Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 670, Sun Jun 18 13:10:57 2000 UTC	revision 1078, Tue Feb 19 21:26:48 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
17		Date: 2002/02/19 16:20:00 EST
18		Tag: blume-20020219-cmffi
19		Description:
20
21		A wild mix of changes, some minor, some major:
22
23		* All C FFI-related libraries are now anchored under $c:
24		$/c.cm      --> $c/c.cm
25		$/c-int.cm  --> $c/internals/c-int.cm
26		$/memory.cm --> $c/memory/memory.cm
27
28		* "make" tool (in CM) now treats its argument pathname slightly
29		differently:
30		1. If the native expansion is an absolute name, then before invoking
31		the "make" command on it, CM will apply OS.Path.mkRelative
32		(with relativeTo = OS.FileSys.getDir()) to it.
33		2. The argument will be passed through to subsequent phases of CM
34		processing without "going native".  In particular, if the argument
35		was an anchored path, then "make" will not lose track of that anchor.
36
37		* Compiler backends now "know" their respective C calling conventions
38		instead of having to be told about it by ml-nlffigen.  This relieves
39		ml-nlffigen from one of its burdens.
40
41		* The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
42
43		* Export C_DEBUG and C_Debug from $c/c.cm.
44
45		* C type encoding in ml-nlffi-lib has been improved to model the conceptual
46		subtyping relationship between incomplete pointers and their complete
47		counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
48		with the convention of instantiating 'o with ('t, 'c) obj whenever
49		the pointer target type is complete.  In the incomplete case, 'o
50		will be instantiated with some "'c iobj" -- a type obtained by
51		using one of the functors PointerToIncompleteType or PointerToCompleteType.
52
53		Operations that work on both incomplete and complete pointer types are
54		typed as taking an 'o ptr while operations that require the target to
55		be known are typed as taking some ('t, 'c) obj ptr.
56
57		voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
58		where void is an eqtype without any values.  This makes it possible
59		to work on voidptr values using functions meant to operate on light
60		incomplete pointers.
61
62		* As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
63		been vastly simplified.
64
65		----------------------------------------------------------------------
66		Name: Matthias Blume
67		Date: 2002/02/19 10:48:00 EST
68		Tag: blume-20020219-pqfix
69		Description:
70
71		Applied Chris Okasaki's bug fix for priority queues.
72
73		----------------------------------------------------------------------
74		Name: Matthias Blume
75		Date: 2002/02/15 17:05:00
76		Tag: Release_110_39
77		Description:
78
79		Last-minute retagging is becoming a tradition... :-(
80
81		This is the working release 110.39.
82
83		----------------------------------------------------------------------
84		Name: Matthias Blume
85		Date: 2002/02/15 16:00:00 EST
86		Tag: Release_110_39-orig
87		Description:
88
89		Working release 110.39.  New bootfiles.
90
91		(Update: There was a small bug in the installer so it wouldn't work
92		with all shells.  So I retagged. -Matthias)
93
94		----------------------------------------------------------------------
95		Name: Matthias Blume
96		Date: 2002/02/15 14:17:00 EST
97		Tag: blume-20020215-showbindings
98		Description:
99
100		Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
101		the latter can be useful for exploring what bindings are available at
102		the interactive prompt.  (The first function returns only the list
103		of symbols that are really bound, the second prints those but also the
104		ones that CM's autoloading mechanism knows about.)
105
106		----------------------------------------------------------------------
107		Name: Matthias Blume
108		Date: 2002/02/15 12:08:00 EST
109		Tag: blume-20020215-iptrs
110		Description:
111
112		Two improvements to ml-nlffigen:
113
114		1. Write files only if they do not exist or if their current contents
115		do not coincide with what's being written.  (That is, avoid messing
116		with the time stamps unless absolutely necessary.)
117
118		2. Implement a "repository" mechanism for generated files related
119		to "incomplete pointer types".   See the README file for details.
120
121		----------------------------------------------------------------------
122		Name: Matthias Blume
123		Date: 2002/02/14 11:50:00 EST
124		Tag: blume-20020214-quote
125		Description:
126
127		Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
128		because of the new and improved tag generation scheme.  (Thanks to Allen
129		Leung for pointing it out.)
130
131		----------------------------------------------------------------------
132		Name: Lal George
133		Date: 2002/02/14 09:55:27 EST 2002
134		Tag: george-20020214-isabelle-bug
135		Description:
136
137		Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
138		of Isabelle on the x86.
