Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 572, Thu Mar 9 02:43:06 2000 UTC	revision 1016, Tue Jan 15 23:10:06 2002 UTC
#	Line 8	Line 8
8	The form of an entry should be:	The form of an entry should be:
9
10	Name:	Name:
11	Date:	Date: yyyy/mm/dd
12	Tag: <post-commit CVS tag>	Tag: <post-commit CVS tag>
13	Description:	Description:
14		----------------------------------------------------------------------
15		Name: Lal George
16		Date: 2002/01/15
17		Tag: <none>
18		Description:
19
20		1. Since COPY instructions are no longer native to the architecture,
21		a generic functor can be used to implement the expandCopies function.
22
23		2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
24		TEXT segment.
25
26		----------------------------------------------------------------------
27		Name: Matthias Blume
28		Date: 2002/01/15
29		Tag: blume-20020115-ffiupdates
30		Description:
31
32		1. Fix for bug resulting in single-precision float values being returned
33		incorrectly from FFI calls.
34
35		2. Small modifications to C FFI API:
36
37		- memory-allocation routines return straight objects (no options)
38		and raise an exception in out-of-memory situations
39		- unsafe extensions to cast between function pointers and pointers
40		from/to ints
41		- added structure C_Debug as an alternative to structure C where
42		pointer-dereferencing (|*| and |*!) always check for null-pointers
43		- added open_lib' to DynLinkage;  open_lib' works like open_lib
44		but also takes a (possibly empty) list of existing library handles
45		that the current library depends on
46
47		----------------------------------------------------------------------
48		Name: Matthias Blume
49		Date: 2002/01/10
50		Tag: blume-20020110-newffigen
51		Description:
52
53		1. Updates to portable graph code.
54
55		2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
56		more scalable now so that even huge interfaces such as the one
57		for GTK compile in finite time and space. :-)
58		See src/ml-nlffigen/README for details on what's new.
59
60		----------------------------------------------------------------------
61		Name: Lal George
62		Date: 2001/01/09 14:31:35 EST 2002
63		Tag: george-20011206-rm-native-copy
64		Description:
65
66		Removed the native COPY and FCOPY instructions
67		from all the architectures and replaced it with the
68		explicit COPY instruction from the previous commit.
69
70		It is now possible to simplify many of the optimizations
71		modules that manipulate copies. This has not been
72		done in this change.
73
74		----------------------------------------------------------------------
75		Name: Lal George
76		Date: 2001/12/06 16:50:13 EST 2001
77		Tag: george-20011206-mlrisc-instruction
78		Description:
79
80		Changed the representation of instructions from being fully abstract
81		to being partially concrete. That is to say:
82
83		from
84		type instruction
85
86		to
87		type instr                              (* machine instruction *)
88
89		datatype instruction =
90		LIVE of {regs: C.cellset, spilled: C.cellset}
91		| KILL of {regs: C.cellset, spilled: C.cellset}
92		| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
93		| ANNOTATION of {i: instruction, a: Annotations.annotation}
94		| INSTR of instr
95
96		This makes the handling of certain special instructions that appear on
97		all architectures easier and uniform.
98
99		LIVE and KILL say that a list of registers are live or killed at the
100		program point where they appear. No spill code is generated when an
101		element of the 'regs' field is spilled, but the register is moved to
102		the 'spilled' (which is present, more for debugging than anything else).
103
104		LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
105		We used to generate:
106
107		DEFFREG f1
108		f1 := f2 + f3
109		trapb
110
111		but now generate:
112
113		f1 := f2 + f3
114		trapb
115		LIVE {regs=[f1,f2,f3], spilled=[]}
116
117		Furthermore, the DEFFREG (hack) required that all floating point instruction
118		use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
119		defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
120		in a cleaner alpha implementation. (Hopefully, intel will not get rid of
121		this architecture).
122
123		COPYXXX is intended to replace the parallel COPY and FCOPY  available on
124		all the architectures. This will result in further simplification of the
125		register allocator that must be aware of them for coalescing purposes, and
126		will also simplify certain aspects of the machine description that provides
127		callbacks related to parallel copies.
128
129		ANNOTATION should be obvious, and now INSTR represents the honest to God
130		machine instruction set!
131
132		The <arch>/instructions/<arch>Instr.sml files define certain utility
133		functions for making porting easier -- essentially converting upper case
134		to lower case. All machine instructions (of type instr) are in upper case,
135		and the lower case form generates an MLRISC instruction. For example on
136		the alpha we have:
137
138		datatype instr =
139		LDA of {r:cell, b:cell, d:operand}
140		| ...
141
142		val lda : {r:cell, b:cell, d:operand} -> instruction
143		...
144
145		where lda is just (INSTR o LDA), etc.
146
147		----------------------------------------------------------------------
148		Name: Matthias Blume
149		Date: 2001/11/22 21:40:00 EST
150		Tag: Release_110_37
151		Description:
152
153		Release 110.37.  This time for real.
154
155		----------------------------------------------------------------------
156		Name: Matthias Blume
157		Date: 2001/11/21 16:35:00 EST
158		Tag: blume-20011121-foot-in-mouth
159		Description:
160
161		Removed the "Release_110_37" tag because of a serious bug.
162		This will be re-tagged once the bug is fixed.
163
164		----------------------------------------------------------------------
165		Name: Matthias Blume
166		Date: 2001/11/21 16:14:00 EST
167		Tag: blume-20011121-forgottenfile
168		Description:
169
170		Forgot to add a file.  (Just a .tex-file -- part of
171		the CM manual source.)
172
173		----------------------------------------------------------------------
174		Name: Matthias Blume
175		Date: 2001/11/21 16:10:00 EST
176		Tag: blume-20011121-invalid_110_37
177		Description:
178
179		Note: I removed the original tag "Release_110_37" from this commit
180		because we found a serious bug in all non-x86 backends.
181		- Matthias
182
183		1. Modifications to the SML/NJ code generator and to the runtime system
184		so that code object name strings are directly inserted into code
185		objects at code generation time.  The only business the runtime system
186		has with this is now to read the name strings on occasions.
187		(The encoding of the name string has also changed somewhat.)
188
189		2. CM now implements a simple "set calculus" for specifying export lists.
190		In particular, it is now possible to refer to the export lists of
191		other libraries/groups/sources and form unions as well as differences.
192		See the latest CM manual for details.
193
194		3. An separate notion of "proxy" libraries has again be eliminated from
195		CM's model.  (Proxy libraries are now simply a special case of using
196		the export list calculus.)
197
198		4. Some of the existing libraries now take advantage of the new set
199		calculus.
200		(Notice that not all libraries have been converted because some
201		of the existing .cm-files are supposed to be backward compatible
202		with 110.0.x.)
203
204		5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
205		"General.exnMessage"!  The former relies on a certain hook to be
206		initialized, and that often does not happen in the stand-alone case.)
207
208		----------------------------------------------------------------------
209		Name: Lal George
210		Date: 2001/11/21  13:56:18 EST
211		Tag: george-2001121-pseudo-ops
212		Description:
213
214		Implemented a complete redesign of MLRISC pseudo-ops. Now there
215		ought to never be any question of incompatabilities with
216		pseudo-op syntax expected by host assemblers.
217
218		For now, only modules supporting GAS syntax are implemented
219		but more should follow, such as MASM, and vendor assembler
220		syntax, e.g. IBM as, Sun as, etc.
221
222		----------------------------------------------------------------------
223		Name: Matthias Blume
224		Date: 2001/11/14 11:52:00 EST
225		Tag: blume-20011114-srcname
226		Description:
227
228		1. Routed the name of the current source file to mlriscgen where it
229		should be directly emitted into the code object.  (This last part
230		is yet to be done.)
231
232		2. Some cleanup of the pgraph code to make it match the proposal that
233		I put out the other day.  (The proposal notwithstanding, things are
234		still in flux here.)
235
236		----------------------------------------------------------------------
237		Name: Lal George
238		Date: 2001/11/14 09:44:04 EST
239		Tag:
240		Description:
241
242		Fix for a backpatching bug reported by Allen.
243
244		Because the boundary between short and long span-dependent
245		instructions is +/- 128, there are an astounding number of
246		span-dependent instructions whose size is over estimated.
247
248		Allen came up with the idea of letting the size of span
249		dependent instructions be non-monotonic, for a maxIter
250		number of times, after which the size must be monotonically
251		increasing.
252
253		This table shows the number of span-dependent instructions
254		whose size was over-estimated as a function of maxIter, for the
255		file Parse/parse/ml.grm.sml:
256
257		maxIter            # of instructions:
258		10                      687
259		20                      438
260		30                      198
261		40                        0
262
263		In compiling the compiler, there is no significant difference in
264		compilation speed between maxIter=10 and maxIter=40. Actually,
265		my measurements showed that maxIter=40 was a tad faster than
266		maxIter=10! Also 96% of the  files in the compiler reach a fix
267		point within 13 iterations, so fixing maxIter at 40, while high,
268		is okay.
269
270		----------------------------------------------------------------------
271		Name: Matthias Blume
272		Date: 2001/10/31 15:25:00 EST
273		Tag: blume-20011031-pgraph
274		Description:
275
276		CKIT:
277		* Changed the "Function" constructor of type Ast.ctype to carry optional
278		argument identifiers.
279		* Changed the return type of TypeUtil.getFunction accordingly.
280		* Type equality ignores the argument names.
281		* TypeUtil.composite tries to preserve argument names but gives up quickly
282		if there is a mismatch.
283
284		installation script:
285		* attempts to use "curl" if available (unless "wget" is available as well)
286
287		CM:
288		* has an experimental implementation of "portable graphs" which I will
289		soon propose as an implementation-independent library format
290		* there are also new libraries $/pgraph.cm and $/pgraph-util.cm
291
292		NLFFI-LIB:
293		* some cleanup (all cosmetic)
294
295		NLFFIGEN:
296		* temporarily disabled the mechanism that suppresses ML output for
297		C definitions whose identifiers start with an underscore character
298		* generate val bindings for enum constants
299		* user can request that only one style (light or heavy) is being used;
300		default is to use both (command-line arguments: -heavy and -light)
301		* fixed bug in handling of function types involving incomplete pointers
302		* generate ML entry points that take record arguments (i.e., using
303		named arguments) for C functions that have a prototype with named
304		arguments
305		(see changes to CKIT)
306
307		----------------------------------------------------------------------
308		Name: Allen Leung
309		Date: 2001/10/27 20:34:00 EDT
310		Tag: leunga-20011027-x86-fast-fp-call
311		Description:
312
313		Fixed the bug described in blume-20010920-slowfp.
314
315		The fix involves
316		1. generating FCOPYs in FSTP in ia32-svid
317		2. marking a CALL with the appropriate annotation
318
319		----------------------------------------------------------------------
320		Name: Matthias Blume
321		Date: 2001/10/16 11:32:00 EDT
322		Tag: blume-20011016-netbsd
323		Description:
324
325		Underscore patch from Chris Richards (fixing problem with compiling
326		runtime system under recent NetBSD).
327
328		----------------------------------------------------------------------
329		Name: Allen Leung
330		Date: 2001/10/12 17:18:32 EDT 2001
331		Tag: leung-20011012-x86-printflowgraph
332		Description:
333
334		X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
335
336		----------------------------------------------------------------------
337		Name: Lal George
338		Date: 2001/10/11 23:51:34 EDT
339		Tag: george-20011011-too-many-instrs
340		Description:
341
342		The representation of a program point never expected to see more
343		than 65536 instructions in a basic block!
344
345		----------------------------------------------------------------------
346		Name: Lal George
347		Date: 2001/10/09 09:41:37 EDT
348		Tag: george-20011008-mlrisc-labels
349		Description:
350
351		Changed the machine description files to support printing of
352		local and global labels in assembly code, based on host assembler
353		conventions.
354
355		----------------------------------------------------------------------
356		Name: Matthias Blume
357		Date: 2001/09/25 15:25:00 EDT
358		Tag: blume-20010925-exninfo
359		Description:
360
361		I provided a non-hook implementation of exnName (at the toplevel) and
362		made the "dummy" implementation of exnMessage (at the toplevel) more
363		useful: if nothing gets "hooked in", then at least you are going to
364		see the exception name and a message indicating why you don't see more.
365
366		[For the time being, programs that need exnMessage and want to use
367		ml-build should either use General.exnMessage (strongly recommended) or
368		refer to structure General at some other point so that CM sees a
369		static dependency.]
370
371		[Similar remarks go for "print" and "use":  If you want to use their
372		functionality in stand-alone programs generated by ml-build, then use
373		TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
374
375		----------------------------------------------------------------------
376		Name: Matthias Blume
377		Date: 2001/09/20 17:28:00 EDT
378		Tag: blume-20010920-slowfp
379		Description:
380
381		Allen says that x86-fast-fp is not safe yet, so I turned it off again...
382
383		----------------------------------------------------------------------
384		Name: Matthias Blume
385		Date: 2001/09/20 17:20:00 EDT
386		Tag: blume-20010920-canonicalpaths
387		Description:
388
389		0. Updated the BOOT file (something that I forgot to do earlier).
390
391		1. Small internal change to CM so that it avoids "/../" in filenames
392		as much as possible (but only where it is safe).
393
394		2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
395		that arguments that contain delimiters are passed through correctly.
396		This change also means that all "special" arguments of the form
397		@SMLxxx... must come first.
398
399		3. Changed install script to put relative anchor names for tool commands
400		into pathconfig.
401
402		----------------------------------------------------------------------
403		Name: Matthias Blume
404		>>>>>>> 1.169
405		Date: 2001/09/18 15:35:00 EDT
406		Tag: blume-20010918-readme11036
407		Description:
408
409		Added README files.
410
411		----------------------------------------------------------------------
412		Name: Matthias Blume
413		Date: 2001/09/18 11:45:00 EDT
414		Tag: Release_110_36 (retag)
415		Description:
416
417		Fixed mistake in config/preloads. Retagged as 110.36.
418
419		----------------------------------------------------------------------
420		Name: Matthias Blume
421		Date: 2001/09/18 09:40:00 EDT
422		Tag: Release_110_36_orig (tag changed)
423		Description:
424
425		New version (110.36).  New bootfiles.
426
427		----------------------------------------------------------------------
428		Name: Matthias Blume
429		Date: 2001/09/14 16:15:00 EDT
430		Tag: blume-20010914-x86fastfp
431		Description:
432
433		John committed some changes that Allen made, in particular a (hopefully)
434		correctly working version of the x86-fp module.
435
436		I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
437		flag to "true".  Everything seems to compile to a fixpoint ok, and
438		"mandelbrot" speeds up by about 15%.
439
440		----------------------------------------------------------------------
441		Name: Matthias Blume
442		Date: 2001/09/13 11:20:00 EDT
443		Tag: blume-20010913-minimal
444		Description:
445
446		1. Stefan Monnier's patch to fix a miscompilation problem that
447		was brought to light by John Reppy's work on Moby.
448
449		2. Implemented a minimal "structure Compiler" that contains just
450		"version" and "architecture".  The minimal version will be
451		available when the full version is not.  This is for backward-
452		compatibility with code that wants to test Compiler.version.
453
454		----------------------------------------------------------------------
455		Name: Matthias Blume
456		Date: 2001/08/28 14:03:00 EDT
457		Tag: blume-20010828-ml-lex
458		Description:
459
460		Fix for bug 1581, received from Neophytos Michael.
461
462		----------------------------------------------------------------------
463		Name: Matthias Blume
464		Date: 2001/08/27 11:20:00 EDT
465		Tag: blume-20010827-readme11035
466		Description:
467
468		Fleshed out the README file for 110.35.
469
470		----------------------------------------------------------------------
471		Name: Matthias Blume
472		Date: 2001/08/24 17:10:00 EDT
473		Tag: Release_110_35
474		Description:
475
476		New version number (110.35).  New bootfiles.
477
478		----------------------------------------------------------------------
479		Name: Lal George
480		Date: 2001/08/24 13:47:18 EDT 2001
481		Tag: george-20010824-MLRISC-graphs
482		Description:
483
484		removed clusters from MLRISC completely and replaced with graphs.
485
486		----------------------------------------------------------------------
487		Name: Matthias Blume
488		Date: 2001/08/23 17:50:00 EDT
489		Tag: blume-20010823-toplevel
490		Description:
491
492		- some reorganization of the code that implements various kinds of
493		environments in the compiler (static, dynamic, symbolic, combined)
494		- re-implemented the EnvRef module so that evalStream works properly
495		(if the stream contains references to "use", "CM.make", etc.)
496		- cleaned up evalloop.sml and interact.sml (but they need more cleaning)
497
498		----------------------------------------------------------------------
499		Name: Matthias Blume
500		Date: 2001/08/20 15:50 EDT
501		Tag: blume20010820-slipup
502		Description:
503
504		I forgot to commit a few files.  Here they are...
505
506		----------------------------------------------------------------------
507		Name: Matthias Blume
508		Date: 2001/08/20 15:35:00 EDT
509		Tag: blume-20010820-debugprof
510		Description:
511
512		!!!! NEW BOOTFILES !!!!
513
514		This is another round of reorganizing the compiler sources.  This
515		time the main goal was to factor out all the "instrumentation"
516		passes (for profiling and backtracing) into their own library.
