Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 845, Thu Jun 21 16:25:49 2001 UTC	revision 1040, Mon Jan 28 16:55:24 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: 2002/01/28 12:00:00
18		Tag: Release_110_38
19		Description:
20
21		This time for real!!!
22
23		----------------------------------------------------------------------
24		Name: Matthias Blume
25		Date: 2002/01/28 10:56:00 EST
26		Tag: blume-20020128-retraction
27		Description:
28
29		0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
30		with blume-Release_110_38-retracted.)
31
32		1. Fixed a problem with incorrect rounding modes in real64.sml.
33		(Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
34
35		2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
36		and enums fixed.  The naming of corresponding ML identifiers should
37		now be consistent again.
38
39		----------------------------------------------------------------------
40		Name: Allen Leung
41		Date: 2002/01/27
42		Tag: leunga-20020127-nowhere
43		Description:
44
45		Added a target called nowhere in the configuration scripts.
46		Enabling this will build the MLRISC 'nowhere' tool (for translating
47		programs with where-clauses into legal SML code) during installation.
48
49		----------------------------------------------------------------------
50		Name: Matthias Blume
51		Date: 2002/01/25 21:27:00 EST
52		Tag: blume-Release_110_38-retracted
53		Description:
54
55		Call it a (working) release!  Version is 110.38. Bootfiles are ready.
56
57		README will be added later.
58
59		!!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
60		(Release_110_38) removed.  Reason: Last-minute bug fixes.
61
62		----------------------------------------------------------------------
63		Name: Matthias Blume
64		Date: 2002/01/25
65		Tag: blume-20020125-ffi
66		Description:
67
68		A large number of tweaks and improvements to ml-nlffi-lib and
69		ml-nlffigen:
70
71		- ML represenation types have been streamlined
72		- getter and setter functions work with concrete values, not abstract
73		ones where possible
74		- ml-nlffigen command line more flexible (see README file there)
75		- some bugs have been fixed (hopefully)
76
77		----------------------------------------------------------------------
78		Name: Lal George
79		Date: 2002/01/24
80		Tag: george-20020124-risc-ra-interface
81		Description:
82
83		There is a dramatic simplification in the interface to the
84		register allocator for RISC architectures as a result of making
85		parallel copy instructions explicit.
86
87		----------------------------------------------------------------------
88		Name: Matthias Blume
89		Date: 2002/01/22
90		Tag: blume-20020122-x86-ccalls
91		Description:
92
93		Bug fix for c-calls on x86 (having to do with how char- and
94		short-arguments are being handled).
95
96		----------------------------------------------------------------------
97		Name: Matthias Blume
98		Date: 2002/01/21
99		Tag: blume-20020121-ff
100		Description:
101
102		Another day of fiddling with the FFI...
103
104		1. Bug fix/workaround:  CKIT does not complain about negative array
105		dimensions, so ml-nlffigen has to guard itself against this possibility.
106		(Otherwise a negative dimension would send it into an infinite loop.)
107
108		2. Some of the abstract types (light objects, light pointers, most "base"
109		types) in structure C are now eqtypes.
110
111		3. Added constructors and test functions for NULL function pointers.
112
113		----------------------------------------------------------------------
114		Name: Matthias Blume
115		Date: 2002/01/18
116		Tag: blume-20020118-ready-for-new-release
117		Description:
118
119		Made config/srcarchiveurl point to a new place.  (Will provide boot
120		files shortly.)
121
122		Maybe we christen this to be 110.38?
123
124		----------------------------------------------------------------------
125		Name: Matthias Blume
126		Date: 2002/01/18
127		Tag: blume-20020118-more-ffifiddle
128		Description:
129
130		Today's FFI fiddling:
131
132		- Provided a structure CGetSet with "convenient" versions of C.Get.* and
133		C.Set.* that use concrete (MLRep.*) arguments and results instead
134		of abstract ones.
135
136		- Provided word-style bit operations etc. for "int" representation
137		types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
138		and Long.
139
140		----------------------------------------------------------------------
141		Name: Matthias Blume
142		Date: 2002/01/18
143		Tag: blume-20020118-use-x86-fp
144		Description:
145
146		Now that x86-fast-fp seems to be working, I turned it back on again
147		by default.  (Seems to work fine now, even with the FFI.)
148
149		Other than that, I added some documentation about the FFI to
150		src/ml-nlffigen/README and updated the FFI test examples in
151		src/ml-nlffi-lib/Tests/*.
152
153		----------------------------------------------------------------------
154		Name: Allen Leung
155		Date: 2002/01/17
156		Tag: leunga-20020117-x86-fast-fp-call
157		Description:
158
159		1. Fixed a problem with handling return fp values when x86's fast fp
160		mode is turned on.
161
162		2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
163		of %f32.