139
140		From Allen:
141		-----------
142		I've found the problem:
143
144		in ra-core.sml, I use the counter "blocked" to keep track of the
145		true number of elements in the freeze queue.  When the counter goes
146		to zero, I skip examining the queue.  But I've messed up the
147		bookkeeping in combine():
148
149		else ();
150		case !ucol of
151		PSEUDO => (if !cntv > 0 then
152		(if !cntu > 0 then blocked := !blocked - 1 else ();
153		^^^^^^^^^^^^^^^^^^^^^^^
154		moveu := mergeMoveList(!movev, !moveu)
155		)
156		else ();
157
158		combine() is called to coalesce two nodes u and v.
159		I think I was thinking that if the move counts of u and v are both
160		greater than zero then after they are coalesced then one node is
161		removed from the freeze queue.  Apparently I was thinking that
162		both u and v are of low degree, but that's clearly not necessarily true.
163
164
165		02/12/2002:
166		Here's the patch.  HOL now compiles.
167
168		I don't know how this impact on performance (compile
169		time or runtime).  This bug caused the RA (especially on the x86)
170		to go thru the potential spill phase when there are still nodes on the
171		freeze queue.
172
173
174
175
176		----------------------------------------------------------------------
177		Name: Matthias Blume
178		Date: 2002/02/13 22:40:00 EST
179		Tag: blume-20020213-fptr-rtti
180		Description:
181
182		Fixed a bug in ml-nlffigen that was introduced with one of the previous
183		updates.
184
185		----------------------------------------------------------------------
186		Name: Matthias Blume
187		Date: 2002/02/13 16:41:00 EST
188		Tag: blume-20020213-cmlpq
189		Description:
190
191		Added new priority queue export symbols (which have just been added to
192		smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
193		not compile and the installer would choke.)
194
195		----------------------------------------------------------------------
196		Name: Matthias Blume
197		Date: 2002/02/13 16:15:00 EST
198		Tag: blume-20020213-various
199		Description:
200
201		1. More tweaks to ml-nlffigen:
202
203		- better internal datastructures (resulting in slight speedup)
204		- "-match" option requires exact match
205		- "localized" gensym counters (untagged structs/unions nested within
206		other structs/unions or within typedefs get a fresh counter; their
207		tag will be prefixed by a concatenation of their parents' tags)
208		- bug fixes (related to calculation of transitive closure of types
209		to be included in the output)
210
211		2. Minor Basis updates:
212
213		- added implementations for List.collate and Option.app
214
215		----------------------------------------------------------------------
216		Name: Matthias Blume
217		Date: 2002/02/11 15:55:00 EST
218		Tag: blume-20020211-gensym
219		Description:
220
221		Added a "-gensym" option to command line of ml-nlffigen.  This can be
222		used to specify a "stem" -- a string that is inserted in all "gensym'd"
223		names (ML structure names that correspond to unnamed C structs, unions,
224		and enums), so that separate runs of ml-nlffigen do not clash.
225
226		----------------------------------------------------------------------
227		Name: Matthias Blume
228		Date: 2002/02/11 12:05:00 EST
229		Tag: blume-20020211-gensml
230		Description:
231
232		A quick fix for a problem with GenSML (in the pgraph-util library):
233		Make generation of toplevel "local" optional.  (Strictly speaking,
234		signature definitions within "local" are not legal SML.)
235
236		Other than that: updates to INSTALL and cm/TODO.
237
238		----------------------------------------------------------------------
239		Name: Matthias Blume
240		Date: 2002/02/08 15:00:00 EST
241		Tag: blume-20020208-uniquepid
242		Description:
243
244		0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
245
246		1. The installer (config/install.sh) has gotten smarter:
247
248		- Configuration options are a bit easier to specify now
249		(in config/targets).
250		- Bug in recognizing .tar.bz2 files fixed.
251		- Installer automatically resolves dependencies between
252		configuration options (e.g., if you ask for eXene, you will
253		also get cml -- regardless whether you asked for it or not).
254		- Installer can run in "quieter mode" by setting the environment
255		variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
256		"completely silent", though.
257		- Build HashCons library as part of smlnj-lib.
258
259		2. A new scheme for assigning persistent identifiers to compilation
260		units (and, by extension, to types etc.) has been put into place.