517		The difficulty was to do it in such a way that it does not depend
518		on elaborate.cm but only on elabdata.cm.
519
520		Therefore there have been further changes to both elaborate.cm and
521		elabdata.cm -- more "generic" things have been moved from the former
522		to the latter.  As a result, I was forced to split the assignment
523		of numbers indicating "primtyc"s into two portions: SML-generic and
524		SML/NJ-specific.  Since it would have been awkward to maintain,
525		I bit the bullet and actually _changed_ the mapping between these
526		numbers and primtycs.  The bottom line of this is that you need
527		a new set of bin- and bootfiles.
528
529		I have built new bootfiles for all architectures, so doing a fresh
530		checkout and config/install.sh should be all you need.
531
532		The newly created library's name is
533
534		$smlnj/viscomp/debugprof.cm
535
536		and its sources live under
537
538		src/compiler/DebugProf
539
540		----------------------------------------------------------------------
541		Name: Matthias Blume
542		Date: 2001/08/15 17:15:00 EDT
543		Tag: blume-20010815-compreorg
544		Description:
545
546		This is a first cut at reorganizing the CM libraries that make up the
547		core of the compiler.  The idea is to separate out pieces that could
548		be used independently by tools, e.g., the parser, the typechecker, etc.
549
550		The current status is a step in this direction, but it is not quite
551		satisfactory yet.  Expect more changes in the future.
552
553		Here is the current (new) organization...
554
555		What used to be $smlnj/viscomp/core.cm is now divided into
556		six CM libraries:
557
558		$smlnj/viscomp/basics.cm
559		/parser.cm
560		/elabdata.cm
561		/elaborate.cm
562		/execute.cm
563		/core.cm
564
565		The CM files for these libraries live under src/system/smlnj/viscomp.
566		All these libraries are proxy libraries that contain precisely
567		one CM library component.  Here are the locations of the components
568		(all within the src/compiler tree):
569
570		Basics/basics.cm
571		Parse/parser.cm
572		ElabData/elabdata.cm
573		Elaborator/elaborate.cm
574		Execution/execute.cm
575		core.cm
576
577		[This organization is the same that has been used already
578		for a while for the architecture-specific parts of the visible
579		compiler and for the old version of core.cm.]
580
581		As you will notice, many source files have been moved from their
582		respective original locations to a new home in one of the above
583		subtrees.
584
585		The division of labor between the new libraries is the following:
586
587		basics.cm:
588		- Simple, basic definitions that pertain to many (or all) of
589		the other libraries.
590		parser.cm:
591		- The SML parser, producing output of type Ast.dec.
592		- The type family for Ast is also defined and exported here.
593		elabdata.cm:
594		- The datatypes that describe input and output of the elaborator.
595		This includes types, absyn, and static environments.
596		elaborator.cm:
597		- The SML/NJ type checker and elaborator.
598		This maps an Ast.dec (with a given static environment) to
599		an Absyn.dec (with a new static environment).
600		- This libraries implements certain modules that used to be
601		structures as functors (to remove dependencies on FLINT).
602		execute.cm:
603		- Everything having to do with executing binary code objects.
604		- Dynamic environments.
605		core.cm:
606		- SML/NJ-specific instantiations of the elaborator and MLRISC.
607		- Top-level modules.
608		- FLINT (this should eventually become its own library)
609
610		Notes:
611
612		I am not 100% happy with the way I separated the elaborator (and its
613		data structures) from FLINT.  Two instances of the same problem:
614
615		1. Data structures contain certain fields that carry FLINT-specific
616		information.  I hacked around this using exn and the property list
617		module from smlnj-lib.  But the fact that there are middle-end
618		specific fields around at all is a bit annoying.
619
620		2. The elaborator calculates certain FLINT-related information.  I tried
621		to make this as abstract as I could using functorization, but, again,
622		the fact that the elaborator has to perform calculations on behalf
623		of the middle-end at all is not nice.
624
625		3. Having to used exn and property lists is unfortunate because it
626		weakens type checking.  The other alternative (parameterizing
627		nearly *everything*) is not appealing, though.
628
629		I removed the "rebinding =" warning hack because due to the new organization
630		it was awkward to maintain it.  As a result, the compiler now issues some of
631		these warnings when compiling init.cmi during bootstrap compilation. On
632		the plus side, you also get a warning when you do, for example:
633		val op = = Int32.+
634		which was not the case up to now.
635
636		I placed "assign" and "deref" into the _Core structure so that the
637		code that deals with the "lazy" keyword can find them there.  This
638		removes the need for having access to the primitive environment
639		during elaboration.
640
641		----------------------------------------------------------------------
642		Name: Matthias Blume
643		Date: 2001/08/13
644		Tag: blume-20010813-closures
645		Description:
646
647		This fix was sent to us by Zhong Shao.  It is supposed to improve the
648		performance of certain loops by avoiding needless closure allocation.
649
650		----------------------------------------------------------------------
651		Name: Lal George
652		Date: 2001/07/31 10:03:23 EDT 2001
653		Tag: george-20010731-x86-fmalloc
654		Description: Fixed bug in x86 calls
655
656		There was a bug where call instructions would mysteriously
657		vanish. The call instruction had to be one that returned
658		a floating point value.
659
660		----------------------------------------------------------------------
661		Name: Lal George
662		Date: 2001/07/19 16:36:29 EDT 2001
663		Tag: george-20010719-simple-cells
664		Description:
665
666		I have dramatically simplified the interface for CELLS in MLRISC.
667
668		In summary, the cells interface is broken up into three parts:
669
670		1. CellsBasis : CELLS_BASIS
671
672		CellsBasis is a top level structure and common for all
673		architectures.  it contains the definitions of basic datatypes
674		and utility  functions over these types.
675
676		2. functor Cells() : CELLS
677
678		Cells generates an interface for CELLS that incorporates the
679		specific resources on the target architecture, such as the
680		presence of special register classes, their number and size,
681		and various useful substructures.
682
683		3. <ARCH>CELLS
684
685		e.g. SparcCells: SPARCCELLS
686
687		<ARCH>CELLS usually contains additional bindings for special
688		registers  on the architecture, such as:
689
690		val r0 : cell           (* register zero *)
691		val y : cell            (* Y register *)
692		val psr : cell          (* processor status register *)
693		...
694
695		The structure returned by applying the Cells functor is opened
696		in this interface.
697
698		The main implication of all this is that the datatypes for cells is
699		split between CellsBasis and CELLS -- a fairly simple change for user
700		code.
701
702		In the old scheme the CELLS interface had a definitional binding of
703		the form:
704
705		signature CELLS = sig
706
707		structure CellsBasis = CellsBasis
708
709		...
710
711		end
712
713		With all the sharing constraints that goes on in MLRISC, this old
714		design  quickly leads to errors such as:
715
716		"structure definition spec inside of sharing ... "
717
718
719		and appears to require an unacceptable amount of sharing and where
720		constraint hackery.
721
722		I think this error message (the interaction of definitional specs and
723		sharing) requires more explanation on our web page.
724
725		----------------------------------------------------------------------
726		Name: Matthias Blume
727		Date: 2001/07/19 15:00:00 EDT
728		Tag: blume-20010719-libreorg
729		Description:
730
731		This update puts together a fairly extensive but straightforward change
732		to the way the libraries that implement the interactive system are
733		organized:
734
735		The biggest change is the elimination of structure Compiler.  As a
736		replacement for this structure, there is now a CM library
737		(known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
738		that exports all the substructures of the original structure Compiler
739		directly.  So instead of saying Compiler.Foo.bar one now simply
740		says Foo.bar.  (The CM libraries actually export a collection of
741		structures that is richer than the collection of substructures of
742		structure Compiler.)
743
744		To make the transition smooth, there is a separate library called
745		$smlnj/compiler/compiler.cm which puts together and exports the
746		original structure Compiler (or at least something very close to it).
747
748		There are five members of the original structure Compiler
749		that are not exported directly but which instead became members
750		of a new structure Backend (described by signature BACKEND).  These are:
751		structure Profile (: PROFILE), structure Compile (: COMPILE), structure
752		Interact (: INTERACT), structure Machine (: MACHINE), and val
753		architecture (: string).
754
755		Structure Compiler.Version has become structure CompilerVersion.
756
757		Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
758		by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
759		or x86, respectively.
760		Each of these exports the same frontend structures that
761		$smlnj/compiler.cm exports.  But they do not have a structure Backend
762		and instead export some structure <Arch>Backend where <Arch> is Alpha32,
763		Hppa, PPC, Sparc, or X86, respectively.
764
765		Library $smlnj/compiler/all.cm exports the union of the exports of
766		$smlnj/compiler/<arch>.cm
767
768		There are no structures <Arch>Compiler anymore, use
769		$smlnj/compiler/<arch>.cm instead.
770
771		Library host-compiler-0.cm is gone.  Instead, the internal library
772		that instantiates CM is now called cm0.cm.  Selection of the host
773		compiler (backend) is no longer done here but. (Responsibility for it
774		now lies with $smlnj/compiler/current.cm.  This seems to be more
775		logical.)
776
777		Many individual files have been moved or renamed.  Some files have
778		been split into multiple files, and some "dead" files have been deleted.
779
780		Aside from these changes to library organization, there are also changes
781		to the way the code itself is organized:
782
783		Structure Binfile has been re-implemented in such a way that it no
784		longer needs any knowledge of the compiler.  It exclusively deals
785		with the details of binfile layout.  It no longer invokes the
786		compiler (for the purpose of creating new prospective binfile
787		content), and it no longer has any knowledge of how to interpret
788		pickles.
789
790		Structure Compile (: COMPILE) has been stripped down to the bare
791		essentials of compilation.  It no longer deals with linking/execution.
792		The interface has been cleaned up considerably.
793
794		Utility routines for dealing with linking and execution have been
795		moved into their own substructures.
796
797		(The ultimate goal of these changes is to provide a light-weight
798		binfile loader/linker (at least for, e.g., stable libraries) that
799		does not require CM or the compiler to be present.)
800
801		CM documentation has been updated to reflect the changes to library
802		organization.
803
804		----------------------------------------------------------------------
805		Name: Matthias Blume
806		Date: 2001/07/10 17:30:00 EDT
807		Tag: Release_110_34
808		Description:
809
810		Minor tweak to 110.34 (re-tagged):
811
812		- README.html file added to CVS repository
813		- runtime compiles properly under FreeBSD 3.X and 4.X
814
815		----------------------------------------------------------------------
816		Name: Matthias Blume
817		Date: 2001/07/10 17:30:00 EDT
818		Tag: Release_110_34
819		Description:
820
821		New version number (110.34). New bootfiles.
822
823		----------------------------------------------------------------------
824		Name: Matthias Blume
825		Date: 2001/07/09 16:00:00 EDT
826		Tag: blume-20010709-more-varargs
827		Description:
828
829		I changed the handling of varargs in ml-nlffigen again:
830		The ellipsis ... will now simply be ignored (with an accompanying warning).
831
832		The immediate effect is that you can actually call a varargs function
833		from ML -- but you can't actually supply any arguments beyond the ones
834		specified explicitly.  (For example, you can call printf with its format
835		string, but you cannot pass additional arguments.)
836
837		This behavior is only marginally more useful than the one before, but
838		it has the advantage that a function or, more importantly, a function
839		type never gets dropped on the floor, thus avoiding follow-up problems with
840		other types that refer to the offending one.
841
842		----------------------------------------------------------------------
843		Name: Matthias Blume
844		Date: 2001/07/09 11:25:00 EDT
845		Tag: blume-20010709-varargs
846		Description:
847
848		1. ckit-lib.cm now exports structure Error
849		2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
850		with a warning accompanied by a source location.  Moreover, it
851		merely skips the offending function or type and proceeds with the
852		rest of its work.u  As a result, one can safely feed C code containing
853		"..." to ml-nlffigen.
854		3. There are some internal improvements to CM, providing slightly
855		more general string substitutions in the tools subsystem.
856
857		----------------------------------------------------------------------
858		Name: Matthias Blume
859		Date: 2001/06/27 15:10:00 EDT
860		Tag: blume-20010627-concur
861		Description:
862
863		Fixed a small bug in CM's handling of parallel compilation.
864		(You could observe the bug by Control-C-interrupting an ordinary
865		CMB.make or CM.stabilize and then attaching some compile servers.
866		The result was that all of a sudden the previously interrupted
867		compilation would continue on its own.  This was because of
868		an over-optimization: CM did not bother to clean out certain queues
869		when no servers were attached "anyway", resulting in the contents
870		of these queues to grab control when new servers did get attached.)
871
872		There is also another minor update to the CM manual.
873
874		----------------------------------------------------------------------
875		Name: Matthias Blume
876		Date: 2001/06/26 16:15:00 EDT
877		Tag: blume-20010626-cmdoc
878		Description:
879
880		Minor typo fixed in CM manual (syntax diagram for libraries).
881
882		----------------------------------------------------------------------
883		Name: Matthias Blume
884		Date: 2001/06/25 22:55:00 EDT
885		Tag: blume-20010625-x86pc
886		Description:
887
888		Fixed a nasty bug in the X86 assembly code that caused signal
889		handlers to fail (crash) randomly.
890
891		----------------------------------------------------------------------
892		Name: Matthias Blume
893		Date: 2001/06/25 12:05:00 EDT
894		Tag: blume-20010625-nlffigen
895		Description:
896
897		This update fixes a number of minor bugs in ml-nlffigen as reported by
898		Nick Carter <nbc@andrew.cmu.edu>.
899
900		1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
901		2. Default names for generated files are now derived from the name of
902		the C file *without its directory*.  In particular, this causes generated
903		files to be placed locally even if the C file is in some system directory.
904		3. Default names for generated signatures and structures are also derived
905		from the C file name without its directory.  This avoids silly things
906		like "structure GL/GL".
907		(Other silly names are still possible because ml-nlffigen does not do
908		a thorough check of whether generated names are legal ML identifiers.
909		When in doubt, use command line arguments to force particular names.)
910
911		----------------------------------------------------------------------
912		Name: Matthias Blume
913		Date: 2001/06/21 12:25:00 EDT
914		Tag: blume-20010621-eXene
915		Description:
916
917		eXene now compiles and (sort of) works again.
918
919		The library name (for version > 110.33) is $/eXene.cm.
920
921		I also added an new example in src/eXene/examples/nbody.  See the
922		README file there for details.
923
924		----------------------------------------------------------------------
925		Name: Matthias Blume
926		Date: 2001/06/20 16:40:00 EDT
927		Tag: blume-20010620-cml
928		Description:
929
930		CML now compiles and works again.
931
932		Libraries (for version > 110.33):
933
934		$cml/cml.cm            Main CML library.
935		$cml/basis.cm          CML's version of $/basis.cm.
936		$cml/cml-internal.cm   Internal helper library.
937		$cml/core-cml.cm       Internal helper library.
938		$cml-lib/trace-cml.cm  Tracing facility.
939		$cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
940
941		The installer (config/install.sh) has been taught how to properly
942		install this stuff.
943
944		----------------------------------------------------------------------
945		Name: Matthias Blume
946		Date: 2001/06/19 17:55:00 EDT
947		Tag: blume-20010619-instantiate
948		Description:
949
950		This un-breaks the fix for bug 1432.
951		(The bug was originally fixed in 110.9 but I broke it again some
952		time after that.)
953
954		----------------------------------------------------------------------
955		Name: Matthias Blume
956		Date: 2001/06/19 17:25:00 EDT
957		Tag: blume-20010619-signals
958		Description:
959
960		This should (hopefully) fix the long-standing signal handling bug.
961		(The runtime system was constructing a continuation record with an
962		incorrect descriptor which would cause the GC to drop data on the floor...)
963
964		----------------------------------------------------------------------
965		Name: Matthias Blume
966		Date: 2001/06/15 15:05:00 EDT
967		Tag: blume-20010615-moresparc
968		Description:
969
970		Here is a short late-hour update related to Sparc c-calls:
971
972		-- made handling of double-word arguments a bit smarter
973
974		-- instruction selection phase tries to collapse certain clumsily
975		constructed ML-Trees; typical example:
976
977		ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
978
979		This currently has no further impact on SML/NJ since mlriscGen does
980		not seem to generate such patterns in the first place, and c-calls
981		(which did generate them in the beginning) has meanwhile been fixed
982		so as to avoid them as well.
983
984		----------------------------------------------------------------------
985		Name: Matthias Blume
986		Date: 2001/06/15 15:05:00 EDT
987		Tag: blume-20010615-sparc
988		Description:
989
990		The purpose of this update is to provide an implementation of NLFFI
991		on Sparc machines.
992
993		Here are the changes in detail:
994
995		* src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
996		the Sparc implementation of the c-calls API.
997		* The Sparc backend of SML/NJ has been modified to uniformely use %fp
998		for accessing the ML frame.  Thus, we have a real frame pointer and
999		can freely modify %sp without need for an omit-frame-ptr phase.
1000		The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
1001		this case.
1002		* ml-nlffigen has been taught to produce code for different architectures
1003		and calling conventions.
1004		* In a way similar to what was done in the x86 case, the Sparc
1005		backend uses its own specific extension to mltree.  (For example,
1006		it needs to be able to generate UNIMP instructions which are part
1007		of the calling convention.)
