Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 578, Tue Mar 14 05:16:29 2000 UTC	revision 701, Thu Aug 31 08:46:43 2000 UTC
#	Line 13	Line 13
13	Description:	Description:
14	----------------------------------------------------------------------	----------------------------------------------------------------------
15	Name: Matthias Blume	Name: Matthias Blume
16		Date: 2000/08/31 18:00:00 JST
17		Tag: blume-20000831-cvsbootfiles
18		Description:
19
20		New URL for boot files (because the 110.29 files on the BL server do
21		now work correctly with my updated install scripts for yacc and lex).
22
23		----------------------------------------------------------------------
24		Name: Matthias Blume
25		Date: 2000/08/08 12:33:00 JST
26		Tag: blume-20000808-manual
27		Description:
28
29		Tiny update to CM manual.
30
31		----------------------------------------------------------------------
32		Name: Allen Leung
33		Date: 2000/08/7 19:31:00
34		Tag: leunga-20000807-a-whole-bunch-of-stuff
35
36		Moby, C--, SSA, x86, machine descriptions etc.  Should only affect C--
37		and Mobdy.
38
39		1.  x86
40
41		a.  Fixes to peephole module by John and Dan.
42		b.  Assembly fix to SETcc by Allen.
43		c.  Fix to c-call by John.
44		d.  Fix to spilling by John.  (This one deals with the missing FSTPT case)
45		e.  Instruction selection optimization to SETcc as suggested by John.
46
47		For example,
48
49		MV(32, x, COND(32, CMP(32, LT, a, b), LI 1, LI 0))
50
51		should generate:
52
53		MOVL a, x
54		SUBL b, x
55		SHRL 31, x
56
57		2.  IR stuff
58
59		A bunch of new DJ-graph related algorithms added.  These
60		speed up SSA construction.
61
62		3.  SSA + Scheduling
63
64		Added code for SSA and scheduling to the repository
65
66		----------------------------------------------------------------------
67		Name: Lal George
68		Date: 2000/07/27 11:53:14 EDT
69
70		Tag: lal-20000727-linux-ppc
71		Description:
72
73		Made changes to support Linux PPC.
74		p.s. I have confirmation that the 110.29 boot files work fine.
75
76		----------------------------------------------------------------------
77		Name: Matthias Blume
78		Date: 2000/07/27 17:40:00 JST
79		Tag: blume-20000727-scripts
80		Description:
81
82		!!!! WARNING !!!!
83		You must recompile the runtime system!
84		!!!! WARNING !!!!
85
86		This is basically another round of script-enhancements:
87
88		1. sml, ml-build, and ml-makedepend accept options -D and -U to define
89		and undefine CM preprocessor symbols.
90
91		2. ml-build avoids generating a new heap image if it finds that the
92		existing one is still ok.  (The condition is that no ML file had to
93		be recompiled and all ML files are found to be older that the heap
94		file.)
95
96		To make this work smoothly, I also hacked the runtime system as
97		well as SMLofNJ.SysInfo to get access to the heap image suffix
98		(.sparc-solaris, ...) that is currently being used.
99
100		Moreover, the signature of CM.mk_standalone has changed.  See the
101		CM manual.
102
103		3. ml-makedepend accepts additional options -n, -a, and -o.  (See the
104		CM manual for details.)
105
106		4. More CM manual updates:
107		- all of the above has been documented.
108		- there is now a section describing the (CM-related) command line
109		arguments that are accepted by the "sml" command
110
111		----------------------------------------------------------------------
112		Name: Matthias Blume
113		Date: 2000/07/25 16:20:00 JST
114		Tag: blume-20000725-makedepend
115		Description:
116
117		Added a script called ml-makedepend.  This can be used in makefiles
118		for Unix' make in a way very similar to the "makedepend" command for
119		C.
120
121		The script internally uses function CM.sources.
122
123		Synopsis:
124
125		ml-makedepend [-f makefile] cmfile targetname
126
127		The default for the makefile is "makefile" (or "Makefile" should
128		"makefile" not exist).
129
130		ml-makedepend adds a cmfile/targetname-specific section to this
131		makefile (after removing the previous version of this section).  The
132		section contains a single dependency specification with targetname on
133		the LHS (targetname is an arbitrary name), and a list of files derived
134		from the cmfile on the RHS.  Some of the files on the RHS are
135		ARCH/OPSYS-specific.  Therefore, ml-makedepend inserts references to
136		"make" variables $(ARCH) and $(OPSYS) in place of the corresponding
137		path names.  The makefile writer is responsible for making sure that
138		these variables have correct at the time "make" is invoked.
139
140		----------------------------------------------------------------------
141		Name: Matthias Blume
142		Date: 2000/07/22 23:30:00 JST
143		Tag: blume-20000722-urlupdate
144		Description:
145
146		Changed BOOT and config/srcarchiveurl to point to BL server:
147
148		ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.29/
149
150		----------------------------------------------------------------------
151		Name: Matthias Blume
152		Date: 2000/07/18 18:00:00 JST
153		Tag: blume-20000718-Version_110_29
154		Description:
155
156		1. Updated src/compiler/TopLevel/main/version.sml to version 110.29
157
158		2. Updated config/version to 110.29
159
160		3. Updated config/srcarchiveurl
161
162		3. New boot files!
163		ftp://ftp.cs.princeton.edu/pub/people/blume/sml/110.29-autofetch
164
165		----------------------------------------------------------------------
166		Name: Matthias Blume
167		Date: 2000/07/11 13:58:00 JST
168		Tag: blume-20000711-doctypo
169		Description:
170
171		Fixed a few typos in CM manual.
172
173		----------------------------------------------------------------------
174		Name: Allen Leung
175		Date: 2000/06/15 00:38:00
176		Tag: leunga-20000704-sparc-x86
177
178		1. x86 peephole improvement sp += k; sp -= k => nop  [from John]
179		2. fix to x86 RET bug [found by Dan Grossman]
180		3. sparc assembly bug fix for ticc instructions [found by Fermin]
181
182		Affects c-- and moby only
183
184		----------------------------------------------------------------------
185		Name: Matthias Blume
186		Date: 2000/07/04 15:26:00
187		Tag: blume-20000704-trigger
188		Description:
189
190		1. Improvements to CM manual.
191		2. SMLofNJ.Internals.BTrace.trigger reinstated as an alternative way
192		of getting a back-trace.  The function, when called, raises an
193		internal exception which explicitly carries the full back-trace history,
194		so it is unaffected by any intervening handle-raise pairs ("trivial"
195		or not).  The interactive loop will print that history once it arrives
196		at top level.
197		Short of having all exceptions implicitly carry the full history, the
198		recommended way of using this facility is:
199		- compile your program with instrumentation "on"
200		- run it, when it raises an exception, look at the history
201		- if the history is "cut off" because of some handler, go and modify
202		your program so that it explicitly calls BTrace.trigger
203		- recompile (still instrumented), and rerun; look at the full history
204
205		----------------------------------------------------------------------
206		Name: Matthias Blume
207		Date: 2000/07/03 15:36:00 JST
208		Tag: blume-20000702-manual
209		Description:
210
211		Small corrections and updates to CM manual.
212
213		----------------------------------------------------------------------
214		Name: Matthias Blume
215		Date: 2000/06/29 16:04:00 JST
216		Tag: blume-20000629-yacctool
217		Description:
218
219		Changes:
220
221		1. Class "mlyacc" now takes separate arguments to pass options to
222		generated .sml- and .sig-files independently.
223		2. Corresponding CM manual updates.
224		3. BTrace module now also reports call sites.  (However, for loop clusters
225		it only shows from where the cluster was entered.)  There are associated
226		modifications to core.sml, internals.{sig,sml}, btrace.sml, and btimp.sml.
227
228		----------------------------------------------------------------------
229		Name: Matthias Blume
230		Date: 2000/06/27 16:51:00 JST
231		Tag: blume-20000627-noweb
232		Description:
233
234		Changes:
235
236		1. Implemented "subdir" and "witness" options for noweb tool.
237		This caused some slight internal changes in CM's tool implementation.
238		2. Fixed bug in "tool plugin" mechanism.  This is essentially cleaning
239		some remaining issues from earlier path anchor changes.
240		3. Updated CM manual accordingly.
241
242		4. Changed implementation of back-tracing so that I now consider it
243		ready for prime-time.
244
245		In particular, you don't have to explicitly trigger the back-trace
246		anymore.  Instead, if you are running BTrace-instrumented code and
247		there is an uncaught exception (regardless of whether or not it was
248		raised in instrumented code), the top-level evalloop will print
249		the back-trace.
250
251		Features:
252
253		- Instrumented and uninstrumented code work together seemlessly.
254		(Of course, uninstrumented code is never mentioned in actual
255		back-traces.)
256
257		- Asymptotic time- and space-complexity of instrumented code is
258		equal to that of uninstrumented code.  (This means that
259		tail-recursion is preserved by the instrumentation phase.)
