Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 701, Thu Aug 31 08:46:43 2000 UTC	revision 1009, Wed Jan 9 19:44:22 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: 2001/01/09 14:31:35 EST 2002
17		Tag: george-20011206-rm-native-copy
18		Description:
19
20		Removed the native COPY and FCOPY instructions
21		from all the architectures and replaced it with the
22		explicit COPY instruction from the previous commit.
23
24		It is now possible to simplify many of the optimizations
25		modules that manipulate copies. This has not been
26		done in this change.
27
28		----------------------------------------------------------------------
29		Name: Lal George
30		Date: 2001/12/06 16:50:13 EST 2001
31		Tag: george-20011206-mlrisc-instruction
32		Description:
33
34		Changed the representation of instructions from being fully abstract
35		to being partially concrete. That is to say:
36
37		from
38		type instruction
39
40		to
41		type instr                              (* machine instruction *)
42
43		datatype instruction =
44		LIVE of {regs: C.cellset, spilled: C.cellset}
45		| KILL of {regs: C.cellset, spilled: C.cellset}
46		| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
47		| ANNOTATION of {i: instruction, a: Annotations.annotation}
48		| INSTR of instr
49
50		This makes the handling of certain special instructions that appear on
51		all architectures easier and uniform.
52
53		LIVE and KILL say that a list of registers are live or killed at the
54		program point where they appear. No spill code is generated when an
55		element of the 'regs' field is spilled, but the register is moved to
56		the 'spilled' (which is present, more for debugging than anything else).
57
58		LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
59		We used to generate:
60
61		DEFFREG f1
62		f1 := f2 + f3
63		trapb
64
65		but now generate:
66
67		f1 := f2 + f3
68		trapb
69		LIVE {regs=[f1,f2,f3], spilled=[]}
70
71		Furthermore, the DEFFREG (hack) required that all floating point instruction
72		use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
73		defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
74		in a cleaner alpha implementation. (Hopefully, intel will not get rid of
75		this architecture).
76
77		COPYXXX is intended to replace the parallel COPY and FCOPY  available on
78		all the architectures. This will result in further simplification of the
79		register allocator that must be aware of them for coalescing purposes, and
80		will also simplify certain aspects of the machine description that provides
81		callbacks related to parallel copies.
82
83		ANNOTATION should be obvious, and now INSTR represents the honest to God
84		machine instruction set!
85
86		The <arch>/instructions/<arch>Instr.sml files define certain utility
87		functions for making porting easier -- essentially converting upper case
88		to lower case. All machine instructions (of type instr) are in upper case,
89		and the lower case form generates an MLRISC instruction. For example on
90		the alpha we have:
91
92		datatype instr =
93		LDA of {r:cell, b:cell, d:operand}
94		| ...
95
96		val lda : {r:cell, b:cell, d:operand} -> instruction
97		...
98
99		where lda is just (INSTR o LDA), etc.
100
101		----------------------------------------------------------------------
102		Name: Matthias Blume
103		Date: 2001/11/22 21:40:00 EST
104		Tag: Release_110_37
105		Description:
106
107		Release 110.37.  This time for real.
108
109		----------------------------------------------------------------------
110		Name: Matthias Blume
111		Date: 2001/11/21 16:35:00 EST
112		Tag: blume-20011121-foot-in-mouth
113		Description:
114
115		Removed the "Release_110_37" tag because of a serious bug.
116		This will be re-tagged once the bug is fixed.
117
118		----------------------------------------------------------------------
119		Name: Matthias Blume
120		Date: 2001/11/21 16:14:00 EST
121		Tag: blume-20011121-forgottenfile
122		Description:
123
124		Forgot to add a file.  (Just a .tex-file -- part of
125		the CM manual source.)
126
127		----------------------------------------------------------------------
128		Name: Matthias Blume
129		Date: 2001/11/21 16:10:00 EST
130		Tag: blume-20011121-invalid_110_37
131		Description:
132
133		Note: I removed the original tag "Release_110_37" from this commit
134		because we found a serious bug in all non-x86 backends.
135		- Matthias
136
137		1. Modifications to the SML/NJ code generator and to the runtime system
138		so that code object name strings are directly inserted into code
139		objects at code generation time.  The only business the runtime system
140		has with this is now to read the name strings on occasions.
141		(The encoding of the name string has also changed somewhat.)
142
143		2. CM now implements a simple "set calculus" for specifying export lists.
144		In particular, it is now possible to refer to the export lists of
145		other libraries/groups/sources and form unions as well as differences.
146		See the latest CM manual for details.
147
148		3. An separate notion of "proxy" libraries has again be eliminated from
149		CM's model.  (Proxy libraries are now simply a special case of using
150		the export list calculus.)
151
152		4. Some of the existing libraries now take advantage of the new set
153		calculus.
154		(Notice that not all libraries have been converted because some
155		of the existing .cm-files are supposed to be backward compatible
156		with 110.0.x.)
157
158		5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
159		"General.exnMessage"!  The former relies on a certain hook to be
160		initialized, and that often does not happen in the stand-alone case.)
161
162		----------------------------------------------------------------------
163		Name: Lal George
164		Date: 2001/11/21  13:56:18 EST
165		Tag: george-2001121-pseudo-ops
166		Description:
167
168		Implemented a complete redesign of MLRISC pseudo-ops. Now there
169		ought to never be any question of incompatabilities with
170		pseudo-op syntax expected by host assemblers.
171
172		For now, only modules supporting GAS syntax are implemented
173		but more should follow, such as MASM, and vendor assembler
174		syntax, e.g. IBM as, Sun as, etc.
175
176		----------------------------------------------------------------------
177		Name: Matthias Blume
178		Date: 2001/11/14 11:52:00 EST
179		Tag: blume-20011114-srcname
180		Description:
181
182		1. Routed the name of the current source file to mlriscgen where it
183		should be directly emitted into the code object.  (This last part
184		is yet to be done.)
185
186		2. Some cleanup of the pgraph code to make it match the proposal that
187		I put out the other day.  (The proposal notwithstanding, things are
188		still in flux here.)
189
190		----------------------------------------------------------------------
191		Name: Lal George
192		Date: 2001/11/14 09:44:04 EST
193		Tag:
194		Description:
195
196		Fix for a backpatching bug reported by Allen.
197
198		Because the boundary between short and long span-dependent
199		instructions is +/- 128, there are an astounding number of
200		span-dependent instructions whose size is over estimated.
201
202		Allen came up with the idea of letting the size of span
203		dependent instructions be non-monotonic, for a maxIter
204		number of times, after which the size must be monotonically
205		increasing.
206
207		This table shows the number of span-dependent instructions
208		whose size was over-estimated as a function of maxIter, for the
209		file Parse/parse/ml.grm.sml:
210
211		maxIter            # of instructions:
212		10                      687
213		20                      438
214		30                      198
215		40                        0
216
217		In compiling the compiler, there is no significant difference in
218		compilation speed between maxIter=10 and maxIter=40. Actually,
219		my measurements showed that maxIter=40 was a tad faster than
220		maxIter=10! Also 96% of the  files in the compiler reach a fix
221		point within 13 iterations, so fixing maxIter at 40, while high,
222		is okay.
223
224		----------------------------------------------------------------------
225		Name: Matthias Blume
226		Date: 2001/10/31 15:25:00 EST
227		Tag: blume-20011031-pgraph
228		Description:
229
230		CKIT:
231		* Changed the "Function" constructor of type Ast.ctype to carry optional
232		argument identifiers.
233		* Changed the return type of TypeUtil.getFunction accordingly.
234		* Type equality ignores the argument names.
235		* TypeUtil.composite tries to preserve argument names but gives up quickly
236		if there is a mismatch.
237
238		installation script:
239		* attempts to use "curl" if available (unless "wget" is available as well)
240
241		CM:
242		* has an experimental implementation of "portable graphs" which I will
243		soon propose as an implementation-independent library format
244		* there are also new libraries $/pgraph.cm and $/pgraph-util.cm
245
246		NLFFI-LIB:
247		* some cleanup (all cosmetic)
248
249		NLFFIGEN:
250		* temporarily disabled the mechanism that suppresses ML output for
251		C definitions whose identifiers start with an underscore character
252		* generate val bindings for enum constants
253		* user can request that only one style (light or heavy) is being used;
254		default is to use both (command-line arguments: -heavy and -light)
255		* fixed bug in handling of function types involving incomplete pointers
256		* generate ML entry points that take record arguments (i.e., using
257		named arguments) for C functions that have a prototype with named
258		arguments
259		(see changes to CKIT)
260
261		----------------------------------------------------------------------
262		Name: Allen Leung
263		Date: 2001/10/27 20:34:00 EDT
264		Tag: leunga-20011027-x86-fast-fp-call
265		Description:
266
267		Fixed the bug described in blume-20010920-slowfp.
268
269		The fix involves
270		1. generating FCOPYs in FSTP in ia32-svid
271		2. marking a CALL with the appropriate annotation
272
273		----------------------------------------------------------------------
274		Name: Matthias Blume
275		Date: 2001/10/16 11:32:00 EDT
276		Tag: blume-20011016-netbsd
277		Description:
278
279		Underscore patch from Chris Richards (fixing problem with compiling
280		runtime system under recent NetBSD).