1008		* ml-nlffi-lib was reorganized to make it more modular (in particular,
1009		to make it easier to plug in new machine- and os-dependent parts).
1010
1011		There are some other fairly unrelated bug fixes and cleanups as well:
1012
1013		* I further hacked the .cm files for MLRISC tools (like MDLGen) so
1014		that they properly share their libraries with existing SML/NJ libraries.
1015		* I fixed a minor cosmetic bug in CM, supressing certain spurious
1016		follow-up error messages.
1017		* Updates to CM/CMB documentation.
1018
1019		TODO items:
1020
1021		* MLRISC should use a different register as its asmTemp on the Sparc.
1022		(The current %o2 is a really bad choice because it is part of the
1023		calling conventions, so things might interfere in unexpected ways.)
1024
1025		----------------------------------------------------------------------
1026		Name: Matthias Blume
1027		Date: 2001/06/07
1028		Tag: blume-20010607-calls
1029		Description:
1030
1031		A number of internal changes related to C calls and calling conventions:
1032
1033		1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
1034		number of bytes popped implicitly (by the callee).  In most cases
1035		this field is 0 but on x86/win32 it is some non-zero value.  This
1036		is information provided for the benefit of the "omit-frameptr" pass.
1037		2. The CALL instruction on the x86 carries a similar "pops" field.
1038		The instruction selection phase copies its value from the ML-Tree
1039		CALL statement.
1040		3. On all other architectures, the instruction selection phase checks
1041		whether "pops=0" and complains if not.
1042		4. The c-calls implementation for x86 now accepts two calling conventions:
1043		"ccall" and "stdcall".  When "ccall" is selected, the caller cleans
1044		up after the call and pops is set to 0.  For "stdcall", the caller
1045		does nothing, leaving the cleanup to the callee; pops is set to
1046		the number of bytes that were pushed onto the stack.
1047		5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
1048		distinguish between "ccall" and "stdcall".
1049		6. The UNIMP instruction has been added to the supported Sparc instruction
1050		set. (This is needed for implementing the official C calling convention
1051		on this architecture.)
1052		7. I fixed some of the .cm files under src/MLRISC/Tools to make them
1053		work with the latest CM.
1054
1055		----------------------------------------------------------------------
1056		Name: Matthias Blume
1057		Date: 2001/06/05 15:10:00 EDT
1058		Tag: blume-20010605-cm-index
1059		Description:
1060
1061		0. The "lambdasplit" parameter for class "sml" in CM has been documented.
1062
1063		1. CM can now generate "index files".  These are human-readable files
1064		that list on a per-.cm-file basis each toplevel symbol defined or
1065		imported.  The location of the index file for
1066		<p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
1067		To enable index-file generation, set CM.Control.generate_index to true
1068		or export an environment-symbol: export CM_GENERATE_INDEX=true.
1069
1070		The CM manual has been updated accordingly.
1071
1072		2. I made some slight modifications to the c-calls API in MLRISC.
1073
1074		a) There is now a callback to support saving/restoring of
1075		dedicated but caller-save registers around the actual call
1076		instruction.
1077		b) One can optionally specify a comment-annotation for the
1078		call instruction.
1079
1080		3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
1081		(For example, the comment annotation shows the C prototype of
1082		the function being called.)
1083
1084		----------------------------------------------------------------------
1085		Name: Matthias Blume
1086		Date: 2001/06/01 13:30:00 EDT
1087		Tag: blume-20010601-nlffi-cleanup
1088		Description:
1089
1090		This is mostly a cleanup of MLFFI stuff:
1091
1092		- some signature files have been put into a more exposed place
1093		- the ugly 'f type parameter is gone (simplifies types tremendously!)
1094		- ml-nlffigen changed accordingly
1095		- tutorial updated
1096
1097		Other changes:
1098
1099		- author's affiliation in CM manual(s) updated
1100		- some more recognized keywords added to Allen's sml.sty
1101
1102		----------------------------------------------------------------------
1103		Name: Matthias Blume
1104		Date: 2001/05/25 15:30:00 EDT
1105		Tag: blume-20010525-iptr
1106		Description:
1107
1108		- put the official 110.33-README (as it appears on the ftp server) under
1109		CVS
1110		- fixed a small bug related to incomplete pointer types in
1111		ml-nlffigen
1112		- small cosmetic change to the ml-nlffi-lib's "arr" type constructor
1113		(it does not need the 'f type parameter)
1114
1115		----------------------------------------------------------------------
1116		Name: Matthias Blume
1117		Date: 2001/05/23 14:30:00 EDT
1118		Tag: Release_110_33
1119		Description:
1120
1121		New version number (110.33).  New bootfiles.
1122
1123		----------------------------------------------------------------------
1124		Name: Matthias Blume
1125		Date: 2001/05/22 18:06:00 EDT
1126		Tag: blume-20010522-targets
1127		Description:
1128
1129		Made install.sh use file config/targets.customized if it exists, falling
1130		back to config/targets if it doesn't.  This way one can have a customized
1131		version of the targets file without touching the "real thing", thus
1132		eliminating the constant fear of accidentally checking something bogus
1133		back into the CVS repository...  (File config/targets.customized must
1134		not be added to the repository!)
1135
1136		----------------------------------------------------------------------
1137		Name: Matthias Blume
1138		Date: 2001/05/22 16:30:00 EDT
1139		Tag: blume-20010522-minitut
1140		Description:
1141
1142		1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
1143		struct returns.
1144		2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
1145		incomplete, preliminary documentation for NLFFI.
1146
1147		----------------------------------------------------------------------
1148		Name: Matthias Blume
1149		Date: 2001/05/14 11:30:00 EDT
1150		Tag: blume-20010514-script
1151		Description:
1152
1153		Some bugs in install script fixed.
1154
1155		In addition to that I also made a slight change to the NLFFI API:
1156		Functors generated by ml-nlffigen now take the dynamic library as a
1157		straight functor argument, not as a suspended one.  (The original
1158		functor code used to force the suspension right away anyway, so there
1159		was nothing gained by this complication of the interface.)
1160
1161		----------------------------------------------------------------------
1162		Name: Matthias Blume
1163		Date: 2001/05/11 14:35:00 EDT
1164		Tag: blume-20010511-ml-nlffi
1165		Description:
1166
1167		I finally took the plunge and added my new FFI code to the main
1168		repository.  For x86-linux it is now ready for prime-time.
1169
1170		There are two new subdirectories of "src":
1171
1172		- ml-nlffi-lib:
1173		The utility library for programs using the FFI interface.
1174		Here is the implementation of $/c.cm and its associated low-level
1175		partners $/c-int.cm and $/memory.cm.
1176		- ml-nlffigen:
1177		A stand-alone program for generating ML glue code from C source
1178		code.
1179
1180		Building ml-nlffigen requires $/ckit-lib.cm.
1181
1182		The config/install.sh script has been updates to do the Right Thing
1183		(hopefully).
1184
1185		Notice that the source tree for the C-Kit will not be put under "src"
1186		but directly under the installation root directory.  (This is the
1187		structure that currently exists on the CVS server when you check out
1188		module "sml".)  Fortunately, config/install.sh knows about this oddity.
1189
1190		Bugs: No documentation yet.
1191
1192		----------------------------------------------------------------------
1193		Name: Matthias Blume
1194		Date: 2001/05/09 16:35:00 EDT
1195		Tag: blume-20010509-cpscontract
1196		Description:
1197
1198		Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
1199		wrapper/unwrapper elimination did not decrement usage counts and some
1200		dead variables got overlooked by the dead-up logic.)
1201
1202		----------------------------------------------------------------------
1203		Name: Lal George
1204		Date: 2001/05/08  17:26:09 EDT
1205		Tag: george-20010508-omit-frameptr
1206		Description:
1207
1208		Changes to implement the omit-frame-pointer optimization to support
1209		raw C calls. For now, there is only support on the Intel x86, but
1210		other architectures will follow as more experience is gained with this.
1211
1212
1213		----------------------------------------------------------------------
1214		Name: Matthias Blume
1215		Date: 2001/05/07 14:40:00 EDT
1216		Tag: blume-20010507-proxies
1217		Description:
1218
1219		I made into "proxy libraries" all libraries that qualify for such a
1220		change.  (A qualifying library is a library that has another library or
1221		groups as its sole member and repeats that member's export list
1222		verbatim.  A proxy library avoids this repetition by omitting its export
1223		list, effectively inheriting the list that its (only) member exports.
1224		See the CM manual for more explanation.)
1225		The main effect is that explicit export lists for these libraries
1226		do not have to be kepts in sync, making maintenance a bit easier.
1227
1228		I also added copyright notices to many .cm-files.
1229
1230		Last but not least, I made a new set of bootfiles.
1231
1232		----------------------------------------------------------------------
1233		Name: Matthias Blume
1234		Date: 2001/05/04 17:00:00 EDT
1235		Tag: blume-20010504-cm-lsplit
1236		Description:
1237
1238		0. John merged pending changes to $/smlnj-lib.cm
1239
1240		1. Allen's previous change accidentally backed out of one of Lal's
1241		earlier changes.  I undid this mistake (re-introducing Lal's change).
1242
1243		2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
1244		within the compiler where applicable.  There is some code simplification
1245		because of that.
1246
1247		3. The "split" phase (in FLINT) is now part of the default list of phases.
1248		Compiler.Control.LambdaSplitting.* can be used to globally control the
1249		lambda-splitting (cross-module-inlining) engine.  In addition to that,
1250		it can now also be controlled on a per-source basis: CM has been taught
1251		a new tool parameter applicable to ML source files.
1252
1253		- To turn lambda-splitting off completely:
1254		local open Compiler.Control.LambdaSplitting in
1255		val _ = set Off
1256		end
1257		- To make "no lambda-splitting" the global default (but allow per-source
1258		overriding); this is the initial setting:
1259		local open Compiler.Control.LambdaSplitting in
1260		val _ = set (Default NONE)
1261		end
1262		- To make "lambda-splitting with aggressiveness a" the global default
1263		(and allow per-source overriding):
1264		local open Compiler.Control.LambdaSplitting in
1265		val _ = set (Default (SOME a))
1266		end
1267
1268		- To turn lambda-splitting off for a given ML souce file (say: a.sml)
1269		write (in the respective .cm-file):
1270		a.sml (lambdasplitting:off)
1271		- To turn lambda-splitting for a.sml on with minimal aggressiveness:
1272		a.sml (lambdasplitting:on)
1273		- To turn lambda-splitting for a.sml on with aggressiveness <a> (where
1274		<a> is a decimal non-negative integer):
1275		a.sml (lambdasplitting:<a>)
1276		- To turn lambda-splitting for a.sml on with maximal aggressiveness:
1277		a.sml (lambdasplitting:infinity)
1278		- To use the global default for a.sml:
1279		a.sml (lambdasplitting:default)
1280		or simply
1281		a.sml
1282
1283		----------------------------------------------------------------------
1284		Name: Allen Leung
1285		Date: 2001/05/04 01:57:00 EDT
1286		Tag: leunga-20010504-sync
1287		Description:
1288
1289		MLRISC features.
1290
1291		1. Fix to CMPXCHG instructions.
1292		2. Changed RA interface to allow annotations in callbacks.
1293		3. Added a new method to the stream interface to allow annotations updates.
1294
1295		----------------------------------------------------------------------
1296		Name: Matthias Blume
1297		Date: 2001/05/01 11:45:00 EDT
1298		Tag: blume-20010501-pcedittmp
1299		Description:
1300
1301		Changed install.sh to use the current working directory instead of
1302		/usr/tmp for a temporary file (pcedittmp).  The previous choice
1303		of /usr/tmp caused trouble with MacOS X because of file premission
1304		problems.
1305
1306		----------------------------------------------------------------------
1307		Name: Matthias Blume
1308		Date: 2001/04/20 11:10:00 EDT
1309		Tag: blume-20010420-inMLflag
1310		Description:
1311
1312		- added vp_limitPtrMask to vproc-state.h
1313		(for use by the raw-C-calls mechanism to implement proper interrupt
1314		handling)
1315		- made the ML compiler aware of various data-structure offsets so it
1316		can generate code for accessing the vp_inML flag and vp_limitPtrMask
1317		- tweaked mlriscGen.sml to have it emit interrupt-handling code for
1318		raw C-calls
1319
1320		----------------------------------------------------------------------
1321		Name: Lal George
1322		Date: 2001/04/20 09:15:28 EDT
1323		Tag: george-20010420-macosX
1324		Description:
1325
1326		- Changes to port to Mac OS X; Darwin.
1327
1328		- In the process I found that sqrt was broken on the PPC, because the
1329		fsqrt instruction is not implemented.
1330
1331		----------------------------------------------------------------------
1332		Name: Matthias Blume
1333		Date: 2001/04/18 12:45:00 EDT
1334		Tag: blume-20010418-ccalls
1335		Description:
1336
1337		- fixed two off-by-4 errors in the x86-specific c-calls implementation
1338		(this bug prevented structure arguments containing pointers from being
1339		passed correctly)
1340		- changed the raw-C-call code in mlriscGen.sml in such a way that
1341		structure arguments are represented as a pointer to the beginning
1342		of the structure (instead of having a series of synthesized arguments,
1343		one for each structure member)
1344
1345		- made makeml script's verbosity level configurable via environment
1346		variable (MAKEML_VERBOSITY)
1347
1348		- eliminated placeholder implementations for f32l, w16s, i16s, and f32s
1349		in rawmem-x86.sml; we are now using the real thing
1350
1351		----------------------------------------------------------------------
1352		Name: Matthias Blume
1353		Date: 2001/03/22 16:25:00 EST
1354		Tag: blume-20010322-bootfiles
1355		Description:
1356
1357		Created a new set of bootfiles (for your automatic installation convenience).
1358
1359		----------------------------------------------------------------------
1360		Name: Matthias Blume
1361		Date: 2001/03/22 15:10:00 EST
1362		Tag: blume-20010322-rawmem-parcm
1363		Description:
1364
1365		1. All "raw memory access" primitives for the new FFI are implemented now
1366		(at least on the x86).
1367		2. Some further cleanup of CM's parallel make mechanism.
1368
1369		----------------------------------------------------------------------
1370		Name: Matthias Blume
1371		Date: 2001/03/19 17:53:00 EST
1372		Tag: blume-20010319-parallel
1373		Description:
1374
1375		Parallel make (using compile servers) now works again.
1376
1377		To this end, CM.stabilize and CMB.make have been modified to work in
1378		two passes when compile servers are attached:
1379		1. Compile everything, do not perform stabilization; this pass
1380		uses compile servers
1381		2. Stabilize everything; this pass does not use compile servers
1382		If there are no compile servers, the two passes are combined into one
1383		(as before).  Splitting the passes increases the inherent parallelism
1384		in the dependency graph because the entire graph including all
1385		libraries is available at the same time.  This, in turn, improves
1386		server utilization.  The downside is that the master process will
1387		have to do some extra work after compilation is done (because for
1388		technical reasons it must re-read all the binfiles during stabilization).
1389
1390		----------------------------------------------------------------------
1391		Name: Matthias Blume
1392		Date: 2001/03/16 12:22:00 EST
1393		Tag: blume-20010316-bootfiles
1394		Description:
1395
1396		Created a new set of bootfiles (for your automatic installation convenience).
1397
1398		----------------------------------------------------------------------
1399		Name: Matthias Blume
1400		Date: 2001/03/16 11:00:00 EST
1401		Tag: blume-20010316-MLTREE-fixup
1402		Description:
1403
1404		This is a minor fixup for an (untagged) earlier commit by Allen.
1405		(A file was missing).
1406
1407		----------------------------------------------------------------------
1408		Name: Allen Leung
1409		Date: Mon Mar  5 18:54:57 EST 2001
1410		Tag: leunga-20010305-cut-support
1411
1412		1. New support for alternative control-flow in MLTREE.
1413		Currently we support
1414
1415		FLOW_TO(CALL ...., [k1,...,kn])
1416
1417		This is needed for 'cuts to' in C-- and try/handle-like constructs
1418		in Moby
1419
1420		New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
1421
1422		2. Register Allocator
1423
1424		Changes in interface [from Fermin, John]
1425
1426		3. Alpha 8-bit SLL support [Fermin]
1427
1428		4. All architectures
1429
1430		A new module (ClusterExpandCopies) for expanding parallel copies.
1431
1432		----------------------------------------------------------------------
1433		Name: Allen Leung
1434		Date: 2001/02/27 23:07:00 EST
1435		Tag: leunga-20010227-minor-stuff
1436
1437		1. Alpha bug fix for CMOVNE
1438		2. Handle mltree COND(..,FCMP ...,...)
1439		3. Bug fix in simplifier
1440
1441		----------------------------------------------------------------------
1442		Name: Matthias Blume
1443		Date: 2001/01/30 17:50:00 EST
1444		Tag: blume-20010130-sync
1445		Description:
1446
1447		This is just a minor update to sync my devel branch with the main brach.
1448		The only visible change is the addition of some README files.
1449
1450		----------------------------------------------------------------------
1451		Name: Matthias Blume
1452		Date: 2001/01/12 23:30:00 JST
1453		Tag: blume-20010112-bootfiles
1454		Description:
1455
1456		Made a new set of bootfiles that goes with the current state of the
1457		repository.