164
165		3. Added a constructor INT32lit to the ast of MLRISC tools.
166
167		----------------------------------------------------------------------
168		Name: Matthias Blume
169		Date: 2002/01/16
170		Tag: blume-20020116-ffifiddle
171		Description:
172
173		More fiddling with the FFI interface:
174
175		- Make constness 'c instead of rw wherever possible.  This eliminates
176		the need for certain explicit coercions.  (However, due to ML's
177		value polymorphism, there will still be many cases where explicit
178		coercions are necessary.  Phantom types are not the whole answer
179		to modeling a subtyping relationship in ML.)
180
181		- ro/rw coersions for pointers added.  (Avoids the detour through */&.)
182
183		- "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
184		clumsy workaround for varargs problem.)
185
186		----------------------------------------------------------------------
187		Name: Lal George
188		Date: 2002/01/15
189		Tag: <none>
190		Description:
191
192		1. Since COPY instructions are no longer native to the architecture,
193		a generic functor can be used to implement the expandCopies function.
194
195		2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
196		TEXT segment.
197
198		----------------------------------------------------------------------
199		Name: Matthias Blume
200		Date: 2002/01/15
201		Tag: blume-20020115-ffiupdates
202		Description:
203
204		1. Fix for bug resulting in single-precision float values being returned
205		incorrectly from FFI calls.
206
207		2. Small modifications to C FFI API:
208
209		- memory-allocation routines return straight objects (no options)
210		and raise an exception in out-of-memory situations
211		- unsafe extensions to cast between function pointers and pointers
212		from/to ints
213		- added structure C_Debug as an alternative to structure C where
214		pointer-dereferencing (|*| and |*!) always check for null-pointers
215		- added open_lib' to DynLinkage;  open_lib' works like open_lib
216		but also takes a (possibly empty) list of existing library handles
217		that the current library depends on
218
219		----------------------------------------------------------------------
220		Name: Matthias Blume
221		Date: 2002/01/10
222		Tag: blume-20020110-newffigen
223		Description:
224
225		1. Updates to portable graph code.
226
227		2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
228		more scalable now so that even huge interfaces such as the one
229		for GTK compile in finite time and space. :-)
230		See src/ml-nlffigen/README for details on what's new.
231
232		----------------------------------------------------------------------
233		Name: Lal George
234		Date: 2001/01/09 14:31:35 EST 2002
235		Tag: george-20011206-rm-native-copy
236		Description:
237
238		Removed the native COPY and FCOPY instructions
239		from all the architectures and replaced it with the
240		explicit COPY instruction from the previous commit.
241
242		It is now possible to simplify many of the optimizations
243		modules that manipulate copies. This has not been
244		done in this change.
245
246		----------------------------------------------------------------------
247		Name: Lal George
248		Date: 2001/12/06 16:50:13 EST 2001
249		Tag: george-20011206-mlrisc-instruction
250		Description:
251
252		Changed the representation of instructions from being fully abstract
253		to being partially concrete. That is to say:
254
255		from
256		type instruction
257
258		to
259		type instr                              (* machine instruction *)
260
261		datatype instruction =
262		LIVE of {regs: C.cellset, spilled: C.cellset}
263		| KILL of {regs: C.cellset, spilled: C.cellset}
264		| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
265		| ANNOTATION of {i: instruction, a: Annotations.annotation}
266		| INSTR of instr
267
268		This makes the handling of certain special instructions that appear on
269		all architectures easier and uniform.
270
271		LIVE and KILL say that a list of registers are live or killed at the
272		program point where they appear. No spill code is generated when an
273		element of the 'regs' field is spilled, but the register is moved to
274		the 'spilled' (which is present, more for debugging than anything else).
275
276		LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
277		We used to generate:
278
279		DEFFREG f1
280		f1 := f2 + f3
281		trapb
282
283		but now generate:
284
285		f1 := f2 + f3
286		trapb
287		LIVE {regs=[f1,f2,f3], spilled=[]}
288
289		Furthermore, the DEFFREG (hack) required that all floating point instruction
290		use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
291		defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
292		in a cleaner alpha implementation. (Hopefully, intel will not get rid of
293		this architecture).
294
295		COPYXXX is intended to replace the parallel COPY and FCOPY  available on
296		all the architectures. This will result in further simplification of the
297		register allocator that must be aware of them for coalescing purposes, and
298		will also simplify certain aspects of the machine description that provides
299		callbacks related to parallel copies.
300
301		ANNOTATION should be obvious, and now INSTR represents the honest to God
302		machine instruction set!
303
304		The <arch>/instructions/<arch>Instr.sml files define certain utility
305		functions for making porting easier -- essentially converting upper case
306		to lower case. All machine instructions (of type instr) are in upper case,
307		and the lower case form generates an MLRISC instruction. For example on
308		the alpha we have:
309
310		datatype instr =
311		LDA of {r:cell, b:cell, d:operand}
312		| ...