260
261		- Modules whose code has been instrumented in different sessions
262		work together without problem.
263
264		- There is no penalty whatsoever on uninstrumented code.
265
266		- There is no penalty on "raise" expressions, even in
267		instrumented code.
268
269		A potential bug (or perhaps it is a feature, too):
270
271		A back-trace reaches no further than the outermost instrumented
272		non-trivial "raise".  Here, a "trivial" raise is one that is the
273		sole RHS of a "handle" rule.  Thus, back-traces reach trough
274
275		<exp> handle e => raise e
276
277		and even
278
279		<exp> handle Foo => raise Bar
280
281		and, of course, through
282
283		<exp> handle Foo => ...
284
285		if the exception was not Foo.
286
287		Back-traces always reach right through any un-instrumented code
288		including any of its "handle" expressions, trivial or not.
289
290		To try this out, do the following:
291
292		- Erase all existing binfiles for your program.
293		(You may keep binfiles for those modules where you think you
294		definitely don't need back-tracing.)
295		- Turn on back-trace instrumentation:
296		SMLofNJ.Internals.BTrace.mode (SOME true);
297		- Recompile your program.  (I.e., run "CM.make" or "use".)
298		- You may now turn instrumentation off again (if you want):
299		SMLofNJ.Internals.BTrace.mode (SOME false);
300		- Run your program as usual.  If it raises an exception that
301		reaches the interactive toplevel, then a back-trace will
302		automatically be printed.  After that, the toplevel loop
303		will print the exception history as usual.
304
305		----------------------------------------------------------------------
306		Name: Matthias Blume
307		Date: 2000/06/26 09:56:46 JST
308		Tag: blume-20000626-setup
309		Description:
310
311		CM: - setup-parameter to "sml" added; this can be used to run arbitrary
312		ML code before and after compiling a file (e.g., to set compiler
313		flags)
314
315		Compiler: - improved btrace API (in core.sml, internals.{sig,sml})
316		- associated changes to btrace.sml (BTrace instrumentation pass)
317		- cleaner implementation of btimp.sml (BTrace tracing and report
318		module)
319
320		CM manual: * new path encoding documented
321		* description of setup-parameter to "sml" added
322
323		The biggest user-visible change to back-tracing is that it is no
324		longer necessary to compile all traced modules within the same
325		session.  (This was a real limitation.)
326
327		----------------------------------------------------------------------
328		Name: Matthias Blume
329		Date: 2000/06/24 12:40:00 JST
330		Tag: blume-20000624-startup
331		Description:
332
333		Fixes startup slowdown problem.  (I was calling SrcPath.sync a _tad_
334		bit too often -- to put it mildly. :)
335
336		----------------------------------------------------------------------
337		Name: Matthias Blume
338		Date: 2000/06/23 18:20:00 JST
339		Tag: blume-20000623-btrace
340		Description:
341
342		This updates adds a backtrace facility to aid programmers in debugging
343		their programs.  This involves the following changes:
344
345		1. Module system/smlnj/init/core.sml (structure _Core) now has hooks for
346		keeping track of the current call stack.  When programs are compiled
347		in a special mode, the compiler will insert calls to these hooks
348		into the user program.
349		"Hook" means that it is possible for different implementations of
350		back-tracing to register themselves (at different times).
351
352		2. compiler/MiscUtil/profile/btrace.sml implements the annotation phase
353		as an Absyn.dec->Absyn.dec rewrite.  Normally this phase is turned off.
354		It can be turned on using this call:
355		SMLofNJ.Internals.BTrace.mode (SOME true);
356		Turning it off again:
357		SMLofNJ.Internals.BTrace.mode (SOME false);
358		Querying the current status:
359		SMLofNJ.Internals.BTrace.mode NONE;
360		Annotated programs are about twice as big as normal ones, and they
361		run a factor of 2 to 4 slower with a dummy back-trace plugin (one
362		where all hooks do nothing).  The slowdown with a plugin that is
363		actually useful (such as the one supplied by default) is even greater,
364		but in the case of the default plugin it is still only an constant
365		factor (amortized).
366
367		3. system/Basis/Implementation/NJ/internals.{sig,sml} have been augmented
368		with a sub-structure BTrace for controlling back-tracing.  In particular,
369		the above-mentioned function "mode" controls whether the annotation
370		phase is invoked by the compiler.  Another important function is
371		"trigger": when called it aborts the current execution and causes
372		the top-level loop to print a full back-trace.
373
374		4. compiler/MiscUtil/profile/btimp.sml is the current default plugin
375		for back-tracing.  It keeps track of the dynamic call stack and in
376		addition to that it keeps a partial history at each "level" of that
377		stack.  For example, if a tail-calls b, b tail-calls c, and c tail-calls
378		d and b (at separate times, dynamically), then the report will show:
379
380		GOTO   d
381		/c
382		GOTO  \b
383		CALL   a
384
385		This shows that there was an initial non-tail call of a, then a
386		tail-call to b or c, looping behavior in a cluster of functions that
387		consist of b and c, and then a goto from that cluster (i.e., either from
388		b or from c) to d.
389
390		Note that (depending on the user program) the amount of information
391		that the back-trace module has to keep track of at each level is bounded
392		by a constant.  Thus, the whole implementation has the same asymptotical
393		complexity as the original program (both in space and in time).
394
395		5. compiler/TopLevel/interact/evalloop.sml has been modified to
396		handle the special exception SMLofNJ.Internals.BTrace.BTrace
397		which is raised by the "trigger" function mentioned above.
398
399		Notes on usage:
400
401		- Annotated code works well together with unannotated code:
402		Unannotated calls simply do not show up at all in the backtrace.
403
404		- It is not a good idea to let modules that were annotated during
405		different sessions run at the same time.  This is because the compiler
406		chooses small integers to identify individual functions, and there
407		will be clashes if different modules were compiled in separate sessions.
408		(Nothing will crash, and you will even be told about the clashes, but
409		back-trace information will in general not be useful.)
410
411		- Back-tracing can be confused by callcc and capture.
412
413		- The only way of getting a back-trace right now is to explicitly
414		invoke the "trigger" function from your user program.  Eventually, we
415		should make every exception carry back-trace information (if
416		available).  But since this creates more overhead at "raise"-time
417		(similar to the current exnHistory overhead), I have not yet
418		implemented this.  (The implementation will be rather easy.)  With
419		exceptions carrying back-trace information, this facility will be even
420		more useful because users don't need to modify their programs...
421
422		- While it is possible to compile the compiler with back-trace
423		annotations turned on (I did it to get some confidence in
424		correctness), you must make absolutely sure that core.sml and
425		btimp.sml are compiled WITHOUT annotation!  (core.sml cannot actually
426		be compiled with annotation because there is no core access yet, but
427		if you compile btimp.sml with annotation, then the system will go into
428		an infinite recursion and crash.)
429		Since CM currently does not know about BTrace, the only way to turn
430		annotations on and off for different modules of the compiler is to
431		interrupt CMB.make, change the settings, and re-invoke it.  Of course,
432		this is awkward and clumsy.
433
434		Sample sessions:
435
436		Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
437		- SMLofNJ.Internals.BTrace.mode (SOME true);
438		[autoloading]
439		[autoloading done]
440		val it = false : bool
441		- structure X = struct
442		-     fun main n = let
443		-         fun a (x, 0) = d x
444		-           | a (x, n) = b (x, n - 1)
445		-         and b (x, n) = c (x, n)
446		-         and c (x, n) = a (x, n)
447		-         and d x = e (x, 3)
448		-         and e (x, 0) = f x
449		-           | e (x, n) = e (x, n - 1)
450		-         and f 0 = SMLofNJ.Internals.BTrace.trigger ()
451		-           | f n = n * g (n - 1)
452		-         and g n = a (n, 3)
453		-     in
454		-         f n
455		-     end
456		- end;
457		structure X : sig val main : int -> int end
458		- X.main 3;
459		*** BACK-TRACE ***
460		GOTO   stdIn:4.2-13.20: X.main[2].f
461		GOTO-( stdIn:4.2-13.20: X.main[2].e
462		GOTO   stdIn:4.2-13.20: X.main[2].d
463		/ stdIn:4.2-13.20: X.main[2].a
464		| stdIn:4.2-13.20: X.main[2].b
465		GOTO-\ stdIn:4.2-13.20: X.main[2].c
466		CALL   stdIn:4.2-13.20: X.main[2].g
467		GOTO   stdIn:4.2-13.20: X.main[2].f
468		GOTO-( stdIn:4.2-13.20: X.main[2].e
469		GOTO   stdIn:4.2-13.20: X.main[2].d
470		/ stdIn:4.2-13.20: X.main[2].a
471		| stdIn:4.2-13.20: X.main[2].b
472		GOTO-\ stdIn:4.2-13.20: X.main[2].c
473		CALL   stdIn:4.2-13.20: X.main[2].g
474		GOTO   stdIn:4.2-13.20: X.main[2].f
475		GOTO-( stdIn:4.2-13.20: X.main[2].e
476		GOTO   stdIn:4.2-13.20: X.main[2].d
477		/ stdIn:4.2-13.20: X.main[2].a
478		| stdIn:4.2-13.20: X.main[2].b
479		GOTO-\ stdIn:4.2-13.20: X.main[2].c
480		CALL   stdIn:4.2-13.20: X.main[2].g
481		GOTO   stdIn:4.2-13.20: X.main[2].f
482		CALL   stdIn:2.15-17.4: X.main[2]
483		-
484
485		(Note that because of a FLINt bug the above code currently does not
486		compile without BTrace turned on.)