281
282		----------------------------------------------------------------------
283		Name: Allen Leung
284		Date: 2001/10/12 17:18:32 EDT 2001
285		Tag: leung-20011012-x86-printflowgraph
286		Description:
287
288		X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
289
290		----------------------------------------------------------------------
291		Name: Lal George
292		Date: 2001/10/11 23:51:34 EDT
293		Tag: george-20011011-too-many-instrs
294		Description:
295
296		The representation of a program point never expected to see more
297		than 65536 instructions in a basic block!
298
299		----------------------------------------------------------------------
300		Name: Lal George
301		Date: 2001/10/09 09:41:37 EDT
302		Tag: george-20011008-mlrisc-labels
303		Description:
304
305		Changed the machine description files to support printing of
306		local and global labels in assembly code, based on host assembler
307		conventions.
308
309		----------------------------------------------------------------------
310		Name: Matthias Blume
311		Date: 2001/09/25 15:25:00 EDT
312		Tag: blume-20010925-exninfo
313		Description:
314
315		I provided a non-hook implementation of exnName (at the toplevel) and
316		made the "dummy" implementation of exnMessage (at the toplevel) more
317		useful: if nothing gets "hooked in", then at least you are going to
318		see the exception name and a message indicating why you don't see more.
319
320		[For the time being, programs that need exnMessage and want to use
321		ml-build should either use General.exnMessage (strongly recommended) or
322		refer to structure General at some other point so that CM sees a
323		static dependency.]
324
325		[Similar remarks go for "print" and "use":  If you want to use their
326		functionality in stand-alone programs generated by ml-build, then use
327		TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
328
329		----------------------------------------------------------------------
330		Name: Matthias Blume
331		Date: 2001/09/20 17:28:00 EDT
332		Tag: blume-20010920-slowfp
333		Description:
334
335		Allen says that x86-fast-fp is not safe yet, so I turned it off again...
336
337		----------------------------------------------------------------------
338		Name: Matthias Blume
339		Date: 2001/09/20 17:20:00 EDT
340		Tag: blume-20010920-canonicalpaths
341		Description:
342
343		0. Updated the BOOT file (something that I forgot to do earlier).
344
345		1. Small internal change to CM so that it avoids "/../" in filenames
346		as much as possible (but only where it is safe).
347
348		2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
349		that arguments that contain delimiters are passed through correctly.
350		This change also means that all "special" arguments of the form
351		@SMLxxx... must come first.
352
353		3. Changed install script to put relative anchor names for tool commands
354		into pathconfig.
355
356		----------------------------------------------------------------------
357		Name: Matthias Blume
358		>>>>>>> 1.169
359		Date: 2001/09/18 15:35:00 EDT
360		Tag: blume-20010918-readme11036
361		Description:
362
363		Added README files.
364
365		----------------------------------------------------------------------
366		Name: Matthias Blume
367		Date: 2001/09/18 11:45:00 EDT
368		Tag: Release_110_36 (retag)
369		Description:
370
371		Fixed mistake in config/preloads. Retagged as 110.36.
372
373		----------------------------------------------------------------------
374		Name: Matthias Blume
375		Date: 2001/09/18 09:40:00 EDT
376		Tag: Release_110_36_orig (tag changed)
377		Description:
378
379		New version (110.36).  New bootfiles.
380
381		----------------------------------------------------------------------
382		Name: Matthias Blume
383		Date: 2001/09/14 16:15:00 EDT
384		Tag: blume-20010914-x86fastfp
385		Description:
386
387		John committed some changes that Allen made, in particular a (hopefully)
388		correctly working version of the x86-fp module.
389
390		I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
391		flag to "true".  Everything seems to compile to a fixpoint ok, and
392		"mandelbrot" speeds up by about 15%.
393
394		----------------------------------------------------------------------
395		Name: Matthias Blume
396		Date: 2001/09/13 11:20:00 EDT
397		Tag: blume-20010913-minimal
398		Description:
399
400		1. Stefan Monnier's patch to fix a miscompilation problem that
401		was brought to light by John Reppy's work on Moby.
402
403		2. Implemented a minimal "structure Compiler" that contains just
404		"version" and "architecture".  The minimal version will be
405		available when the full version is not.  This is for backward-
406		compatibility with code that wants to test Compiler.version.
407
408		----------------------------------------------------------------------
409		Name: Matthias Blume
410		Date: 2001/08/28 14:03:00 EDT
411		Tag: blume-20010828-ml-lex
412		Description:
413
414		Fix for bug 1581, received from Neophytos Michael.
415
416		----------------------------------------------------------------------
417		Name: Matthias Blume
418		Date: 2001/08/27 11:20:00 EDT
419		Tag: blume-20010827-readme11035
420		Description:
421
422		Fleshed out the README file for 110.35.
423
424		----------------------------------------------------------------------
425		Name: Matthias Blume
426		Date: 2001/08/24 17:10:00 EDT
427		Tag: Release_110_35
428		Description:
429
430		New version number (110.35).  New bootfiles.
431
432		----------------------------------------------------------------------
433		Name: Lal George
434		Date: 2001/08/24 13:47:18 EDT 2001
435		Tag: george-20010824-MLRISC-graphs
436		Description:
437
438		removed clusters from MLRISC completely and replaced with graphs.
439
440		----------------------------------------------------------------------
441		Name: Matthias Blume
442		Date: 2001/08/23 17:50:00 EDT
443		Tag: blume-20010823-toplevel
444		Description:
445
446		- some reorganization of the code that implements various kinds of
447		environments in the compiler (static, dynamic, symbolic, combined)
448		- re-implemented the EnvRef module so that evalStream works properly
449		(if the stream contains references to "use", "CM.make", etc.)
450		- cleaned up evalloop.sml and interact.sml (but they need more cleaning)
451
452		----------------------------------------------------------------------
453		Name: Matthias Blume
454		Date: 2001/08/20 15:50 EDT
455		Tag: blume20010820-slipup
456		Description:
457
458		I forgot to commit a few files.  Here they are...
459
460		----------------------------------------------------------------------
461		Name: Matthias Blume
462		Date: 2001/08/20 15:35:00 EDT
463		Tag: blume-20010820-debugprof
464		Description:
465
466		!!!! NEW BOOTFILES !!!!
467
468		This is another round of reorganizing the compiler sources.  This
469		time the main goal was to factor out all the "instrumentation"
470		passes (for profiling and backtracing) into their own library.
471		The difficulty was to do it in such a way that it does not depend
472		on elaborate.cm but only on elabdata.cm.
473
474		Therefore there have been further changes to both elaborate.cm and
475		elabdata.cm -- more "generic" things have been moved from the former
476		to the latter.  As a result, I was forced to split the assignment
477		of numbers indicating "primtyc"s into two portions: SML-generic and
478		SML/NJ-specific.  Since it would have been awkward to maintain,
479		I bit the bullet and actually _changed_ the mapping between these
480		numbers and primtycs.  The bottom line of this is that you need
481		a new set of bin- and bootfiles.
482
483		I have built new bootfiles for all architectures, so doing a fresh
484		checkout and config/install.sh should be all you need.
485
486		The newly created library's name is
487
488		$smlnj/viscomp/debugprof.cm
489
490		and its sources live under
491
492		src/compiler/DebugProf
493
494		----------------------------------------------------------------------
495		Name: Matthias Blume
496		Date: 2001/08/15 17:15:00 EDT
497		Tag: blume-20010815-compreorg
498		Description:
499
500		This is a first cut at reorganizing the CM libraries that make up the
501		core of the compiler.  The idea is to separate out pieces that could
502		be used independently by tools, e.g., the parser, the typechecker, etc.
503
504		The current status is a step in this direction, but it is not quite
505		satisfactory yet.  Expect more changes in the future.