313
314		val lda : {r:cell, b:cell, d:operand} -> instruction
315		...
316
317		where lda is just (INSTR o LDA), etc.
318
319		----------------------------------------------------------------------
320		Name: Matthias Blume
321		Date: 2001/11/22 21:40:00 EST
322		Tag: Release_110_37
323		Description:
324
325		Release 110.37.  This time for real.
326
327		----------------------------------------------------------------------
328		Name: Matthias Blume
329		Date: 2001/11/21 16:35:00 EST
330		Tag: blume-20011121-foot-in-mouth
331		Description:
332
333		Removed the "Release_110_37" tag because of a serious bug.
334		This will be re-tagged once the bug is fixed.
335
336		----------------------------------------------------------------------
337		Name: Matthias Blume
338		Date: 2001/11/21 16:14:00 EST
339		Tag: blume-20011121-forgottenfile
340		Description:
341
342		Forgot to add a file.  (Just a .tex-file -- part of
343		the CM manual source.)
344
345		----------------------------------------------------------------------
346		Name: Matthias Blume
347		Date: 2001/11/21 16:10:00 EST
348		Tag: blume-20011121-invalid_110_37
349		Description:
350
351		Note: I removed the original tag "Release_110_37" from this commit
352		because we found a serious bug in all non-x86 backends.
353		- Matthias
354
355		1. Modifications to the SML/NJ code generator and to the runtime system
356		so that code object name strings are directly inserted into code
357		objects at code generation time.  The only business the runtime system
358		has with this is now to read the name strings on occasions.
359		(The encoding of the name string has also changed somewhat.)
360
361		2. CM now implements a simple "set calculus" for specifying export lists.
362		In particular, it is now possible to refer to the export lists of
363		other libraries/groups/sources and form unions as well as differences.
364		See the latest CM manual for details.
365
366		3. An separate notion of "proxy" libraries has again be eliminated from
367		CM's model.  (Proxy libraries are now simply a special case of using
368		the export list calculus.)
369
370		4. Some of the existing libraries now take advantage of the new set
371		calculus.
372		(Notice that not all libraries have been converted because some
373		of the existing .cm-files are supposed to be backward compatible
374		with 110.0.x.)
375
376		5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
377		"General.exnMessage"!  The former relies on a certain hook to be
378		initialized, and that often does not happen in the stand-alone case.)
379
380		----------------------------------------------------------------------
381		Name: Lal George
382		Date: 2001/11/21  13:56:18 EST
383		Tag: george-2001121-pseudo-ops
384		Description:
385
386		Implemented a complete redesign of MLRISC pseudo-ops. Now there
387		ought to never be any question of incompatabilities with
388		pseudo-op syntax expected by host assemblers.
389
390		For now, only modules supporting GAS syntax are implemented
391		but more should follow, such as MASM, and vendor assembler
392		syntax, e.g. IBM as, Sun as, etc.
393
394		----------------------------------------------------------------------
395		Name: Matthias Blume
396		Date: 2001/11/14 11:52:00 EST
397		Tag: blume-20011114-srcname
398		Description:
399
400		1. Routed the name of the current source file to mlriscgen where it
401		should be directly emitted into the code object.  (This last part
402		is yet to be done.)
403
404		2. Some cleanup of the pgraph code to make it match the proposal that
405		I put out the other day.  (The proposal notwithstanding, things are
406		still in flux here.)
407
408		----------------------------------------------------------------------
409		Name: Lal George
410		Date: 2001/11/14 09:44:04 EST
411		Tag:
412		Description:
413
414		Fix for a backpatching bug reported by Allen.
415
416		Because the boundary between short and long span-dependent
417		instructions is +/- 128, there are an astounding number of
418		span-dependent instructions whose size is over estimated.
419
420		Allen came up with the idea of letting the size of span
421		dependent instructions be non-monotonic, for a maxIter
422		number of times, after which the size must be monotonically
423		increasing.
424
425		This table shows the number of span-dependent instructions
426		whose size was over-estimated as a function of maxIter, for the
427		file Parse/parse/ml.grm.sml:
428
429		maxIter            # of instructions:
430		10                      687
431		20                      438
432		30                      198
433		40                        0
434
435		In compiling the compiler, there is no significant difference in
436		compilation speed between maxIter=10 and maxIter=40. Actually,
437		my measurements showed that maxIter=40 was a tad faster than
438		maxIter=10! Also 96% of the  files in the compiler reach a fix
439		point within 13 iterations, so fixing maxIter at 40, while high,
440		is okay.