1458
1459		----------------------------------------------------------------------
1460		Name: Matthias Blume
1461		Date: 2001/01/12 21:20:00 JST
1462		Tag: blume-20010112-sync
1463		Description:
1464
1465		I am just flushing out some minor changes that had accumulated in
1466		my private branch in order to sync with the main tree.  (This is
1467		mainly because I had CVS trouble when trying to merge _into_ my
1468		private branch.)
1469
1470		Most people should be completely unaffected by this.
1471
1472		----------------------------------------------------------------------
1473		Name: Allen Leung
1474		Date: Thu Jan 11 21:03:00 EST 2001
1475		Tag: leunga-20010111-labexp=mltree
1476		Description:
1477
1478		1.  Removed the type LabelExp and replace it by MLTree.
1479		2.  Rewritten mltree-simplify with the pattern matcher tool.
1480		3.  There were some bugs in alpha code generator which would break
1481		64-bit code generation.
1482		4.  Redo the tools to generate code with the
1483		5.  The CM files in MLRISC (and in src/system/smlnj/MLRISC)
1484		are now generated by perl scripts.
1485
1486		----------------------------------------------------------------------
1487		Name: Matthias Blume
1488		Date: 2001/01/10 21:55:00 JST
1489		Tag: blume-20010110-rcc
1490		Description:
1491
1492		The RCC stuff now seems to work (but only on the x86).
1493		This required hacking of the c-calls interface (and -implementation) in
1494		MLRISC.
1495
1496		Normal compiler users should be unaffected.
1497
1498		----------------------------------------------------------------------
1499		Name: Matthias Blume
1500		Date: 2001/01/09 01:20:00 JST
1501		Tag: blume-20010109-rcc
1502		Description:
1503
1504		This is a fairly big patch, flushing out a large number of pending
1505		changes that I made to my development copy over the last couple of days.
1506
1507		Of practical relevance at this moment is a workaround for a pickling
1508		bug that Allen ran into the other day.  The cause of the bug itself is
1509		still unknown and it might be hard to fix it properly, but the
1510		workaround has some merits of its own (namely somewhat reducing pickling
1511		overhead for certain libraries).  Therefore, I think this solution should
1512		be satisfactory at this time.
1513
1514		The rest of the changes (i.e., the vast majority) has to do with my
1515		ongoing efforts of providing direct support for C function calls from
1516		ML.  At the moment there is a new primop "RAW_CCALL", typing magic
1517		in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new
1518		case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to
1519		translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to
1520		handle RCC.
1521
1522		The last part (the changes to mlriscGen.sml) are still known to be
1523		wrong on the x86 and not implemented on all other architectures.  But
1524		the infrastructure is in place. I had to change a few functor
1525		signatures in the backend to be able to route the CCalls interface
1526		from MLRISC there, and I had to specialize the mltree type (on the
1527		x86) to include the necessary extensions. (The extensions themselves
1528		were already there and redy to go in MLRISC/x86).
1529
1530		Everything should be very happy as soon as someone helps me with
1531		mlriscGen.sml...
1532
1533		In any case, nothing of this should matter to anyone as long as the
1534		new primop is not being used (which is going to be the case unless you
1535		find it where I hid it :). The rest of the compiler is completely
1536		unaffected.
1537
1538		----------------------------------------------------------------------
1539		Name: Matthias Blume
1540		Date: 2001/01/05 00:30:00 JST
1541		Tag: blume-20010105-primops
1542		Description:
1543
1544		Added some experimental support for work that I am doing right now.
1545		These changes mostly concern added primops, but there is also a new
1546		experimental C library in the runtime system (but currently not enabled
1547		anywhere except on Linux/X86).
1548
1549		In the course of adding primops (and playing with them), I discovered that
1550		Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly
1551		broken.  (Zhong was very right he labeled this stuff as "major gross hack".)
1552		To recover, I made type information in INL_PRIM mandatory and changed
1553		prim.sml as well as built-in.sml accordingly.  The InLine structure now
1554		has complete, correct type information (i.e., no bottom types).
1555
1556		Since all these changes mean that we need new binfiles, I also bumped the
1557		version number to 110.32.1.
1558
1559		----------------------------------------------------------------------
1560		Name: Matthias Blume
1561		Date: 2000/12/30 22:10:00 JST
1562		Tag: blume-20001230-various
1563		Description:
1564
1565		Added proxy libraries for MLRISC and let MLRISC libraries refer
1566		to each other using path anchors.  (See CM manual for explanation.)
1567
1568		Updated CM documentation.
1569
1570		Fixed some bugs in CM.
1571
1572		Implemented "proxy" libraries (= syntactic sugar for CM).
1573
1574		Added "-quiet" option to makeml and changed runtime system accordingly.
1575
1576		Added cleanup handler for exportML to reset timers and compiler stats.
1577
1578		----------------------------------------------------------------------
1579		Name: Lal George
1580		Date: 2000/12/22 22:22:58 EST 2000
1581		Tag: Release_110_32
1582		Description:
1583
1584		Infinite precision used throughout MLRISC.
1585		see MLRISC/mltree/machine-int.sig
1586
1587		----------------------------------------------------------------------
1588		Name: Matthias Blume
1589		Date: 2000/12/22 23:16:00 JST
1590		Tag: blume-20001222-warn
1591		Description:
1592
1593		Corrected wording and formatting of some CM warning message which I
1594		broke in my previous patch.
1595
1596		----------------------------------------------------------------------
1597		Name: Matthias Blume
1598		Date: 2000/12/22 21:20:00 JST
1599		Tag: blume-20001222-anchorenv
1600		Description:
1601
1602		Fixed CM's handling of anchor environments in connection with CMB.make.
1603
1604		----------------------------------------------------------------------
1605		Name: Matthias Blume
1606		Date: 2000/12/22 13:15:00 JST
1607		Tag: blume-20001222-cleanup
1608		Description:
1609
1610		Removed src/cm/ffi which does not (and did not) belong here.
1611
1612		----------------------------------------------------------------------
1613		Name: Matthias Blume
1614		Date: 2000/12/21 23:55:00 JST
1615		Tag: blume-20001221-exn
1616		Description:
1617
1618		Probably most important: CM no longer silently swallows all exceptions
1619		in the compiler.
1620		Plus: some other minor CM changes.  For example, CM now reports some
1621		sizes for generated binfiles (code, data, envpickle, lambdapickle).
1622
1623		----------------------------------------------------------------------
1624		Name: Matthias Blume
1625		Date: 2000/12/15 00:01:05 JST
1626		Tag: blume-20001215-dirtool
1627		Description:
1628
1629		- "dir" tool added.
1630		- improvements and cleanup to Tools structure
1631		- documentation updates
1632
1633		----------------------------------------------------------------------
1634		Name: Allen Leung
1635		Date: Thu Dec 14 03:45:24 EST 2000
1636		Description:
1637		Tag:  leunga-20001214-int-inf
1638		Description:
1639
1640		In IntInf, added these standard functions, which are missing from our
1641		implementation:
1642
1643		andb : int * int -> int
1644		xorb : int * int -> int
1645		orb  : int * int -> int
1646		notb : int -> int
1647		<<   : int * word -> int
1648		~>>  : int * word -> int
1649
1650		Not tested, I hope they are correct.
1651
1652		----------------------------------------------------------------------
1653		Name: Allen Leung
1654		Date: Fri Dec  8 19:23:26 EST 2000
1655		Description:
1656		Tag:  leunga-20001208-nowhere
1657		Description:
1658
1659		Slight improvements to the 'nowhere' tool to handle OR-patterns,
1660		to generate better error messages etc.  Plus a brief manual.
1661
1662		----------------------------------------------------------------------
1663		Name: Lal George
1664		Date: 2000/12/08 09:54:02 EST 2000
1665		Tag: Release_110_31
1666		Description:
1667
1668		- Version 110.31
1669		----------------------------------------------------------------------
1670		Name: Allen Leung
1671		Date: Thu Dec  7 22:01:04 EST 2000
1672		Tag:  leunga-20001207-cell-monster-hack
1673		Description:
1674
1675		Major MLRISC internal changes.  Affect all clients.
1676		Summary:
1677
1678		1.  Type CELLS.cell = int is now replaced by a datatype.
1679		As a result, the old regmap is now gone.  Almost all interfaces
1680		in MLRISC change as a consequence.
1681
1682		2.  A new brand version of machine description tool (v3.0) that generates
1683		modules expecting the new interface.  The old version is removed.
1684
1685		3.  The RA interface has been further abstracted into two new functors.
1686		RISC_RA and X86RA.  These functors have much simpler interfaces.
1687		[See also directory MLRISC/demo.]
1688
1689		4.  Some other new source->source code generation tools are available:
1690
1691		a. MLRISC/Tools/RewriteGen -- generate rewriters from rules.
1692		b. MLRISC/Tools/WhereGen -- expands conditional pattern matching rules.
1693		I use this tool to generate the peephole optimizers---with the new
1694		cell type changes, peephole rules are becoming difficult to write
1695		without conditional pattern matching.
1696
1697		5.  More Intmap -> IntHashTable change.  Previous changes by Matthias didn't
1698		cover the entire MLRISC source tree so many things broke.
1699
1700		6.  CM files have been moved to the subdirectory MLRISC/cm.
1701		They are moved because there are a lot of them and they clutter up the
1702		root dir.
1703
1704		7.  More detailed documentation to come...
1705
1706		NOTE: To rebuild from 110.30 (ftp distribution), you'll have to do
1707		a makeml -rebuild first.  This is because of other other
1708		changes that Matthias has made (see below).
1709
1710
1711		----------------------------------------------------------------------
1712		Name: Matthias Blume
1713		Date: 2000/11/30 23:12:00 JST
1714		Tag: blume-20001130-filereorg
1715		Description:
1716
1717		Some manual updates and some file reorganizations in CM.
1718
1719		----------------------------------------------------------------------
1720		Name: Matthias Blume
1721		Date: 2000/11/24 17:45:00 JST
1722		Tag: blume-20001124-link
1723		Description:
1724
1725		Drastically improved link traversal code for the case that the dynamic
1726		value was already loaded at bootstrap time.  As a result, CM and CMB
1727		now both load blazingly fast -- even on a very slow machine.  Also,
1728		memory consumption has been further reduced by this.
1729
1730		Warning: The format of the PIDMAP file has changed.  THerefore, to
1731		bootstrap you have to do this:
1732
1733		1. Run CMB.make
1734		2. Make a symbolic link for the boot directory:
1735		ln -s sml.boot.ARCH-OS xxx
1736		3. "Rebuild" the boot directory:
1737		./makeml -boot xxx -rebuild sml ; rm xxx
1738		4. Boot normally:
1739		./makeml
1740
1741		----------------------------------------------------------------------
1742		Name: Matthias Blume
1743		Date: 2000/11/21 21:20:00 JST
1744		Tag: blume-20001121-tools
1745		Description:
1746
1747		Continued hacking on autoloading problem -- with success this time.
1748		Also changed tool-plugin mechanism.  See new CM manual.
1749
1750		----------------------------------------------------------------------
1751		Name: Matthias Blume
1752		Date: 2000/11/19 14:30:00 JST
1753		Tag:  blume-20001119-autoload
1754		Description:
1755
1756		Some hacking to make autoloading faster.  Success for CMB, no success
1757		so far for CM.  There is a reduced structure CM' that autoloads faster.
1758		(This is a temporary, non-documented hack to be eliminated again when
1759		the general problem is solved.)
1760
1761		----------------------------------------------------------------------
1762		Name: Matthias Blume
1763		Date: 2000/11/17 14:10:00 JST
1764		Tag: blume-20001117-pickle-lib
1765		Description:
1766
1767		1. Eliminated comp-lib.cm
1768		2. Made pickle-lib.cm
1769		3. Eliminated all uses of intset.sml (from comp-lib.cm)
1770		4. Replaced all uses of intmap.{sig,sml} (from comp-lib.cm) with
1771		equivalent constructs from smlnj-lib.cm (INtHashTable).
1772		5. Point 4. also goes for those uses of intmap.* in MLRISC.
1773		Duplicated intmap modules thrown out.
1774		6. Hunted down all duplicated SCC code and replaced it with
1775		equivalent stuff (GraphSCCFn from smlnj-lib.cm).
1776		7. Rewrote Feedback module.
1777		8. Moved sortedlist.sml into viscomp-lib.cm.  Eventually it
1778		should be thrown out and equivalent modules from smlnj-lib.cm
1779		should be used (IntRedBlackSet, IntListSet, ...).
1780
1781		Confirmed that compiler compiles to fixpoint.
1782
1783		----------------------------------------------------------------------
1784		Name: Allen Leung
1785		Date: 2000/11/10 18:00:00
1786		Tag: leunga-20001110-new-x86-fp
1787
1788		A new x86 floating point code generator has been added.
1789		By default this is turned off.  To turn this on, do:
1790
1791		CM.autoload "$smlnj/compiler.cm";
1792		Compiler.Control.MLRISC.getFlag "x86-fast-fp" := true;
1793
1794		Changes:
1795
1796		1.  Changed FTAN to FPTAN so that the assembly output is correct.
1797		2.  Changed the extension callback for FTANGENT to generate:
1798
1799		fptan
1800		fstp  %st(0)
1801		instead of
1802		fptan
1803		fstpl ftempmem
1804
1805		3.  Numerous assembly fixes for x86.
1806
1807		5.  Cleaned up the machine code output module x86/x86MC.sml and added
1808		support for a whole bunch of instructions and addressing modes:
1809
1810		fadd/fsub/fsubr/fmul/fdiv/fdivr  %st, %st(n)
1811		faddp/fsubp/fsubrp/fmulp/fdivp/fdivrp  %st, %st(n)
1812		fadd/fsub/fsubr/fmul/fdiv/fdivr  %st(n), %st
1813		fiadd/fisub/fisubr/fimul/fidiv/fidivr mem
1814		fxch %st(n)
1815		fld %st(n)
1816		fst %st(n)
1817		fst mem
1818		fstp %st(n)
1819		fucom %st(n)
1820		fucomp %st(n)
1821
1822		All these are now generated when the fast fp mode is turned on.
1823
1824		6.  Removed the dedicated registers %st(0), ..., %st(7) from X86CpsRegs
1825
1826		----------------------------------------------------------------------
1827		Name: Matthias Blume
1828		Date: 2000/11/09 11:20:00 JST
1829		Tag: blume-20001109-scc
1830		Description:
1831
1832		Eliminated some code duplication:
1833
1834		1. Added "where" clause to GraphSCCFn in SML/NJ Library.
1835		(Otherwise the functor is useless.)
1836		2. Used GraphSCCFn where SCCUtilFun was used previously.
1837		3. Got rid of SCCUtilFun (in comp-lib.cm).
1838
1839		----------------------------------------------------------------------
1840		Name: Lal George
1841		Date: 2000/11/06 09:02:21 EST 2000
1842		Tag: Release_110_30
1843		Description:
1844
1845		- Version 110.30
1846		----------------------------------------------------------------------
1847		Name: Matthias Blume
1848		Date: 2000/11/04 14:45:00
1849		Tag: blume-20001104-mlbuild
1850		Description:
1851
1852		- Made ml-build faster on startup.
1853		- Documentation fixes.
1854
1855		----------------------------------------------------------------------
1856		Name: Matthias Blume
1857		Date: 2000/11/02 17:00:00 JST
1858		Tag: blume-20001102-condcomp
1859		Description:
1860
1861		- Small tweaks to pickler -- new BOOTFILES!
1862		- Version bumped to 110.29.2.
1863		- Added conditional compilation facility to init.cmi (see comment there).
1864		----------------------------------------------------------------------
1865		Name: Allen Leung
1866		Date: 2000/10/23 19:31:00
1867		Tag: leunga-20001023-demo-ra
1868
1869		1. Minor RA changes that improves spilling on x86 (affects Moby and C-- only)
1870		2. Test programs for the graph library updated
1871		3. Some new MLRISC demo programs added
1872
1873		----------------------------------------------------------------------
1874		Name: Matthias Blume
1875		Date: 2000/08/31 22:15:00 JST
1876		Tag: blume-20001017-errmsg
1877		Description:
1878
1879		More error message grief: Where there used to be no messages, there
1880		now were some that had bogus error regions.  Fixed.
1881
1882		----------------------------------------------------------------------
1883		Name: Matthias Blume
1884		Date: 2000/08/31 17:30:00 JST
1885		Tag: blume-20001017-v110p29p1
1886		Description:
1887
1888		I made a version 110.29.1 with new bootfiles.
1889
1890		Changes:  Modified pickler/unpickler for faster and leaner unpickling.
1891		CM documentation changes and a small bugfix in CM's error reporting.
1892
1893		----------------------------------------------------------------------
1894		Name: Lal George
1895		Date: 2000/09/27 14:42:35 EDT
1896		Tag: george-20000927-nodestatus
1897		Description:
1898
1899		Changed the type of the nodestatus, so that:
1900
1901		SPILLED(~1)             is now SPILLED
1902		SPILLED(m) where m>=0   is now MEMREG(m)
1903		SPILLED(s) where s<~1   is now SPILL_LOC(~s)
1904
1905		----------------------------------------------------------------------
1906		Name: Matthias Blume
1907		Date: 2000/09/07 14:45:00 JST
1908		Tag: blume-20000907-cmerrmsg
1909		Description:
1910
1911		Small tweak to CM to avoid getting ML syntax error messages twice.
1912
1913		----------------------------------------------------------------------
1914		Name: Matthias Blume
1915		Date: 2000/08/31 18:00:00 JST
1916		Tag: blume-20000831-cvsbootfiles
1917		Description:
1918
1919		New URL for boot files (because the 110.29 files on the BL server do
1920		now work correctly with my updated install scripts for yacc and lex).