487
488		Here is another example, using my modified Tiger compiler:
489
490		Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
491		- SMLofNJ.Internals.BTrace.mode (SOME true);
492		[autoloading]
493		[autoloading done]
494		val it = false : bool
495		- CM.make "sources.cm";
496		[autoloading]
497		...
498		[autoloading done]
499		[scanning sources.cm]
500		[parsing (sources.cm):parse.sml]
501		[creating directory CM/SKEL ...]
502		[parsing (sources.cm):tiger.lex.sml]
503		...
504		[wrote CM/sparc-unix/semant.sml]
505		[compiling (sources.cm):main.sml]
506		[wrote CM/sparc-unix/main.sml]
507		[New bindings added.]
508		val it = true : bool
509		- Main.compile ("../testcases/merge.tig", "foo.out");
510		*** BACK-TRACE ***
511		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trvar
512		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
513		CALL   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
514		GOTO   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
515		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
516		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
517		CALL   lib/semant.sml:488.3-505.6: SemantFun[2].transDec.trdec[2].transBody[2]
518		/ lib/semant.sml:411.65-543.8: SemantFun[2].transDec
519		CALL-\ lib/semant.sml:413.2-540.9: SemantFun[2].transDec.trdec[2]
520		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
521		CALL   lib/semant.sml:8.52-558.4: SemantFun[2].transProg[2]
522		CALL   main.sml:1.18-118.4: Main.compile[2]
523		-
524
525		----------------------------------------------------------------------
526		Name: Matthias Blumen
527		Date: 2000/06/21 18:00:00 JST
528		Tag: blume-20000621-manual
529		Description:
530
531		CM manual update: Path environments documented.
532
533		----------------------------------------------------------------------
534		Name: Matthias Blume
535		Date: 2000/06/19 13:40:00
536		Tag: blume-20000619-manual
537		Description:
538
539		CM manual and system/README update.  This only covers the fact that
540		there are no more implicit anchors.  (Path environments and the "bind"
541		option to "cm" have yet to be documented.)
542
543		----------------------------------------------------------------------
544		Name: Matthias Blume
545		Date: 2000/06/19 11:05:00 JST
546		Tag: blume-20000619-chdir-bugfix
547		Description:
548
549		Fixed a bug in new SrcPath module that sometimes led to a bad chDir call.
550
551		----------------------------------------------------------------------
552		Name: Matthias Blume
553		Date: 2000/06/18 22:00:10 JST
554		Tag: blume-20000618-implicit-anchors-really-gone
555		Description:
556
557		I updates the previous HISTORY entry where I forgot to mention that
558		implicit anchors are no longer with us.
559
560		The current update also gets rid of the (now useless) controller
561		CM.Control.implicit_anchors.
562
563		----------------------------------------------------------------------
564		Name: Matthias Blume
565		Date: 2000/06/16 17:30:00 JST
566		Tag: blume-20000616-anchorenv
567		Description:
568
569		This patch implements the long anticipated (just kidding :) "anchor
570		environment" mechanism.  In the course of doing this, I also
571		re-implemented CM's internal "SrcPath" module from scratch.  The new
572		one should be more robust in certain boundary cases.  In any case, it
573		is a lot cleaner than its predecessor (IMHO).
574
575		This time, although there is yet another boot file format change, I
576		kept the unpickler backward-compatible.  As a result, no new bootfiles
577		are necessary and bootstrapping is straightforward.  (You cannot read
578		new bootfiles into an old system, but the other way around is no
579		problem.)
580
581		Visible changes:
582
583		** 0. Implicit path anchors (without the leading $-symbol) are no
584		longer recognized at all. This means that such path names are not
585		illegal either.  For example, the name basis.cm simply refers to a
586		local file called "basis.cm" (i.e, the name is an ordinary path
587		relative to .cm-files directory).  Or, to put it differently, only
588		names that start with $ are anchored paths.
589
590		** 1. The $<singlearc> abbreviation for $/<singlearc> has finally
591		vanished.
592
593		John (Reppy) had critizised this as soon as I originally proposed and
594		implemented it, but at that time I did not really deeply believe
595		him. :) Now I came full-circle because I need the $<singlearc> syntax
596		in another place where it cannot be seen as an abbreviation for
597		$/<singlearc>.  To avoid the confusion, $<singlearc> now means what it
598		seems to mean (i.e., it "expands" into the corresponding anchor
599		value).
600
601		However, when paths are used as members in CM description files, it
602		continues to be true that there must be at least another arc after the
603		anchor.  This is now enforced separately during semantic analysis
604		(i.e., from a lexical/syntactical point of view, the notation is ok.)
605
606		** 2. The "cm" class now accepts an option "bind".  The option's value
607		is a sub-option list of precisely two items -- one labeled "anchor"
608		and the other one labeled "value".  As you might expect, "anchor" is
609		used to specify an anchor name to be bound, and "value" specifies what
610		the anchor is being bound to.
611
612		The value must be a directory name and can be given in either standard
613		syntax (including the possibility that it is itself an anchored path)
614		or native syntax.
615
616		Examples:
617
618		foo.cm (bind:(anchor:bar value:$mystuff/bar))
619		lib.cm (bind:(anchor:a value:"H:\\x\\y\\z"))  (* only works under windows *)
620
621		and so on.
622
623		The meaning of this is that the .cm-file will be processed with an
624		augmented anchor environment where the given anchor(s) is/are bound to
625		the given values(s).
626
627		The rationale for having this feature is this: Suppose you are trying
628		to use two different (already stable) libraries a.cm and b.cm (that
629		you perhaps didn't write yourself).  Further, suppose each of these
630		two libraries internally uses its own auxiliary library $aux/lib.cm.
631		Normally you would now have a problem because the anchor "lib" can not
632		be bound to more than one value globally.  Therefore, the project that
633		uses both a.cm and b.cm must locally redirect the anchor to some other
634		place:
635
636		a.cm (bind:(anchor:lib value:/usr/lib/smlnj/a-stuff))
637		b.cm (bind:(anchor:lib value:/usr/lib/smlnj/b-stuff))
638
639		This hard-wires $lib/aux.cm to /usr/lib/smlnj/a-stuff/aux.cm or
640		/usr/lib/smlnj/b-stuff/aux.cm, respectively.
641
642		Hard-wiring path names is a bit inflexible (and CM will verbosely warn
643		you when you do so at the time of CM.stabilize).  Therefore, you can
644		also use an anchored path as the value:
645
646		a.cm (bind:(anchor:lib value:$a-lib))
647		b.cm (bind:(anchor:lib value:$b-lib))
648
649		Now you can globally configure (using the usual CM.Anchor.anchor or
650		pathconfig machinery) bindings for "a-lib" and "b-lib".  Since "lib"
651		itself is always locally bound, setting it globally is no longer
652		meaningful or necessary (but it does not hurt either).  In fact, "lib"
653		can still be used as a global anchor for separate purposes.  As a
654		matter of fact, one can locally define "lib" in terms of a global
655		"lib":
656
657		a.cm (bind:(anchor:lib value:$lib/a))
658		b.cm (bind:(anchor:lib value:$lib/b))
659
660		** 3: The encoding of path names has changed.  This affects the way
661		path names are shown in CM's progress report and also the internal
662		protocol encoding used for parallel make.
663
664		The encoding now uses one or more ':'-separated segments.  Each
665		segments corresponds to a file that has been specified relative to the
666		file given by its preceding segment.  The first segment is either
667		relative to the CWD, absolute, or anchored.  Each segment itself is
668		basically a Unix pathname; all segments but the first are relative.
669
670		Example:
671
672		$foo/bar/baz.cm:a/b/c.sml
673
674		This path denotes the file bar/a/b/c.sml relative to the directory
675		denoted by anchor "foo".  Notice that the encoding also includes
676		baz.cm which is the .cm-file that listed a/b/c.sml.  As usual, such
677		paths are resolved relative to the .cm-files directory, so baz.cm must
678		be ignored to get the "real" pathname.
679
680		To make this fact more obvious, CM puts the names of such "virtual
681		arcs" into parentheses when they appear in progress reports. (No
682		parentheses will appear in the internal protocol encoding.)  Thus,
683		what you really see is:
684
685		$foo/bar/(baz.cm):a/b/c.sml
686
687		I find this notation to be much more informative than before.
688
689		Another new feature of the encoding is that special characters
690		including parentheses, colons, (back)slashes, and white space are
691		written as \ddd (where ddd is the decimal encoding of the character).