506
507		Here is the current (new) organization...
508
509		What used to be $smlnj/viscomp/core.cm is now divided into
510		six CM libraries:
511
512		$smlnj/viscomp/basics.cm
513		/parser.cm
514		/elabdata.cm
515		/elaborate.cm
516		/execute.cm
517		/core.cm
518
519		The CM files for these libraries live under src/system/smlnj/viscomp.
520		All these libraries are proxy libraries that contain precisely
521		one CM library component.  Here are the locations of the components
522		(all within the src/compiler tree):
523
524		Basics/basics.cm
525		Parse/parser.cm
526		ElabData/elabdata.cm
527		Elaborator/elaborate.cm
528		Execution/execute.cm
529		core.cm
530
531		[This organization is the same that has been used already
532		for a while for the architecture-specific parts of the visible
533		compiler and for the old version of core.cm.]
534
535		As you will notice, many source files have been moved from their
536		respective original locations to a new home in one of the above
537		subtrees.
538
539		The division of labor between the new libraries is the following:
540
541		basics.cm:
542		- Simple, basic definitions that pertain to many (or all) of
543		the other libraries.
544		parser.cm:
545		- The SML parser, producing output of type Ast.dec.
546		- The type family for Ast is also defined and exported here.
547		elabdata.cm:
548		- The datatypes that describe input and output of the elaborator.
549		This includes types, absyn, and static environments.
550		elaborator.cm:
551		- The SML/NJ type checker and elaborator.
552		This maps an Ast.dec (with a given static environment) to
553		an Absyn.dec (with a new static environment).
554		- This libraries implements certain modules that used to be
555		structures as functors (to remove dependencies on FLINT).
556		execute.cm:
557		- Everything having to do with executing binary code objects.
558		- Dynamic environments.
559		core.cm:
560		- SML/NJ-specific instantiations of the elaborator and MLRISC.
561		- Top-level modules.
562		- FLINT (this should eventually become its own library)
563
564		Notes:
565
566		I am not 100% happy with the way I separated the elaborator (and its
567		data structures) from FLINT.  Two instances of the same problem:
568
569		1. Data structures contain certain fields that carry FLINT-specific
570		information.  I hacked around this using exn and the property list
571		module from smlnj-lib.  But the fact that there are middle-end
572		specific fields around at all is a bit annoying.
573
574		2. The elaborator calculates certain FLINT-related information.  I tried
575		to make this as abstract as I could using functorization, but, again,
576		the fact that the elaborator has to perform calculations on behalf
577		of the middle-end at all is not nice.
578
579		3. Having to used exn and property lists is unfortunate because it
580		weakens type checking.  The other alternative (parameterizing
581		nearly *everything*) is not appealing, though.
582
583		I removed the "rebinding =" warning hack because due to the new organization
584		it was awkward to maintain it.  As a result, the compiler now issues some of
585		these warnings when compiling init.cmi during bootstrap compilation. On
586		the plus side, you also get a warning when you do, for example:
587		val op = = Int32.+
588		which was not the case up to now.
589
590		I placed "assign" and "deref" into the _Core structure so that the
591		code that deals with the "lazy" keyword can find them there.  This
592		removes the need for having access to the primitive environment
593		during elaboration.
594
595		----------------------------------------------------------------------
596		Name: Matthias Blume
597		Date: 2001/08/13
598		Tag: blume-20010813-closures
599		Description:
600
601		This fix was sent to us by Zhong Shao.  It is supposed to improve the
602		performance of certain loops by avoiding needless closure allocation.
603
604		----------------------------------------------------------------------
605		Name: Lal George
606		Date: 2001/07/31 10:03:23 EDT 2001
607		Tag: george-20010731-x86-fmalloc
608		Description: Fixed bug in x86 calls
609
610		There was a bug where call instructions would mysteriously
611		vanish. The call instruction had to be one that returned
612		a floating point value.
613
614		----------------------------------------------------------------------
615		Name: Lal George
616		Date: 2001/07/19 16:36:29 EDT 2001
617		Tag: george-20010719-simple-cells
618		Description:
619
620		I have dramatically simplified the interface for CELLS in MLRISC.
621
622		In summary, the cells interface is broken up into three parts:
623
624		1. CellsBasis : CELLS_BASIS
625
626		CellsBasis is a top level structure and common for all
627		architectures.  it contains the definitions of basic datatypes
628		and utility  functions over these types.
629
630		2. functor Cells() : CELLS
631
632		Cells generates an interface for CELLS that incorporates the
633		specific resources on the target architecture, such as the
634		presence of special register classes, their number and size,
635		and various useful substructures.
636
637		3. <ARCH>CELLS
638
639		e.g. SparcCells: SPARCCELLS
640
641		<ARCH>CELLS usually contains additional bindings for special
642		registers  on the architecture, such as:
643
644		val r0 : cell           (* register zero *)
645		val y : cell            (* Y register *)
646		val psr : cell          (* processor status register *)
647		...
648
649		The structure returned by applying the Cells functor is opened
650		in this interface.
651
652		The main implication of all this is that the datatypes for cells is
653		split between CellsBasis and CELLS -- a fairly simple change for user
654		code.
655
656		In the old scheme the CELLS interface had a definitional binding of
657		the form:
658
659		signature CELLS = sig
660
661		structure CellsBasis = CellsBasis
662
663		...
664
665		end
666
667		With all the sharing constraints that goes on in MLRISC, this old
668		design  quickly leads to errors such as:
669
670		"structure definition spec inside of sharing ... "
671
672
673		and appears to require an unacceptable amount of sharing and where
674		constraint hackery.
675
676		I think this error message (the interaction of definitional specs and
677		sharing) requires more explanation on our web page.
678
679		----------------------------------------------------------------------
680		Name: Matthias Blume
681		Date: 2001/07/19 15:00:00 EDT
682		Tag: blume-20010719-libreorg
683		Description:
684
685		This update puts together a fairly extensive but straightforward change
686		to the way the libraries that implement the interactive system are
687		organized:
688
689		The biggest change is the elimination of structure Compiler.  As a
690		replacement for this structure, there is now a CM library
691		(known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
692		that exports all the substructures of the original structure Compiler
693		directly.  So instead of saying Compiler.Foo.bar one now simply
694		says Foo.bar.  (The CM libraries actually export a collection of
695		structures that is richer than the collection of substructures of
696		structure Compiler.)
697
698		To make the transition smooth, there is a separate library called
699		$smlnj/compiler/compiler.cm which puts together and exports the
700		original structure Compiler (or at least something very close to it).
701
702		There are five members of the original structure Compiler
703		that are not exported directly but which instead became members
704		of a new structure Backend (described by signature BACKEND).  These are:
705		structure Profile (: PROFILE), structure Compile (: COMPILE), structure
706		Interact (: INTERACT), structure Machine (: MACHINE), and val
707		architecture (: string).
708
709		Structure Compiler.Version has become structure CompilerVersion.
710
711		Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
712		by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
713		or x86, respectively.
714		Each of these exports the same frontend structures that
715		$smlnj/compiler.cm exports.  But they do not have a structure Backend
716		and instead export some structure <Arch>Backend where <Arch> is Alpha32,
717		Hppa, PPC, Sparc, or X86, respectively.
718
719		Library $smlnj/compiler/all.cm exports the union of the exports of
720		$smlnj/compiler/<arch>.cm
721
722		There are no structures <Arch>Compiler anymore, use
723		$smlnj/compiler/<arch>.cm instead.
724
725		Library host-compiler-0.cm is gone.  Instead, the internal library
726		that instantiates CM is now called cm0.cm.  Selection of the host
727		compiler (backend) is no longer done here but. (Responsibility for it
728		now lies with $smlnj/compiler/current.cm.  This seems to be more
729		logical.)
730
731		Many individual files have been moved or renamed.  Some files have
732		been split into multiple files, and some "dead" files have been deleted.
733
734		Aside from these changes to library organization, there are also changes
735		to the way the code itself is organized:
736
737		Structure Binfile has been re-implemented in such a way that it no
738		longer needs any knowledge of the compiler.  It exclusively deals
739		with the details of binfile layout.  It no longer invokes the
740		compiler (for the purpose of creating new prospective binfile
741		content), and it no longer has any knowledge of how to interpret
742		pickles.
743
744		Structure Compile (: COMPILE) has been stripped down to the bare
745		essentials of compilation.  It no longer deals with linking/execution.
746		The interface has been cleaned up considerably.
747
748		Utility routines for dealing with linking and execution have been
749		moved into their own substructures.
750
751		(The ultimate goal of these changes is to provide a light-weight
752		binfile loader/linker (at least for, e.g., stable libraries) that
753		does not require CM or the compiler to be present.)
754
755		CM documentation has been updated to reflect the changes to library
756		organization.