441
442		----------------------------------------------------------------------
443		Name: Matthias Blume
444		Date: 2001/10/31 15:25:00 EST
445		Tag: blume-20011031-pgraph
446		Description:
447
448		CKIT:
449		* Changed the "Function" constructor of type Ast.ctype to carry optional
450		argument identifiers.
451		* Changed the return type of TypeUtil.getFunction accordingly.
452		* Type equality ignores the argument names.
453		* TypeUtil.composite tries to preserve argument names but gives up quickly
454		if there is a mismatch.
455
456		installation script:
457		* attempts to use "curl" if available (unless "wget" is available as well)
458
459		CM:
460		* has an experimental implementation of "portable graphs" which I will
461		soon propose as an implementation-independent library format
462		* there are also new libraries $/pgraph.cm and $/pgraph-util.cm
463
464		NLFFI-LIB:
465		* some cleanup (all cosmetic)
466
467		NLFFIGEN:
468		* temporarily disabled the mechanism that suppresses ML output for
469		C definitions whose identifiers start with an underscore character
470		* generate val bindings for enum constants
471		* user can request that only one style (light or heavy) is being used;
472		default is to use both (command-line arguments: -heavy and -light)
473		* fixed bug in handling of function types involving incomplete pointers
474		* generate ML entry points that take record arguments (i.e., using
475		named arguments) for C functions that have a prototype with named
476		arguments
477		(see changes to CKIT)
478
479		----------------------------------------------------------------------
480		Name: Allen Leung
481		Date: 2001/10/27 20:34:00 EDT
482		Tag: leunga-20011027-x86-fast-fp-call
483		Description:
484
485		Fixed the bug described in blume-20010920-slowfp.
486
487		The fix involves
488		1. generating FCOPYs in FSTP in ia32-svid
489		2. marking a CALL with the appropriate annotation
490
491		----------------------------------------------------------------------
492		Name: Matthias Blume
493		Date: 2001/10/16 11:32:00 EDT
494		Tag: blume-20011016-netbsd
495		Description:
496
497		Underscore patch from Chris Richards (fixing problem with compiling
498		runtime system under recent NetBSD).
499
500		----------------------------------------------------------------------
501		Name: Allen Leung
502		Date: 2001/10/12 17:18:32 EDT 2001
503		Tag: leung-20011012-x86-printflowgraph
504		Description:
505
506		X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
507
508		----------------------------------------------------------------------
509		Name: Lal George
510		Date: 2001/10/11 23:51:34 EDT
511		Tag: george-20011011-too-many-instrs
512		Description:
513
514		The representation of a program point never expected to see more
515		than 65536 instructions in a basic block!
516
517		----------------------------------------------------------------------
518		Name: Lal George
519		Date: 2001/10/09 09:41:37 EDT
520		Tag: george-20011008-mlrisc-labels
521		Description:
522
523		Changed the machine description files to support printing of
524		local and global labels in assembly code, based on host assembler
525		conventions.
526
527		----------------------------------------------------------------------
528		Name: Matthias Blume
529		Date: 2001/09/25 15:25:00 EDT
530		Tag: blume-20010925-exninfo
531		Description:
532
533		I provided a non-hook implementation of exnName (at the toplevel) and
534		made the "dummy" implementation of exnMessage (at the toplevel) more
535		useful: if nothing gets "hooked in", then at least you are going to
536		see the exception name and a message indicating why you don't see more.
537
538		[For the time being, programs that need exnMessage and want to use
539		ml-build should either use General.exnMessage (strongly recommended) or
540		refer to structure General at some other point so that CM sees a
541		static dependency.]
542
543		[Similar remarks go for "print" and "use":  If you want to use their
544		functionality in stand-alone programs generated by ml-build, then use
545		TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
546
547		----------------------------------------------------------------------
548		Name: Matthias Blume
549		Date: 2001/09/20 17:28:00 EDT
550		Tag: blume-20010920-slowfp
551		Description:
552
553		Allen says that x86-fast-fp is not safe yet, so I turned it off again...
554
555		----------------------------------------------------------------------
556		Name: Matthias Blume
557		Date: 2001/09/20 17:20:00 EDT
558		Tag: blume-20010920-canonicalpaths
559		Description:
560
561		0. Updated the BOOT file (something that I forgot to do earlier).
562
563		1. Small internal change to CM so that it avoids "/../" in filenames
564		as much as possible (but only where it is safe).
565
566		2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
567		that arguments that contain delimiters are passed through correctly.
568		This change also means that all "special" arguments of the form
569		@SMLxxx... must come first.
570
571		3. Changed install script to put relative anchor names for tool commands
572		into pathconfig.
