Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 873, Mon Jul 9 15:26:20 2001 UTC	revision 1047, Tue Jan 29 21:31:50 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	----------------------------------------------------------------------	----------------------------------------------------------------------
16	Name: Matthias Blume	Name: Matthias Blume
17		Date: 2002/01/29
18		Tag: blume-20020129-INSTALL
19		Description:
20
21		- Added somewhat detailed installation instructions (file INSTALL).
22		- Fixed curl-detection bug in config/install.sh.
23		- It is now possible to select the URL getter using the URLGETTER
24		environment variable:
25
26		not set / "unknown"      --> automatic detection (script tries wget,
27		curl, and lynx)
28		"wget" / "curl" / "lynx" --> use the specified program (script "knows"
29		how to properly invoke them)
30		other                    --> use $URLGETTER directly, it must take
31		precisely two command-line arguments
32		(source URL and destination file name)
33
34		----------------------------------------------------------------------
35		Name: Matthias Blume
36		Date: 2002/01/28
37		Tag: blume-20020128-sparc-ccalls
38		Description:
39
40		- Fixed problem with calculation of "used" registers in sparc-c-calls.
41		- Make use of the allocParam argument in sparc-c-calls.
42
43		----------------------------------------------------------------------
44		Name: Matthias Blume
45		Date: 2002/01/28
46		Tag: blume-20020128-allocParam
47		Description:
48
49		John Reppy:  Changes c-calls API to accept client-callback for
50		allocating extra stack space.
51		me: Corresponding changes to mlriscGen (using a dummy argument that
52		does not change the current behavior).
53
54		----------------------------------------------------------------------
55		Name: Matthias Blume
56		Date: 2002/01/28 12:00:00
57		Tag: Release_110_38
58		Description:
59
60		This time for real!!!
61
62		----------------------------------------------------------------------
63		Name: Matthias Blume
64		Date: 2002/01/28 10:56:00 EST
65		Tag: blume-20020128-retraction
66		Description:
67
68		0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
69		with blume-Release_110_38-retracted.)
70
71		1. Fixed a problem with incorrect rounding modes in real64.sml.
72		(Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
73
74		2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
75		and enums fixed.  The naming of corresponding ML identifiers should
76		now be consistent again.
77
78		----------------------------------------------------------------------
79		Name: Allen Leung
80		Date: 2002/01/27
81		Tag: leunga-20020127-nowhere
82		Description:
83
84		Added a target called nowhere in the configuration scripts.
85		Enabling this will build the MLRISC 'nowhere' tool (for translating
86		programs with where-clauses into legal SML code) during installation.
87
88		----------------------------------------------------------------------
89		Name: Matthias Blume
90		Date: 2002/01/25 21:27:00 EST
91		Tag: blume-Release_110_38-retracted
92		Description:
93
94		Call it a (working) release!  Version is 110.38. Bootfiles are ready.
95
96		README will be added later.
97
98		!!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
99		(Release_110_38) removed.  Reason: Last-minute bug fixes.
100
101		----------------------------------------------------------------------
102		Name: Matthias Blume
103		Date: 2002/01/25
104		Tag: blume-20020125-ffi
105		Description:
106
107		A large number of tweaks and improvements to ml-nlffi-lib and
108		ml-nlffigen:
109
110		- ML represenation types have been streamlined
111		- getter and setter functions work with concrete values, not abstract
112		ones where possible
113		- ml-nlffigen command line more flexible (see README file there)
114		- some bugs have been fixed (hopefully)
115
116		----------------------------------------------------------------------
117		Name: Lal George
118		Date: 2002/01/24
119		Tag: george-20020124-risc-ra-interface
120		Description:
121
122		There is a dramatic simplification in the interface to the
123		register allocator for RISC architectures as a result of making
124		parallel copy instructions explicit.
125
126		----------------------------------------------------------------------
127		Name: Matthias Blume
128		Date: 2002/01/22
129		Tag: blume-20020122-x86-ccalls
130		Description:
131
132		Bug fix for c-calls on x86 (having to do with how char- and
133		short-arguments are being handled).
134
135		----------------------------------------------------------------------
136		Name: Matthias Blume
137		Date: 2002/01/21
138		Tag: blume-20020121-ff
139		Description:
140
141		Another day of fiddling with the FFI...
142
143		1. Bug fix/workaround:  CKIT does not complain about negative array
144		dimensions, so ml-nlffigen has to guard itself against this possibility.
145		(Otherwise a negative dimension would send it into an infinite loop.)
146
147		2. Some of the abstract types (light objects, light pointers, most "base"
148		types) in structure C are now eqtypes.