757
758		----------------------------------------------------------------------
759		Name: Matthias Blume
760		Date: 2001/07/10 17:30:00 EDT
761		Tag: Release_110_34
762		Description:
763
764		Minor tweak to 110.34 (re-tagged):
765
766		- README.html file added to CVS repository
767		- runtime compiles properly under FreeBSD 3.X and 4.X
768
769		----------------------------------------------------------------------
770		Name: Matthias Blume
771		Date: 2001/07/10 17:30:00 EDT
772		Tag: Release_110_34
773		Description:
774
775		New version number (110.34). New bootfiles.
776
777		----------------------------------------------------------------------
778		Name: Matthias Blume
779		Date: 2001/07/09 16:00:00 EDT
780		Tag: blume-20010709-more-varargs
781		Description:
782
783		I changed the handling of varargs in ml-nlffigen again:
784		The ellipsis ... will now simply be ignored (with an accompanying warning).
785
786		The immediate effect is that you can actually call a varargs function
787		from ML -- but you can't actually supply any arguments beyond the ones
788		specified explicitly.  (For example, you can call printf with its format
789		string, but you cannot pass additional arguments.)
790
791		This behavior is only marginally more useful than the one before, but
792		it has the advantage that a function or, more importantly, a function
793		type never gets dropped on the floor, thus avoiding follow-up problems with
794		other types that refer to the offending one.
795
796		----------------------------------------------------------------------
797		Name: Matthias Blume
798		Date: 2001/07/09 11:25:00 EDT
799		Tag: blume-20010709-varargs
800		Description:
801
802		1. ckit-lib.cm now exports structure Error
803		2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
804		with a warning accompanied by a source location.  Moreover, it
805		merely skips the offending function or type and proceeds with the
806		rest of its work.u  As a result, one can safely feed C code containing
807		"..." to ml-nlffigen.
808		3. There are some internal improvements to CM, providing slightly
809		more general string substitutions in the tools subsystem.
810
811		----------------------------------------------------------------------
812		Name: Matthias Blume
813		Date: 2001/06/27 15:10:00 EDT
814		Tag: blume-20010627-concur
815		Description:
816
817		Fixed a small bug in CM's handling of parallel compilation.
818		(You could observe the bug by Control-C-interrupting an ordinary
819		CMB.make or CM.stabilize and then attaching some compile servers.
820		The result was that all of a sudden the previously interrupted
821		compilation would continue on its own.  This was because of
822		an over-optimization: CM did not bother to clean out certain queues
823		when no servers were attached "anyway", resulting in the contents
824		of these queues to grab control when new servers did get attached.)
825
826		There is also another minor update to the CM manual.
827
828		----------------------------------------------------------------------
829		Name: Matthias Blume
830		Date: 2001/06/26 16:15:00 EDT
831		Tag: blume-20010626-cmdoc
832		Description:
833
834		Minor typo fixed in CM manual (syntax diagram for libraries).
835
836		----------------------------------------------------------------------
837		Name: Matthias Blume
838		Date: 2001/06/25 22:55:00 EDT
839		Tag: blume-20010625-x86pc
840		Description:
841
842		Fixed a nasty bug in the X86 assembly code that caused signal
843		handlers to fail (crash) randomly.
844
845		----------------------------------------------------------------------
846		Name: Matthias Blume
847		Date: 2001/06/25 12:05:00 EDT
848		Tag: blume-20010625-nlffigen
849		Description:
850
851		This update fixes a number of minor bugs in ml-nlffigen as reported by
852		Nick Carter <nbc@andrew.cmu.edu>.
853
854		1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
855		2. Default names for generated files are now derived from the name of
856		the C file *without its directory*.  In particular, this causes generated
857		files to be placed locally even if the C file is in some system directory.
858		3. Default names for generated signatures and structures are also derived
859		from the C file name without its directory.  This avoids silly things
860		like "structure GL/GL".
861		(Other silly names are still possible because ml-nlffigen does not do
862		a thorough check of whether generated names are legal ML identifiers.
863		When in doubt, use command line arguments to force particular names.)
864
865		----------------------------------------------------------------------
866		Name: Matthias Blume
867		Date: 2001/06/21 12:25:00 EDT
868		Tag: blume-20010621-eXene
869		Description:
870
871		eXene now compiles and (sort of) works again.
872
873		The library name (for version > 110.33) is $/eXene.cm.
874
875		I also added an new example in src/eXene/examples/nbody.  See the
876		README file there for details.
877
878		----------------------------------------------------------------------
879		Name: Matthias Blume
880		Date: 2001/06/20 16:40:00 EDT
881		Tag: blume-20010620-cml
882		Description:
883
884		CML now compiles and works again.
885
886		Libraries (for version > 110.33):
887
888		$cml/cml.cm            Main CML library.
889		$cml/basis.cm          CML's version of $/basis.cm.
890		$cml/cml-internal.cm   Internal helper library.
891		$cml/core-cml.cm       Internal helper library.
892		$cml-lib/trace-cml.cm  Tracing facility.
893		$cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
894
895		The installer (config/install.sh) has been taught how to properly
896		install this stuff.
897
898		----------------------------------------------------------------------
899		Name: Matthias Blume
900		Date: 2001/06/19 17:55:00 EDT
901		Tag: blume-20010619-instantiate
902		Description:
903
904		This un-breaks the fix for bug 1432.
905		(The bug was originally fixed in 110.9 but I broke it again some
906		time after that.)
907
908		----------------------------------------------------------------------
909		Name: Matthias Blume
910		Date: 2001/06/19 17:25:00 EDT
911		Tag: blume-20010619-signals
912		Description:
913
914		This should (hopefully) fix the long-standing signal handling bug.
915		(The runtime system was constructing a continuation record with an
916		incorrect descriptor which would cause the GC to drop data on the floor...)
917
918		----------------------------------------------------------------------
919		Name: Matthias Blume
920		Date: 2001/06/15 15:05:00 EDT
921		Tag: blume-20010615-moresparc
922		Description:
923
924		Here is a short late-hour update related to Sparc c-calls:
925
926		-- made handling of double-word arguments a bit smarter
927
928		-- instruction selection phase tries to collapse certain clumsily
929		constructed ML-Trees; typical example:
930
931		ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
932
933		This currently has no further impact on SML/NJ since mlriscGen does
934		not seem to generate such patterns in the first place, and c-calls
935		(which did generate them in the beginning) has meanwhile been fixed
936		so as to avoid them as well.
937
938		----------------------------------------------------------------------
939		Name: Matthias Blume
940		Date: 2001/06/15 15:05:00 EDT
941		Tag: blume-20010615-sparc
942		Description:
943
944		The purpose of this update is to provide an implementation of NLFFI
945		on Sparc machines.
946
947		Here are the changes in detail:
948
949		* src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
950		the Sparc implementation of the c-calls API.
951		* The Sparc backend of SML/NJ has been modified to uniformely use %fp
952		for accessing the ML frame.  Thus, we have a real frame pointer and
953		can freely modify %sp without need for an omit-frame-ptr phase.
954		The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
955		this case.
956		* ml-nlffigen has been taught to produce code for different architectures
957		and calling conventions.
958		* In a way similar to what was done in the x86 case, the Sparc
959		backend uses its own specific extension to mltree.  (For example,
960		it needs to be able to generate UNIMP instructions which are part
961		of the calling convention.)
962		* ml-nlffi-lib was reorganized to make it more modular (in particular,
963		to make it easier to plug in new machine- and os-dependent parts).
964
965		There are some other fairly unrelated bug fixes and cleanups as well:
966
967		* I further hacked the .cm files for MLRISC tools (like MDLGen) so
968		that they properly share their libraries with existing SML/NJ libraries.
969		* I fixed a minor cosmetic bug in CM, supressing certain spurious
970		follow-up error messages.
971		* Updates to CM/CMB documentation.
972
973		TODO items:
974
975		* MLRISC should use a different register as its asmTemp on the Sparc.
976		(The current %o2 is a really bad choice because it is part of the
977		calling conventions, so things might interfere in unexpected ways.)
978
979		----------------------------------------------------------------------
980		Name: Matthias Blume
981		Date: 2001/06/07
982		Tag: blume-20010607-calls
983		Description:
984
985		A number of internal changes related to C calls and calling conventions:
986
987		1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
988		number of bytes popped implicitly (by the callee).  In most cases
989		this field is 0 but on x86/win32 it is some non-zero value.  This
990		is information provided for the benefit of the "omit-frameptr" pass.
991		2. The CALL instruction on the x86 carries a similar "pops" field.
992		The instruction selection phase copies its value from the ML-Tree
993		CALL statement.
994		3. On all other architectures, the instruction selection phase checks
995		whether "pops=0" and complains if not.