573
574		----------------------------------------------------------------------
575		Name: Matthias Blume
576		>>>>>>> 1.169
577		Date: 2001/09/18 15:35:00 EDT
578		Tag: blume-20010918-readme11036
579		Description:
580
581		Added README files.
582
583		----------------------------------------------------------------------
584		Name: Matthias Blume
585		Date: 2001/09/18 11:45:00 EDT
586		Tag: Release_110_36 (retag)
587		Description:
588
589		Fixed mistake in config/preloads. Retagged as 110.36.
590
591		----------------------------------------------------------------------
592		Name: Matthias Blume
593		Date: 2001/09/18 09:40:00 EDT
594		Tag: Release_110_36_orig (tag changed)
595		Description:
596
597		New version (110.36).  New bootfiles.
598
599		----------------------------------------------------------------------
600		Name: Matthias Blume
601		Date: 2001/09/14 16:15:00 EDT
602		Tag: blume-20010914-x86fastfp
603		Description:
604
605		John committed some changes that Allen made, in particular a (hopefully)
606		correctly working version of the x86-fp module.
607
608		I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
609		flag to "true".  Everything seems to compile to a fixpoint ok, and
610		"mandelbrot" speeds up by about 15%.
611
612		----------------------------------------------------------------------
613		Name: Matthias Blume
614		Date: 2001/09/13 11:20:00 EDT
615		Tag: blume-20010913-minimal
616		Description:
617
618		1. Stefan Monnier's patch to fix a miscompilation problem that
619		was brought to light by John Reppy's work on Moby.
620
621		2. Implemented a minimal "structure Compiler" that contains just
622		"version" and "architecture".  The minimal version will be
623		available when the full version is not.  This is for backward-
624		compatibility with code that wants to test Compiler.version.
625
626		----------------------------------------------------------------------
627		Name: Matthias Blume
628		Date: 2001/08/28 14:03:00 EDT
629		Tag: blume-20010828-ml-lex
630		Description:
631
632		Fix for bug 1581, received from Neophytos Michael.
633
634		----------------------------------------------------------------------
635		Name: Matthias Blume
636		Date: 2001/08/27 11:20:00 EDT
637		Tag: blume-20010827-readme11035
638		Description:
639
640		Fleshed out the README file for 110.35.
641
642		----------------------------------------------------------------------
643		Name: Matthias Blume
644		Date: 2001/08/24 17:10:00 EDT
645		Tag: Release_110_35
646		Description:
647
648		New version number (110.35).  New bootfiles.
649
650		----------------------------------------------------------------------
651		Name: Lal George
652		Date: 2001/08/24 13:47:18 EDT 2001
653		Tag: george-20010824-MLRISC-graphs
654		Description:
655
656		removed clusters from MLRISC completely and replaced with graphs.
657
658		----------------------------------------------------------------------
659		Name: Matthias Blume
660		Date: 2001/08/23 17:50:00 EDT
661		Tag: blume-20010823-toplevel
662		Description:
663
664		- some reorganization of the code that implements various kinds of
665		environments in the compiler (static, dynamic, symbolic, combined)
666		- re-implemented the EnvRef module so that evalStream works properly
667		(if the stream contains references to "use", "CM.make", etc.)
668		- cleaned up evalloop.sml and interact.sml (but they need more cleaning)
669
670		----------------------------------------------------------------------
671		Name: Matthias Blume
672		Date: 2001/08/20 15:50 EDT
673		Tag: blume20010820-slipup
674		Description:
675
676		I forgot to commit a few files.  Here they are...
677
678		----------------------------------------------------------------------
679		Name: Matthias Blume
680		Date: 2001/08/20 15:35:00 EDT
681		Tag: blume-20010820-debugprof
682		Description:
683
684		!!!! NEW BOOTFILES !!!!
685
686		This is another round of reorganizing the compiler sources.  This
687		time the main goal was to factor out all the "instrumentation"
688		passes (for profiling and backtracing) into their own library.
689		The difficulty was to do it in such a way that it does not depend
690		on elaborate.cm but only on elabdata.cm.
691
692		Therefore there have been further changes to both elaborate.cm and
693		elabdata.cm -- more "generic" things have been moved from the former
694		to the latter.  As a result, I was forced to split the assignment
695		of numbers indicating "primtyc"s into two portions: SML-generic and
696		SML/NJ-specific.  Since it would have been awkward to maintain,
697		I bit the bullet and actually _changed_ the mapping between these
698		numbers and primtycs.  The bottom line of this is that you need
699		a new set of bin- and bootfiles.
700
701		I have built new bootfiles for all architectures, so doing a fresh
702		checkout and config/install.sh should be all you need.
703
704		The newly created library's name is
705
706		$smlnj/viscomp/debugprof.cm
707
708		and its sources live under
709
710		src/compiler/DebugProf
711
712		----------------------------------------------------------------------
713		Name: Matthias Blume
714		Date: 2001/08/15 17:15:00 EDT
715		Tag: blume-20010815-compreorg
716		Description:
717
718		This is a first cut at reorganizing the CM libraries that make up the
719		core of the compiler.  The idea is to separate out pieces that could
720		be used independently by tools, e.g., the parser, the typechecker, etc.