261		This fixes a long-standing bug where types and even dynamic values
262		can get internally confused, thereby compromising type safety
263		(abstraction) and dynamic correctness.  See
264
265		http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
266
267		for an example of how things could go wrong until now.
268
269		The downside of the new scheme is that pids are not quite as
270		persistent as they used to be: CM will generate a fresh pid
271		for every compilation unit that it thinks it sees for the first
272		time.  That means that if you compile starting from a clean, fresh
273		source tree at two different times, you end up with different
274		binaries.
275
276		Cutoff recompilation, however, has not been compromised because
277		CM keeps pid information in special caches between runs.
278
279		----------------------------------------------------------------------
280		Name: Lal George
281		Date: 2002/02/07 15:34:13 EST 2002
282		Tag: <none>
283		Description:
284
285		Compilers that generate assembly code may produce  global labels
286		whose value is resolved at link time. The various peephole optimization
287		modules did not take this in account.
288
289		TODO. The Labels.addrOf function should really return an option
290		type so that clients are forced to deal with this issue, rather
291		than an exception being raised.
292
293		----------------------------------------------------------------------
294		Name: Lal George
295		Date: 2002/02/06 13:55:02 EST
296		Tag: george-20020206-ra-breakup
297		Description:
298
299		1. A bug fix from Allen.
300
301		A typo causes extra fstp %st(0)'s to be generated at compensation
302		edges, which might cause stack underflow traps at runtime.  This
303		occurs in fft where there are extraneous fstps right before the 'into'
304		trap instruction (in this case they are harmless since none of the
305		integers overflow.)
306
307		2. Pulled out various utility modules that were embedded in the modules
308		of the register allocator. I need these modules for other purposes, but
309		they are not complete enough to put into a library (just yet).
310		----------------------------------------------------------------------
311		Name: Matthias Blume
312		Date: 2002/01/31 16:05:00 EST
313		Tag: blume-20020131-sparc-ccalls
314		Description:
315
316		1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
317		of extra stack space by mistake.  Fixed.
318
319		2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
320
321		----------------------------------------------------------------------
322		Name: Allen Leung
323		Date: 2002/01/30
324		Tag: leunga-20020130-nowhere-bug-fix
325		Description:
326
327		MLRISC bug fixes:
328		1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
329		2. MachineInt.fromString was negating its value.
330
331		----------------------------------------------------------------------
332		Name: Matthias Blume
333		Date: 2002/01/29
334		Tag: blume-20020129-INSTALL
335		Description:
336
337		- Added somewhat detailed installation instructions (file INSTALL).
338		- Fixed curl-detection bug in config/install.sh.
339		- It is now possible to select the URL getter using the URLGETTER
340		environment variable:
341
342		not set / "unknown"      --> automatic detection (script tries wget,
343		curl, and lynx)
344		"wget" / "curl" / "lynx" --> use the specified program (script "knows"
345		how to properly invoke them)
346		other                    --> use $URLGETTER directly, it must take
347		precisely two command-line arguments
348		(source URL and destination file name)
349
350		----------------------------------------------------------------------
351		Name: Matthias Blume
352		Date: 2002/01/28
353		Tag: blume-20020128-sparc-ccalls
354		Description:
355
356		- Fixed problem with calculation of "used" registers in sparc-c-calls.
357		- Make use of the allocParam argument in sparc-c-calls.
358
359		----------------------------------------------------------------------
360		Name: Matthias Blume
361		Date: 2002/01/28
362		Tag: blume-20020128-allocParam
363		Description:
364
365		John Reppy:  Changes c-calls API to accept client-callback for
366		allocating extra stack space.
367		me: Corresponding changes to mlriscGen (using a dummy argument that
368		does not change the current behavior).
369
370		----------------------------------------------------------------------
371		Name: Matthias Blume
372		Date: 2002/01/28 12:00:00
373		Tag: Release_110_38
374		Description:
375
376		This time for real!!!
377
378		----------------------------------------------------------------------
379		Name: Matthias Blume
380		Date: 2002/01/28 10:56:00 EST
381		Tag: blume-20020128-retraction
382		Description:
383
384		0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
385		with blume-Release_110_38-retracted.)
386
387		1. Fixed a problem with incorrect rounding modes in real64.sml.
388		(Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
389
390		2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
391		and enums fixed.  The naming of corresponding ML identifiers should
392		now be consistent again.