692
693		*** The CM manual still needs to be updated.
694
695		----------------------------------------------------------------------
696		Name: Allen Leung
697		Date: 2000/06/15 00:38:00
698		Tag: leunga-20000615-x86-peephole
699
700		x86 Peephole fix by Fermin.  Affects c-- and moby only.
701
702		----------------------------------------------------------------------
703		Name: Matthias Blume
704		Date: 2000/06/12 11:40:00
705		Tag: blume-20000612-parmakefix
706		Description:
707
708		More cleanup after changing the file naming scheme: This time I
709		repaired the parallel make mechanism for CMB.make which I broke earlier.
710
711		----------------------------------------------------------------------
712		Name: Allen Leung
713		Date: 2000/06/09 01:25:00
714		Tag: leunga-20000609-various
715
716		None of these things should affect normal SML/NJ operations
717
718		1. Peephole improvements provided by Fermin (c--)
719		2. New annotation DEFUSE for adding extra dependence (moby)
720		3. New X86 LOCK instructions (moby)
721		4. New machine description language for reservation tables (scheduling)
722		5. Fixes to various optimization/analysis modules (branch chaining, dominator
723		trees etc.)
724		6. I've changed the CM files so that they can work with versions
725		110.0.6, 110.25 and 110.28
726
727		----------------------------------------------------------------------
728		Name: Matthias Blume
729		Date: 2000/06/09 12:40:00
730		Tag: blume-20000609-log
731		Description:
732
733		- Removed all(?) remaining RCS Log entries from sources.
734
735		- Fixed bug in ml-yacc and ml-lex sources (use explicit anchors for
736		anchored paths).
737
738		----------------------------------------------------------------------
739		Name: Matthias Blume
740		Date: 2000/06/07 17:00:00 JST
741		Tag: blume-20000607-no-implicit-anchors
742		Description:
743
744		1. This update changes the default setting for
745		CM.Control.implicit_anchors from true to false.  This means that
746		implicit anchors are no longer permitted by default.  I also tried to
747		make sure that nothing else still relies on implicit anchors.
748		(This is the next step on the schedule towards a CM that does not even
749		have the notion of implicit anchors anymore.)
750
751		2. More CM manual updates.
752
753		3. I managed to track down and fix the pickling bug I mentioned last
754		time.  Because of the previously existing workaround, this entails no
755		immediate practical changes.
756
757		----------------------------------------------------------------------
758		Name: Matthias Blume
759		Date: 2000/06/06 11:15:00 JST
760		Tag: blume-20000606-lazierpickle
761		Description:
762
763		!!!! NEW BOOT FILES !!!!
764
765		* The main purpose of this update is to make library pickles lazier in
766		order to reduce the initial space penalty for autoloading a library.
767		As a result, it is now possible to have $smlnj/compiler.cm
768		pre-registered.  This should take care of the many complaints or
769		inquiries about missing structure Compiler.  This required changes to
770		CM's internal data structures and small tweaks to some algorithms.
771
772		As a neat additional effect, it is no longer necessary (for the sake
773		of lean heap image files) to distinguish between a "minimal" CM and a
774		"full" CM.  Now, there is only one CM (i.e., the "full" version:
775		$smlnj/cm.cm aka $smlnj/cm/full.cm), and it is always available at the
776		interactive top level. ($smlnj/cm/minimal.cm is gone.)
777
778		To make the life of compiler-hackers easier, "makeml" now also
779		pre-registers $smlnj/cmb.cm (aka $smlnj/cmb/current.cm).  In other
780		words, after you bootstrap a new sml for the first time, you will not
781		have to autoload $smlnj/cmb.cm again afterwards.  (The first time
782		around you will still have to do it, though.)
783
784		* A second change consists of major updates to the CM manual.  There
785		are now several appendices with summary information and also a full
786		specification of the CM description file syntax.
787
788		* In directory src/system I added the script "allcross".  This script
789		invokes sml and cross-compiles the compiler for all supported
790		architectures.  (Useful when providing a new set of boot files.)
791
792		* There seems to be a latent bug in my "lazy pickles" mechanism.  I
793		added a small tweak to pickle-util.sml to work around this problem,
794		but it is not a proper fix yet.  I will investigate further.  (The
795		effect of the bug was an inflation of library pickle size.)
796
797		* Version number increased to 110.28.1 (to avoid compatibility problems).
798
799		----------------------------------------------------------------------
800		Name: Allen Leung
801		Date: 2000/05/25 17:28 EDT
802		Tag: leunga-20000525-ra
803		Description:
804
805		Fixed a bug in freezing phase of the register allocator.
806
807		----------------------------------------------------------------------
808		Name: Allen Leung
809		Date: 2000/05/15 22:53 EDT
810		Tag: leunga-20000515-alpha-x86-ra
811		Description:
812
813		1. Alpha
814
815		Slight cleanup.  Removed the instruction SGNXL
816
817		2. X86
818
819		Added the following instructions to the instruction set:
820
821		ROLx, RORx,
822		BTx, BTSx, BTLx, BTRx,
823		XCHGx, and variants with the LOCK prefix
824
825		3. Register Allocation
826
827		The module ra-rewrite-with-renaming has been improved.
828
829		These have no effect on SML/NJ.
830
831		----------------------------------------------------------------------
832		Name: Matthias Blume
833		Date: 2000/05/15 16:20:00 JST
834		Tag: blume-20000515-lightrebuild
835		Description:
836
837		1. I added an alternative to "-rebuild" to "makeml".  The difference is
838		that prior to calling CMB.make' the CM-variable "LIGHT" will be
839		defined.  In effect, the command will not build any cross-compiler
840		backends and therefore finish more quickly.
841
842		The "fixpt" script also takes a "-light" switch to be able to use
843		this new facility while compiling for a fixpoint.
844
845		2. I replaced all mentions of anchored paths in group owner specifications
846		with simple relative paths (usually starting with "..").
847		The rationale is that a library's internal workings should not be
848		compromised by the lack of some anchor.  (An anchor is necessary
849		for someone who wants to refer to the library by an anchored path,
850		but it should not be necessary to build the same library in the first
851		place.)
852
853		3. I changed the way CM's tool mechanism determines the shell command
854		string used for things like ml-yacc etc. so that it does not break
855		when CM.Control.implicit_anchors is turned off.
856
857		----------------------------------------------------------------------
858		Name: Matthias Blume
859		Date: 2000/05/12 18:20:00 JST
860		Tag: blume-20000512-ml-build
861		Description:
862
863		Fixed a bug in config/_ml-build that prevented ml-yacc and ml-lex from
864		getting installed properly (by config/install.sh).
865
866		----------------------------------------------------------------------
867		Name: Matthias Blume
868		Date: 2000/05/12 17:30:00 JST
869		Tag: blume-20000512-anchors
870		Description:
871
872		!!! NEW BOOT FILES !!!
873
874		This change is in preparation of fading out support for "implicitly
875		anchored path names".  I went through all sources and used the
876		explicit (and relatively new) $-notation.  See system/README and the
877		CM manual for more info on this.
878
879		I also modified the anchoring scheme for some things such as "smlnj",
880		"MLRISC", "cm", etc. to take advantage of the fact that explicit
881		anchors are more expressive: anchor name and first arc do not have to
882		coincide.  This entails the following user-visible change:
883
884		You have to write $smlnj/foo/bar instead of smlnj/foo/bar.  In
885		particular, when you fire up sml with a command-line argument, say,
886		e.g.:
887
888		sml '$smlnj/cmb.cm'
889
890		At the ML toplevel prompt:
891
892		CM.autoload "$smlnj/cmb.cm";
893
894		There is also a new controller in CM.Control that can be used to turn
895		off all remaining support for implicit anchors by saying:
896
897		CM.autoload "$smlnj/
898		#set CM.Control.implicit_anchors false;
899
900		This causes CM to reject implicitly anchored paths.  This is (for the
901		time being) less permissive than the "final" version where there will
902		be no more such implicit anchors and relative paths will be just that:
903		relative.
904
905		The next step (version after next version?) will be to make the
906		default for CM.Control.implicit_anchors false.  After the dust has
907		settled, I can then produce the "final" version of this...
908
909		Note: Since bootstrapping is a bit tricky, I provided new boot files.
910
911		----------------------------------------------------------------------
912		Name: Matthias Blume
913		Date: 2000/05/11 16:30:00 JST
914		Tag: blume-20000511-sources
915		Description:
916
917		The main change is that I added function CM.sources as a generalized
918		version of the earlier CM.makedepend.  This entails the following
919		additional changes:
920
921		- CM.makedepend has been dropped.
922
923		- CM manual has been updated.
924
925		- TOOLS signature and API have been changed.