149
150		3. Added constructors and test functions for NULL function pointers.
151
152		----------------------------------------------------------------------
153		Name: Matthias Blume
154		Date: 2002/01/18
155		Tag: blume-20020118-ready-for-new-release
156		Description:
157
158		Made config/srcarchiveurl point to a new place.  (Will provide boot
159		files shortly.)
160
161		Maybe we christen this to be 110.38?
162
163		----------------------------------------------------------------------
164		Name: Matthias Blume
165		Date: 2002/01/18
166		Tag: blume-20020118-more-ffifiddle
167		Description:
168
169		Today's FFI fiddling:
170
171		- Provided a structure CGetSet with "convenient" versions of C.Get.* and
172		C.Set.* that use concrete (MLRep.*) arguments and results instead
173		of abstract ones.
174
175		- Provided word-style bit operations etc. for "int" representation
176		types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
177		and Long.
178
179		----------------------------------------------------------------------
180		Name: Matthias Blume
181		Date: 2002/01/18
182		Tag: blume-20020118-use-x86-fp
183		Description:
184
185		Now that x86-fast-fp seems to be working, I turned it back on again
186		by default.  (Seems to work fine now, even with the FFI.)
187
188		Other than that, I added some documentation about the FFI to
189		src/ml-nlffigen/README and updated the FFI test examples in
190		src/ml-nlffi-lib/Tests/*.
191
192		----------------------------------------------------------------------
193		Name: Allen Leung
194		Date: 2002/01/17
195		Tag: leunga-20020117-x86-fast-fp-call
196		Description:
197
198		1. Fixed a problem with handling return fp values when x86's fast fp
199		mode is turned on.
200
201		2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
202		of %f32.
203
204		3. Added a constructor INT32lit to the ast of MLRISC tools.
205
206		----------------------------------------------------------------------
207		Name: Matthias Blume
208		Date: 2002/01/16
209		Tag: blume-20020116-ffifiddle
210		Description:
211
212		More fiddling with the FFI interface:
213
214		- Make constness 'c instead of rw wherever possible.  This eliminates
215		the need for certain explicit coercions.  (However, due to ML's
216		value polymorphism, there will still be many cases where explicit
217		coercions are necessary.  Phantom types are not the whole answer
218		to modeling a subtyping relationship in ML.)
219
220		- ro/rw coersions for pointers added.  (Avoids the detour through */&.)
221
222		- "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
223		clumsy workaround for varargs problem.)
224
225		----------------------------------------------------------------------
226		Name: Lal George
227		Date: 2002/01/15
228		Tag: <none>
229		Description:
230
231		1. Since COPY instructions are no longer native to the architecture,
232		a generic functor can be used to implement the expandCopies function.
233
234		2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
235		TEXT segment.
236
237		----------------------------------------------------------------------
238		Name: Matthias Blume
239		Date: 2002/01/15
240		Tag: blume-20020115-ffiupdates
241		Description:
242
243		1. Fix for bug resulting in single-precision float values being returned
244		incorrectly from FFI calls.
245
246		2. Small modifications to C FFI API:
247
248		- memory-allocation routines return straight objects (no options)
249		and raise an exception in out-of-memory situations
250		- unsafe extensions to cast between function pointers and pointers
251		from/to ints
252		- added structure C_Debug as an alternative to structure C where
253		pointer-dereferencing (|*| and |*!) always check for null-pointers
254		- added open_lib' to DynLinkage;  open_lib' works like open_lib
255		but also takes a (possibly empty) list of existing library handles
256		that the current library depends on
257
258		----------------------------------------------------------------------
259		Name: Matthias Blume
260		Date: 2002/01/10
261		Tag: blume-20020110-newffigen
262		Description:
263
264		1. Updates to portable graph code.
265
266		2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
267		more scalable now so that even huge interfaces such as the one
268		for GTK compile in finite time and space. :-)
269		See src/ml-nlffigen/README for details on what's new.
270
271		----------------------------------------------------------------------
272		Name: Lal George
273		Date: 2001/01/09 14:31:35 EST 2002
274		Tag: george-20011206-rm-native-copy
275		Description:
276
277		Removed the native COPY and FCOPY instructions
278		from all the architectures and replaced it with the
279		explicit COPY instruction from the previous commit.
280
281		It is now possible to simplify many of the optimizations
282		modules that manipulate copies. This has not been
283		done in this change.