393
394		----------------------------------------------------------------------
395		Name: Allen Leung
396		Date: 2002/01/27
397		Tag: leunga-20020127-nowhere
398		Description:
399
400		Added a target called nowhere in the configuration scripts.
401		Enabling this will build the MLRISC 'nowhere' tool (for translating
402		programs with where-clauses into legal SML code) during installation.
403
404		----------------------------------------------------------------------
405		Name: Matthias Blume
406		Date: 2002/01/25 21:27:00 EST
407		Tag: blume-Release_110_38-retracted
408		Description:
409
410		Call it a (working) release!  Version is 110.38. Bootfiles are ready.
411
412		README will be added later.
413
414		!!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
415		(Release_110_38) removed.  Reason: Last-minute bug fixes.
416
417		----------------------------------------------------------------------
418		Name: Matthias Blume
419		Date: 2002/01/25
420		Tag: blume-20020125-ffi
421		Description:
422
423		A large number of tweaks and improvements to ml-nlffi-lib and
424		ml-nlffigen:
425
426		- ML represenation types have been streamlined
427		- getter and setter functions work with concrete values, not abstract
428		ones where possible
429		- ml-nlffigen command line more flexible (see README file there)
430		- some bugs have been fixed (hopefully)
431
432		----------------------------------------------------------------------
433		Name: Lal George
434		Date: 2002/01/24
435		Tag: george-20020124-risc-ra-interface
436		Description:
437
438		There is a dramatic simplification in the interface to the
439		register allocator for RISC architectures as a result of making
440		parallel copy instructions explicit.
441
442		----------------------------------------------------------------------
443		Name: Matthias Blume
444		Date: 2002/01/22
445		Tag: blume-20020122-x86-ccalls
446		Description:
447
448		Bug fix for c-calls on x86 (having to do with how char- and
449		short-arguments are being handled).
450
451		----------------------------------------------------------------------
452		Name: Matthias Blume
453		Date: 2002/01/21
454		Tag: blume-20020121-ff
455		Description:
456
457		Another day of fiddling with the FFI...
458
459		1. Bug fix/workaround:  CKIT does not complain about negative array
460		dimensions, so ml-nlffigen has to guard itself against this possibility.
461		(Otherwise a negative dimension would send it into an infinite loop.)
462
463		2. Some of the abstract types (light objects, light pointers, most "base"
464		types) in structure C are now eqtypes.
465
466		3. Added constructors and test functions for NULL function pointers.
467
468		----------------------------------------------------------------------
469		Name: Matthias Blume
470		Date: 2002/01/18
471		Tag: blume-20020118-ready-for-new-release
472		Description:
473
474		Made config/srcarchiveurl point to a new place.  (Will provide boot
475		files shortly.)
476
477		Maybe we christen this to be 110.38?
478
479		----------------------------------------------------------------------
480		Name: Matthias Blume
481		Date: 2002/01/18
482		Tag: blume-20020118-more-ffifiddle
483		Description:
484
485		Today's FFI fiddling:
486
487		- Provided a structure CGetSet with "convenient" versions of C.Get.* and
488		C.Set.* that use concrete (MLRep.*) arguments and results instead
489		of abstract ones.
490
491		- Provided word-style bit operations etc. for "int" representation
492		types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
493		and Long.
494
495		----------------------------------------------------------------------
496		Name: Matthias Blume
497		Date: 2002/01/18
498		Tag: blume-20020118-use-x86-fp
499		Description:
500
501		Now that x86-fast-fp seems to be working, I turned it back on again
502		by default.  (Seems to work fine now, even with the FFI.)
503
504		Other than that, I added some documentation about the FFI to
505		src/ml-nlffigen/README and updated the FFI test examples in
506		src/ml-nlffi-lib/Tests/*.
507
508		----------------------------------------------------------------------
509		Name: Allen Leung
510		Date: 2002/01/17
511		Tag: leunga-20020117-x86-fast-fp-call
512		Description:
513
514		1. Fixed a problem with handling return fp values when x86's fast fp
515		mode is turned on.
516
517		2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
518		of %f32.
519
520		3. Added a constructor INT32lit to the ast of MLRISC tools.
521
522		----------------------------------------------------------------------
523		Name: Matthias Blume
524		Date: 2002/01/16
525		Tag: blume-20020116-ffifiddle
526		Description:
527
528		More fiddling with the FFI interface:
529
530		- Make constness 'c instead of rw wherever possible.  This eliminates
531		the need for certain explicit coercions.  (However, due to ML's
532		value polymorphism, there will still be many cases where explicit
533		coercions are necessary.  Phantom types are not the whole answer
534		to modeling a subtyping relationship in ML.)