1921
1922		----------------------------------------------------------------------
1923		Name: Matthias Blume
1924		Date: 2000/08/08 12:33:00 JST
1925		Tag: blume-20000808-manual
1926		Description:
1927
1928		Tiny update to CM manual.
1929
1930		----------------------------------------------------------------------
1931		Name: Allen Leung
1932		Date: 2000/08/7 19:31:00
1933		Tag: leunga-20000807-a-whole-bunch-of-stuff
1934
1935		Moby, C--, SSA, x86, machine descriptions etc.  Should only affect C--
1936		and Mobdy.
1937
1938		1.  x86
1939
1940		a.  Fixes to peephole module by John and Dan.
1941		b.  Assembly fix to SETcc by Allen.
1942		c.  Fix to c-call by John.
1943		d.  Fix to spilling by John.  (This one deals with the missing FSTPT case)
1944		e.  Instruction selection optimization to SETcc as suggested by John.
1945
1946		For example,
1947
1948		MV(32, x, COND(32, CMP(32, LT, a, b), LI 1, LI 0))
1949
1950		should generate:
1951
1952		MOVL a, x
1953		SUBL b, x
1954		SHRL 31, x
1955
1956		2.  IR stuff
1957
1958		A bunch of new DJ-graph related algorithms added.  These
1959		speed up SSA construction.
1960
1961		3.  SSA + Scheduling
1962
1963		Added code for SSA and scheduling to the repository
1964
1965		----------------------------------------------------------------------
1966		Name: Lal George
1967		Date: 2000/07/27 11:53:14 EDT
1968
1969		Tag: lal-20000727-linux-ppc
1970		Description:
1971
1972		Made changes to support Linux PPC.
1973		p.s. I have confirmation that the 110.29 boot files work fine.
1974
1975		----------------------------------------------------------------------
1976		Name: Matthias Blume
1977		Date: 2000/07/27 17:40:00 JST
1978		Tag: blume-20000727-scripts
1979		Description:
1980
1981		!!!! WARNING !!!!
1982		You must recompile the runtime system!
1983		!!!! WARNING !!!!
1984
1985		This is basically another round of script-enhancements:
1986
1987		1. sml, ml-build, and ml-makedepend accept options -D and -U to define
1988		and undefine CM preprocessor symbols.
1989
1990		2. ml-build avoids generating a new heap image if it finds that the
1991		existing one is still ok.  (The condition is that no ML file had to
1992		be recompiled and all ML files are found to be older that the heap
1993		file.)
1994
1995		To make this work smoothly, I also hacked the runtime system as
1996		well as SMLofNJ.SysInfo to get access to the heap image suffix
1997		(.sparc-solaris, ...) that is currently being used.
1998
1999		Moreover, the signature of CM.mk_standalone has changed.  See the
2000		CM manual.
2001
2002		3. ml-makedepend accepts additional options -n, -a, and -o.  (See the
2003		CM manual for details.)
2004
2005		4. More CM manual updates:
2006		- all of the above has been documented.
2007		- there is now a section describing the (CM-related) command line
2008		arguments that are accepted by the "sml" command
2009
2010		----------------------------------------------------------------------
2011		Name: Matthias Blume
2012		Date: 2000/07/25 16:20:00 JST
2013		Tag: blume-20000725-makedepend
2014		Description:
2015
2016		Added a script called ml-makedepend.  This can be used in makefiles
2017		for Unix' make in a way very similar to the "makedepend" command for
2018		C.
2019
2020		The script internally uses function CM.sources.
2021
2022		Synopsis:
2023
2024		ml-makedepend [-f makefile] cmfile targetname
2025
2026		The default for the makefile is "makefile" (or "Makefile" should
2027		"makefile" not exist).
2028
2029		ml-makedepend adds a cmfile/targetname-specific section to this
2030		makefile (after removing the previous version of this section).  The
2031		section contains a single dependency specification with targetname on
2032		the LHS (targetname is an arbitrary name), and a list of files derived
2033		from the cmfile on the RHS.  Some of the files on the RHS are
2034		ARCH/OPSYS-specific.  Therefore, ml-makedepend inserts references to
2035		"make" variables $(ARCH) and $(OPSYS) in place of the corresponding
2036		path names.  The makefile writer is responsible for making sure that
2037		these variables have correct at the time "make" is invoked.
2038
2039		----------------------------------------------------------------------
2040		Name: Matthias Blume
2041		Date: 2000/07/22 23:30:00 JST
2042		Tag: blume-20000722-urlupdate
2043		Description:
2044
2045		Changed BOOT and config/srcarchiveurl to point to BL server:
2046
2047		ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.29/
2048
2049		----------------------------------------------------------------------
2050		Name: Matthias Blume
2051		Date: 2000/07/18 18:00:00 JST
2052		Tag: blume-20000718-Version_110_29
2053		Description:
2054
2055		1. Updated src/compiler/TopLevel/main/version.sml to version 110.29
2056
2057		2. Updated config/version to 110.29
2058
2059		3. Updated config/srcarchiveurl
2060
2061		3. New boot files!
2062		ftp://ftp.cs.princeton.edu/pub/people/blume/sml/110.29-autofetch
2063
2064		----------------------------------------------------------------------
2065		Name: Matthias Blume
2066		Date: 2000/07/11 13:58:00 JST
2067		Tag: blume-20000711-doctypo
2068		Description:
2069
2070		Fixed a few typos in CM manual.
2071
2072		----------------------------------------------------------------------
2073		Name: Allen Leung
2074		Date: 2000/06/15 00:38:00
2075		Tag: leunga-20000704-sparc-x86
2076
2077		1. x86 peephole improvement sp += k; sp -= k => nop  [from John]
2078		2. fix to x86 RET bug [found by Dan Grossman]
2079		3. sparc assembly bug fix for ticc instructions [found by Fermin]
2080
2081		Affects c-- and moby only
2082
2083		----------------------------------------------------------------------
2084		Name: Matthias Blume
2085		Date: 2000/07/04 15:26:00
2086		Tag: blume-20000704-trigger
2087		Description:
2088
2089		1. Improvements to CM manual.
2090		2. SMLofNJ.Internals.BTrace.trigger reinstated as an alternative way
2091		of getting a back-trace.  The function, when called, raises an
2092		internal exception which explicitly carries the full back-trace history,
2093		so it is unaffected by any intervening handle-raise pairs ("trivial"
2094		or not).  The interactive loop will print that history once it arrives
2095		at top level.
2096		Short of having all exceptions implicitly carry the full history, the
2097		recommended way of using this facility is:
2098		- compile your program with instrumentation "on"
2099		- run it, when it raises an exception, look at the history
2100		- if the history is "cut off" because of some handler, go and modify
2101		your program so that it explicitly calls BTrace.trigger
2102		- recompile (still instrumented), and rerun; look at the full history
2103
2104		----------------------------------------------------------------------
2105		Name: Matthias Blume
2106		Date: 2000/07/03 15:36:00 JST
2107		Tag: blume-20000702-manual
2108		Description:
2109
2110		Small corrections and updates to CM manual.
2111
2112		----------------------------------------------------------------------
2113		Name: Matthias Blume
2114		Date: 2000/06/29 16:04:00 JST
2115		Tag: blume-20000629-yacctool
2116		Description:
2117
2118		Changes:
2119
2120		1. Class "mlyacc" now takes separate arguments to pass options to
2121		generated .sml- and .sig-files independently.
2122		2. Corresponding CM manual updates.
2123		3. BTrace module now also reports call sites.  (However, for loop clusters
2124		it only shows from where the cluster was entered.)  There are associated
2125		modifications to core.sml, internals.{sig,sml}, btrace.sml, and btimp.sml.
2126
2127		----------------------------------------------------------------------
2128		Name: Matthias Blume
2129		Date: 2000/06/27 16:51:00 JST
2130		Tag: blume-20000627-noweb
2131		Description:
2132
2133		Changes:
2134
2135		1. Implemented "subdir" and "witness" options for noweb tool.
2136		This caused some slight internal changes in CM's tool implementation.
2137		2. Fixed bug in "tool plugin" mechanism.  This is essentially cleaning
2138		some remaining issues from earlier path anchor changes.
2139		3. Updated CM manual accordingly.
2140
2141		4. Changed implementation of back-tracing so that I now consider it
2142		ready for prime-time.
2143
2144		In particular, you don't have to explicitly trigger the back-trace
2145		anymore.  Instead, if you are running BTrace-instrumented code and
2146		there is an uncaught exception (regardless of whether or not it was
2147		raised in instrumented code), the top-level evalloop will print
2148		the back-trace.
2149
2150		Features:
2151
2152		- Instrumented and uninstrumented code work together seemlessly.
2153		(Of course, uninstrumented code is never mentioned in actual
2154		back-traces.)
2155
2156		- Asymptotic time- and space-complexity of instrumented code is
2157		equal to that of uninstrumented code.  (This means that
2158		tail-recursion is preserved by the instrumentation phase.)
2159
2160		- Modules whose code has been instrumented in different sessions
2161		work together without problem.
2162
2163		- There is no penalty whatsoever on uninstrumented code.
2164
2165		- There is no penalty on "raise" expressions, even in
2166		instrumented code.
2167
2168		A potential bug (or perhaps it is a feature, too):
2169
2170		A back-trace reaches no further than the outermost instrumented
2171		non-trivial "raise".  Here, a "trivial" raise is one that is the
2172		sole RHS of a "handle" rule.  Thus, back-traces reach trough
2173
2174		<exp> handle e => raise e
2175
2176		and even
2177
2178		<exp> handle Foo => raise Bar
2179
2180		and, of course, through
2181
2182		<exp> handle Foo => ...
2183
2184		if the exception was not Foo.
2185
2186		Back-traces always reach right through any un-instrumented code
2187		including any of its "handle" expressions, trivial or not.
2188
2189		To try this out, do the following:
2190
2191		- Erase all existing binfiles for your program.
2192		(You may keep binfiles for those modules where you think you
2193		definitely don't need back-tracing.)
2194		- Turn on back-trace instrumentation:
2195		SMLofNJ.Internals.BTrace.mode (SOME true);
2196		- Recompile your program.  (I.e., run "CM.make" or "use".)
2197		- You may now turn instrumentation off again (if you want):
2198		SMLofNJ.Internals.BTrace.mode (SOME false);
2199		- Run your program as usual.  If it raises an exception that
2200		reaches the interactive toplevel, then a back-trace will
2201		automatically be printed.  After that, the toplevel loop
2202		will print the exception history as usual.
2203
2204		----------------------------------------------------------------------
2205		Name: Matthias Blume
2206		Date: 2000/06/26 09:56:46 JST
2207		Tag: blume-20000626-setup
2208		Description:
2209
2210		CM: - setup-parameter to "sml" added; this can be used to run arbitrary
2211		ML code before and after compiling a file (e.g., to set compiler
2212		flags)
2213
2214		Compiler: - improved btrace API (in core.sml, internals.{sig,sml})
2215		- associated changes to btrace.sml (BTrace instrumentation pass)
2216		- cleaner implementation of btimp.sml (BTrace tracing and report
2217		module)
2218
2219		CM manual: * new path encoding documented
2220		* description of setup-parameter to "sml" added
2221
2222		The biggest user-visible change to back-tracing is that it is no
2223		longer necessary to compile all traced modules within the same
2224		session.  (This was a real limitation.)
2225
2226		----------------------------------------------------------------------
2227		Name: Matthias Blume
2228		Date: 2000/06/24 12:40:00 JST
2229		Tag: blume-20000624-startup
2230		Description:
2231
2232		Fixes startup slowdown problem.  (I was calling SrcPath.sync a _tad_
2233		bit too often -- to put it mildly. :)
2234
2235		----------------------------------------------------------------------
2236		Name: Matthias Blume
2237		Date: 2000/06/23 18:20:00 JST
2238		Tag: blume-20000623-btrace
2239		Description:
2240
2241		This updates adds a backtrace facility to aid programmers in debugging
2242		their programs.  This involves the following changes:
2243
2244		1. Module system/smlnj/init/core.sml (structure _Core) now has hooks for
2245		keeping track of the current call stack.  When programs are compiled
2246		in a special mode, the compiler will insert calls to these hooks
2247		into the user program.
2248		"Hook" means that it is possible for different implementations of
2249		back-tracing to register themselves (at different times).
2250
2251		2. compiler/MiscUtil/profile/btrace.sml implements the annotation phase
2252		as an Absyn.dec->Absyn.dec rewrite.  Normally this phase is turned off.
2253		It can be turned on using this call:
2254		SMLofNJ.Internals.BTrace.mode (SOME true);
2255		Turning it off again:
2256		SMLofNJ.Internals.BTrace.mode (SOME false);
2257		Querying the current status:
2258		SMLofNJ.Internals.BTrace.mode NONE;
2259		Annotated programs are about twice as big as normal ones, and they
2260		run a factor of 2 to 4 slower with a dummy back-trace plugin (one
2261		where all hooks do nothing).  The slowdown with a plugin that is
2262		actually useful (such as the one supplied by default) is even greater,
2263		but in the case of the default plugin it is still only an constant
2264		factor (amortized).
2265
2266		3. system/Basis/Implementation/NJ/internals.{sig,sml} have been augmented
2267		with a sub-structure BTrace for controlling back-tracing.  In particular,
2268		the above-mentioned function "mode" controls whether the annotation
2269		phase is invoked by the compiler.  Another important function is
2270		"trigger": when called it aborts the current execution and causes
2271		the top-level loop to print a full back-trace.
2272
2273		4. compiler/MiscUtil/profile/btimp.sml is the current default plugin
2274		for back-tracing.  It keeps track of the dynamic call stack and in
2275		addition to that it keeps a partial history at each "level" of that
2276		stack.  For example, if a tail-calls b, b tail-calls c, and c tail-calls
2277		d and b (at separate times, dynamically), then the report will show:
2278
2279		GOTO   d
2280		/c
2281		GOTO  \b
2282		CALL   a
2283
2284		This shows that there was an initial non-tail call of a, then a
2285		tail-call to b or c, looping behavior in a cluster of functions that
2286		consist of b and c, and then a goto from that cluster (i.e., either from
2287		b or from c) to d.
2288
2289		Note that (depending on the user program) the amount of information
2290		that the back-trace module has to keep track of at each level is bounded
2291		by a constant.  Thus, the whole implementation has the same asymptotical
2292		complexity as the original program (both in space and in time).
2293
2294		5. compiler/TopLevel/interact/evalloop.sml has been modified to
2295		handle the special exception SMLofNJ.Internals.BTrace.BTrace
2296		which is raised by the "trigger" function mentioned above.
2297
2298		Notes on usage:
2299
2300		- Annotated code works well together with unannotated code:
2301		Unannotated calls simply do not show up at all in the backtrace.
2302
2303		- It is not a good idea to let modules that were annotated during
2304		different sessions run at the same time.  This is because the compiler
2305		chooses small integers to identify individual functions, and there
2306		will be clashes if different modules were compiled in separate sessions.
2307		(Nothing will crash, and you will even be told about the clashes, but
2308		back-trace information will in general not be useful.)
2309
2310		- Back-tracing can be confused by callcc and capture.
2311
2312		- The only way of getting a back-trace right now is to explicitly
2313		invoke the "trigger" function from your user program.  Eventually, we
2314		should make every exception carry back-trace information (if
2315		available).  But since this creates more overhead at "raise"-time
2316		(similar to the current exnHistory overhead), I have not yet
2317		implemented this.  (The implementation will be rather easy.)  With
2318		exceptions carrying back-trace information, this facility will be even
2319		more useful because users don't need to modify their programs...
2320
2321		- While it is possible to compile the compiler with back-trace
2322		annotations turned on (I did it to get some confidence in
2323		correctness), you must make absolutely sure that core.sml and
2324		btimp.sml are compiled WITHOUT annotation!  (core.sml cannot actually
2325		be compiled with annotation because there is no core access yet, but
2326		if you compile btimp.sml with annotation, then the system will go into
2327		an infinite recursion and crash.)
2328		Since CM currently does not know about BTrace, the only way to turn
2329		annotations on and off for different modules of the compiler is to
2330		interrupt CMB.make, change the settings, and re-invoke it.  Of course,
2331		this is awkward and clumsy.
