388  6.843 -6.65% 1182+257   981+300  120.142 120.345 -0.17%
1696		ray  2.332  2.338 -0.29%  346+398   402+494   3.593  3.540   1.51%
1697		simple  9.912  9.903  0.08% 1475+941  1579+1168  3.057  3.178  -3.83%
1698		tsp  1.623  1.532  5.98%  266+200   250+211   8.045  7.878   2.12%
1699		vliw 33.947 35.470 -4.29% 2629+2774 2877+3171  2.072  1.890   9.61%
1700		----------------------------------------------------------------------------
1701		Average                     1.22%                                     3.36%
1702
1703		----------------------------------------------------------------------
1704	Name: Allen Leung	Name: Allen Leung
1705	Date: 2000/03/23 16:25:00	Date: 2000/03/23 16:25:00
1706	Tag: leunga-20000323-fix_x86_alpha	Tag: leunga-20000323-fix_x86_alpha
#	Line 66	Line 1755
1755
1756	1. Tools.registerStdShellCmdTool (from smlnj/cm/tool.cm) takes an	1. Tools.registerStdShellCmdTool (from smlnj/cm/tool.cm) takes an
1757	additional argument called "template" which is an optional string that	additional argument called "template" which is an optional string that
1758	specifiel the layout of the tool command line.  See the CM manual for	specifies the layout of the tool command line.  See the CM manual for
1759	explanation.	explanation.
1760
1761	2. A special-purpose tool can be "regisitered" by simply dropping the	2. A special-purpose tool can be "registered" by simply dropping the
1762	corresponding <...>-tool.cm (and/or <...>-ext.cm) into the same	corresponding <...>-tool.cm (and/or <...>-ext.cm) into the same
1763	directory where the .cm file lives that uses this tool.  (The	directory where the .cm file lives that uses this tool.  (The
1764	behavior/misfeature until now was to look for the tool description	behavior/misfeature until now was to look for the tool description
#	Line 113	Line 1802
1802	rebuild.  Having sets of the form <base><k>.{bin,boot}.<arch>-unix for	rebuild.  Having sets of the form <base><k>.{bin,boot}.<arch>-unix for
1803	<k>=1,2,... is normally not a good idea when invoking fixpt.  However,	<k>=1,2,... is normally not a good idea when invoking fixpt.  However,
1804	they might be the result of an earlier partial run of fixpt (which	they might be the result of an earlier partial run of fixpt (which
1805	perhaps got accidentially killed).  In this case, fixpt will quickly	perhaps got accidentally killed).  In this case, fixpt will quickly
1806	move through what exists before continuing where it left off earlier,	move through what exists before continuing where it left off earlier,
1807	and, thus, saves a lot of time.	and, thus, saves a lot of time.
1808
#	Line 163	Line 1852
1852	it from that remote directory.	it from that remote directory.
1853	This should simplify installation further:  For machines that have	This should simplify installation further:  For machines that have
1854	access to the internet, just fetch <version>-config.tgz, unpack it,	access to the internet, just fetch <version>-config.tgz, unpack it,
1855	edit config/targets, and go (run config/install.sh).  The scipt will	edit config/targets, and go (run config/install.sh).  The script will
1856	fetch everything else that it might need all by itself.	fetch everything else that it might need all by itself.
1857
1858	For CVS users, this mechanism is not relevant for source archives, but	For CVS users, this mechanism is not relevant for source archives, but

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.583   changed lines   Added in v.697

---

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