926
927		----------------------------------------------------------------------
928		Name: Allen Leung
929		Date: 2000/05/10 21:17 EDT
930		Tag: leunga-20000510-moby-c--ssa
931		Description:
932
933		Various bug fixes and new features for C--, Moby and MLRISC optimizations.
934		None of these affect SML/NJ.
935
936		1. Register Allocation
937
938		a. A new ra spilling module (ra/ra-spill-with-renaming) is implemented.
939		This module tries to remove local (i.e. basic block level) redundancies
940		during spilling.
941
942		b. A new framework for performing region based register allocation.
943		Not yet entirely functional.
944
945		2. X86
946
947		a. DefUse for POP was missing the stack pointer [found by Lal]
948		b. Reload for CALL was incorrect in X86Spill [found by John]
949		c. Various fixes in X86Spill so that it can be used correctly for
950		the new spilling module.
951
952		3. SSA/IR
953
954		a. New module ir/dj-dataflow.sml implements elimination based
955		data flow analysis.
956
957		4. MLRiscGen
958
959		a. Fix for gc type annotation
960
961		5. MDGen
962
963		Various fixes for machine description -> ml code translation.  For ssa
964		only.
965
966		----------------------------------------------------------------------
967		Name: Allen Leung
968		Date: 2000/05/08 22:17 EDT
969		Tag: leunga-20000508-labexp
970		Description:
971
972		Fermin has found a few assembly problems with constant expressions
973		generated in LabelExp.  Mostly, the problems involve extra parentheses,
974		which choke on dumb assemblers.  This is his fix.
975
976		----------------------------------------------------------------------
977		Name: Dave MacQueen
978		Date: 2000/04/09 14:00 EDT
979		Tag: dbm-20000502-Version_110_28
980		Description:
981
982		1. Updated src/compiler/TopLevel/main/version.sml to version 110.28
983
984		2. Updated config/version to 110.28
985
986		3. Updated config/srcarchiveurl
987
988		3. New boot files!
989		ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.28/
990
991		----------------------------------------------------------------------
992		Name: Matthias Blume
993		Date: 2000/05/01 19:05:00 JST
994		Tag: blume-20000501-noweb
995		Description:
996
997		A new noweb tool has been added.  The existing system is entirely
998		unaffected by this, but some CM users have asked for renewed noweb
999		support.  Everything is documented in the CM manual.
1000
1001		New (plugin) libraries:
1002
1003		noweb-tool.cm
1004		nw-ext.cm
1005
1006		----------------------------------------------------------------------
1007		Name: Dave MacQueen
1008		Date: 2000/04/30 12:40PM EDT
1009		Tag: dbm-20000430-bug_fixes
1010		Description:
1011
1012		1. Fix for bug 1498
1013		smlnj/src/system/Basis/Implementation/Unsafe/object.sig
1014		smlnj/src/system/Basis/Implementation/Unsafe/object.sml
1015		added toRealArray function
1016		smlnj/src/compiler/MiscUtil/print/ppobj.sml
1017		added check for tag Obj.RealArray to array printing case in ppObj
1018
1019		2. Fix for bug 1510
1020		smlnj/src/compiler/Semant/types/typesutil.sml
1021		fixed definition of dummyargs (used by equalTycon) so that
1022		dummy args are distinct types
1023
1024		----------------------------------------------------------------------
1025		Name: Matthias Blume
1026		Date: 2000/04/30 01:00:00 JST
1027		Tag: blume-20000430-versions
1028		Description:
1029
1030		1. CM version numbering added.  This is an implementation of Lal's
1031		proposal for adding version numbers and version checking to .cm
1032		files.  Lal said that his proposal was just that -- a proposal.
1033		For the time being I went ahead and implemented it so that people
1034		can comment on it.  Everything is completely backward-compatible
1035		(except for the stable library format, i.e., new bootfiles!).
1036
1037		As usual, see the CM manual for details.
1038
1039		2. An alternative syntax for anchored paths has been implemented.
1040		Dave has recently voiced the same concerns that I had when I did
1041		this, so there should be some support.  My take is that eventually
1042		I will let support for the current syntax (where anchors are
1043		"implicit") fade out in favor of the new, explicit syntax.
1044		In order to be backward-compatible, both old and new syntax are
1045		currently supported.
1046
1047		Again, see the CM manual for details.
1048
1049		3. Parallel make is trying to be slightly smarter:  When the master
1050		process finds a "bottleneck", i.e., when there is only one
1051		compilation unit that can be compiled and everybody else is
1052		waiting on it, then it will simply compile it directly instead
1053		of clumsily telling one of the slaves to do it.
1054
1055		4. Support for "unsharing" added.  This is necessary in order to be
1056		able to have two different versions of the same library running
1057		at the same time (e.g., for trying out a new MLRISC while still
1058		having the old MLRISC linked into the current compiler, etc.)
1059		See the CM manual.
1060
1061		5. Simple "makedepend" functionality added for generating Makefile
1062		dependency information.  (This is rather crude at the moment.
1063		Expect some changes here in the future.)
1064
1065		6. ".fun" added as a recognized suffix for ML files. Also documented
1066		explicitly in the manual that the fallback behavior (unknown suffix
1067		-> ML file) is not an official feature!
1068
1069		7. Small changes to the pickler for stable libraries.
1070
1071		8. Several internal changes to CM (for cleanup/improvement).
1072
1073
1074		!!!! NEW BINFILES !!!!
1075
1076		----------------------------------------------------------------------
1077		Name: Matthias Blume
1078		Date: 2000/04/28 17:30:00 JST
1079		Tag: blume-20000428-pathconfig
1080		Description:
1081
1082		1. I changed config/install.sh to remove duplicate entries from the
1083		lib/pathconfig file at the end.  Moreover, the final version of
1084		lib/pathconfig is sorted alphabetically.  The same (sorting) is done
1085		in src/system/installml.
1086
1087		2. The config/install.sh script now consistently uses relative
1088		pathnames in lib/pathconfig whenever the anchor is in the lib
1089		directory.  (So far this was true for the libraries that come
1090		pre-compiled and bundled as part of the bootfiles but not for
1091		libraries that are compiled by the script itself.)
1092
1093		----------------------------------------------------------------------
1094		Name: Matthias Blume
1095		Date: 2000/04/26 13:10:00 JST
1096		Tag: blume-20000426-fun_suffix
1097		Description:
1098
1099		Added ".fun" as a recognized file name suffix (for ML code).
1100
1101		----------------------------------------------------------------------
1102		Name: Allen Leung
1103		Date: 2000/04/25 17:00:00 EST
1104		Tag: leunga-20000425-alpha-ra
1105		Description:
1106
1107		1. Alpha
1108
1109		PSEUDOARITH was missing in AlphaRewrite.  This causes an endless loop
1110		in C--.
1111
1112		2. RA
1113
1114		Added a flag "ra-dump-size" to print out the size of the flowgraph
1115		and the interference graph.
1116
1117		----------------------------------------------------------------------
1118		Name: Dave MacQueen
1119		Date: 2000/04/25/
1120		Tag: dbm-20000425-mlyacc_doc_examples
1121		Description:
1122		Updated mlyacc.tex sections 5 and 7 for SML '97 and CM.
1123		Updated all three examples in src/ml-yacc/examples to run
1124		under 110.* using CM.make.
1125
1126		----------------------------------------------------------------------
1127		Name: Allen Leung
1128		Date: 2000/04/20 23:04:00 EST
1129		Tag: leunga-20000420-ssa-c---stuff
1130		Description:
1131
1132		This update synchronizes my repository with Yale's.  Most of these
1133		changes, however, do not affect SML/NJ at all (the RA is an exception).
1134
1135		1. Register Allocator
1136
1137		a. An improvement in the interference graph construction:
1138		Given a copy
1139
1140		s <- t
1141
1142		no interference edge between s and t is added for this definition of s.
1143
1144		b. I've added two new spill heuristic modules that Fermin and I developed
1145		(in the new library RA.cm). These are unused in SML/NJ but maybe
1146		useful for others (Moby?)
1147
1148		2. X86
1149
1150		a. Various fixes in the backend provided by Fermin [C--] and Lal.
1151
1152		3. Alpha
1153
1154		a. Added the BSR instruction and code generation that goes with it [C--]
1155		b. Other fixes too numerous to recount provided by Fermin [C--]
1156
1157		4. Regmaps
1158
1159		a. The regmaps are not initialized with the identity physical bindings
1160		at creation time.  This is unneeded.
1161
1162		5. MLRISC Optimizations
1163
1164		a. The DJ-Graph module can now compute the iterated dominance frontiers
1165		intersects with liveness incrementally in linear time! Woohoo!
1166		This is now used in my new SSA construction algorithm.
1167
1168		b. THe branch reorganization module is now smarter about linear chains of
1169		basic blocks.
1170
1171
1172		----------------------------------------------------------------------
1173		Name: Matthias Blume
1174		Date: 2000/04/12 13:52:00 JST
1175		Tag: blume_main_v110p27_1
1176		Description:
1177
1178		Changed install.sh script to handle archive files without version number
1179		and to use "boot.<arch>-<os>" instead of "sml.boot.<arch>-<os>" for the
1180		name of the boot file archive.