284
285		----------------------------------------------------------------------
286		Name: Lal George
287		Date: 2001/12/06 16:50:13 EST 2001
288		Tag: george-20011206-mlrisc-instruction
289		Description:
290
291		Changed the representation of instructions from being fully abstract
292		to being partially concrete. That is to say:
293
294		from
295		type instruction
296
297		to
298		type instr                              (* machine instruction *)
299
300		datatype instruction =
301		LIVE of {regs: C.cellset, spilled: C.cellset}
302		| KILL of {regs: C.cellset, spilled: C.cellset}
303		| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
304		| ANNOTATION of {i: instruction, a: Annotations.annotation}
305		| INSTR of instr
306
307		This makes the handling of certain special instructions that appear on
308		all architectures easier and uniform.
309
310		LIVE and KILL say that a list of registers are live or killed at the
311		program point where they appear. No spill code is generated when an
312		element of the 'regs' field is spilled, but the register is moved to
313		the 'spilled' (which is present, more for debugging than anything else).
314
315		LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
316		We used to generate:
317
318		DEFFREG f1
319		f1 := f2 + f3
320		trapb
321
322		but now generate:
323
324		f1 := f2 + f3
325		trapb
326		LIVE {regs=[f1,f2,f3], spilled=[]}
327
328		Furthermore, the DEFFREG (hack) required that all floating point instruction
329		use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
330		defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
331		in a cleaner alpha implementation. (Hopefully, intel will not get rid of
332		this architecture).
333
334		COPYXXX is intended to replace the parallel COPY and FCOPY  available on
335		all the architectures. This will result in further simplification of the
336		register allocator that must be aware of them for coalescing purposes, and
337		will also simplify certain aspects of the machine description that provides
338		callbacks related to parallel copies.
339
340		ANNOTATION should be obvious, and now INSTR represents the honest to God
341		machine instruction set!
342
343		The <arch>/instructions/<arch>Instr.sml files define certain utility
344		functions for making porting easier -- essentially converting upper case
345		to lower case. All machine instructions (of type instr) are in upper case,
346		and the lower case form generates an MLRISC instruction. For example on
347		the alpha we have:
348
349		datatype instr =
350		LDA of {r:cell, b:cell, d:operand}
351		| ...
352
353		val lda : {r:cell, b:cell, d:operand} -> instruction
354		...
355
356		where lda is just (INSTR o LDA), etc.
357
358		----------------------------------------------------------------------
359		Name: Matthias Blume
360		Date: 2001/11/22 21:40:00 EST
361		Tag: Release_110_37
362		Description:
363
364		Release 110.37.  This time for real.
365
366		----------------------------------------------------------------------
367		Name: Matthias Blume
368		Date: 2001/11/21 16:35:00 EST
369		Tag: blume-20011121-foot-in-mouth
370		Description:
371
372		Removed the "Release_110_37" tag because of a serious bug.
373		This will be re-tagged once the bug is fixed.
374
375		----------------------------------------------------------------------
376		Name: Matthias Blume
377		Date: 2001/11/21 16:14:00 EST
378		Tag: blume-20011121-forgottenfile
379		Description:
380
381		Forgot to add a file.  (Just a .tex-file -- part of
382		the CM manual source.)
383
384		----------------------------------------------------------------------
385		Name: Matthias Blume
386		Date: 2001/11/21 16:10:00 EST
387		Tag: blume-20011121-invalid_110_37
388		Description:
389
390		Note: I removed the original tag "Release_110_37" from this commit
391		because we found a serious bug in all non-x86 backends.
392		- Matthias
393
394		1. Modifications to the SML/NJ code generator and to the runtime system
395		so that code object name strings are directly inserted into code
396		objects at code generation time.  The only business the runtime system
397		has with this is now to read the name strings on occasions.
398		(The encoding of the name string has also changed somewhat.)
399
400		2. CM now implements a simple "set calculus" for specifying export lists.
401		In particular, it is now possible to refer to the export lists of
402		other libraries/groups/sources and form unions as well as differences.
403		See the latest CM manual for details.
404
405		3. An separate notion of "proxy" libraries has again be eliminated from
406		CM's model.  (Proxy libraries are now simply a special case of using
407		the export list calculus.)
408
409		4. Some of the existing libraries now take advantage of the new set
410		calculus.
411		(Notice that not all libraries have been converted because some
412		of the existing .cm-files are supposed to be backward compatible
413		with 110.0.x.)
414
415		5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
416		"General.exnMessage"!  The former relies on a certain hook to be
417		initialized, and that often does not happen in the stand-alone case.)
418
419		----------------------------------------------------------------------
420		Name: Lal George
421		Date: 2001/11/21  13:56:18 EST
422		Tag: george-2001121-pseudo-ops
423		Description:
424
425		Implemented a complete redesign of MLRISC pseudo-ops. Now there
426		ought to never be any question of incompatabilities with
427		pseudo-op syntax expected by host assemblers.