535
536		- ro/rw coersions for pointers added.  (Avoids the detour through */&.)
537
538		- "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
539		clumsy workaround for varargs problem.)
540
541		----------------------------------------------------------------------
542		Name: Lal George
543		Date: 2002/01/15
544		Tag: <none>
545		Description:
546
547		1. Since COPY instructions are no longer native to the architecture,
548		a generic functor can be used to implement the expandCopies function.
549
550		2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
551		TEXT segment.
552
553		----------------------------------------------------------------------
554		Name: Matthias Blume
555		Date: 2002/01/15
556		Tag: blume-20020115-ffiupdates
557		Description:
558
559		1. Fix for bug resulting in single-precision float values being returned
560		incorrectly from FFI calls.
561
562		2. Small modifications to C FFI API:
563
564		- memory-allocation routines return straight objects (no options)
565		and raise an exception in out-of-memory situations
566		- unsafe extensions to cast between function pointers and pointers
567		from/to ints
568		- added structure C_Debug as an alternative to structure C where
569		pointer-dereferencing (|*| and |*!) always check for null-pointers
570		- added open_lib' to DynLinkage;  open_lib' works like open_lib
571		but also takes a (possibly empty) list of existing library handles
572		that the current library depends on
573
574		----------------------------------------------------------------------
575		Name: Matthias Blume
576		Date: 2002/01/10
577		Tag: blume-20020110-newffigen
578		Description:
579
580		1. Updates to portable graph code.
581
582		2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
583		more scalable now so that even huge interfaces such as the one
584		for GTK compile in finite time and space. :-)
585		See src/ml-nlffigen/README for details on what's new.
586
587		----------------------------------------------------------------------
588		Name: Lal George
589		Date: 2001/01/09 14:31:35 EST 2002
590		Tag: george-20011206-rm-native-copy
591		Description:
592
593		Removed the native COPY and FCOPY instructions
594		from all the architectures and replaced it with the
595		explicit COPY instruction from the previous commit.
596
597		It is now possible to simplify many of the optimizations
598		modules that manipulate copies. This has not been
599		done in this change.
600
601		----------------------------------------------------------------------
602		Name: Lal George
603		Date: 2001/12/06 16:50:13 EST 2001
604		Tag: george-20011206-mlrisc-instruction
605		Description:
606
607		Changed the representation of instructions from being fully abstract
608		to being partially concrete. That is to say:
609
610		from
611		type instruction
612
613		to
614		type instr                              (* machine instruction *)
615
616		datatype instruction =
617		LIVE of {regs: C.cellset, spilled: C.cellset}
618		| KILL of {regs: C.cellset, spilled: C.cellset}
619		| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
620		| ANNOTATION of {i: instruction, a: Annotations.annotation}
621		| INSTR of instr
622
623		This makes the handling of certain special instructions that appear on
624		all architectures easier and uniform.
625
626		LIVE and KILL say that a list of registers are live or killed at the
627		program point where they appear. No spill code is generated when an
628		element of the 'regs' field is spilled, but the register is moved to
629		the 'spilled' (which is present, more for debugging than anything else).
630
631		LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
632		We used to generate:
633
634		DEFFREG f1
635		f1 := f2 + f3
636		trapb
637
638		but now generate:
639
640		f1 := f2 + f3
641		trapb
642		LIVE {regs=[f1,f2,f3], spilled=[]}
643
644		Furthermore, the DEFFREG (hack) required that all floating point instruction
645		use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
646		defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
647		in a cleaner alpha implementation. (Hopefully, intel will not get rid of
648		this architecture).
649
650		COPYXXX is intended to replace the parallel COPY and FCOPY  available on
651		all the architectures. This will result in further simplification of the
652		register allocator that must be aware of them for coalescing purposes, and
653		will also simplify certain aspects of the machine description that provides
654		callbacks related to parallel copies.
655
656		ANNOTATION should be obvious, and now INSTR represents the honest to God
657		machine instruction set!
658
659		The <arch>/instructions/<arch>Instr.sml files define certain utility
660		functions for making porting easier -- essentially converting upper case
661		to lower case. All machine instructions (of type instr) are in upper case,
662		and the lower case form generates an MLRISC instruction. For example on
663		the alpha we have:
664
665		datatype instr =
666		LDA of {r:cell, b:cell, d:operand}
667		| ...