1181
1182		----------------------------------------------------------------------
1183		Name: Dave MacQueen
1184		Date: 2000/04/09 14:00 EDT
1185		Tag: dbm-20000410-Version_110_27
1186		Description:
1187
1188		1. Updated src/compiler/TopLevel/main/version.sml to version 110.27
1189
1190		2. Updated src/config/version to 110.27
1191
1192		3. New boot files!
1193
1194		----------------------------------------------------------------------
1195		Name: Allen Leung
1196		Date: 2000/04/09 19:09:00 EST
1197		Tag: leunga-20000409-misc
1198		Description:
1199
1200		1.  Yet another fix for x86 assembly for idivl, imull, mull and friends.
1201
1202		2.  Miscellaneous improvements to MLRISC (unused in sml/nj)
1203
1204		----------------------------------------------------------------------
1205		Name: Stefan
1206		Date: 2000/04/07 10:00:00 EDT
1207		Tag: monnier-20000406-branch-handling
1208		Description:
1209
1210		Improved handling of branches (mostly those generated from
1211		polymorphic equality), removed switchoff and changed the
1212		default optimization settings (more cpsopt and less flintopt).
1213
1214		----------------------------------------------------------------------
1215		Name: Allen Leung
1216		Date: 2000/04/06 01:30:00 EST
1217		Tag: leunga-20000406-peephole-x86-SSA-2
1218		Description:
1219
1220		Forgot a few files.
1221
1222		----------------------------------------------------------------------
1223		Name: Allen Leung
1224		Date: 2000/04/06 00:36:00 EST
1225		Tag: leunga-20000406-peephole-x86-SSA
1226		Description:
1227
1228		1.  New Peephole code
1229
1230		2.  Minor improvement to X86 instruction selection
1231
1232		3.  Various fixes to SSA and machine description -> code translator
1233
1234		----------------------------------------------------------------------
1235		Name: Matthias Blume
1236		Date: 2000/04/05 12:30:00 JST
1237		Tag: blume_main_v110p26p2_3
1238		Description:
1239
1240		This update just merges three minor cosmetic updates to CM's sources
1241		to get ready for the 110.27 code freeze on Friday.  No functionality
1242		has changed.
1243
1244		----------------------------------------------------------------------
1245		Name: Allen Leung
1246		Date: 2000/04/04 19:39:00 EST
1247		Tag: leunga-20000404-x86-asm
1248		Description:
1249
1250		1.  Fixed a problem in X86 assembly.
1251
1252		Things like
1253
1254		jmp %eax
1255		jmp (%eax)
1256
1257		should be output as
1258
1259		jmp *%eax
1260		jmp *(%eax)
1261
1262		2.  Assembly output
1263
1264		Added a new flag
1265
1266		"asm-indent-copies" (default to false)
1267
1268		When this flag is on, parallel copies will be indented an extra level.
1269
1270		----------------------------------------------------------------------
1271		Name: Allen Leung
1272		Date: 2000/04/04 03:18:00 EST
1273		Tag: leunga-20000404-C--Moby
1274		Description:
1275
1276		All of these fixes are related to C--, Moby, and my own optimization
1277		stuff; so they shouldn't affect SML/NJ.
1278
1279		1.  X86
1280
1281		Various fixes related floating point, and extensions.
1282
1283		2.  Alpha
1284
1285		Some extra patterns related to loads with signed/zero extension
1286		provided by Fermin.
1287
1288		3.  Assembly
1289
1290		When generating assembly, resolve the value of client defined constants,
1291		instead of generating symbolic values.  This is controlled by the
1292		new flag "asm-resolve-constants", which is default to true.
1293
1294		4.  Machine Descriptions
1295
1296		a. The precedence parser was slightly broken when parsing infixr symbols.
1297		b. The type generalizing code had the bound variables reversed, resulting
1298		in a problem during arity raising.
1299		c. Various fixes in machine descriptions.
1300
1301		----------------------------------------------------------------------
1302		Name: Matthias Blume
1303		Date: 2000/04/03 16:05:00 JST
1304		Tag: blume_main_v110p26p2_2
1305		Description:
1306
1307		I eliminated coreEnv from compInfo.  Access to the "Core" structure is
1308		now done via the ordinary static environment that is context to each
1309		compilation unit.
1310
1311		To this end, I arranged that instead of "structure Core" as "structure
1312		_Core" is bound in the pervasive environment.  Core access is done via
1313		_Core (which can never be accidentally rebound because _Core is not a
1314		legal surface-syntax symbol).
1315
1316		The current solution is much cleaner because the core environment is
1317		now simply part of the pervasive environment which is part of every
1318		compilation unit's context anyway.  In particular, this eliminates all
1319		special-case handling that was necessary until now in order to deal
1320		with dynamic and symbolic parts of the core environment.
1321
1322		Remaining hackery (to bind the "magic" symbol _Core) is localized in the
1323		compilation manager's bootstrap compiler (actually: in the "init group"
1324		handling).  See the comments in src/system/smlnj/init/init.cmi for
1325		more details.
1326
1327		I also tried to track down all mentions of "Core" (as string argument
1328		to Symbol.strSymbol) in the compiler and replaced them with a
1329		reference to the new CoreSym.coreSym.  Seems cleaner since the actual
1330		name appears in one place only.
1331
1332		Binfile and bootfile format have not changed, but the switchover from
1333		the old "init.cmi" to the new one is a bit tricky, so I supplied new
1334		bootfiles anyway.
1335
1336		----------------------------------------------------------------------
1337		Name: Allen Leung
1338		Date: 2000/04/02 21:17:00 EST
1339		Tag: leunga-20000402-mltree
1340		Description:
1341
1342		1. Renamed the constructor CALL in MLTREE by popular demand.
1343		2. Added a bunch of files from my repository.  These are currently
1344		used by other non-SMLNJ backends.
1345
1346		----------------------------------------------------------------------
1347		Name: Allen Leung
1348		Date: 2000/03/31 21:15:00 EST
1349		Tag: leunga-20000331-aliasing
1350		Description:
1351
1352		This update contains a rewritten (and hopefully more correct) module
1353		for extracting aliasing information from CPS.
1354
1355		To turn on this feature:
1356
1357		Compiler.Control.CG.memDisambiguate := true
1358
1359		To pretty print the region information with assembly
1360
1361		Compiler.Control.MLRISC.getFlag "asm-show-region" := true;
1362
1363		To control how many levels of aliasing information are printed, use:
1364
1365		Compiler.Control.MLRISC.getInt "points-to-show-level" := n
1366
1367		The default of n is 3.
1368
1369		----------------------------------------------------------------------
1370		Name: David MacQueen
1371		Date: 2000/03/31 11:15:00 EST
1372		Tag: dbm-20000331-runtime_fix
1373		Description:
1374
1375		This update contains:
1376
1377		1. runtime/c-lib/c-libraries.c
1378		includes added in revision 1.2 caused compilation errors on hppa-hpux
1379
1380		2. fix for bug 1556
1381		system/Basis/Implementation/NJ/internal-signals.sml
1382
1383		----------------------------------------------------------------------
1384		Name: Matthias Blume
1385		Date: 2000/03/31 18:00:00 JST
1386		Tag: blume_main_v110p26p2_1
1387		Description:
1388
1389		This update contains:
1390
1391		1. A small change to CM's handling of stable libraries:
1392		CM now maintains one "global" modmap that is used for all stable
1393		libraries.  The use of such a global modmap maximizes sharing and
1394		minimizes the need for re-traversing parts of environments during
1395		modmap construction.  (However, this has minor impact since modmap
1396		construction seems to account for just one percent or less of total
1397		compile time.)
1398
1399		2. I added a "genmap" phase to the statistics.  This is where I got the
1400		"one percent" number (see above).
1401
1402		3. CM's new tool parameter mechanism just became _even_ better. :)
1403		- The parser understands named parameters and recursive options.
1404		- The "make" and "shell" tools use these new features.
1405		(This makes it a lot easier to cascade these tools.)
1406		- There is a small syntax change: named parameters use a
1407
1408		<name> : ( <option> ... )            or
1409		<name> : <string>
1410
1411		syntax.  Previously, named parameters were implemented in an
1412		ad-hoc fashion by each tool individually (by parsing strings)
1413		and had the form
1414
1415		<name>=<string>
1416
1417		See the CM manual for a full description of these issues.
1418
1419		----------------------------------------------------------------------
1420		Name: Matthias Blume
1421		Date: 2000/03/30 18:00:00 JST
1422		Tag: blume_main_v110p26p2_0
1423		Description:
1424
1425		!!!!! WARNING !!!!!!
1426		!!  New binfiles  !!
1427		!!!!!!!!!!!!!!!!!!!!