996		4. The c-calls implementation for x86 now accepts two calling conventions:
997		"ccall" and "stdcall".  When "ccall" is selected, the caller cleans
998		up after the call and pops is set to 0.  For "stdcall", the caller
999		does nothing, leaving the cleanup to the callee; pops is set to
1000		the number of bytes that were pushed onto the stack.
1001		5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
1002		distinguish between "ccall" and "stdcall".
1003		6. The UNIMP instruction has been added to the supported Sparc instruction
1004		set. (This is needed for implementing the official C calling convention
1005		on this architecture.)
1006		7. I fixed some of the .cm files under src/MLRISC/Tools to make them
1007		work with the latest CM.
1008
1009		----------------------------------------------------------------------
1010		Name: Matthias Blume
1011		Date: 2001/06/05 15:10:00 EDT
1012		Tag: blume-20010605-cm-index
1013		Description:
1014
1015		0. The "lambdasplit" parameter for class "sml" in CM has been documented.
1016
1017		1. CM can now generate "index files".  These are human-readable files
1018		that list on a per-.cm-file basis each toplevel symbol defined or
1019		imported.  The location of the index file for
1020		<p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
1021		To enable index-file generation, set CM.Control.generate_index to true
1022		or export an environment-symbol: export CM_GENERATE_INDEX=true.
1023
1024		The CM manual has been updated accordingly.
1025
1026		2. I made some slight modifications to the c-calls API in MLRISC.
1027
1028		a) There is now a callback to support saving/restoring of
1029		dedicated but caller-save registers around the actual call
1030		instruction.
1031		b) One can optionally specify a comment-annotation for the
1032		call instruction.
1033
1034		3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
1035		(For example, the comment annotation shows the C prototype of
1036		the function being called.)
1037
1038		----------------------------------------------------------------------
1039		Name: Matthias Blume
1040		Date: 2001/06/01 13:30:00 EDT
1041		Tag: blume-20010601-nlffi-cleanup
1042		Description:
1043
1044		This is mostly a cleanup of MLFFI stuff:
1045
1046		- some signature files have been put into a more exposed place
1047		- the ugly 'f type parameter is gone (simplifies types tremendously!)
1048		- ml-nlffigen changed accordingly
1049		- tutorial updated
1050
1051		Other changes:
1052
1053		- author's affiliation in CM manual(s) updated
1054		- some more recognized keywords added to Allen's sml.sty
1055
1056		----------------------------------------------------------------------
1057		Name: Matthias Blume
1058		Date: 2001/05/25 15:30:00 EDT
1059		Tag: blume-20010525-iptr
1060		Description:
1061
1062		- put the official 110.33-README (as it appears on the ftp server) under
1063		CVS
1064		- fixed a small bug related to incomplete pointer types in
1065		ml-nlffigen
1066		- small cosmetic change to the ml-nlffi-lib's "arr" type constructor
1067		(it does not need the 'f type parameter)
1068
1069		----------------------------------------------------------------------
1070		Name: Matthias Blume
1071		Date: 2001/05/23 14:30:00 EDT
1072		Tag: Release_110_33
1073		Description:
1074
1075		New version number (110.33).  New bootfiles.
1076
1077		----------------------------------------------------------------------
1078		Name: Matthias Blume
1079		Date: 2001/05/22 18:06:00 EDT
1080		Tag: blume-20010522-targets
1081		Description:
1082
1083		Made install.sh use file config/targets.customized if it exists, falling
1084		back to config/targets if it doesn't.  This way one can have a customized
1085		version of the targets file without touching the "real thing", thus
1086		eliminating the constant fear of accidentally checking something bogus
1087		back into the CVS repository...  (File config/targets.customized must
1088		not be added to the repository!)
1089
1090		----------------------------------------------------------------------
1091		Name: Matthias Blume
1092		Date: 2001/05/22 16:30:00 EDT
1093		Tag: blume-20010522-minitut
1094		Description:
1095
1096		1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
1097		struct returns.
1098		2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
1099		incomplete, preliminary documentation for NLFFI.
1100
1101		----------------------------------------------------------------------
1102		Name: Matthias Blume
1103		Date: 2001/05/14 11:30:00 EDT
1104		Tag: blume-20010514-script
1105		Description:
1106
1107		Some bugs in install script fixed.
1108
1109		In addition to that I also made a slight change to the NLFFI API:
1110		Functors generated by ml-nlffigen now take the dynamic library as a
1111		straight functor argument, not as a suspended one.  (The original
1112		functor code used to force the suspension right away anyway, so there
1113		was nothing gained by this complication of the interface.)
1114
1115		----------------------------------------------------------------------
1116		Name: Matthias Blume
1117		Date: 2001/05/11 14:35:00 EDT
1118		Tag: blume-20010511-ml-nlffi
1119		Description:
1120
1121		I finally took the plunge and added my new FFI code to the main
1122		repository.  For x86-linux it is now ready for prime-time.
1123
1124		There are two new subdirectories of "src":
1125
1126		- ml-nlffi-lib:
1127		The utility library for programs using the FFI interface.
1128		Here is the implementation of $/c.cm and its associated low-level
1129		partners $/c-int.cm and $/memory.cm.
1130		- ml-nlffigen:
1131		A stand-alone program for generating ML glue code from C source
1132		code.
1133
1134		Building ml-nlffigen requires $/ckit-lib.cm.
1135
1136		The config/install.sh script has been updates to do the Right Thing
1137		(hopefully).
1138
1139		Notice that the source tree for the C-Kit will not be put under "src"
1140		but directly under the installation root directory.  (This is the
1141		structure that currently exists on the CVS server when you check out
1142		module "sml".)  Fortunately, config/install.sh knows about this oddity.
1143
1144		Bugs: No documentation yet.
1145
1146		----------------------------------------------------------------------
1147		Name: Matthias Blume
1148		Date: 2001/05/09 16:35:00 EDT
1149		Tag: blume-20010509-cpscontract
1150		Description:
1151
1152		Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
1153		wrapper/unwrapper elimination did not decrement usage counts and some
1154		dead variables got overlooked by the dead-up logic.)
1155
1156		----------------------------------------------------------------------
1157		Name: Lal George
1158		Date: 2001/05/08  17:26:09 EDT
1159		Tag: george-20010508-omit-frameptr
1160		Description:
1161
1162		Changes to implement the omit-frame-pointer optimization to support
1163		raw C calls. For now, there is only support on the Intel x86, but
1164		other architectures will follow as more experience is gained with this.
1165
1166
1167		----------------------------------------------------------------------
1168		Name: Matthias Blume
1169		Date: 2001/05/07 14:40:00 EDT
1170		Tag: blume-20010507-proxies
1171		Description:
1172
1173		I made into "proxy libraries" all libraries that qualify for such a
1174		change.  (A qualifying library is a library that has another library or
1175		groups as its sole member and repeats that member's export list
1176		verbatim.  A proxy library avoids this repetition by omitting its export
1177		list, effectively inheriting the list that its (only) member exports.
1178		See the CM manual for more explanation.)
1179		The main effect is that explicit export lists for these libraries
1180		do not have to be kepts in sync, making maintenance a bit easier.
1181
1182		I also added copyright notices to many .cm-files.
1183
1184		Last but not least, I made a new set of bootfiles.
1185
1186		----------------------------------------------------------------------
1187		Name: Matthias Blume
1188		Date: 2001/05/04 17:00:00 EDT
1189		Tag: blume-20010504-cm-lsplit
1190		Description:
1191
1192		0. John merged pending changes to $/smlnj-lib.cm
1193
1194		1. Allen's previous change accidentally backed out of one of Lal's
1195		earlier changes.  I undid this mistake (re-introducing Lal's change).
1196
1197		2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
1198		within the compiler where applicable.  There is some code simplification
1199		because of that.
1200
1201		3. The "split" phase (in FLINT) is now part of the default list of phases.
1202		Compiler.Control.LambdaSplitting.* can be used to globally control the
1203		lambda-splitting (cross-module-inlining) engine.  In addition to that,
1204		it can now also be controlled on a per-source basis: CM has been taught
1205		a new tool parameter applicable to ML source files.