428
429		For now, only modules supporting GAS syntax are implemented
430		but more should follow, such as MASM, and vendor assembler
431		syntax, e.g. IBM as, Sun as, etc.
432
433		----------------------------------------------------------------------
434		Name: Matthias Blume
435		Date: 2001/11/14 11:52:00 EST
436		Tag: blume-20011114-srcname
437		Description:
438
439		1. Routed the name of the current source file to mlriscgen where it
440		should be directly emitted into the code object.  (This last part
441		is yet to be done.)
442
443		2. Some cleanup of the pgraph code to make it match the proposal that
444		I put out the other day.  (The proposal notwithstanding, things are
445		still in flux here.)
446
447		----------------------------------------------------------------------
448		Name: Lal George
449		Date: 2001/11/14 09:44:04 EST
450		Tag:
451		Description:
452
453		Fix for a backpatching bug reported by Allen.
454
455		Because the boundary between short and long span-dependent
456		instructions is +/- 128, there are an astounding number of
457		span-dependent instructions whose size is over estimated.
458
459		Allen came up with the idea of letting the size of span
460		dependent instructions be non-monotonic, for a maxIter
461		number of times, after which the size must be monotonically
462		increasing.
463
464		This table shows the number of span-dependent instructions
465		whose size was over-estimated as a function of maxIter, for the
466		file Parse/parse/ml.grm.sml:
467
468		maxIter            # of instructions:
469		10                      687
470		20                      438
471		30                      198
472		40                        0
473
474		In compiling the compiler, there is no significant difference in
475		compilation speed between maxIter=10 and maxIter=40. Actually,
476		my measurements showed that maxIter=40 was a tad faster than
477		maxIter=10! Also 96% of the  files in the compiler reach a fix
478		point within 13 iterations, so fixing maxIter at 40, while high,
479		is okay.
480
481		----------------------------------------------------------------------
482		Name: Matthias Blume
483		Date: 2001/10/31 15:25:00 EST
484		Tag: blume-20011031-pgraph
485		Description:
486
487		CKIT:
488		* Changed the "Function" constructor of type Ast.ctype to carry optional
489		argument identifiers.
490		* Changed the return type of TypeUtil.getFunction accordingly.
491		* Type equality ignores the argument names.
492		* TypeUtil.composite tries to preserve argument names but gives up quickly
493		if there is a mismatch.
494
495		installation script:
496		* attempts to use "curl" if available (unless "wget" is available as well)
497
498		CM:
499		* has an experimental implementation of "portable graphs" which I will
500		soon propose as an implementation-independent library format
501		* there are also new libraries $/pgraph.cm and $/pgraph-util.cm
502
503		NLFFI-LIB:
504		* some cleanup (all cosmetic)
505
506		NLFFIGEN:
507		* temporarily disabled the mechanism that suppresses ML output for
508		C definitions whose identifiers start with an underscore character
509		* generate val bindings for enum constants
510		* user can request that only one style (light or heavy) is being used;
511		default is to use both (command-line arguments: -heavy and -light)
512		* fixed bug in handling of function types involving incomplete pointers
513		* generate ML entry points that take record arguments (i.e., using
514		named arguments) for C functions that have a prototype with named
515		arguments
516		(see changes to CKIT)
517
518		----------------------------------------------------------------------
519		Name: Allen Leung
520		Date: 2001/10/27 20:34:00 EDT
521		Tag: leunga-20011027-x86-fast-fp-call
522		Description:
523
524		Fixed the bug described in blume-20010920-slowfp.
525
526		The fix involves
527		1. generating FCOPYs in FSTP in ia32-svid
528		2. marking a CALL with the appropriate annotation
529
530		----------------------------------------------------------------------
531		Name: Matthias Blume
532		Date: 2001/10/16 11:32:00 EDT
533		Tag: blume-20011016-netbsd
534		Description:
535
536		Underscore patch from Chris Richards (fixing problem with compiling
537		runtime system under recent NetBSD).
538
539		----------------------------------------------------------------------
540		Name: Allen Leung
541		Date: 2001/10/12 17:18:32 EDT 2001
542		Tag: leung-20011012-x86-printflowgraph
543		Description:
544
545		X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
546
547		----------------------------------------------------------------------
548		Name: Lal George
549		Date: 2001/10/11 23:51:34 EDT
550		Tag: george-20011011-too-many-instrs
551		Description:
552
553		The representation of a program point never expected to see more
554		than 65536 instructions in a basic block!