1428
1429		This update contains:
1430
1431		1. Moderate changes to CM:
1432
1433		- Changes to CM's tools mechanism.  In particular, it is now possible
1434		to have tools that accept additional "command line" parameters
1435		(specified in the .cm file at each instance where the tool's class is
1436		used).
1437
1438		This was done to accommodate the new "make" and "shell" tools which
1439		facilitate fairly seamless hookup to portions of code managed using
1440		Makefiles or Shell scripts.
1441
1442		There are no classes "shared" or "private" anymore.  Instead, the
1443		sharing annotation is now a parameter to the "sml" class.
1444
1445		There is a bit of generic machinery for implementing one's own
1446		tools that accept command-line parameters.  However, I am not yet fully
1447		satisfied with that part, so expect changes here in the future.
1448
1449		All existing tools are described in the CM manual.
1450
1451		- Slightly better error handling.  (CM now suppresses many followup
1452		error messages that tended to be more annoying than helpful.)
1453
1454		2. Major changes to the compiler's static environment data structures.
1455
1456		- no CMStaticEnv anymore.
1457		- no CMEnv, no "BareEnvironment" (actually, _only_ BareEnvironment,
1458		but it is called Environment), no conversions between different
1459		kinds of static environments
1460
1461		- There is still a notion of a "modmap", but such modmaps are generated
1462		on demand at the time when they are needed.  This sounds slow, but I
1463		sped up the code that generates modmaps enough for this not to lead to
1464		a slowdown of the compiler (at least I didn't detect any).
1465
1466		- To facilitate rapid modmap generation, static environments now
1467		contain an (optional) "modtree" structure.  Modtree annotations are
1468		constructed by the unpickler during unpickling.  (This means that
1469		the elaborator does not have to worry about modtrees at all.)
1470		Modtrees have the advantage that they are compositional in the same
1471		way as the environment data structure itself is compositional.
1472		As a result, modtrees never hang on to parts of an environment that
1473		has already been rendered "stale" by filtering or rebinding.
1474
1475		- I went through many, many trials and errors before arriving at the
1476		current solution.  (The initial idea of "linkpaths" did not work.)
1477		But the result of all this is that I have touched a lot of files that
1478		depend on the "modules" and "types" data structures (most of the
1479		elaborator). There were a lot of changes during my "linkpath" trials
1480		that could have been reverted to their original state but weren't.
1481		Please, don't be too harsh on me for messing with this code a bit more
1482		than what was strictly necessary...  (I _did_ resist the tempation
1483		of doing any "global reformatting" to avoid an untimely death at
1484		Dave's hands. :)
1485
1486		- One positive aspect of the previous point:  At least I made sure that
1487		all files that I touched now compile without warnings (other than
1488		"polyEqual").
1489
1490		- compiler now tends to run "leaner" (i.e., ties up less memory in
1491		redundant modmaps)
1492
1493		----------------------------------------------------------------------
1494		Name: Allen Leung
1495		Date: 2000/03/29 18:00:00
1496		Tag: leunga-20000327-mlriscGen_hppa_alpha_x86
1497		Boot files (optional): ftp://react-ilp.cs.nyu.edu/leunga/110.26.1-sml.boot.x86-unix-20000330.tar.gz
1498		Description:
1499
1500		This update contains *MAJOR* changes to the way code is generated from CPS
1501		in the module mlriscGen, and in various backend modules.
1502
1503		CHANGES
1504		=======
1505
1506		1. MLRiscGen: forward propagation fix.
1507
1508		There was a bug in forward propagation introduced at about the same time
1509		as the MLRISC x86 backend, which prohibits coalescing to be
1510		performed effectively in loops.
1511
1512		Effect: speed up of loops in RISC architectures.
1513		By itself, this actually slowed down certain benchmarks on the x86.
1514
1515		2. MLRiscGen:  forward propagating addresses from consing.
1516
1517		I've changed the way consing code is generated.  Basically I separated
1518		out the initialization part:
1519
1520		store tag,   offset(allocptr)
1521		store elem1, offset+4(allocptr)
1522		store elem2, offset+8(allocptr)
1523		...
1524		store elemn, offset+4n(allocptr)
1525
1526		and the address computation part:
1527
1528		celladdr <- offset+4+alloctpr
1529
1530		and move the address computation part
1531
1532		Effect:  register pressure is generally lower as a result.  This
1533		makes compilation of certain expressions much faster, such as
1534		long lists with non-trivial elements.
1535
1536		[(0,0), (0,0), .... (0,0)]
1537
1538		3. MLRiscGen: base pointer elimination.
1539
1540		As part of the linkage mechanism, we generate the sequence:
1541
1542		L:  ...  <- start of the code fragment
1543
1544		L1:
1545		base pointer <- linkreg - L1 + L
1546
1547		The base pointer was then used for computing relocatable addresses
1548		in the code fragment.  Frequently (such as in lots of continuations)
1549		this is not needed.  We now eliminate this sequence whenever possible.
1550
1551		For compile time efficiency, I'm using a very stupid local heuristic.
1552		But in general, this should be done as a control flow analysis.
1553
1554		Effect:  Smaller code size.  Speed up of most programs.
1555
1556		4. Hppa back end
1557
1558		Long jumps in span dependence resolution used to depend on the existence
1559		of the base pointer.
1560
1561		A jump to a long label L was expanded into the following sequence:
1562
1563		LDIL %hi(L-8192), %r29
1564		LDO  %lo(L-8192)(%r29), %r29
1565		ADD  %r29, baseptr, %r29
1566		BV,n %r0(%r29)
1567
1568		In the presence of change (3) above, this will not work.  I've changed
1569		it so that the following sequence of instructions are generated, which
1570		doesn't mention the base pointer at all:
1571
1572		BL,n  L', %r29           /* branch and link, L' + 4 -> %r29 */
1573		L':  ADDIL L-(L'+4), %r29     /* Compute address of L */
1574		BV,n  %r0(%r29)          /* Jump */
1575
1576		5. Alpha back end
1577
1578		New alpha instructions LDB/LDW have been added, as per Fermin's
1579		suggestions.   This is unrelated to all other changes.
1580
1581		6. X86 back end
1582
1583		I've changed andl to testl in the floating point test sequence
1584		whenever appropriate.  The Intel optimization guide states that
1585		testl is preferable to andl.
1586
1587		7. RA (x86 only)
1588
1589		I've improved the spill propagation algorithm, using an approximation
1590		of maximal weighted independent sets.   This seems to be necessary to
1591		alleviate the negative effect in light of the slow down in (1).
1592
1593		I'll write down the algorithm one of these days.
1594
1595		8. MLRiscGen: frequencies
1596
1597		I've added an annotation that states that all call gc blocks have zero
1598		execution frequencies.  This improves register allocation on the x86.
1599
1600		BENCHMARKS
1601		==========
1602
1603		I've only perform the comparison on 110.25.
1604
1605		The platforms are:
1606
1607		HPPA  A four processor HP machine (E9000) with 5G of memory.
1608		X86   A 300Hhz Pentium II with 128M of memory, and
1609		SPARC An Ultra sparc 2 with 512M of memory.
1610
1611		I used the following parameters for the SML benchmarks:
1612
1613		@SMLalloc
1614		HPPA    256k
1615		SPARC   512k
1616		X86     256k
1617
1618		COMPILATION TIME
1619		----------------
1620		Here are the numbers comparing the compilation times of the compilers.
1621		I've only compared 110.25 compiling the new sources versus
1622		a fixpoint version of the new compiler compiling the same.
1623
1624		110.25                                  New
1625		Total  Time in RA  Spill+Reload   Total  Time In RA Spill+Reload
1626		HPPA   627s    116s        2684+3584     599s    95s       1003+1879
1627		SPARC  892s    173s        2891+3870     708s    116s      1004+1880
1628		X86    999s    315s       94006+130691   987s    296s    108877+141957
1629
1630		110.25         New
1631		Code Size      Code Size
1632		HPPA   8596736         8561421
1633		SPARC  8974299         8785143
1634		X86    9029180         8716783
1635
1636		So in summary, things are at least as good as before.   Dramatic
1637		reduction in compilation is obtained on the Sparc; I can't explain it,
1638		but it is reproducible.  Perhaps someone should try to reproduce this
1639		on their own machines.