668
669		val lda : {r:cell, b:cell, d:operand} -> instruction
670		...
671
672		where lda is just (INSTR o LDA), etc.
673
674		----------------------------------------------------------------------
675		Name: Matthias Blume
676		Date: 2001/11/22 21:40:00 EST
677		Tag: Release_110_37
678		Description:
679
680		Release 110.37.  This time for real.
681
682		----------------------------------------------------------------------
683		Name: Matthias Blume
684		Date: 2001/11/21 16:35:00 EST
685		Tag: blume-20011121-foot-in-mouth
686		Description:
687
688		Removed the "Release_110_37" tag because of a serious bug.
689		This will be re-tagged once the bug is fixed.
690
691		----------------------------------------------------------------------
692		Name: Matthias Blume
693		Date: 2001/11/21 16:14:00 EST
694		Tag: blume-20011121-forgottenfile
695		Description:
696
697		Forgot to add a file.  (Just a .tex-file -- part of
698		the CM manual source.)
699
700		----------------------------------------------------------------------
701		Name: Matthias Blume
702		Date: 2001/11/21 16:10:00 EST
703		Tag: blume-20011121-invalid_110_37
704		Description:
705
706		Note: I removed the original tag "Release_110_37" from this commit
707		because we found a serious bug in all non-x86 backends.
708		- Matthias
709
710		1. Modifications to the SML/NJ code generator and to the runtime system
711		so that code object name strings are directly inserted into code
712		objects at code generation time.  The only business the runtime system
713		has with this is now to read the name strings on occasions.
714		(The encoding of the name string has also changed somewhat.)
715
716		2. CM now implements a simple "set calculus" for specifying export lists.
717		In particular, it is now possible to refer to the export lists of
718		other libraries/groups/sources and form unions as well as differences.
719		See the latest CM manual for details.
720
721		3. An separate notion of "proxy" libraries has again be eliminated from
722		CM's model.  (Proxy libraries are now simply a special case of using
723		the export list calculus.)
724
725		4. Some of the existing libraries now take advantage of the new set
726		calculus.
727		(Notice that not all libraries have been converted because some
728		of the existing .cm-files are supposed to be backward compatible
729		with 110.0.x.)
730
731		5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
732		"General.exnMessage"!  The former relies on a certain hook to be
733		initialized, and that often does not happen in the stand-alone case.)
734
735		----------------------------------------------------------------------
736		Name: Lal George
737		Date: 2001/11/21  13:56:18 EST
738		Tag: george-2001121-pseudo-ops
739		Description:
740
741		Implemented a complete redesign of MLRISC pseudo-ops. Now there
742		ought to never be any question of incompatabilities with
743		pseudo-op syntax expected by host assemblers.
744
745		For now, only modules supporting GAS syntax are implemented
746		but more should follow, such as MASM, and vendor assembler
747		syntax, e.g. IBM as, Sun as, etc.
748
749		----------------------------------------------------------------------
750		Name: Matthias Blume
751		Date: 2001/11/14 11:52:00 EST
752		Tag: blume-20011114-srcname
753		Description:
754
755		1. Routed the name of the current source file to mlriscgen where it
756		should be directly emitted into the code object.  (This last part
757		is yet to be done.)
758
759		2. Some cleanup of the pgraph code to make it match the proposal that
760		I put out the other day.  (The proposal notwithstanding, things are
761		still in flux here.)
762
763		----------------------------------------------------------------------
764		Name: Lal George
765		Date: 2001/11/14 09:44:04 EST
766		Tag:
767		Description:
768
769		Fix for a backpatching bug reported by Allen.
770
771		Because the boundary between short and long span-dependent
772		instructions is +/- 128, there are an astounding number of
773		span-dependent instructions whose size is over estimated.
774
775		Allen came up with the idea of letting the size of span
776		dependent instructions be non-monotonic, for a maxIter
777		number of times, after which the size must be monotonically
778		increasing.
779
780		This table shows the number of span-dependent instructions
781		whose size was over-estimated as a function of maxIter, for the
782		file Parse/parse/ml.grm.sml:
783
784		maxIter            # of instructions:
785		10                      687
786		20                      438
787		30                      198
788		40                        0
789
790		In compiling the compiler, there is no significant difference in
791		compilation speed between maxIter=10 and maxIter=40. Actually,
792		my measurements showed that maxIter=40 was a tad faster than
793		maxIter=10! Also 96% of the  files in the compiler reach a fix
794		point within 13 iterations, so fixing maxIter at 40, while high,
795		is okay.