1206
1207		- To turn lambda-splitting off completely:
1208		local open Compiler.Control.LambdaSplitting in
1209		val _ = set Off
1210		end
1211		- To make "no lambda-splitting" the global default (but allow per-source
1212		overriding); this is the initial setting:
1213		local open Compiler.Control.LambdaSplitting in
1214		val _ = set (Default NONE)
1215		end
1216		- To make "lambda-splitting with aggressiveness a" the global default
1217		(and allow per-source overriding):
1218		local open Compiler.Control.LambdaSplitting in
1219		val _ = set (Default (SOME a))
1220		end
1221
1222		- To turn lambda-splitting off for a given ML souce file (say: a.sml)
1223		write (in the respective .cm-file):
1224		a.sml (lambdasplitting:off)
1225		- To turn lambda-splitting for a.sml on with minimal aggressiveness:
1226		a.sml (lambdasplitting:on)
1227		- To turn lambda-splitting for a.sml on with aggressiveness <a> (where
1228		<a> is a decimal non-negative integer):
1229		a.sml (lambdasplitting:<a>)
1230		- To turn lambda-splitting for a.sml on with maximal aggressiveness:
1231		a.sml (lambdasplitting:infinity)
1232		- To use the global default for a.sml:
1233		a.sml (lambdasplitting:default)
1234		or simply
1235		a.sml
1236
1237		----------------------------------------------------------------------
1238		Name: Allen Leung
1239		Date: 2001/05/04 01:57:00 EDT
1240		Tag: leunga-20010504-sync
1241		Description:
1242
1243		MLRISC features.
1244
1245		1. Fix to CMPXCHG instructions.
1246		2. Changed RA interface to allow annotations in callbacks.
1247		3. Added a new method to the stream interface to allow annotations updates.
1248
1249		----------------------------------------------------------------------
1250		Name: Matthias Blume
1251		Date: 2001/05/01 11:45:00 EDT
1252		Tag: blume-20010501-pcedittmp
1253		Description:
1254
1255		Changed install.sh to use the current working directory instead of
1256		/usr/tmp for a temporary file (pcedittmp).  The previous choice
1257		of /usr/tmp caused trouble with MacOS X because of file premission
1258		problems.
1259
1260		----------------------------------------------------------------------
1261		Name: Matthias Blume
1262		Date: 2001/04/20 11:10:00 EDT
1263		Tag: blume-20010420-inMLflag
1264		Description:
1265
1266		- added vp_limitPtrMask to vproc-state.h
1267		(for use by the raw-C-calls mechanism to implement proper interrupt
1268		handling)
1269		- made the ML compiler aware of various data-structure offsets so it
1270		can generate code for accessing the vp_inML flag and vp_limitPtrMask
1271		- tweaked mlriscGen.sml to have it emit interrupt-handling code for
1272		raw C-calls
1273
1274		----------------------------------------------------------------------
1275		Name: Lal George
1276		Date: 2001/04/20 09:15:28 EDT
1277		Tag: george-20010420-macosX
1278		Description:
1279
1280		- Changes to port to Mac OS X; Darwin.
1281
1282		- In the process I found that sqrt was broken on the PPC, because the
1283		fsqrt instruction is not implemented.
1284
1285		----------------------------------------------------------------------
1286		Name: Matthias Blume
1287		Date: 2001/04/18 12:45:00 EDT
1288		Tag: blume-20010418-ccalls
1289		Description:
1290
1291		- fixed two off-by-4 errors in the x86-specific c-calls implementation
1292		(this bug prevented structure arguments containing pointers from being
1293		passed correctly)
1294		- changed the raw-C-call code in mlriscGen.sml in such a way that
1295		structure arguments are represented as a pointer to the beginning
1296		of the structure (instead of having a series of synthesized arguments,
1297		one for each structure member)
1298
1299		- made makeml script's verbosity level configurable via environment
1300		variable (MAKEML_VERBOSITY)
1301
1302		- eliminated placeholder implementations for f32l, w16s, i16s, and f32s
1303		in rawmem-x86.sml; we are now using the real thing
1304
1305		----------------------------------------------------------------------
1306		Name: Matthias Blume
1307		Date: 2001/03/22 16:25:00 EST
1308		Tag: blume-20010322-bootfiles
1309		Description:
1310
1311		Created a new set of bootfiles (for your automatic installation convenience).
1312
1313		----------------------------------------------------------------------
1314		Name: Matthias Blume
1315		Date: 2001/03/22 15:10:00 EST
1316		Tag: blume-20010322-rawmem-parcm
1317		Description:
1318
1319		1. All "raw memory access" primitives for the new FFI are implemented now
1320		(at least on the x86).
1321		2. Some further cleanup of CM's parallel make mechanism.
1322
1323		----------------------------------------------------------------------
1324		Name: Matthias Blume
1325		Date: 2001/03/19 17:53:00 EST
1326		Tag: blume-20010319-parallel
1327		Description:
1328
1329		Parallel make (using compile servers) now works again.
1330
1331		To this end, CM.stabilize and CMB.make have been modified to work in
1332		two passes when compile servers are attached:
1333		1. Compile everything, do not perform stabilization; this pass
1334		uses compile servers
1335		2. Stabilize everything; this pass does not use compile servers
1336		If there are no compile servers, the two passes are combined into one
1337		(as before).  Splitting the passes increases the inherent parallelism
1338		in the dependency graph because the entire graph including all
1339		libraries is available at the same time.  This, in turn, improves
1340		server utilization.  The downside is that the master process will
1341		have to do some extra work after compilation is done (because for
1342		technical reasons it must re-read all the binfiles during stabilization).
1343
1344		----------------------------------------------------------------------
1345		Name: Matthias Blume
1346		Date: 2001/03/16 12:22:00 EST
1347		Tag: blume-20010316-bootfiles
1348		Description:
1349
1350		Created a new set of bootfiles (for your automatic installation convenience).
1351
1352		----------------------------------------------------------------------
1353		Name: Matthias Blume
1354		Date: 2001/03/16 11:00:00 EST
1355		Tag: blume-20010316-MLTREE-fixup
1356		Description:
1357
1358		This is a minor fixup for an (untagged) earlier commit by Allen.
1359		(A file was missing).
1360
1361		----------------------------------------------------------------------
1362		Name: Allen Leung
1363		Date: Mon Mar  5 18:54:57 EST 2001
1364		Tag: leunga-20010305-cut-support
1365
1366		1. New support for alternative control-flow in MLTREE.
1367		Currently we support
1368
1369		FLOW_TO(CALL ...., [k1,...,kn])
1370
1371		This is needed for 'cuts to' in C-- and try/handle-like constructs
1372		in Moby
1373
1374		New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
1375
1376		2. Register Allocator
1377
1378		Changes in interface [from Fermin, John]
1379
1380		3. Alpha 8-bit SLL support [Fermin]
1381
1382		4. All architectures
1383
1384		A new module (ClusterExpandCopies) for expanding parallel copies.
1385
1386		----------------------------------------------------------------------
1387		Name: Allen Leung
1388		Date: 2001/02/27 23:07:00 EST
1389		Tag: leunga-20010227-minor-stuff
1390
1391		1. Alpha bug fix for CMOVNE
1392		2. Handle mltree COND(..,FCMP ...,...)
1393		3. Bug fix in simplifier
1394
1395		----------------------------------------------------------------------
1396		Name: Matthias Blume
1397		Date: 2001/01/30 17:50:00 EST
1398		Tag: blume-20010130-sync
1399		Description:
1400
1401		This is just a minor update to sync my devel branch with the main brach.
1402		The only visible change is the addition of some README files.
1403
1404		----------------------------------------------------------------------
1405		Name: Matthias Blume
1406		Date: 2001/01/12 23:30:00 JST
1407		Tag: blume-20010112-bootfiles
1408		Description:
1409
1410		Made a new set of bootfiles that goes with the current state of the
1411		repository.
1412
1413		----------------------------------------------------------------------
1414		Name: Matthias Blume
1415		Date: 2001/01/12 21:20:00 JST
1416		Tag: blume-20010112-sync
1417		Description:
1418
1419		I am just flushing out some minor changes that had accumulated in
1420		my private branch in order to sync with the main tree.  (This is
1421		mainly because I had CVS trouble when trying to merge _into_ my
1422		private branch.)
1423
1424		Most people should be completely unaffected by this.
1425
1426		----------------------------------------------------------------------
1427		Name: Allen Leung
1428		Date: Thu Jan 11 21:03:00 EST 2001
1429		Tag: leunga-20010111-labexp=mltree
1430		Description:
1431
1432		1.  Removed the type LabelExp and replace it by MLTree.
1433		2.  Rewritten mltree-simplify with the pattern matcher tool.
1434		3.  There were some bugs in alpha code generator which would break
1435		64-bit code generation.
1436		4.  Redo the tools to generate code with the
1437		5.  The CM files in MLRISC (and in src/system/smlnj/MLRISC)
1438		are now generated by perl scripts.