2332
2333		Sample sessions:
2334
2335		Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
2336		- SMLofNJ.Internals.BTrace.mode (SOME true);
2337		[autoloading]
2338		[autoloading done]
2339		val it = false : bool
2340		- structure X = struct
2341		-     fun main n = let
2342		-         fun a (x, 0) = d x
2343		-           | a (x, n) = b (x, n - 1)
2344		-         and b (x, n) = c (x, n)
2345		-         and c (x, n) = a (x, n)
2346		-         and d x = e (x, 3)
2347		-         and e (x, 0) = f x
2348		-           | e (x, n) = e (x, n - 1)
2349		-         and f 0 = SMLofNJ.Internals.BTrace.trigger ()
2350		-           | f n = n * g (n - 1)
2351		-         and g n = a (n, 3)
2352		-     in
2353		-         f n
2354		-     end
2355		- end;
2356		structure X : sig val main : int -> int end
2357		- X.main 3;
2358		*** BACK-TRACE ***
2359		GOTO   stdIn:4.2-13.20: X.main[2].f
2360		GOTO-( stdIn:4.2-13.20: X.main[2].e
2361		GOTO   stdIn:4.2-13.20: X.main[2].d
2362		/ stdIn:4.2-13.20: X.main[2].a
2363		| stdIn:4.2-13.20: X.main[2].b
2364		GOTO-\ stdIn:4.2-13.20: X.main[2].c
2365		CALL   stdIn:4.2-13.20: X.main[2].g
2366		GOTO   stdIn:4.2-13.20: X.main[2].f
2367		GOTO-( stdIn:4.2-13.20: X.main[2].e
2368		GOTO   stdIn:4.2-13.20: X.main[2].d
2369		/ stdIn:4.2-13.20: X.main[2].a
2370		| stdIn:4.2-13.20: X.main[2].b
2371		GOTO-\ stdIn:4.2-13.20: X.main[2].c
2372		CALL   stdIn:4.2-13.20: X.main[2].g
2373		GOTO   stdIn:4.2-13.20: X.main[2].f
2374		GOTO-( stdIn:4.2-13.20: X.main[2].e
2375		GOTO   stdIn:4.2-13.20: X.main[2].d
2376		/ stdIn:4.2-13.20: X.main[2].a
2377		| stdIn:4.2-13.20: X.main[2].b
2378		GOTO-\ stdIn:4.2-13.20: X.main[2].c
2379		CALL   stdIn:4.2-13.20: X.main[2].g
2380		GOTO   stdIn:4.2-13.20: X.main[2].f
2381		CALL   stdIn:2.15-17.4: X.main[2]
2382		-
2383
2384		(Note that because of a FLINt bug the above code currently does not
2385		compile without BTrace turned on.)
2386
2387		Here is another example, using my modified Tiger compiler:
2388
2389		Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
2390		- SMLofNJ.Internals.BTrace.mode (SOME true);
2391		[autoloading]
2392		[autoloading done]
2393		val it = false : bool
2394		- CM.make "sources.cm";
2395		[autoloading]
2396		...
2397		[autoloading done]
2398		[scanning sources.cm]
2399		[parsing (sources.cm):parse.sml]
2400		[creating directory CM/SKEL ...]
2401		[parsing (sources.cm):tiger.lex.sml]
2402		...
2403		[wrote CM/sparc-unix/semant.sml]
2404		[compiling (sources.cm):main.sml]
2405		[wrote CM/sparc-unix/main.sml]
2406		[New bindings added.]
2407		val it = true : bool
2408		- Main.compile ("../testcases/merge.tig", "foo.out");
2409		*** BACK-TRACE ***
2410		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trvar
2411		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
2412		CALL   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
2413		GOTO   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
2414		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
2415		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
2416		CALL   lib/semant.sml:488.3-505.6: SemantFun[2].transDec.trdec[2].transBody[2]
2417		/ lib/semant.sml:411.65-543.8: SemantFun[2].transDec
2418		CALL-\ lib/semant.sml:413.2-540.9: SemantFun[2].transDec.trdec[2]
2419		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
2420		CALL   lib/semant.sml:8.52-558.4: SemantFun[2].transProg[2]
2421		CALL   main.sml:1.18-118.4: Main.compile[2]
2422		-
2423
2424		----------------------------------------------------------------------
2425		Name: Matthias Blumen
2426		Date: 2000/06/21 18:00:00 JST
2427		Tag: blume-20000621-manual
2428		Description:
2429
2430		CM manual update: Path environments documented.
2431
2432		----------------------------------------------------------------------
2433		Name: Matthias Blume
2434		Date: 2000/06/19 13:40:00
2435		Tag: blume-20000619-manual
2436		Description:
2437
2438		CM manual and system/README update.  This only covers the fact that
2439		there are no more implicit anchors.  (Path environments and the "bind"
2440		option to "cm" have yet to be documented.)
2441
2442		----------------------------------------------------------------------
2443		Name: Matthias Blume
2444		Date: 2000/06/19 11:05:00 JST
2445		Tag: blume-20000619-chdir-bugfix
2446		Description:
2447
2448		Fixed a bug in new SrcPath module that sometimes led to a bad chDir call.
2449
2450		----------------------------------------------------------------------
2451		Name: Matthias Blume
2452		Date: 2000/06/18 22:00:10 JST
2453		Tag: blume-20000618-implicit-anchors-really-gone
2454		Description:
2455
2456		I updates the previous HISTORY entry where I forgot to mention that
2457		implicit anchors are no longer with us.
2458
2459		The current update also gets rid of the (now useless) controller
2460		CM.Control.implicit_anchors.
2461
2462		----------------------------------------------------------------------
2463		Name: Matthias Blume
2464		Date: 2000/06/16 17:30:00 JST
2465		Tag: blume-20000616-anchorenv
2466		Description:
2467
2468		This patch implements the long anticipated (just kidding :) "anchor
2469		environment" mechanism.  In the course of doing this, I also
2470		re-implemented CM's internal "SrcPath" module from scratch.  The new
2471		one should be more robust in certain boundary cases.  In any case, it
2472		is a lot cleaner than its predecessor (IMHO).
2473
2474		This time, although there is yet another boot file format change, I
2475		kept the unpickler backward-compatible.  As a result, no new bootfiles
2476		are necessary and bootstrapping is straightforward.  (You cannot read
2477		new bootfiles into an old system, but the other way around is no
2478		problem.)
2479
2480		Visible changes:
2481
2482		** 0. Implicit path anchors (without the leading $-symbol) are no
2483		longer recognized at all. This means that such path names are not
2484		illegal either.  For example, the name basis.cm simply refers to a
2485		local file called "basis.cm" (i.e, the name is an ordinary path
2486		relative to .cm-files directory).  Or, to put it differently, only
2487		names that start with $ are anchored paths.
2488
2489		** 1. The $<singlearc> abbreviation for $/<singlearc> has finally
2490		vanished.
2491
2492		John (Reppy) had critizised this as soon as I originally proposed and
2493		implemented it, but at that time I did not really deeply believe
2494		him. :) Now I came full-circle because I need the $<singlearc> syntax
2495		in another place where it cannot be seen as an abbreviation for
2496		$/<singlearc>.  To avoid the confusion, $<singlearc> now means what it
2497		seems to mean (i.e., it "expands" into the corresponding anchor
2498		value).
2499
2500		However, when paths are used as members in CM description files, it
2501		continues to be true that there must be at least another arc after the
2502		anchor.  This is now enforced separately during semantic analysis
2503		(i.e., from a lexical/syntactical point of view, the notation is ok.)
2504
2505		** 2. The "cm" class now accepts an option "bind".  The option's value
2506		is a sub-option list of precisely two items -- one labeled "anchor"
2507		and the other one labeled "value".  As you might expect, "anchor" is
2508		used to specify an anchor name to be bound, and "value" specifies what
2509		the anchor is being bound to.
2510
2511		The value must be a directory name and can be given in either standard
2512		syntax (including the possibility that it is itself an anchored path)
2513		or native syntax.
2514
2515		Examples:
2516
2517		foo.cm (bind:(anchor:bar value:$mystuff/bar))
2518		lib.cm (bind:(anchor:a value:"H:\\x\\y\\z"))  (* only works under windows *)
2519
2520		and so on.
2521
2522		The meaning of this is that the .cm-file will be processed with an
2523		augmented anchor environment where the given anchor(s) is/are bound to
2524		the given values(s).
2525
2526		The rationale for having this feature is this: Suppose you are trying
2527		to use two different (already stable) libraries a.cm and b.cm (that
2528		you perhaps didn't write yourself).  Further, suppose each of these
2529		two libraries internally uses its own auxiliary library $aux/lib.cm.
2530		Normally you would now have a problem because the anchor "lib" can not
2531		be bound to more than one value globally.  Therefore, the project that
2532		uses both a.cm and b.cm must locally redirect the anchor to some other
2533		place:
2534
2535		a.cm (bind:(anchor:lib value:/usr/lib/smlnj/a-stuff))
2536		b.cm (bind:(anchor:lib value:/usr/lib/smlnj/b-stuff))
2537
2538		This hard-wires $lib/aux.cm to /usr/lib/smlnj/a-stuff/aux.cm or
2539		/usr/lib/smlnj/b-stuff/aux.cm, respectively.
2540
2541		Hard-wiring path names is a bit inflexible (and CM will verbosely warn
2542		you when you do so at the time of CM.stabilize).  Therefore, you can
2543		also use an anchored path as the value:
2544
2545		a.cm (bind:(anchor:lib value:$a-lib))
2546		b.cm (bind:(anchor:lib value:$b-lib))
2547
2548		Now you can globally configure (using the usual CM.Anchor.anchor or
2549		pathconfig machinery) bindings for "a-lib" and "b-lib".  Since "lib"
2550		itself is always locally bound, setting it globally is no longer
2551		meaningful or necessary (but it does not hurt either).  In fact, "lib"
2552		can still be used as a global anchor for separate purposes.  As a
2553		matter of fact, one can locally define "lib" in terms of a global
2554		"lib":
2555
2556		a.cm (bind:(anchor:lib value:$lib/a))
2557		b.cm (bind:(anchor:lib value:$lib/b))
2558
2559		** 3: The encoding of path names has changed.  This affects the way
2560		path names are shown in CM's progress report and also the internal
2561		protocol encoding used for parallel make.
2562
2563		The encoding now uses one or more ':'-separated segments.  Each
2564		segments corresponds to a file that has been specified relative to the
2565		file given by its preceding segment.  The first segment is either
2566		relative to the CWD, absolute, or anchored.  Each segment itself is
2567		basically a Unix pathname; all segments but the first are relative.
2568
2569		Example:
2570
2571		$foo/bar/baz.cm:a/b/c.sml
2572
2573		This path denotes the file bar/a/b/c.sml relative to the directory
2574		denoted by anchor "foo".  Notice that the encoding also includes
2575		baz.cm which is the .cm-file that listed a/b/c.sml.  As usual, such
2576		paths are resolved relative to the .cm-files directory, so baz.cm must
2577		be ignored to get the "real" pathname.
2578
2579		To make this fact more obvious, CM puts the names of such "virtual
2580		arcs" into parentheses when they appear in progress reports. (No
2581		parentheses will appear in the internal protocol encoding.)  Thus,
2582		what you really see is:
2583
2584		$foo/bar/(baz.cm):a/b/c.sml
2585
2586		I find this notation to be much more informative than before.
2587
2588		Another new feature of the encoding is that special characters
2589		including parentheses, colons, (back)slashes, and white space are
2590		written as \ddd (where ddd is the decimal encoding of the character).
2591
2592		*** The CM manual still needs to be updated.
2593
2594		----------------------------------------------------------------------
2595		Name: Allen Leung
2596		Date: 2000/06/15 00:38:00
2597		Tag: leunga-20000615-x86-peephole
2598
2599		x86 Peephole fix by Fermin.  Affects c-- and moby only.
2600
2601		----------------------------------------------------------------------
2602		Name: Matthias Blume
2603		Date: 2000/06/12 11:40:00
2604		Tag: blume-20000612-parmakefix
2605		Description:
2606
2607		More cleanup after changing the file naming scheme: This time I
2608		repaired the parallel make mechanism for CMB.make which I broke earlier.
2609
2610		----------------------------------------------------------------------
2611		Name: Allen Leung
2612		Date: 2000/06/09 01:25:00
2613		Tag: leunga-20000609-various
2614
2615		None of these things should affect normal SML/NJ operations
2616
2617		1. Peephole improvements provided by Fermin (c--)
2618		2. New annotation DEFUSE for adding extra dependence (moby)
2619		3. New X86 LOCK instructions (moby)
2620		4. New machine description language for reservation tables (scheduling)
2621		5. Fixes to various optimization/analysis modules (branch chaining, dominator
2622		trees etc.)
2623		6. I've changed the CM files so that they can work with versions
2624		110.0.6, 110.25 and 110.28
2625
2626		----------------------------------------------------------------------
2627		Name: Matthias Blume
2628		Date: 2000/06/09 12:40:00
2629		Tag: blume-20000609-log
2630		Description:
2631
2632		- Removed all(?) remaining RCS Log entries from sources.
2633
2634		- Fixed bug in ml-yacc and ml-lex sources (use explicit anchors for
2635		anchored paths).
2636
2637		----------------------------------------------------------------------
2638		Name: Matthias Blume
2639		Date: 2000/06/07 17:00:00 JST
2640		Tag: blume-20000607-no-implicit-anchors
2641		Description:
2642
2643		1. This update changes the default setting for
2644		CM.Control.implicit_anchors from true to false.  This means that
2645		implicit anchors are no longer permitted by default.  I also tried to
2646		make sure that nothing else still relies on implicit anchors.
2647		(This is the next step on the schedule towards a CM that does not even
2648		have the notion of implicit anchors anymore.)
2649
2650		2. More CM manual updates.
2651
2652		3. I managed to track down and fix the pickling bug I mentioned last
2653		time.  Because of the previously existing workaround, this entails no
2654		immediate practical changes.
2655
2656		----------------------------------------------------------------------
2657		Name: Matthias Blume
2658		Date: 2000/06/06 11:15:00 JST
2659		Tag: blume-20000606-lazierpickle
2660		Description:
2661
2662		!!!! NEW BOOT FILES !!!!
2663
2664		* The main purpose of this update is to make library pickles lazier in
2665		order to reduce the initial space penalty for autoloading a library.
2666		As a result, it is now possible to have $smlnj/compiler.cm
2667		pre-registered.  This should take care of the many complaints or
2668		inquiries about missing structure Compiler.  This required changes to
2669		CM's internal data structures and small tweaks to some algorithms.
2670
2671		As a neat additional effect, it is no longer necessary (for the sake
2672		of lean heap image files) to distinguish between a "minimal" CM and a
2673		"full" CM.  Now, there is only one CM (i.e., the "full" version:
2674		$smlnj/cm.cm aka $smlnj/cm/full.cm), and it is always available at the
2675		interactive top level. ($smlnj/cm/minimal.cm is gone.)
2676
2677		To make the life of compiler-hackers easier, "makeml" now also
2678		pre-registers $smlnj/cmb.cm (aka $smlnj/cmb/current.cm).  In other
2679		words, after you bootstrap a new sml for the first time, you will not
2680		have to autoload $smlnj/cmb.cm again afterwards.  (The first time
2681		around you will still have to do it, though.)
2682
2683		* A second change consists of major updates to the CM manual.  There
2684		are now several appendices with summary information and also a full
2685		specification of the CM description file syntax.
2686
2687		* In directory src/system I added the script "allcross".  This script
2688		invokes sml and cross-compiles the compiler for all supported
2689		architectures.  (Useful when providing a new set of boot files.)
2690
2691		* There seems to be a latent bug in my "lazy pickles" mechanism.  I
2692		added a small tweak to pickle-util.sml to work around this problem,
2693		but it is not a proper fix yet.  I will investigate further.  (The
2694		effect of the bug was an inflation of library pickle size.)
2695
2696		* Version number increased to 110.28.1 (to avoid compatibility problems).
2697
2698		----------------------------------------------------------------------
2699		Name: Allen Leung
2700		Date: 2000/05/25 17:28 EDT
2701		Tag: leunga-20000525-ra
2702		Description:
2703
2704		Fixed a bug in freezing phase of the register allocator.
2705
2706		----------------------------------------------------------------------
2707		Name: Allen Leung
2708		Date: 2000/05/15 22:53 EDT
2709		Tag: leunga-20000515-alpha-x86-ra
2710		Description:
2711
2712		1. Alpha
2713
2714		Slight cleanup.  Removed the instruction SGNXL
2715
2716		2. X86
2717
2718		Added the following instructions to the instruction set:
2719
2720		ROLx, RORx,
2721		BTx, BTSx, BTLx, BTRx,
2722		XCHGx, and variants with the LOCK prefix
2723
2724		3. Register Allocation
2725
2726		The module ra-rewrite-with-renaming has been improved.
2727
2728		These have no effect on SML/NJ.
2729
2730		----------------------------------------------------------------------
2731		Name: Matthias Blume
2732		Date: 2000/05/15 16:20:00 JST
2733		Tag: blume-20000515-lightrebuild
2734		Description:
2735
2736		1. I added an alternative to "-rebuild" to "makeml".  The difference is
2737		that prior to calling CMB.make' the CM-variable "LIGHT" will be
2738		defined.  In effect, the command will not build any cross-compiler
2739		backends and therefore finish more quickly.
2740
2741		The "fixpt" script also takes a "-light" switch to be able to use
2742		this new facility while compiling for a fixpoint.
2743
2744		2. I replaced all mentions of anchored paths in group owner specifications
2745		with simple relative paths (usually starting with "..").
2746		The rationale is that a library's internal workings should not be
2747		compromised by the lack of some anchor.  (An anchor is necessary
2748		for someone who wants to refer to the library by an anchored path,
2749		but it should not be necessary to build the same library in the first
2750		place.)