721
722		The current status is a step in this direction, but it is not quite
723		satisfactory yet.  Expect more changes in the future.
724
725		Here is the current (new) organization...
726
727		What used to be $smlnj/viscomp/core.cm is now divided into
728		six CM libraries:
729
730		$smlnj/viscomp/basics.cm
731		/parser.cm
732		/elabdata.cm
733		/elaborate.cm
734		/execute.cm
735		/core.cm
736
737		The CM files for these libraries live under src/system/smlnj/viscomp.
738		All these libraries are proxy libraries that contain precisely
739		one CM library component.  Here are the locations of the components
740		(all within the src/compiler tree):
741
742		Basics/basics.cm
743		Parse/parser.cm
744		ElabData/elabdata.cm
745		Elaborator/elaborate.cm
746		Execution/execute.cm
747		core.cm
748
749		[This organization is the same that has been used already
750		for a while for the architecture-specific parts of the visible
751		compiler and for the old version of core.cm.]
752
753		As you will notice, many source files have been moved from their
754		respective original locations to a new home in one of the above
755		subtrees.
756
757		The division of labor between the new libraries is the following:
758
759		basics.cm:
760		- Simple, basic definitions that pertain to many (or all) of
761		the other libraries.
762		parser.cm:
763		- The SML parser, producing output of type Ast.dec.
764		- The type family for Ast is also defined and exported here.
765		elabdata.cm:
766		- The datatypes that describe input and output of the elaborator.
767		This includes types, absyn, and static environments.
768		elaborator.cm:
769		- The SML/NJ type checker and elaborator.
770		This maps an Ast.dec (with a given static environment) to
771		an Absyn.dec (with a new static environment).
772		- This libraries implements certain modules that used to be
773		structures as functors (to remove dependencies on FLINT).
774		execute.cm:
775		- Everything having to do with executing binary code objects.
776		- Dynamic environments.
777		core.cm:
778		- SML/NJ-specific instantiations of the elaborator and MLRISC.
779		- Top-level modules.
780		- FLINT (this should eventually become its own library)
781
782		Notes:
783
784		I am not 100% happy with the way I separated the elaborator (and its
785		data structures) from FLINT.  Two instances of the same problem:
786
787		1. Data structures contain certain fields that carry FLINT-specific
788		information.  I hacked around this using exn and the property list
789		module from smlnj-lib.  But the fact that there are middle-end
790		specific fields around at all is a bit annoying.
791
792		2. The elaborator calculates certain FLINT-related information.  I tried
793		to make this as abstract as I could using functorization, but, again,
794		the fact that the elaborator has to perform calculations on behalf
795		of the middle-end at all is not nice.
796
797		3. Having to used exn and property lists is unfortunate because it
798		weakens type checking.  The other alternative (parameterizing
799		nearly *everything*) is not appealing, though.
800
801		I removed the "rebinding =" warning hack because due to the new organization
802		it was awkward to maintain it.  As a result, the compiler now issues some of
803		these warnings when compiling init.cmi during bootstrap compilation. On
804		the plus side, you also get a warning when you do, for example:
805		val op = = Int32.+
806		which was not the case up to now.
807
808		I placed "assign" and "deref" into the _Core structure so that the
809		code that deals with the "lazy" keyword can find them there.  This
810		removes the need for having access to the primitive environment
811		during elaboration.
812
813		----------------------------------------------------------------------
814		Name: Matthias Blume
815		Date: 2001/08/13
816		Tag: blume-20010813-closures
817		Description:
818
819		This fix was sent to us by Zhong Shao.  It is supposed to improve the
820		performance of certain loops by avoiding needless closure allocation.
821
822		----------------------------------------------------------------------
823		Name: Lal George
824		Date: 2001/07/31 10:03:23 EDT 2001
825		Tag: george-20010731-x86-fmalloc
826		Description: Fixed bug in x86 calls
827
828		There was a bug where call instructions would mysteriously
829		vanish. The call instruction had to be one that returned
830		a floating point value.
831
832		----------------------------------------------------------------------
833		Name: Lal George
834		Date: 2001/07/19 16:36:29 EDT 2001
835		Tag: george-20010719-simple-cells
836		Description:
837
838		I have dramatically simplified the interface for CELLS in MLRISC.
839
840		In summary, the cells interface is broken up into three parts:
841
842		1. CellsBasis : CELLS_BASIS
843
844		CellsBasis is a top level structure and common for all
845		architectures.  it contains the definitions of basic datatypes
846		and utility  functions over these types.
847
848		2. functor Cells() : CELLS
849
850		Cells generates an interface for CELLS that incorporates the
851		specific resources on the target architecture, such as the
852		presence of special register classes, their number and size,
853		and various useful substructures.