796
797		----------------------------------------------------------------------
798		Name: Matthias Blume
799		Date: 2001/10/31 15:25:00 EST
800		Tag: blume-20011031-pgraph
801		Description:
802
803		CKIT:
804		* Changed the "Function" constructor of type Ast.ctype to carry optional
805		argument identifiers.
806		* Changed the return type of TypeUtil.getFunction accordingly.
807		* Type equality ignores the argument names.
808		* TypeUtil.composite tries to preserve argument names but gives up quickly
809		if there is a mismatch.
810
811		installation script:
812		* attempts to use "curl" if available (unless "wget" is available as well)
813
814		CM:
815		* has an experimental implementation of "portable graphs" which I will
816		soon propose as an implementation-independent library format
817		* there are also new libraries $/pgraph.cm and $/pgraph-util.cm
818
819		NLFFI-LIB:
820		* some cleanup (all cosmetic)
821
822		NLFFIGEN:
823		* temporarily disabled the mechanism that suppresses ML output for
824		C definitions whose identifiers start with an underscore character
825		* generate val bindings for enum constants
826		* user can request that only one style (light or heavy) is being used;
827		default is to use both (command-line arguments: -heavy and -light)
828		* fixed bug in handling of function types involving incomplete pointers
829		* generate ML entry points that take record arguments (i.e., using
830		named arguments) for C functions that have a prototype with named
831		arguments
832		(see changes to CKIT)
833
834		----------------------------------------------------------------------
835		Name: Allen Leung
836		Date: 2001/10/27 20:34:00 EDT
837		Tag: leunga-20011027-x86-fast-fp-call
838		Description:
839
840		Fixed the bug described in blume-20010920-slowfp.
841
842		The fix involves
843		1. generating FCOPYs in FSTP in ia32-svid
844		2. marking a CALL with the appropriate annotation
845
846		----------------------------------------------------------------------
847		Name: Matthias Blume
848		Date: 2001/10/16 11:32:00 EDT
849		Tag: blume-20011016-netbsd
850		Description:
851
852		Underscore patch from Chris Richards (fixing problem with compiling
853		runtime system under recent NetBSD).
854
855		----------------------------------------------------------------------
856		Name: Allen Leung
857		Date: 2001/10/12 17:18:32 EDT 2001
858		Tag: leung-20011012-x86-printflowgraph
859		Description:
860
861		X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
862
863		----------------------------------------------------------------------
864		Name: Lal George
865		Date: 2001/10/11 23:51:34 EDT
866		Tag: george-20011011-too-many-instrs
867		Description:
868
869		The representation of a program point never expected to see more
870		than 65536 instructions in a basic block!
871
872		----------------------------------------------------------------------
873		Name: Lal George
874		Date: 2001/10/09 09:41:37 EDT
875		Tag: george-20011008-mlrisc-labels
876		Description:
877
878		Changed the machine description files to support printing of
879		local and global labels in assembly code, based on host assembler
880		conventions.
881
882		----------------------------------------------------------------------
883		Name: Matthias Blume
884		Date: 2001/09/25 15:25:00 EDT
885		Tag: blume-20010925-exninfo
886		Description:
887
888		I provided a non-hook implementation of exnName (at the toplevel) and
889		made the "dummy" implementation of exnMessage (at the toplevel) more
890		useful: if nothing gets "hooked in", then at least you are going to
891		see the exception name and a message indicating why you don't see more.
892
893		[For the time being, programs that need exnMessage and want to use
894		ml-build should either use General.exnMessage (strongly recommended) or
895		refer to structure General at some other point so that CM sees a
896		static dependency.]
897
898		[Similar remarks go for "print" and "use":  If you want to use their
899		functionality in stand-alone programs generated by ml-build, then use
900		TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
901
902		----------------------------------------------------------------------
903		Name: Matthias Blume
904		Date: 2001/09/20 17:28:00 EDT
905		Tag: blume-20010920-slowfp
906		Description:
907
908		Allen says that x86-fast-fp is not safe yet, so I turned it off again...
909
910		----------------------------------------------------------------------
911		Name: Matthias Blume
912		Date: 2001/09/20 17:20:00 EDT
913		Tag: blume-20010920-canonicalpaths
914		Description:
915
916		0. Updated the BOOT file (something that I forgot to do earlier).