2751
2752		3. I changed the way CM's tool mechanism determines the shell command
2753		string used for things like ml-yacc etc. so that it does not break
2754		when CM.Control.implicit_anchors is turned off.
2755
2756		----------------------------------------------------------------------
2757		Name: Matthias Blume
2758		Date: 2000/05/12 18:20:00 JST
2759		Tag: blume-20000512-ml-build
2760		Description:
2761
2762		Fixed a bug in config/_ml-build that prevented ml-yacc and ml-lex from
2763		getting installed properly (by config/install.sh).
2764
2765		----------------------------------------------------------------------
2766		Name: Matthias Blume
2767		Date: 2000/05/12 17:30:00 JST
2768		Tag: blume-20000512-anchors
2769		Description:
2770
2771		!!! NEW BOOT FILES !!!
2772
2773		This change is in preparation of fading out support for "implicitly
2774		anchored path names".  I went through all sources and used the
2775		explicit (and relatively new) $-notation.  See system/README and the
2776		CM manual for more info on this.
2777
2778		I also modified the anchoring scheme for some things such as "smlnj",
2779		"MLRISC", "cm", etc. to take advantage of the fact that explicit
2780		anchors are more expressive: anchor name and first arc do not have to
2781		coincide.  This entails the following user-visible change:
2782
2783		You have to write $smlnj/foo/bar instead of smlnj/foo/bar.  In
2784		particular, when you fire up sml with a command-line argument, say,
2785		e.g.:
2786
2787		sml '$smlnj/cmb.cm'
2788
2789		At the ML toplevel prompt:
2790
2791		CM.autoload "$smlnj/cmb.cm";
2792
2793		There is also a new controller in CM.Control that can be used to turn
2794		off all remaining support for implicit anchors by saying:
2795
2796		CM.autoload "$smlnj/
2797		#set CM.Control.implicit_anchors false;
2798
2799		This causes CM to reject implicitly anchored paths.  This is (for the
2800		time being) less permissive than the "final" version where there will
2801		be no more such implicit anchors and relative paths will be just that:
2802		relative.
2803
2804		The next step (version after next version?) will be to make the
2805		default for CM.Control.implicit_anchors false.  After the dust has
2806		settled, I can then produce the "final" version of this...
2807
2808		Note: Since bootstrapping is a bit tricky, I provided new boot files.
2809
2810		----------------------------------------------------------------------
2811		Name: Matthias Blume
2812		Date: 2000/05/11 16:30:00 JST
2813		Tag: blume-20000511-sources
2814		Description:
2815
2816		The main change is that I added function CM.sources as a generalized
2817		version of the earlier CM.makedepend.  This entails the following
2818		additional changes:
2819
2820		- CM.makedepend has been dropped.
2821
2822		- CM manual has been updated.
2823
2824		- TOOLS signature and API have been changed.
2825
2826		----------------------------------------------------------------------
2827		Name: Allen Leung
2828		Date: 2000/05/10 21:17 EDT
2829		Tag: leunga-20000510-moby-c--ssa
2830		Description:
2831
2832		Various bug fixes and new features for C--, Moby and MLRISC optimizations.
2833		None of these affect SML/NJ.
2834
2835		1. Register Allocation
2836
2837		a. A new ra spilling module (ra/ra-spill-with-renaming) is implemented.
2838		This module tries to remove local (i.e. basic block level) redundancies
2839		during spilling.
2840
2841		b. A new framework for performing region based register allocation.
2842		Not yet entirely functional.
2843
2844		2. X86
2845
2846		a. DefUse for POP was missing the stack pointer [found by Lal]
2847		b. Reload for CALL was incorrect in X86Spill [found by John]
2848		c. Various fixes in X86Spill so that it can be used correctly for
2849		the new spilling module.
2850
2851		3. SSA/IR
2852
2853		a. New module ir/dj-dataflow.sml implements elimination based
2854		data flow analysis.
2855
2856		4. MLRiscGen
2857
2858		a. Fix for gc type annotation
2859
2860		5. MDGen
2861
2862		Various fixes for machine description -> ml code translation.  For ssa
2863		only.
2864
2865		----------------------------------------------------------------------
2866		Name: Allen Leung
2867		Date: 2000/05/08 22:17 EDT
2868		Tag: leunga-20000508-labexp
2869		Description:
2870
2871		Fermin has found a few assembly problems with constant expressions
2872		generated in LabelExp.  Mostly, the problems involve extra parentheses,
2873		which choke on dumb assemblers.  This is his fix.
2874
2875		----------------------------------------------------------------------
2876		Name: Dave MacQueen
2877		Date: 2000/04/09 14:00 EDT
2878		Tag: dbm-20000502-Version_110_28
2879		Description:
2880
2881		1. Updated src/compiler/TopLevel/main/version.sml to version 110.28
2882
2883		2. Updated config/version to 110.28
2884
2885		3. Updated config/srcarchiveurl
2886
2887		3. New boot files!
2888		ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.28/
2889
2890		----------------------------------------------------------------------
2891		Name: Matthias Blume
2892		Date: 2000/05/01 19:05:00 JST
2893		Tag: blume-20000501-noweb
2894		Description:
2895
2896		A new noweb tool has been added.  The existing system is entirely
2897		unaffected by this, but some CM users have asked for renewed noweb
2898		support.  Everything is documented in the CM manual.
2899
2900		New (plugin) libraries:
2901
2902		noweb-tool.cm
2903		nw-ext.cm
2904
2905		----------------------------------------------------------------------
2906		Name: Dave MacQueen
2907		Date: 2000/04/30 12:40PM EDT
2908		Tag: dbm-20000430-bug_fixes
2909		Description:
2910
2911		1. Fix for bug 1498
2912		smlnj/src/system/Basis/Implementation/Unsafe/object.sig
2913		smlnj/src/system/Basis/Implementation/Unsafe/object.sml
2914		added toRealArray function
2915		smlnj/src/compiler/MiscUtil/print/ppobj.sml
2916		added check for tag Obj.RealArray to array printing case in ppObj
2917
2918		2. Fix for bug 1510
2919		smlnj/src/compiler/Semant/types/typesutil.sml
2920		fixed definition of dummyargs (used by equalTycon) so that
2921		dummy args are distinct types
2922
2923		----------------------------------------------------------------------
2924		Name: Matthias Blume
2925		Date: 2000/04/30 01:00:00 JST
2926		Tag: blume-20000430-versions
2927		Description:
2928
2929		1. CM version numbering added.  This is an implementation of Lal's
2930		proposal for adding version numbers and version checking to .cm
2931		files.  Lal said that his proposal was just that -- a proposal.
2932		For the time being I went ahead and implemented it so that people
2933		can comment on it.  Everything is completely backward-compatible
2934		(except for the stable library format, i.e., new bootfiles!).
2935
2936		As usual, see the CM manual for details.
2937
2938		2. An alternative syntax for anchored paths has been implemented.
2939		Dave has recently voiced the same concerns that I had when I did
2940		this, so there should be some support.  My take is that eventually
2941		I will let support for the current syntax (where anchors are
2942		"implicit") fade out in favor of the new, explicit syntax.
2943		In order to be backward-compatible, both old and new syntax are
2944		currently supported.
2945
2946		Again, see the CM manual for details.
2947
2948		3. Parallel make is trying to be slightly smarter:  When the master
2949		process finds a "bottleneck", i.e., when there is only one
2950		compilation unit that can be compiled and everybody else is
2951		waiting on it, then it will simply compile it directly instead
2952		of clumsily telling one of the slaves to do it.
2953
2954		4. Support for "unsharing" added.  This is necessary in order to be
2955		able to have two different versions of the same library running
2956		at the same time (e.g., for trying out a new MLRISC while still
2957		having the old MLRISC linked into the current compiler, etc.)
2958		See the CM manual.
2959
2960		5. Simple "makedepend" functionality added for generating Makefile
2961		dependency information.  (This is rather crude at the moment.
2962		Expect some changes here in the future.)
2963
2964		6. ".fun" added as a recognized suffix for ML files. Also documented
2965		explicitly in the manual that the fallback behavior (unknown suffix
2966		-> ML file) is not an official feature!
2967
2968		7. Small changes to the pickler for stable libraries.
2969
2970		8. Several internal changes to CM (for cleanup/improvement).
2971
2972
2973		!!!! NEW BINFILES !!!!
2974
2975		----------------------------------------------------------------------
2976		Name: Matthias Blume
2977		Date: 2000/04/28 17:30:00 JST
2978		Tag: blume-20000428-pathconfig
2979		Description:
2980
2981		1. I changed config/install.sh to remove duplicate entries from the
2982		lib/pathconfig file at the end.  Moreover, the final version of
2983		lib/pathconfig is sorted alphabetically.  The same (sorting) is done
2984		in src/system/installml.
2985
2986		2. The config/install.sh script now consistently uses relative
2987		pathnames in lib/pathconfig whenever the anchor is in the lib
2988		directory.  (So far this was true for the libraries that come
2989		pre-compiled and bundled as part of the bootfiles but not for
2990		libraries that are compiled by the script itself.)
2991
2992		----------------------------------------------------------------------
2993		Name: Matthias Blume
2994		Date: 2000/04/26 13:10:00 JST
2995		Tag: blume-20000426-fun_suffix
2996		Description:
2997
2998		Added ".fun" as a recognized file name suffix (for ML code).
2999
3000		----------------------------------------------------------------------
3001		Name: Allen Leung
3002		Date: 2000/04/25 17:00:00 EST
3003		Tag: leunga-20000425-alpha-ra
3004		Description:
3005
3006		1. Alpha
3007
3008		PSEUDOARITH was missing in AlphaRewrite.  This causes an endless loop
3009		in C--.
3010
3011		2. RA
3012
3013		Added a flag "ra-dump-size" to print out the size of the flowgraph
3014		and the interference graph.
3015
3016		----------------------------------------------------------------------
3017		Name: Dave MacQueen
3018		Date: 2000/04/25/
3019		Tag: dbm-20000425-mlyacc_doc_examples
3020		Description:
3021		Updated mlyacc.tex sections 5 and 7 for SML '97 and CM.
3022		Updated all three examples in src/ml-yacc/examples to run
3023		under 110.* using CM.make.
3024
3025		----------------------------------------------------------------------
3026		Name: Allen Leung
3027		Date: 2000/04/20 23:04:00 EST
3028		Tag: leunga-20000420-ssa-c---stuff
3029		Description:
3030
3031		This update synchronizes my repository with Yale's.  Most of these
3032		changes, however, do not affect SML/NJ at all (the RA is an exception).
3033
3034		1. Register Allocator
3035
3036		a. An improvement in the interference graph construction:
3037		Given a copy
3038
3039		s <- t
3040
3041		no interference edge between s and t is added for this definition of s.
3042
3043		b. I've added two new spill heuristic modules that Fermin and I developed
3044		(in the new library RA.cm). These are unused in SML/NJ but maybe
3045		useful for others (Moby?)
3046
3047		2. X86
3048
3049		a. Various fixes in the backend provided by Fermin [C--] and Lal.
3050
3051		3. Alpha
3052
3053		a. Added the BSR instruction and code generation that goes with it [C--]
3054		b. Other fixes too numerous to recount provided by Fermin [C--]
3055
3056		4. Regmaps
3057
3058		a. The regmaps are not initialized with the identity physical bindings
3059		at creation time.  This is unneeded.
3060
3061		5. MLRISC Optimizations
3062
3063		a. The DJ-Graph module can now compute the iterated dominance frontiers
3064		intersects with liveness incrementally in linear time! Woohoo!
3065		This is now used in my new SSA construction algorithm.
3066
3067		b. THe branch reorganization module is now smarter about linear chains of
3068		basic blocks.
3069
3070
3071		----------------------------------------------------------------------
3072		Name: Matthias Blume
3073		Date: 2000/04/12 13:52:00 JST
3074		Tag: blume_main_v110p27_1
3075		Description:
3076
3077		Changed install.sh script to handle archive files without version number
3078		and to use "boot.<arch>-<os>" instead of "sml.boot.<arch>-<os>" for the
3079		name of the boot file archive.
3080
3081		----------------------------------------------------------------------
3082		Name: Dave MacQueen
3083		Date: 2000/04/09 14:00 EDT
3084		Tag: dbm-20000410-Version_110_27
3085		Description:
3086
3087		1. Updated src/compiler/TopLevel/main/version.sml to version 110.27
3088
3089		2. Updated src/config/version to 110.27
3090
3091		3. New boot files!
3092
3093		----------------------------------------------------------------------
3094		Name: Allen Leung
3095		Date: 2000/04/09 19:09:00 EST
3096		Tag: leunga-20000409-misc
3097		Description:
3098
3099		1.  Yet another fix for x86 assembly for idivl, imull, mull and friends.
3100
3101		2.  Miscellaneous improvements to MLRISC (unused in sml/nj)
3102
3103		----------------------------------------------------------------------
3104		Name: Stefan
3105		Date: 2000/04/07 10:00:00 EDT
3106		Tag: monnier-20000406-branch-handling
3107		Description:
3108
3109		Improved handling of branches (mostly those generated from
3110		polymorphic equality), removed switchoff and changed the
3111		default optimization settings (more cpsopt and less flintopt).
3112
3113		----------------------------------------------------------------------
3114		Name: Allen Leung
3115		Date: 2000/04/06 01:30:00 EST
3116		Tag: leunga-20000406-peephole-x86-SSA-2
3117		Description:
3118
3119		Forgot a few files.
3120
3121		----------------------------------------------------------------------
3122		Name: Allen Leung
3123		Date: 2000/04/06 00:36:00 EST
3124		Tag: leunga-20000406-peephole-x86-SSA
3125		Description:
3126
3127		1.  New Peephole code
3128
3129		2.  Minor improvement to X86 instruction selection
3130
3131		3.  Various fixes to SSA and machine description -> code translator
3132
3133		----------------------------------------------------------------------
3134		Name: Matthias Blume
3135		Date: 2000/04/05 12:30:00 JST
3136		Tag: blume_main_v110p26p2_3
3137		Description:
3138
3139		This update just merges three minor cosmetic updates to CM's sources
3140		to get ready for the 110.27 code freeze on Friday.  No functionality
3141		has changed.
3142
3143		----------------------------------------------------------------------
3144		Name: Allen Leung
3145		Date: 2000/04/04 19:39:00 EST
3146		Tag: leunga-20000404-x86-asm
3147		Description:
3148
3149		1.  Fixed a problem in X86 assembly.
3150
3151		Things like
3152
3153		jmp %eax
3154		jmp (%eax)
3155
3156		should be output as
3157
3158		jmp *%eax
3159		jmp *(%eax)
3160
3161		2.  Assembly output
3162
3163		Added a new flag
3164
3165		"asm-indent-copies" (default to false)
3166
3167		When this flag is on, parallel copies will be indented an extra level.
3168
3169		----------------------------------------------------------------------
3170		Name: Allen Leung
3171		Date: 2000/04/04 03:18:00 EST
3172		Tag: leunga-20000404-C--Moby
3173		Description:
3174
3175		All of these fixes are related to C--, Moby, and my own optimization
3176		stuff; so they shouldn't affect SML/NJ.
3177
3178		1.  X86
3179
3180		Various fixes related floating point, and extensions.
3181
3182		2.  Alpha
3183
3184		Some extra patterns related to loads with signed/zero extension
3185		provided by Fermin.
3186
3187		3.  Assembly
3188
3189		When generating assembly, resolve the value of client defined constants,
3190		instead of generating symbolic values.  This is controlled by the
3191		new flag "asm-resolve-constants", which is default to true.
3192
3193		4.  Machine Descriptions
3194
3195		a. The precedence parser was slightly broken when parsing infixr symbols.
3196		b. The type generalizing code had the bound variables reversed, resulting
3197		in a problem during arity raising.
3198		c. Various fixes in machine descriptions.
3199
3200		----------------------------------------------------------------------
3201		Name: Matthias Blume
3202		Date: 2000/04/03 16:05:00 JST
3203		Tag: blume_main_v110p26p2_2
3204		Description:
3205
3206		I eliminated coreEnv from compInfo.  Access to the "Core" structure is
3207		now done via the ordinary static environment that is context to each
3208		compilation unit.
3209
3210		To this end, I arranged that instead of "structure Core" as "structure
3211		_Core" is bound in the pervasive environment.  Core access is done via
3212		_Core (which can never be accidentally rebound because _Core is not a
3213		legal surface-syntax symbol).
3214
3215		The current solution is much cleaner because the core environment is
3216		now simply part of the pervasive environment which is part of every
3217		compilation unit's context anyway.  In particular, this eliminates all
3218		special-case handling that was necessary until now in order to deal
3219		with dynamic and symbolic parts of the core environment.
3220
3221		Remaining hackery (to bind the "magic" symbol _Core) is localized in the
3222		compilation manager's bootstrap compiler (actually: in the "init group"
3223		handling).  See the comments in src/system/smlnj/init/init.cmi for
3224		more details.
3225
3226		I also tried to track down all mentions of "Core" (as string argument
3227		to Symbol.strSymbol) in the compiler and replaced them with a
3228		reference to the new CoreSym.coreSym.  Seems cleaner since the actual
3229		name appears in one place only.
3230
3231		Binfile and bootfile format have not changed, but the switchover from
3232		the old "init.cmi" to the new one is a bit tricky, so I supplied new
3233		bootfiles anyway.