854
855		3. <ARCH>CELLS
856
857		e.g. SparcCells: SPARCCELLS
858
859		<ARCH>CELLS usually contains additional bindings for special
860		registers  on the architecture, such as:
861
862		val r0 : cell           (* register zero *)
863		val y : cell            (* Y register *)
864		val psr : cell          (* processor status register *)
865		...
866
867		The structure returned by applying the Cells functor is opened
868		in this interface.
869
870		The main implication of all this is that the datatypes for cells is
871		split between CellsBasis and CELLS -- a fairly simple change for user
872		code.
873
874		In the old scheme the CELLS interface had a definitional binding of
875		the form:
876
877		signature CELLS = sig
878
879		structure CellsBasis = CellsBasis
880
881		...
882
883		end
884
885		With all the sharing constraints that goes on in MLRISC, this old
886		design  quickly leads to errors such as:
887
888		"structure definition spec inside of sharing ... "
889
890
891		and appears to require an unacceptable amount of sharing and where
892		constraint hackery.
893
894		I think this error message (the interaction of definitional specs and
895		sharing) requires more explanation on our web page.
896
897		----------------------------------------------------------------------
898		Name: Matthias Blume
899		Date: 2001/07/19 15:00:00 EDT
900		Tag: blume-20010719-libreorg
901		Description:
902
903		This update puts together a fairly extensive but straightforward change
904		to the way the libraries that implement the interactive system are
905		organized:
906
907		The biggest change is the elimination of structure Compiler.  As a
908		replacement for this structure, there is now a CM library
909		(known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
910		that exports all the substructures of the original structure Compiler
911		directly.  So instead of saying Compiler.Foo.bar one now simply
912		says Foo.bar.  (The CM libraries actually export a collection of
913		structures that is richer than the collection of substructures of
914		structure Compiler.)
915
916		To make the transition smooth, there is a separate library called
917		$smlnj/compiler/compiler.cm which puts together and exports the
918		original structure Compiler (or at least something very close to it).
919
920		There are five members of the original structure Compiler
921		that are not exported directly but which instead became members
922		of a new structure Backend (described by signature BACKEND).  These are:
923		structure Profile (: PROFILE), structure Compile (: COMPILE), structure
924		Interact (: INTERACT), structure Machine (: MACHINE), and val
925		architecture (: string).
926
927		Structure Compiler.Version has become structure CompilerVersion.
928
929		Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
930		by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
931		or x86, respectively.
932		Each of these exports the same frontend structures that
933		$smlnj/compiler.cm exports.  But they do not have a structure Backend
934		and instead export some structure <Arch>Backend where <Arch> is Alpha32,
935		Hppa, PPC, Sparc, or X86, respectively.
936
937		Library $smlnj/compiler/all.cm exports the union of the exports of
938		$smlnj/compiler/<arch>.cm
939
940		There are no structures <Arch>Compiler anymore, use
941		$smlnj/compiler/<arch>.cm instead.
942
943		Library host-compiler-0.cm is gone.  Instead, the internal library
944		that instantiates CM is now called cm0.cm.  Selection of the host
945		compiler (backend) is no longer done here but. (Responsibility for it
946		now lies with $smlnj/compiler/current.cm.  This seems to be more
947		logical.)
948
949		Many individual files have been moved or renamed.  Some files have
950		been split into multiple files, and some "dead" files have been deleted.
951
952		Aside from these changes to library organization, there are also changes
953		to the way the code itself is organized:
954
955		Structure Binfile has been re-implemented in such a way that it no
956		longer needs any knowledge of the compiler.  It exclusively deals
957		with the details of binfile layout.  It no longer invokes the
958		compiler (for the purpose of creating new prospective binfile
959		content), and it no longer has any knowledge of how to interpret
960		pickles.
961
962		Structure Compile (: COMPILE) has been stripped down to the bare
963		essentials of compilation.  It no longer deals with linking/execution.
964		The interface has been cleaned up considerably.
965
966		Utility routines for dealing with linking and execution have been
967		moved into their own substructures.
968
969		(The ultimate goal of these changes is to provide a light-weight
970		binfile loader/linker (at least for, e.g., stable libraries) that
971		does not require CM or the compiler to be present.)
972
973		CM documentation has been updated to reflect the changes to library
974		organization.