555
556		----------------------------------------------------------------------
557		Name: Lal George
558		Date: 2001/10/09 09:41:37 EDT
559		Tag: george-20011008-mlrisc-labels
560		Description:
561
562		Changed the machine description files to support printing of
563		local and global labels in assembly code, based on host assembler
564		conventions.
565
566		----------------------------------------------------------------------
567		Name: Matthias Blume
568		Date: 2001/09/25 15:25:00 EDT
569		Tag: blume-20010925-exninfo
570		Description:
571
572		I provided a non-hook implementation of exnName (at the toplevel) and
573		made the "dummy" implementation of exnMessage (at the toplevel) more
574		useful: if nothing gets "hooked in", then at least you are going to
575		see the exception name and a message indicating why you don't see more.
576
577		[For the time being, programs that need exnMessage and want to use
578		ml-build should either use General.exnMessage (strongly recommended) or
579		refer to structure General at some other point so that CM sees a
580		static dependency.]
581
582		[Similar remarks go for "print" and "use":  If you want to use their
583		functionality in stand-alone programs generated by ml-build, then use
584		TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
585
586		----------------------------------------------------------------------
587		Name: Matthias Blume
588		Date: 2001/09/20 17:28:00 EDT
589		Tag: blume-20010920-slowfp
590		Description:
591
592		Allen says that x86-fast-fp is not safe yet, so I turned it off again...
593
594		----------------------------------------------------------------------
595		Name: Matthias Blume
596		Date: 2001/09/20 17:20:00 EDT
597		Tag: blume-20010920-canonicalpaths
598		Description:
599
600		0. Updated the BOOT file (something that I forgot to do earlier).
601
602		1. Small internal change to CM so that it avoids "/../" in filenames
603		as much as possible (but only where it is safe).
604
605		2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
606		that arguments that contain delimiters are passed through correctly.
607		This change also means that all "special" arguments of the form
608		@SMLxxx... must come first.
609
610		3. Changed install script to put relative anchor names for tool commands
611		into pathconfig.
612
613		----------------------------------------------------------------------
614		Name: Matthias Blume
615		>>>>>>> 1.169
616		Date: 2001/09/18 15:35:00 EDT
617		Tag: blume-20010918-readme11036
618		Description:
619
620		Added README files.
621
622		----------------------------------------------------------------------
623		Name: Matthias Blume
624		Date: 2001/09/18 11:45:00 EDT
625		Tag: Release_110_36 (retag)
626		Description:
627
628		Fixed mistake in config/preloads. Retagged as 110.36.
629
630		----------------------------------------------------------------------
631		Name: Matthias Blume
632		Date: 2001/09/18 09:40:00 EDT
633		Tag: Release_110_36_orig (tag changed)
634		Description:
635
636		New version (110.36).  New bootfiles.
637
638		----------------------------------------------------------------------
639		Name: Matthias Blume
640		Date: 2001/09/14 16:15:00 EDT
641		Tag: blume-20010914-x86fastfp
642		Description:
643
644		John committed some changes that Allen made, in particular a (hopefully)
645		correctly working version of the x86-fp module.
646
647		I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
648		flag to "true".  Everything seems to compile to a fixpoint ok, and
649		"mandelbrot" speeds up by about 15%.
650
651		----------------------------------------------------------------------
652		Name: Matthias Blume
653		Date: 2001/09/13 11:20:00 EDT
654		Tag: blume-20010913-minimal
655		Description:
656
657		1. Stefan Monnier's patch to fix a miscompilation problem that
658		was brought to light by John Reppy's work on Moby.
659
660		2. Implemented a minimal "structure Compiler" that contains just
661		"version" and "architecture".  The minimal version will be
662		available when the full version is not.  This is for backward-
663		compatibility with code that wants to test Compiler.version.
664
665		----------------------------------------------------------------------
666		Name: Matthias Blume
667		Date: 2001/08/28 14:03:00 EDT
668		Tag: blume-20010828-ml-lex
669		Description:
670
671		Fix for bug 1581, received from Neophytos Michael.
672
673		----------------------------------------------------------------------
674		Name: Matthias Blume
675		Date: 2001/08/27 11:20:00 EDT
676		Tag: blume-20010827-readme11035
677		Description:
678
679		Fleshed out the README file for 110.35.
680
681		----------------------------------------------------------------------
682		Name: Matthias Blume
683		Date: 2001/08/24 17:10:00 EDT
684		Tag: Release_110_35
685		Description:
686
687		New version number (110.35).  New bootfiles.
688
689		----------------------------------------------------------------------
690		Name: Lal George
691		Date: 2001/08/24 13:47:18 EDT 2001
692		Tag: george-20010824-MLRISC-graphs
693		Description:
694
695		removed clusters from MLRISC completely and replaced with graphs.