1640
1641		SML BENCHMARKS
1642		--------------
1643
1644		On the average, all benchmarks perform at least as well as before.
1645
1646		HPPA         Compilation Time     Spill+Reload      Run Time
1647		110.25  New            110.25    New   110.25  New
1648
1649		barnesHut  3.158  3.015  4.75%    1+1       0+0   2.980  2.922   2.00%
1650		boyer  6.152  5.708  7.77%    0+0       0+0   0.218  0.213   2.34%
1651		count-graphs  1.168  1.120  4.32%    0+0       0+0  22.705 23.073  -1.60%
1652		fft  0.877  0.792 10.74%    1+3       1+3   0.602  0.587   2.56%
1653		knuthBendix  3.180  2.857 11.32%    0+0       0+0   0.675  0.662   2.02%
1654		lexgen  6.190  5.290 17.01%    0+0       0+0   0.913  0.788  15.86%
1655		life  0.803  0.703 14.22%   25+25      0+0   0.153  0.140   9.52%
1656		logic  2.048  2.007  2.08%    6+6       1+1   4.133  4.008   3.12%
1657		mandelbrot  0.077  0.080 -4.17%    0+0       0+0   0.765  0.712   7.49%
1658		mlyacc 22.932 20.937  9.53%  154+181    32+57  0.468  0.430   8.91%
1659		nucleic  5.183  5.060  2.44%    2+2       0+0   0.125  0.120   4.17%
1660		ratio-regions  3.357  3.142  6.84%    0+0       0+0  116.225 113.173 2.70%
1661		ray  1.283  1.290 -0.52%    0+0       0+0   2.887  2.855   1.11%
1662		simple  6.307  6.032  4.56%   28+30      5+7   3.705  3.658   1.28%
1663		tsp  0.888  0.862  3.09%    0+0       0+0   7.040  6.893   2.13%
1664		vliw 24.378 23.455  3.94%  106+127    25+45  2.758  2.707   1.91%
1665		--------------------------------------------------------------------------
1666		Average                     6.12%                                   4.09%
1667
1668		SPARC        Compilation Time     Spill+Reload      Run Time
1669		110.25  New            110.25    New   110.25  New
1670
1671		barnesHut  3.778  3.592  5.20%    2+2       0+0   3.648  3.453    5.65%
1672		boyer  6.632  6.110  8.54%    0+0       0+0   0.258  0.242    6.90%
1673		count-graphs  1.435  1.325  8.30%    0+0       0+0  33.672 34.737   -3.07%
1674		fft  0.980  0.940  4.26%    3+9       2+6   0.838  0.827    1.41%
1675		knuthBendix  3.590  3.138 14.39%    0+0       0+0   0.962  0.967   -0.52%
1676		lexgen  6.593  6.072  8.59%    1+1       0+0   1.077  1.078   -0.15%
1677		life  0.972  0.868 11.90%   26+26      0+0   0.143  0.140    2.38%
1678		logic  2.525  2.387  5.80%    7+7       1+1   5.625  5.158    9.05%
1679		mandelbrot  0.090  0.093 -3.57%    0+0       0+0   0.855  0.728   17.39%
1680		mlyacc 26.732 23.827 12.19%  162+189    32+57  0.550  0.560   -1.79%
1681		nucleic  6.233  6.197  0.59%    3+3       0+0   0.163  0.173   -5.77%
1682		ratio-regions  3.780  3.507  7.79%    0+0       0+0 133.993 131.035   2.26%
1683		ray  1.595  1.550  2.90%    1+1       0+0   3.440  3.418    0.63%
1684		simple  6.972  6.487  7.48%   29+32      5+7   3.523  3.525   -0.05%
1685		tsp  1.115  1.063  4.86%    0+0       0+0   7.393  7.265    1.77%
1686		vliw 27.765 24.818 11.87%  110+135    25+45  2.265  2.135    6.09%
1687		----------------------------------------------------------------------------
1688		Average                     6.94%                                    2.64%
1689
1690		X86          Compilation Time     Spill+Reload      Run Time
1691		110.25  New            110.25    New   110.25  New
1692
1693		barnesHut  5.530  5.420  2.03%  593+893   597+915   3.532  3.440   2.66%
1694		boyer  8.768  7.747 13.19%  493+199   301+289   0.327  0.297  10.11%
1695		count-graphs  2.040  2.010  1.49%  298+394   315+457  26.578 28.660  -7.26%
1696		fft  1.327  1.302  1.92%  112+209   115+210   1.055  0.962   9.71%
1697		knuthBendix  5.218  5.475 -4.69%  451+598   510+650   0.928  0.932  -0.36%
1698		lexgen  9.970  9.623  3.60% 1014+841  1157+885   0.947  0.928   1.97%
1699		life  1.183  1.183  0.00%  162+182   145+148   0.127  0.103  22.58%
1700		logic  3.285  3.512 -6.45%  514+684   591+836   5.682  5.577   1.88%
1701		mandelbrot  0.147  0.143  2.33%   38+41     33+54    0.703  0.690   1.93%
1702		mlyacc 35.457 32.763  8.22% 3496+4564 3611+4860  0.552  0.550   0.30%
1703		nucleic  7.100  6.888  3.07%  239+168   201+158   0.175  0.173   0.96%
1704		ratio-regions  6.388  6.843 -6.65% 1182+257   981+300  120.142 120.345 -0.17%
1705		ray  2.332  2.338 -0.29%  346+398   402+494   3.593  3.540   1.51%
1706		simple  9.912  9.903  0.08% 1475+941  1579+1168  3.057  3.178  -3.83%
1707		tsp  1.623  1.532  5.98%  266+200   250+211   8.045  7.878   2.12%
1708		vliw 33.947 35.470 -4.29% 2629+2774 2877+3171  2.072  1.890   9.61%
1709		----------------------------------------------------------------------------
1710		Average                     1.22%                                     3.36%
1711
1712		----------------------------------------------------------------------
1713		Name: Allen Leung
1714		Date: 2000/03/23 16:25:00
1715		Tag: leunga-20000323-fix_x86_alpha
1716		Description:
1717
1718		1. X86 fixes/changes
1719
1720		a.  The old code generated for SETcc was completely wrong.
1721		The Intel optimization guide is VERY misleading.
1722
1723		2. ALPHA fixes/changes
1724
1725		a.  Added the instructions LDBU, LDWU, STB, STW as per Fermin's suggestion.
1726		b.  Added a new mode byteWordLoadStores to the functor parameter to Alpha()
1727		c.  Added reassociation code for address computation.
1728
1729		----------------------------------------------------------------------
1730		Name: Allen Leung
1731		Date: 2000/03/22 01:23:00
1732		Tag: leunga-20000322-fix_x86_hppa_ra
1733		Description:
1734
1735		1. X86 fixes/changes
1736
1737		a.  x86Rewrite bug with MUL3 (found by Lal)
1738		b.  Added the instructions FSTS, FSTL
1739
1740		2. PA-RISC fixes/changes
1741
1742		a.  B label should not be a delay slot candidate!  Why did this work?
1743		b.  ADDT(32, REG(32, r), LI n) now generates one instruction instead of two,
1744		as it should be.
1745		c.  The assembly syntax for fstds and fstdd was wrong.
1746		d.  Added the composite instruction COMICLR/LDO, which is the immediate
1747		operand variant of COMCLR/LDO.
1748
1749		3. Generic MLRISC
1750
1751		a.  shuffle.sml rewritten to be slightly more efficient
1752		b.  DIV bug in mltree-simplify fixed (found by Fermin)
1753
1754		4. Register Allocator
1755
1756		a.  I now release the interference graph earlier during spilling.
1757		May improve memory usage.
1758
1759		----------------------------------------------------------------------
1760		Name: Matthias Blume
1761	Date: 2000/03/14 14:15:32	Date: 2000/03/14 14:15:32
1762	Tag: blume_main_v110p26p1_2	Tag: blume_main_v110p26p1_2
1763	Description:	Description:
1764
1765	1. Tools.registerStdShellCmdTool (from smlnj/cm/tool.cm) takes an	1. Tools.registerStdShellCmdTool (from smlnj/cm/tool.cm) takes an
1766	additional argument called "template" which is an optional string that	additional argument called "template" which is an optional string that
1767	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
1768	explanation.	explanation.
1769
1770	2. A special-purpose tool can be "regisitered" by simply dropping the	2. A special-purpose tool can be "registered" by simply dropping the
1771	corresponding <...>-tool.cm (and/or <...>-ext.cm) into the same	corresponding <...>-tool.cm (and/or <...>-ext.cm) into the same
1772	directory where the .cm file lives that uses this tool.  (The	directory where the .cm file lives that uses this tool.  (The
1773	behavior/misfeature until now was to look for the tool description	behavior/misfeature until now was to look for the tool description
#	Line 66	Line 1811
1811	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
1812	<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,
1813	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
1814	perhaps got accidentially killed).  In this case, fixpt will quickly	perhaps got accidentally killed).  In this case, fixpt will quickly
1815	move through what exists before continuing where it left off earlier,	move through what exists before continuing where it left off earlier,
1816	and, thus, saves a lot of time.	and, thus, saves a lot of time.
1817
#	Line 116	Line 1861
1861	it from that remote directory.	it from that remote directory.
1862	This should simplify installation further:  For machines that have	This should simplify installation further:  For machines that have
1863	access to the internet, just fetch <version>-config.tgz, unpack it,	access to the internet, just fetch <version>-config.tgz, unpack it,
1864	edit config/targets, and go (run config/install.sh).  The scipt will	edit config/targets, and go (run config/install.sh).  The script will
1865	fetch everything else that it might need all by itself.	fetch everything else that it might need all by itself.
1866
1867	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.578   changed lines   Added in v.701

---

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