1439
1440		----------------------------------------------------------------------
1441		Name: Matthias Blume
1442		Date: 2001/01/10 21:55:00 JST
1443		Tag: blume-20010110-rcc
1444		Description:
1445
1446		The RCC stuff now seems to work (but only on the x86).
1447		This required hacking of the c-calls interface (and -implementation) in
1448		MLRISC.
1449
1450		Normal compiler users should be unaffected.
1451
1452		----------------------------------------------------------------------
1453		Name: Matthias Blume
1454		Date: 2001/01/09 01:20:00 JST
1455		Tag: blume-20010109-rcc
1456		Description:
1457
1458		This is a fairly big patch, flushing out a large number of pending
1459		changes that I made to my development copy over the last couple of days.
1460
1461		Of practical relevance at this moment is a workaround for a pickling
1462		bug that Allen ran into the other day.  The cause of the bug itself is
1463		still unknown and it might be hard to fix it properly, but the
1464		workaround has some merits of its own (namely somewhat reducing pickling
1465		overhead for certain libraries).  Therefore, I think this solution should
1466		be satisfactory at this time.
1467
1468		The rest of the changes (i.e., the vast majority) has to do with my
1469		ongoing efforts of providing direct support for C function calls from
1470		ML.  At the moment there is a new primop "RAW_CCALL", typing magic
1471		in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new
1472		case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to
1473		translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to
1474		handle RCC.
1475
1476		The last part (the changes to mlriscGen.sml) are still known to be
1477		wrong on the x86 and not implemented on all other architectures.  But
1478		the infrastructure is in place. I had to change a few functor
1479		signatures in the backend to be able to route the CCalls interface
1480		from MLRISC there, and I had to specialize the mltree type (on the
1481		x86) to include the necessary extensions. (The extensions themselves
1482		were already there and redy to go in MLRISC/x86).
1483
1484		Everything should be very happy as soon as someone helps me with
1485		mlriscGen.sml...
1486
1487		In any case, nothing of this should matter to anyone as long as the
1488		new primop is not being used (which is going to be the case unless you
1489		find it where I hid it :). The rest of the compiler is completely
1490		unaffected.
1491
1492		----------------------------------------------------------------------
1493		Name: Matthias Blume
1494		Date: 2001/01/05 00:30:00 JST
1495		Tag: blume-20010105-primops
1496		Description:
1497
1498		Added some experimental support for work that I am doing right now.
1499		These changes mostly concern added primops, but there is also a new
1500		experimental C library in the runtime system (but currently not enabled
1501		anywhere except on Linux/X86).
1502
1503		In the course of adding primops (and playing with them), I discovered that
1504		Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly
1505		broken.  (Zhong was very right he labeled this stuff as "major gross hack".)
1506		To recover, I made type information in INL_PRIM mandatory and changed
1507		prim.sml as well as built-in.sml accordingly.  The InLine structure now
1508		has complete, correct type information (i.e., no bottom types).
1509
1510		Since all these changes mean that we need new binfiles, I also bumped the
1511		version number to 110.32.1.
1512
1513		----------------------------------------------------------------------
1514		Name: Matthias Blume
1515		Date: 2000/12/30 22:10:00 JST
1516		Tag: blume-20001230-various
1517		Description:
1518
1519		Added proxy libraries for MLRISC and let MLRISC libraries refer
1520		to each other using path anchors.  (See CM manual for explanation.)
1521
1522		Updated CM documentation.
1523
1524		Fixed some bugs in CM.
1525
1526		Implemented "proxy" libraries (= syntactic sugar for CM).
1527
1528		Added "-quiet" option to makeml and changed runtime system accordingly.
1529
1530		Added cleanup handler for exportML to reset timers and compiler stats.
1531
1532		----------------------------------------------------------------------
1533		Name: Lal George
1534		Date: 2000/12/22 22:22:58 EST 2000
1535		Tag: Release_110_32
1536		Description:
1537
1538		Infinite precision used throughout MLRISC.
1539		see MLRISC/mltree/machine-int.sig
1540
1541		----------------------------------------------------------------------
1542		Name: Matthias Blume
1543		Date: 2000/12/22 23:16:00 JST
1544		Tag: blume-20001222-warn
1545		Description:
1546
1547		Corrected wording and formatting of some CM warning message which I
1548		broke in my previous patch.
1549
1550		----------------------------------------------------------------------
1551		Name: Matthias Blume
1552		Date: 2000/12/22 21:20:00 JST
1553		Tag: blume-20001222-anchorenv
1554		Description:
1555
1556		Fixed CM's handling of anchor environments in connection with CMB.make.
1557
1558		----------------------------------------------------------------------
1559		Name: Matthias Blume
1560		Date: 2000/12/22 13:15:00 JST
1561		Tag: blume-20001222-cleanup
1562		Description:
1563
1564		Removed src/cm/ffi which does not (and did not) belong here.
1565
1566		----------------------------------------------------------------------
1567		Name: Matthias Blume
1568		Date: 2000/12/21 23:55:00 JST
1569		Tag: blume-20001221-exn
1570		Description:
1571
1572		Probably most important: CM no longer silently swallows all exceptions
1573		in the compiler.
1574		Plus: some other minor CM changes.  For example, CM now reports some
1575		sizes for generated binfiles (code, data, envpickle, lambdapickle).
1576
1577		----------------------------------------------------------------------
1578		Name: Matthias Blume
1579		Date: 2000/12/15 00:01:05 JST
1580		Tag: blume-20001215-dirtool
1581		Description:
1582
1583		- "dir" tool added.
1584		- improvements and cleanup to Tools structure
1585		- documentation updates
1586
1587		----------------------------------------------------------------------
1588		Name: Allen Leung
1589		Date: Thu Dec 14 03:45:24 EST 2000
1590		Description:
1591		Tag:  leunga-20001214-int-inf
1592		Description:
1593
1594		In IntInf, added these standard functions, which are missing from our
1595		implementation:
1596
1597		andb : int * int -> int
1598		xorb : int * int -> int
1599		orb  : int * int -> int
1600		notb : int -> int
1601		<<   : int * word -> int
1602		~>>  : int * word -> int
1603
1604		Not tested, I hope they are correct.
1605
1606		----------------------------------------------------------------------
1607		Name: Allen Leung
1608		Date: Fri Dec  8 19:23:26 EST 2000
1609		Description:
1610		Tag:  leunga-20001208-nowhere
1611		Description:
1612
1613		Slight improvements to the 'nowhere' tool to handle OR-patterns,
1614		to generate better error messages etc.  Plus a brief manual.
1615
1616		----------------------------------------------------------------------
1617		Name: Lal George
1618		Date: 2000/12/08 09:54:02 EST 2000
1619		Tag: Release_110_31
1620		Description:
1621
1622		- Version 110.31
1623		----------------------------------------------------------------------
1624		Name: Allen Leung
1625		Date: Thu Dec  7 22:01:04 EST 2000
1626		Tag:  leunga-20001207-cell-monster-hack
1627		Description:
1628
1629		Major MLRISC internal changes.  Affect all clients.
1630		Summary:
1631
1632		1.  Type CELLS.cell = int is now replaced by a datatype.
1633		As a result, the old regmap is now gone.  Almost all interfaces
1634		in MLRISC change as a consequence.
1635
1636		2.  A new brand version of machine description tool (v3.0) that generates
1637		modules expecting the new interface.  The old version is removed.
1638
1639		3.  The RA interface has been further abstracted into two new functors.
1640		RISC_RA and X86RA.  These functors have much simpler interfaces.
1641		[See also directory MLRISC/demo.]
1642
1643		4.  Some other new source->source code generation tools are available:
1644
1645		a. MLRISC/Tools/RewriteGen -- generate rewriters from rules.
1646		b. MLRISC/Tools/WhereGen -- expands conditional pattern matching rules.
1647		I use this tool to generate the peephole optimizers---with the new
1648		cell type changes, peephole rules are becoming difficult to write
1649		without conditional pattern matching.
1650
1651		5.  More Intmap -> IntHashTable change.  Previous changes by Matthias didn't
1652		cover the entire MLRISC source tree so many things broke.
1653
1654		6.  CM files have been moved to the subdirectory MLRISC/cm.
1655		They are moved because there are a lot of them and they clutter up the
1656		root dir.
1657
1658		7.  More detailed documentation to come...
1659
1660		NOTE: To rebuild from 110.30 (ftp distribution), you'll have to do
1661		a makeml -rebuild first.  This is because of other other
1662		changes that Matthias has made (see below).
1663
1664
1665		----------------------------------------------------------------------
1666		Name: Matthias Blume
1667		Date: 2000/11/30 23:12:00 JST
1668		Tag: blume-20001130-filereorg
1669		Description:
1670
1671		Some manual updates and some file reorganizations in CM.
1672
1673		----------------------------------------------------------------------
1674		Name: Matthias Blume
1675		Date: 2000/11/24 17:45:00 JST
1676		Tag: blume-20001124-link
1677		Description:
1678
1679		Drastically improved link traversal code for the case that the dynamic
1680		value was already loaded at bootstrap time.  As a result, CM and CMB
1681		now both load blazingly fast -- even on a very slow machine.  Also,
1682		memory consumption has been further reduced by this.