696
697		----------------------------------------------------------------------
698		Name: Matthias Blume
699		Date: 2001/08/23 17:50:00 EDT
700		Tag: blume-20010823-toplevel
701		Description:
702
703		- some reorganization of the code that implements various kinds of
704		environments in the compiler (static, dynamic, symbolic, combined)
705		- re-implemented the EnvRef module so that evalStream works properly
706		(if the stream contains references to "use", "CM.make", etc.)
707		- cleaned up evalloop.sml and interact.sml (but they need more cleaning)
708
709		----------------------------------------------------------------------
710		Name: Matthias Blume
711		Date: 2001/08/20 15:50 EDT
712		Tag: blume20010820-slipup
713		Description:
714
715		I forgot to commit a few files.  Here they are...
716
717		----------------------------------------------------------------------
718		Name: Matthias Blume
719		Date: 2001/08/20 15:35:00 EDT
720		Tag: blume-20010820-debugprof
721		Description:
722
723		!!!! NEW BOOTFILES !!!!
724
725		This is another round of reorganizing the compiler sources.  This
726		time the main goal was to factor out all the "instrumentation"
727		passes (for profiling and backtracing) into their own library.
728		The difficulty was to do it in such a way that it does not depend
729		on elaborate.cm but only on elabdata.cm.
730
731		Therefore there have been further changes to both elaborate.cm and
732		elabdata.cm -- more "generic" things have been moved from the former
733		to the latter.  As a result, I was forced to split the assignment
734		of numbers indicating "primtyc"s into two portions: SML-generic and
735		SML/NJ-specific.  Since it would have been awkward to maintain,
736		I bit the bullet and actually _changed_ the mapping between these
737		numbers and primtycs.  The bottom line of this is that you need
738		a new set of bin- and bootfiles.
739
740		I have built new bootfiles for all architectures, so doing a fresh
741		checkout and config/install.sh should be all you need.
742
743		The newly created library's name is
744
745		$smlnj/viscomp/debugprof.cm
746
747		and its sources live under
748
749		src/compiler/DebugProf
750
751		----------------------------------------------------------------------
752		Name: Matthias Blume
753		Date: 2001/08/15 17:15:00 EDT
754		Tag: blume-20010815-compreorg
755		Description:
756
757		This is a first cut at reorganizing the CM libraries that make up the
758		core of the compiler.  The idea is to separate out pieces that could
759		be used independently by tools, e.g., the parser, the typechecker, etc.
760
761		The current status is a step in this direction, but it is not quite
762		satisfactory yet.  Expect more changes in the future.
763
764		Here is the current (new) organization...
765
766		What used to be $smlnj/viscomp/core.cm is now divided into
767		six CM libraries:
768
769		$smlnj/viscomp/basics.cm
770		/parser.cm
771		/elabdata.cm
772		/elaborate.cm
773		/execute.cm
774		/core.cm
775
776		The CM files for these libraries live under src/system/smlnj/viscomp.
777		All these libraries are proxy libraries that contain precisely
778		one CM library component.  Here are the locations of the components
779		(all within the src/compiler tree):
780
781		Basics/basics.cm
782		Parse/parser.cm
783		ElabData/elabdata.cm
784		Elaborator/elaborate.cm
785		Execution/execute.cm
786		core.cm
787
788		[This organization is the same that has been used already
789		for a while for the architecture-specific parts of the visible
790		compiler and for the old version of core.cm.]
791
792		As you will notice, many source files have been moved from their
793		respective original locations to a new home in one of the above
794		subtrees.
795
796		The division of labor between the new libraries is the following:
797
798		basics.cm:
799		- Simple, basic definitions that pertain to many (or all) of
800		the other libraries.
801		parser.cm:
802		- The SML parser, producing output of type Ast.dec.
803		- The type family for Ast is also defined and exported here.
804		elabdata.cm:
805		- The datatypes that describe input and output of the elaborator.
806		This includes types, absyn, and static environments.
807		elaborator.cm:
808		- The SML/NJ type checker and elaborator.
809		This maps an Ast.dec (with a given static environment) to
810		an Absyn.dec (with a new static environment).
811		- This libraries implements certain modules that used to be
812		structures as functors (to remove dependencies on FLINT).
813		execute.cm:
814		- Everything having to do with executing binary code objects.
815		- Dynamic environments.
816		core.cm:
817		- SML/NJ-specific instantiations of the elaborator and MLRISC.
818		- Top-level modules.