975
976		----------------------------------------------------------------------
977		Name: Matthias Blume
978		Date: 2001/07/10 17:30:00 EDT
979		Tag: Release_110_34
980		Description:
981
982		Minor tweak to 110.34 (re-tagged):
983
984		- README.html file added to CVS repository
985		- runtime compiles properly under FreeBSD 3.X and 4.X
986
987		----------------------------------------------------------------------
988		Name: Matthias Blume
989		Date: 2001/07/10 17:30:00 EDT
990		Tag: Release_110_34
991		Description:
992
993		New version number (110.34). New bootfiles.
994
995		----------------------------------------------------------------------
996		Name: Matthias Blume
997		Date: 2001/07/09 16:00:00 EDT
998		Tag: blume-20010709-more-varargs
999		Description:
1000
1001		I changed the handling of varargs in ml-nlffigen again:
1002		The ellipsis ... will now simply be ignored (with an accompanying warning).
1003
1004		The immediate effect is that you can actually call a varargs function
1005		from ML -- but you can't actually supply any arguments beyond the ones
1006		specified explicitly.  (For example, you can call printf with its format
1007		string, but you cannot pass additional arguments.)
1008
1009		This behavior is only marginally more useful than the one before, but
1010		it has the advantage that a function or, more importantly, a function
1011		type never gets dropped on the floor, thus avoiding follow-up problems with
1012		other types that refer to the offending one.
1013
1014		----------------------------------------------------------------------
1015		Name: Matthias Blume
1016		Date: 2001/07/09 11:25:00 EDT
1017		Tag: blume-20010709-varargs
1018		Description:
1019
1020		1. ckit-lib.cm now exports structure Error
1021		2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
1022		with a warning accompanied by a source location.  Moreover, it
1023		merely skips the offending function or type and proceeds with the
1024		rest of its work.u  As a result, one can safely feed C code containing
1025		"..." to ml-nlffigen.
1026		3. There are some internal improvements to CM, providing slightly
1027		more general string substitutions in the tools subsystem.
1028
1029		----------------------------------------------------------------------
1030		Name: Matthias Blume
1031		Date: 2001/06/27 15:10:00 EDT
1032		Tag: blume-20010627-concur
1033		Description:
1034
1035		Fixed a small bug in CM's handling of parallel compilation.
1036		(You could observe the bug by Control-C-interrupting an ordinary
1037		CMB.make or CM.stabilize and then attaching some compile servers.
1038		The result was that all of a sudden the previously interrupted
1039		compilation would continue on its own.  This was because of
1040		an over-optimization: CM did not bother to clean out certain queues
1041		when no servers were attached "anyway", resulting in the contents
1042		of these queues to grab control when new servers did get attached.)
1043
1044		There is also another minor update to the CM manual.
1045
1046		----------------------------------------------------------------------
1047		Name: Matthias Blume
1048		Date: 2001/06/26 16:15:00 EDT
1049		Tag: blume-20010626-cmdoc
1050		Description:
1051
1052		Minor typo fixed in CM manual (syntax diagram for libraries).
1053
1054		----------------------------------------------------------------------
1055		Name: Matthias Blume
1056		Date: 2001/06/25 22:55:00 EDT
1057		Tag: blume-20010625-x86pc
1058		Description:
1059
1060		Fixed a nasty bug in the X86 assembly code that caused signal
1061		handlers to fail (crash) randomly.
1062
1063		----------------------------------------------------------------------
1064		Name: Matthias Blume
1065		Date: 2001/06/25 12:05:00 EDT
1066		Tag: blume-20010625-nlffigen
1067		Description:
1068
1069		This update fixes a number of minor bugs in ml-nlffigen as reported by
1070		Nick Carter <nbc@andrew.cmu.edu>.
1071
1072		1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
1073		2. Default names for generated files are now derived from the name of
1074		the C file *without its directory*.  In particular, this causes generated
1075		files to be placed locally even if the C file is in some system directory.
1076		3. Default names for generated signatures and structures are also derived
1077		from the C file name without its directory.  This avoids silly things
1078		like "structure GL/GL".
1079		(Other silly names are still possible because ml-nlffigen does not do
1080		a thorough check of whether generated names are legal ML identifiers.
1081		When in doubt, use command line arguments to force particular names.)
1082
1083		----------------------------------------------------------------------
1084		Name: Matthias Blume
1085	Date: 2001/06/21 12:25:00 EDT	Date: 2001/06/21 12:25:00 EDT
1086	Tag: blume-20010621-eXene	Tag: blume-20010621-eXene
1087	Description:	Description:
#	Line 2482	Line 3550
3550	elaborator). There were a lot of changes during my "linkpath" trials	elaborator). There were a lot of changes during my "linkpath" trials
3551	that could have been reverted to their original state but weren't.	that could have been reverted to their original state but weren't.
3552	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
3553	than what was strictly necessary...  (I _did_ resist the tempation	than what was strictly necessary...  (I _did_ resist the temptation
3554	of doing any "global reformatting" to avoid an untimely death at	of doing any "global reformatting" to avoid an untimely death at
3555	Dave's hands. :)	Dave's hands. :)
3556

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.845   changed lines   Added in v.1040

---

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