1683
1684		Warning: The format of the PIDMAP file has changed.  THerefore, to
1685		bootstrap you have to do this:
1686
1687		1. Run CMB.make
1688		2. Make a symbolic link for the boot directory:
1689		ln -s sml.boot.ARCH-OS xxx
1690		3. "Rebuild" the boot directory:
1691		./makeml -boot xxx -rebuild sml ; rm xxx
1692		4. Boot normally:
1693		./makeml
1694
1695		----------------------------------------------------------------------
1696		Name: Matthias Blume
1697		Date: 2000/11/21 21:20:00 JST
1698		Tag: blume-20001121-tools
1699		Description:
1700
1701		Continued hacking on autoloading problem -- with success this time.
1702		Also changed tool-plugin mechanism.  See new CM manual.
1703
1704		----------------------------------------------------------------------
1705		Name: Matthias Blume
1706		Date: 2000/11/19 14:30:00 JST
1707		Tag:  blume-20001119-autoload
1708		Description:
1709
1710		Some hacking to make autoloading faster.  Success for CMB, no success
1711		so far for CM.  There is a reduced structure CM' that autoloads faster.
1712		(This is a temporary, non-documented hack to be eliminated again when
1713		the general problem is solved.)
1714
1715		----------------------------------------------------------------------
1716		Name: Matthias Blume
1717		Date: 2000/11/17 14:10:00 JST
1718		Tag: blume-20001117-pickle-lib
1719		Description:
1720
1721		1. Eliminated comp-lib.cm
1722		2. Made pickle-lib.cm
1723		3. Eliminated all uses of intset.sml (from comp-lib.cm)
1724		4. Replaced all uses of intmap.{sig,sml} (from comp-lib.cm) with
1725		equivalent constructs from smlnj-lib.cm (INtHashTable).
1726		5. Point 4. also goes for those uses of intmap.* in MLRISC.
1727		Duplicated intmap modules thrown out.
1728		6. Hunted down all duplicated SCC code and replaced it with
1729		equivalent stuff (GraphSCCFn from smlnj-lib.cm).
1730		7. Rewrote Feedback module.
1731		8. Moved sortedlist.sml into viscomp-lib.cm.  Eventually it
1732		should be thrown out and equivalent modules from smlnj-lib.cm
1733		should be used (IntRedBlackSet, IntListSet, ...).
1734
1735		Confirmed that compiler compiles to fixpoint.
1736
1737		----------------------------------------------------------------------
1738		Name: Allen Leung
1739		Date: 2000/11/10 18:00:00
1740		Tag: leunga-20001110-new-x86-fp
1741
1742		A new x86 floating point code generator has been added.
1743		By default this is turned off.  To turn this on, do:
1744
1745		CM.autoload "$smlnj/compiler.cm";
1746		Compiler.Control.MLRISC.getFlag "x86-fast-fp" := true;
1747
1748		Changes:
1749
1750		1.  Changed FTAN to FPTAN so that the assembly output is correct.
1751		2.  Changed the extension callback for FTANGENT to generate:
1752
1753		fptan
1754		fstp  %st(0)
1755		instead of
1756		fptan
1757		fstpl ftempmem
1758
1759		3.  Numerous assembly fixes for x86.
1760
1761		5.  Cleaned up the machine code output module x86/x86MC.sml and added
1762		support for a whole bunch of instructions and addressing modes:
1763
1764		fadd/fsub/fsubr/fmul/fdiv/fdivr  %st, %st(n)
1765		faddp/fsubp/fsubrp/fmulp/fdivp/fdivrp  %st, %st(n)
1766		fadd/fsub/fsubr/fmul/fdiv/fdivr  %st(n), %st
1767		fiadd/fisub/fisubr/fimul/fidiv/fidivr mem
1768		fxch %st(n)
1769		fld %st(n)
1770		fst %st(n)
1771		fst mem
1772		fstp %st(n)
1773		fucom %st(n)
1774		fucomp %st(n)
1775
1776		All these are now generated when the fast fp mode is turned on.
1777
1778		6.  Removed the dedicated registers %st(0), ..., %st(7) from X86CpsRegs
1779
1780		----------------------------------------------------------------------
1781		Name: Matthias Blume
1782		Date: 2000/11/09 11:20:00 JST
1783		Tag: blume-20001109-scc
1784		Description:
1785
1786		Eliminated some code duplication:
1787
1788		1. Added "where" clause to GraphSCCFn in SML/NJ Library.
1789		(Otherwise the functor is useless.)
1790		2. Used GraphSCCFn where SCCUtilFun was used previously.
1791		3. Got rid of SCCUtilFun (in comp-lib.cm).
1792
1793		----------------------------------------------------------------------
1794		Name: Lal George
1795		Date: 2000/11/06 09:02:21 EST 2000
1796		Tag: Release_110_30
1797		Description:
1798
1799		- Version 110.30
1800		----------------------------------------------------------------------
1801		Name: Matthias Blume
1802		Date: 2000/11/04 14:45:00
1803		Tag: blume-20001104-mlbuild
1804		Description:
1805
1806		- Made ml-build faster on startup.
1807		- Documentation fixes.
1808
1809		----------------------------------------------------------------------
1810		Name: Matthias Blume
1811		Date: 2000/11/02 17:00:00 JST
1812		Tag: blume-20001102-condcomp
1813		Description:
1814
1815		- Small tweaks to pickler -- new BOOTFILES!
1816		- Version bumped to 110.29.2.
1817		- Added conditional compilation facility to init.cmi (see comment there).
1818		----------------------------------------------------------------------
1819		Name: Allen Leung
1820		Date: 2000/10/23 19:31:00
1821		Tag: leunga-20001023-demo-ra
1822
1823		1. Minor RA changes that improves spilling on x86 (affects Moby and C-- only)
1824		2. Test programs for the graph library updated
1825		3. Some new MLRISC demo programs added
1826
1827		----------------------------------------------------------------------
1828		Name: Matthias Blume
1829		Date: 2000/08/31 22:15:00 JST
1830		Tag: blume-20001017-errmsg
1831		Description:
1832
1833		More error message grief: Where there used to be no messages, there
1834		now were some that had bogus error regions.  Fixed.
1835
1836		----------------------------------------------------------------------
1837		Name: Matthias Blume
1838		Date: 2000/08/31 17:30:00 JST
1839		Tag: blume-20001017-v110p29p1
1840		Description:
1841
1842		I made a version 110.29.1 with new bootfiles.
1843
1844		Changes:  Modified pickler/unpickler for faster and leaner unpickling.
1845		CM documentation changes and a small bugfix in CM's error reporting.
1846
1847		----------------------------------------------------------------------
1848		Name: Lal George
1849		Date: 2000/09/27 14:42:35 EDT
1850		Tag: george-20000927-nodestatus
1851		Description:
1852
1853		Changed the type of the nodestatus, so that:
1854
1855		SPILLED(~1)             is now SPILLED
1856		SPILLED(m) where m>=0   is now MEMREG(m)
1857		SPILLED(s) where s<~1   is now SPILL_LOC(~s)
1858
1859		----------------------------------------------------------------------
1860		Name: Matthias Blume
1861		Date: 2000/09/07 14:45:00 JST
1862		Tag: blume-20000907-cmerrmsg
1863		Description:
1864
1865		Small tweak to CM to avoid getting ML syntax error messages twice.
1866
1867		----------------------------------------------------------------------
1868	Name: Matthias Blume	Name: Matthias Blume
1869	Date: 2000/08/31 18:00:00 JST	Date: 2000/08/31 18:00:00 JST
1870	Tag: blume-20000831-cvsbootfiles	Tag: blume-20000831-cvsbootfiles
#	Line 1479	Line 3332
3332	elaborator). There were a lot of changes during my "linkpath" trials	elaborator). There were a lot of changes during my "linkpath" trials
3333	that could have been reverted to their original state but weren't.	that could have been reverted to their original state but weren't.
3334	Please, don't be too harsh on me for messing with this code a bit more	Please, don't be too harsh on me for messing with this code a bit more
3335	than what was strictly necessary...  (I _did_ resist the tempation	than what was strictly necessary...  (I _did_ resist the temptation
3336	of doing any "global reformatting" to avoid an untimely death at	of doing any "global reformatting" to avoid an untimely death at
3337	Dave's hands. :)	Dave's hands. :)
3338

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.701   changed lines   Added in v.1009

---

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