819		- FLINT (this should eventually become its own library)
820
821		Notes:
822
823		I am not 100% happy with the way I separated the elaborator (and its
824		data structures) from FLINT.  Two instances of the same problem:
825
826		1. Data structures contain certain fields that carry FLINT-specific
827		information.  I hacked around this using exn and the property list
828		module from smlnj-lib.  But the fact that there are middle-end
829		specific fields around at all is a bit annoying.
830
831		2. The elaborator calculates certain FLINT-related information.  I tried
832		to make this as abstract as I could using functorization, but, again,
833		the fact that the elaborator has to perform calculations on behalf
834		of the middle-end at all is not nice.
835
836		3. Having to used exn and property lists is unfortunate because it
837		weakens type checking.  The other alternative (parameterizing
838		nearly *everything*) is not appealing, though.
839
840		I removed the "rebinding =" warning hack because due to the new organization
841		it was awkward to maintain it.  As a result, the compiler now issues some of
842		these warnings when compiling init.cmi during bootstrap compilation. On
843		the plus side, you also get a warning when you do, for example:
844		val op = = Int32.+
845		which was not the case up to now.
846
847		I placed "assign" and "deref" into the _Core structure so that the
848		code that deals with the "lazy" keyword can find them there.  This
849		removes the need for having access to the primitive environment
850		during elaboration.
851
852		----------------------------------------------------------------------
853		Name: Matthias Blume
854		Date: 2001/08/13
855		Tag: blume-20010813-closures
856		Description:
857
858		This fix was sent to us by Zhong Shao.  It is supposed to improve the
859		performance of certain loops by avoiding needless closure allocation.
860
861		----------------------------------------------------------------------
862		Name: Lal George
863		Date: 2001/07/31 10:03:23 EDT 2001
864		Tag: george-20010731-x86-fmalloc
865		Description: Fixed bug in x86 calls
866
867		There was a bug where call instructions would mysteriously
868		vanish. The call instruction had to be one that returned
869		a floating point value.
870
871		----------------------------------------------------------------------
872		Name: Lal George
873		Date: 2001/07/19 16:36:29 EDT 2001
874		Tag: george-20010719-simple-cells
875		Description:
876
877		I have dramatically simplified the interface for CELLS in MLRISC.
878
879		In summary, the cells interface is broken up into three parts:
880
881		1. CellsBasis : CELLS_BASIS
882
883		CellsBasis is a top level structure and common for all
884		architectures.  it contains the definitions of basic datatypes
885		and utility  functions over these types.
886
887		2. functor Cells() : CELLS
888
889		Cells generates an interface for CELLS that incorporates the
890		specific resources on the target architecture, such as the
891		presence of special register classes, their number and size,
892		and various useful substructures.
893
894		3. <ARCH>CELLS
895
896		e.g. SparcCells: SPARCCELLS
897
898		<ARCH>CELLS usually contains additional bindings for special
899		registers  on the architecture, such as:
900
901		val r0 : cell           (* register zero *)
902		val y : cell            (* Y register *)
903		val psr : cell          (* processor status register *)
904		...
905
906		The structure returned by applying the Cells functor is opened
907		in this interface.
908
909		The main implication of all this is that the datatypes for cells is
910		split between CellsBasis and CELLS -- a fairly simple change for user
911		code.
912
913		In the old scheme the CELLS interface had a definitional binding of
914		the form:
915
916		signature CELLS = sig
917
918		structure CellsBasis = CellsBasis
919
920		...
921
922		end
923
924		With all the sharing constraints that goes on in MLRISC, this old
925		design  quickly leads to errors such as:
926
927		"structure definition spec inside of sharing ... "
928
929
930		and appears to require an unacceptable amount of sharing and where
931		constraint hackery.
932
933		I think this error message (the interaction of definitional specs and
934		sharing) requires more explanation on our web page.
935
936		----------------------------------------------------------------------
937		Name: Matthias Blume
938		Date: 2001/07/19 15:00:00 EDT
939		Tag: blume-20010719-libreorg
940		Description:
941
942		This update puts together a fairly extensive but straightforward change
943		to the way the libraries that implement the interactive system are
944		organized:
945
946		The biggest change is the elimination of structure Compiler.  As a
947		replacement for this structure, there is now a CM library
948		(known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
949		that exports all the substructures of the original structure Compiler
950		directly.  So instead of saying Compiler.Foo.bar one now simply
951		says Foo.bar.  (The CM libraries actually export a collection of
952		structures that is richer than the collection of substructures of
953		structure Compiler.)
954
955		To make the transition smooth, there is a separate library called
956		$smlnj/compiler/compiler.cm which puts together and exports the
957		original structure Compiler (or at least something very close to it).