3234
3235		----------------------------------------------------------------------
3236		Name: Allen Leung
3237		Date: 2000/04/02 21:17:00 EST
3238		Tag: leunga-20000402-mltree
3239		Description:
3240
3241		1. Renamed the constructor CALL in MLTREE by popular demand.
3242		2. Added a bunch of files from my repository.  These are currently
3243		used by other non-SMLNJ backends.
3244
3245		----------------------------------------------------------------------
3246		Name: Allen Leung
3247		Date: 2000/03/31 21:15:00 EST
3248		Tag: leunga-20000331-aliasing
3249		Description:
3250
3251		This update contains a rewritten (and hopefully more correct) module
3252		for extracting aliasing information from CPS.
3253
3254		To turn on this feature:
3255
3256		Compiler.Control.CG.memDisambiguate := true
3257
3258		To pretty print the region information with assembly
3259
3260		Compiler.Control.MLRISC.getFlag "asm-show-region" := true;
3261
3262		To control how many levels of aliasing information are printed, use:
3263
3264		Compiler.Control.MLRISC.getInt "points-to-show-level" := n
3265
3266		The default of n is 3.
3267
3268		----------------------------------------------------------------------
3269		Name: David MacQueen
3270		Date: 2000/03/31 11:15:00 EST
3271		Tag: dbm-20000331-runtime_fix
3272		Description:
3273
3274		This update contains:
3275
3276		1. runtime/c-lib/c-libraries.c
3277		includes added in revision 1.2 caused compilation errors on hppa-hpux
3278
3279		2. fix for bug 1556
3280		system/Basis/Implementation/NJ/internal-signals.sml
3281
3282		----------------------------------------------------------------------
3283		Name: Matthias Blume
3284		Date: 2000/03/31 18:00:00 JST
3285		Tag: blume_main_v110p26p2_1
3286		Description:
3287
3288		This update contains:
3289
3290		1. A small change to CM's handling of stable libraries:
3291		CM now maintains one "global" modmap that is used for all stable
3292		libraries.  The use of such a global modmap maximizes sharing and
3293		minimizes the need for re-traversing parts of environments during
3294		modmap construction.  (However, this has minor impact since modmap
3295		construction seems to account for just one percent or less of total
3296		compile time.)
3297
3298		2. I added a "genmap" phase to the statistics.  This is where I got the
3299		"one percent" number (see above).
3300
3301		3. CM's new tool parameter mechanism just became _even_ better. :)
3302		- The parser understands named parameters and recursive options.
3303		- The "make" and "shell" tools use these new features.
3304		(This makes it a lot easier to cascade these tools.)
3305		- There is a small syntax change: named parameters use a
3306
3307		<name> : ( <option> ... )            or
3308		<name> : <string>
3309
3310		syntax.  Previously, named parameters were implemented in an
3311		ad-hoc fashion by each tool individually (by parsing strings)
3312		and had the form
3313
3314		<name>=<string>
3315
3316		See the CM manual for a full description of these issues.
3317
3318		----------------------------------------------------------------------
3319		Name: Matthias Blume
3320		Date: 2000/03/30 18:00:00 JST
3321		Tag: blume_main_v110p26p2_0
3322		Description:
3323
3324		!!!!! WARNING !!!!!!
3325		!!  New binfiles  !!
3326		!!!!!!!!!!!!!!!!!!!!
3327
3328		This update contains:
3329
3330		1. Moderate changes to CM:
3331
3332		- Changes to CM's tools mechanism.  In particular, it is now possible
3333		to have tools that accept additional "command line" parameters
3334		(specified in the .cm file at each instance where the tool's class is
3335		used).
3336
3337		This was done to accommodate the new "make" and "shell" tools which
3338		facilitate fairly seamless hookup to portions of code managed using
3339		Makefiles or Shell scripts.
3340
3341		There are no classes "shared" or "private" anymore.  Instead, the
3342		sharing annotation is now a parameter to the "sml" class.
3343
3344		There is a bit of generic machinery for implementing one's own
3345		tools that accept command-line parameters.  However, I am not yet fully
3346		satisfied with that part, so expect changes here in the future.
3347
3348		All existing tools are described in the CM manual.
3349
3350		- Slightly better error handling.  (CM now suppresses many followup
3351		error messages that tended to be more annoying than helpful.)
3352
3353		2. Major changes to the compiler's static environment data structures.
3354
3355		- no CMStaticEnv anymore.
3356		- no CMEnv, no "BareEnvironment" (actually, _only_ BareEnvironment,
3357		but it is called Environment), no conversions between different
3358		kinds of static environments
3359
3360		- There is still a notion of a "modmap", but such modmaps are generated
3361		on demand at the time when they are needed.  This sounds slow, but I
3362		sped up the code that generates modmaps enough for this not to lead to
3363		a slowdown of the compiler (at least I didn't detect any).
3364
3365		- To facilitate rapid modmap generation, static environments now
3366		contain an (optional) "modtree" structure.  Modtree annotations are
3367		constructed by the unpickler during unpickling.  (This means that
3368		the elaborator does not have to worry about modtrees at all.)
3369		Modtrees have the advantage that they are compositional in the same
3370		way as the environment data structure itself is compositional.
3371		As a result, modtrees never hang on to parts of an environment that
3372		has already been rendered "stale" by filtering or rebinding.
3373
3374		- I went through many, many trials and errors before arriving at the
3375		current solution.  (The initial idea of "linkpaths" did not work.)
3376		But the result of all this is that I have touched a lot of files that
3377		depend on the "modules" and "types" data structures (most of the
3378		elaborator). There were a lot of changes during my "linkpath" trials
3379		that could have been reverted to their original state but weren't.
3380		Please, don't be too harsh on me for messing with this code a bit more
3381		than what was strictly necessary...  (I _did_ resist the temptation
3382		of doing any "global reformatting" to avoid an untimely death at
3383		Dave's hands. :)
3384
3385		- One positive aspect of the previous point:  At least I made sure that
3386		all files that I touched now compile without warnings (other than
3387		"polyEqual").
3388
3389		- compiler now tends to run "leaner" (i.e., ties up less memory in
3390		redundant modmaps)
3391
3392		----------------------------------------------------------------------
3393		Name: Allen Leung
3394		Date: 2000/03/29 18:00:00
3395		Tag: leunga-20000327-mlriscGen_hppa_alpha_x86
3396		Boot files (optional): ftp://react-ilp.cs.nyu.edu/leunga/110.26.1-sml.boot.x86-unix-20000330.tar.gz
3397		Description:
3398
3399		This update contains *MAJOR* changes to the way code is generated from CPS
3400		in the module mlriscGen, and in various backend modules.
3401
3402		CHANGES
3403		=======
3404
3405		1. MLRiscGen: forward propagation fix.
3406
3407		There was a bug in forward propagation introduced at about the same time
3408		as the MLRISC x86 backend, which prohibits coalescing to be
3409		performed effectively in loops.
3410
3411		Effect: speed up of loops in RISC architectures.
3412		By itself, this actually slowed down certain benchmarks on the x86.
3413
3414		2. MLRiscGen:  forward propagating addresses from consing.
3415
3416		I've changed the way consing code is generated.  Basically I separated
3417		out the initialization part:
3418
3419		store tag,   offset(allocptr)
3420		store elem1, offset+4(allocptr)
3421		store elem2, offset+8(allocptr)
3422		...
3423		store elemn, offset+4n(allocptr)
3424
3425		and the address computation part:
3426
3427		celladdr <- offset+4+alloctpr
3428
3429		and move the address computation part
3430
3431		Effect:  register pressure is generally lower as a result.  This
3432		makes compilation of certain expressions much faster, such as
3433		long lists with non-trivial elements.
3434
3435		[(0,0), (0,0), .... (0,0)]
3436
3437		3. MLRiscGen: base pointer elimination.
3438
3439		As part of the linkage mechanism, we generate the sequence:
3440
3441		L:  ...  <- start of the code fragment
3442
3443		L1:
3444		base pointer <- linkreg - L1 + L
3445
3446		The base pointer was then used for computing relocatable addresses
3447		in the code fragment.  Frequently (such as in lots of continuations)
3448		this is not needed.  We now eliminate this sequence whenever possible.
3449
3450		For compile time efficiency, I'm using a very stupid local heuristic.
3451		But in general, this should be done as a control flow analysis.
3452
3453		Effect:  Smaller code size.  Speed up of most programs.
3454
3455		4. Hppa back end
3456
3457		Long jumps in span dependence resolution used to depend on the existence
3458		of the base pointer.
3459
3460		A jump to a long label L was expanded into the following sequence:
3461
3462		LDIL %hi(L-8192), %r29
3463		LDO  %lo(L-8192)(%r29), %r29
3464		ADD  %r29, baseptr, %r29
3465		BV,n %r0(%r29)
3466
3467		In the presence of change (3) above, this will not work.  I've changed
3468		it so that the following sequence of instructions are generated, which
3469		doesn't mention the base pointer at all:
3470
3471		BL,n  L', %r29           /* branch and link, L' + 4 -> %r29 */
3472		L':  ADDIL L-(L'+4), %r29     /* Compute address of L */
3473		BV,n  %r0(%r29)          /* Jump */
3474
3475		5. Alpha back end
3476
3477		New alpha instructions LDB/LDW have been added, as per Fermin's
3478		suggestions.   This is unrelated to all other changes.
3479
3480		6. X86 back end
3481
3482		I've changed andl to testl in the floating point test sequence
3483		whenever appropriate.  The Intel optimization guide states that
3484		testl is preferable to andl.
3485
3486		7. RA (x86 only)
3487
3488		I've improved the spill propagation algorithm, using an approximation
3489		of maximal weighted independent sets.   This seems to be necessary to
3490		alleviate the negative effect in light of the slow down in (1).
3491
3492		I'll write down the algorithm one of these days.
3493
3494		8. MLRiscGen: frequencies
3495
3496		I've added an annotation that states that all call gc blocks have zero
3497		execution frequencies.  This improves register allocation on the x86.
3498
3499		BENCHMARKS
3500		==========
3501
3502		I've only perform the comparison on 110.25.
3503
3504		The platforms are:
3505
3506		HPPA  A four processor HP machine (E9000) with 5G of memory.
3507		X86   A 300Hhz Pentium II with 128M of memory, and
3508		SPARC An Ultra sparc 2 with 512M of memory.
3509
3510		I used the following parameters for the SML benchmarks:
3511
3512		@SMLalloc
3513		HPPA    256k
3514		SPARC   512k
3515		X86     256k
3516
3517		COMPILATION TIME
3518		----------------
3519		Here are the numbers comparing the compilation times of the compilers.
3520		I've only compared 110.25 compiling the new sources versus
3521		a fixpoint version of the new compiler compiling the same.
3522
3523		110.25                                  New
3524		Total  Time in RA  Spill+Reload   Total  Time In RA Spill+Reload
3525		HPPA   627s    116s        2684+3584     599s    95s       1003+1879
3526		SPARC  892s    173s        2891+3870     708s    116s      1004+1880
3527		X86    999s    315s       94006+130691   987s    296s    108877+141957
3528
3529		110.25         New
3530		Code Size      Code Size
3531		HPPA   8596736         8561421
3532		SPARC  8974299         8785143
3533		X86    9029180         8716783
3534
3535		So in summary, things are at least as good as before.   Dramatic
3536		reduction in compilation is obtained on the Sparc; I can't explain it,
3537		but it is reproducible.  Perhaps someone should try to reproduce this
3538		on their own machines.
3539
3540		SML BENCHMARKS
3541		--------------
3542
3543		On the average, all benchmarks perform at least as well as before.
3544
3545		HPPA         Compilation Time     Spill+Reload      Run Time
3546		110.25  New            110.25    New   110.25  New
3547
3548		barnesHut  3.158  3.015  4.75%    1+1       0+0   2.980  2.922   2.00%
3549		boyer  6.152  5.708  7.77%    0+0       0+0   0.218  0.213   2.34%
3550		count-graphs  1.168  1.120  4.32%    0+0       0+0  22.705 23.073  -1.60%
3551		fft  0.877  0.792 10.74%    1+3       1+3   0.602  0.587   2.56%
3552		knuthBendix  3.180  2.857 11.32%    0+0       0+0   0.675  0.662   2.02%
3553		lexgen  6.190  5.290 17.01%    0+0       0+0   0.913  0.788  15.86%
3554		life  0.803  0.703 14.22%   25+25      0+0   0.153  0.140   9.52%
3555		logic  2.048  2.007  2.08%    6+6       1+1   4.133  4.008   3.12%
3556		mandelbrot  0.077  0.080 -4.17%    0+0       0+0   0.765  0.712   7.49%
3557		mlyacc 22.932 20.937  9.53%  154+181    32+57  0.468  0.430   8.91%
3558		nucleic  5.183  5.060  2.44%    2+2       0+0   0.125  0.120   4.17%
3559		ratio-regions  3.357  3.142  6.84%    0+0       0+0  116.225 113.173 2.70%
3560		ray  1.283  1.290 -0.52%    0+0       0+0   2.887  2.855   1.11%
3561		simple  6.307  6.032  4.56%   28+30      5+7   3.705  3.658   1.28%
3562		tsp  0.888  0.862  3.09%    0+0       0+0   7.040  6.893   2.13%
3563		vliw 24.378 23.455  3.94%  106+127    25+45  2.758  2.707   1.91%
3564		--------------------------------------------------------------------------
3565		Average                     6.12%                                   4.09%
3566
3567		SPARC        Compilation Time     Spill+Reload      Run Time
3568		110.25  New            110.25    New   110.25  New
3569
3570		barnesHut  3.778  3.592  5.20%    2+2       0+0   3.648  3.453    5.65%
3571		boyer  6.632  6.110  8.54%    0+0       0+0   0.258  0.242    6.90%
3572		count-graphs  1.435  1.325  8.30%    0+0       0+0  33.672 34.737   -3.07%
3573		fft  0.980  0.940  4.26%    3+9       2+6   0.838  0.827    1.41%
3574		knuthBendix  3.590  3.138 14.39%    0+0       0+0   0.962  0.967   -0.52%
3575		lexgen  6.593  6.072  8.59%    1+1       0+0   1.077  1.078   -0.15%
3576		life  0.972  0.868 11.90%   26+26      0+0   0.143  0.140    2.38%
3577		logic  2.525  2.387  5.80%    7+7       1+1   5.625  5.158    9.05%
3578		mandelbrot  0.090  0.093 -3.57%    0+0       0+0   0.855  0.728   17.39%
3579		mlyacc 26.732 23.827 12.19%  162+189    32+57  0.550  0.560   -1.79%
3580		nucleic  6.233  6.197  0.59%    3+3       0+0   0.163  0.173   -5.77%
3581		ratio-regions  3.780  3.507  7.79%    0+0       0+0 133.993 131.035   2.26%
3582		ray  1.595  1.550  2.90%    1+1       0+0   3.440  3.418    0.63%
3583		simple  6.972  6.487  7.48%   29+32      5+7   3.523  3.525   -0.05%
3584		tsp  1.115  1.063  4.86%    0+0       0+0   7.393  7.265    1.77%
3585		vliw 27.765 24.818 11.87%  110+135    25+45  2.265  2.135    6.09%
3586		----------------------------------------------------------------------------
3587		Average                     6.94%                                    2.64%
3588
3589		X86          Compilation Time     Spill+Reload      Run Time
3590		110.25  New            110.25    New   110.25  New
3591
3592		barnesHut  5.530  5.420  2.03%  593+893   597+915   3.532  3.440   2.66%
3593		boyer  8.768  7.747 13.19%  493+199   301+289   0.327  0.297  10.11%
3594		count-graphs  2.040  2.010  1.49%  298+394   315+457  26.578 28.660  -7.26%
3595		fft  1.327  1.302  1.92%  112+209   115+210   1.055  0.962   9.71%
3596		knuthBendix  5.218  5.475 -4.69%  451+598   510+650   0.928  0.932  -0.36%
3597		lexgen  9.970  9.623  3.60% 1014+841  1157+885   0.947  0.928   1.97%
3598		life  1.183  1.183  0.00%  162+182   145+148   0.127  0.103  22.58%
3599		logic  3.285  3.512 -6.45%  514+684   591+836   5.682  5.577   1.88%
3600		mandelbrot  0.147  0.143  2.33%   38+41     33+54    0.703  0.690   1.93%
3601		mlyacc 35.457 32.763  8.22% 3496+4564 3611+4860  0.552  0.550   0.30%
3602		nucleic  7.100  6.888  3.07%  239+168   201+158   0.175  0.173   0.96%
3603		ratio-regions  6.388  6.843 -6.65% 1182+257   981+300  120.142 120.345 -0.17%
3604		ray  2.332  2.338 -0.29%  346+398   402+494   3.593  3.540   1.51%
3605		simple  9.912  9.903  0.08% 1475+941  1579+1168  3.057  3.178  -3.83%
3606		tsp  1.623  1.532  5.98%  266+200   250+211   8.045  7.878   2.12%
3607		vliw 33.947 35.470 -4.29% 2629+2774 2877+3171  2.072  1.890   9.61%
3608		----------------------------------------------------------------------------
3609		Average                     1.22%                                     3.36%
3610