958
959		There are five members of the original structure Compiler
960		that are not exported directly but which instead became members
961		of a new structure Backend (described by signature BACKEND).  These are:
962		structure Profile (: PROFILE), structure Compile (: COMPILE), structure
963		Interact (: INTERACT), structure Machine (: MACHINE), and val
964		architecture (: string).
965
966		Structure Compiler.Version has become structure CompilerVersion.
967
968		Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
969		by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
970		or x86, respectively.
971		Each of these exports the same frontend structures that
972		$smlnj/compiler.cm exports.  But they do not have a structure Backend
973		and instead export some structure <Arch>Backend where <Arch> is Alpha32,
974		Hppa, PPC, Sparc, or X86, respectively.
975
976		Library $smlnj/compiler/all.cm exports the union of the exports of
977		$smlnj/compiler/<arch>.cm
978
979		There are no structures <Arch>Compiler anymore, use
980		$smlnj/compiler/<arch>.cm instead.
981
982		Library host-compiler-0.cm is gone.  Instead, the internal library
983		that instantiates CM is now called cm0.cm.  Selection of the host
984		compiler (backend) is no longer done here but. (Responsibility for it
985		now lies with $smlnj/compiler/current.cm.  This seems to be more
986		logical.)
987
988		Many individual files have been moved or renamed.  Some files have
989		been split into multiple files, and some "dead" files have been deleted.
990
991		Aside from these changes to library organization, there are also changes
992		to the way the code itself is organized:
993
994		Structure Binfile has been re-implemented in such a way that it no
995		longer needs any knowledge of the compiler.  It exclusively deals
996		with the details of binfile layout.  It no longer invokes the
997		compiler (for the purpose of creating new prospective binfile
998		content), and it no longer has any knowledge of how to interpret
999		pickles.
1000
1001		Structure Compile (: COMPILE) has been stripped down to the bare
1002		essentials of compilation.  It no longer deals with linking/execution.
1003		The interface has been cleaned up considerably.
1004
1005		Utility routines for dealing with linking and execution have been
1006		moved into their own substructures.
1007
1008		(The ultimate goal of these changes is to provide a light-weight
1009		binfile loader/linker (at least for, e.g., stable libraries) that
1010		does not require CM or the compiler to be present.)
1011
1012		CM documentation has been updated to reflect the changes to library
1013		organization.
1014
1015		----------------------------------------------------------------------
1016		Name: Matthias Blume
1017		Date: 2001/07/10 17:30:00 EDT
1018		Tag: Release_110_34
1019		Description:
1020
1021		Minor tweak to 110.34 (re-tagged):
1022
1023		- README.html file added to CVS repository
1024		- runtime compiles properly under FreeBSD 3.X and 4.X
1025
1026		----------------------------------------------------------------------
1027		Name: Matthias Blume
1028		Date: 2001/07/10 17:30:00 EDT
1029		Tag: Release_110_34
1030		Description:
1031
1032		New version number (110.34). New bootfiles.
1033
1034		----------------------------------------------------------------------
1035		Name: Matthias Blume
1036		Date: 2001/07/09 16:00:00 EDT
1037		Tag: blume-20010709-more-varargs
1038		Description:
1039
1040		I changed the handling of varargs in ml-nlffigen again:
1041		The ellipsis ... will now simply be ignored (with an accompanying warning).
1042
1043		The immediate effect is that you can actually call a varargs function
1044		from ML -- but you can't actually supply any arguments beyond the ones
1045		specified explicitly.  (For example, you can call printf with its format
1046		string, but you cannot pass additional arguments.)
1047
1048		This behavior is only marginally more useful than the one before, but
1049		it has the advantage that a function or, more importantly, a function
1050		type never gets dropped on the floor, thus avoiding follow-up problems with
1051		other types that refer to the offending one.
1052
1053		----------------------------------------------------------------------
1054		Name: Matthias Blume
1055	Date: 2001/07/09 11:25:00 EDT	Date: 2001/07/09 11:25:00 EDT
1056	Tag: blume-20010709-varargs	Tag: blume-20010709-varargs
1057	Description:	Description:
#	Line 2551	Line 3589
3589	elaborator). There were a lot of changes during my "linkpath" trials	elaborator). There were a lot of changes during my "linkpath" trials
3590	that could have been reverted to their original state but weren't.	that could have been reverted to their original state but weren't.
3591	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
3592	than what was strictly necessary...  (I _did_ resist the tempation	than what was strictly necessary...  (I _did_ resist the temptation
3593	of doing any "global reformatting" to avoid an untimely death at	of doing any "global reformatting" to avoid an untimely death at
3594	Dave's hands. :)	Dave's hands. :)
3595

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.873   changed lines   Added in v.1047

---

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