917
918		1. Small internal change to CM so that it avoids "/../" in filenames
919		as much as possible (but only where it is safe).
920
921		2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
922		that arguments that contain delimiters are passed through correctly.
923		This change also means that all "special" arguments of the form
924		@SMLxxx... must come first.
925
926		3. Changed install script to put relative anchor names for tool commands
927		into pathconfig.
928
929		----------------------------------------------------------------------
930		Name: Matthias Blume
931		Date: 2001/09/18 15:35:00 EDT
932		Tag: blume-20010918-readme11036
933		Description:
934
935		Added README files.
936
937		----------------------------------------------------------------------
938		Name: Matthias Blume
939		Date: 2001/09/18 11:45:00 EDT
940		Tag: Release_110_36 (retag)
941		Description:
942
943		Fixed mistake in config/preloads. Retagged as 110.36.
944
945		----------------------------------------------------------------------
946		Name: Matthias Blume
947		Date: 2001/09/18 09:40:00 EDT
948		Tag: Release_110_36_orig (tag changed)
949		Description:
950
951		New version (110.36).  New bootfiles.
952
953		----------------------------------------------------------------------
954		Name: Matthias Blume
955		Date: 2001/09/14 16:15:00 EDT
956		Tag: blume-20010914-x86fastfp
957		Description:
958
959		John committed some changes that Allen made, in particular a (hopefully)
960		correctly working version of the x86-fp module.
961
962		I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
963		flag to "true".  Everything seems to compile to a fixpoint ok, and
964		"mandelbrot" speeds up by about 15%.
965
966		----------------------------------------------------------------------
967		Name: Matthias Blume
968		Date: 2001/09/13 11:20:00 EDT
969		Tag: blume-20010913-minimal
970		Description:
971
972		1. Stefan Monnier's patch to fix a miscompilation problem that
973		was brought to light by John Reppy's work on Moby.
974
975		2. Implemented a minimal "structure Compiler" that contains just
976		"version" and "architecture".  The minimal version will be
977		available when the full version is not.  This is for backward-
978		compatibility with code that wants to test Compiler.version.
979
980		----------------------------------------------------------------------
981		Name: Matthias Blume
982		Date: 2001/08/28 14:03:00 EDT
983		Tag: blume-20010828-ml-lex
984		Description:
985
986		Fix for bug 1581, received from Neophytos Michael.
987
988		----------------------------------------------------------------------
989		Name: Matthias Blume
990		Date: 2001/08/27 11:20:00 EDT
991		Tag: blume-20010827-readme11035
992		Description:
993
994		Fleshed out the README file for 110.35.
995
996		----------------------------------------------------------------------
997		Name: Matthias Blume
998		Date: 2001/08/24 17:10:00 EDT
999		Tag: Release_110_35
1000		Description:
1001
1002		New version number (110.35).  New bootfiles.
1003
1004		----------------------------------------------------------------------
1005		Name: Lal George
1006		Date: 2001/08/24 13:47:18 EDT 2001
1007		Tag: george-20010824-MLRISC-graphs
1008		Description:
1009
1010		removed clusters from MLRISC completely and replaced with graphs.
1011
1012		----------------------------------------------------------------------
1013		Name: Matthias Blume
1014		Date: 2001/08/23 17:50:00 EDT
1015		Tag: blume-20010823-toplevel
1016		Description:
1017
1018		- some reorganization of the code that implements various kinds of
1019		environments in the compiler (static, dynamic, symbolic, combined)
1020		- re-implemented the EnvRef module so that evalStream works properly
1021		(if the stream contains references to "use", "CM.make", etc.)
1022		- cleaned up evalloop.sml and interact.sml (but they need more cleaning)
1023
1024		----------------------------------------------------------------------
1025		Name: Matthias Blume
1026		Date: 2001/08/20 15:50 EDT
1027		Tag: blume20010820-slipup
1028		Description:
1029
1030		I forgot to commit a few files.  Here they are...
1031
1032		----------------------------------------------------------------------
1033		Name: Matthias Blume
1034		Date: 2001/08/20 15:35:00 EDT
1035		Tag: blume-20010820-debugprof
1036		Description:
1037
1038		!!!! NEW BOOTFILES !!!!
1039
1040		This is another round of reorganizing the compiler sources.  This
1041		time the main goal was to factor out all the "instrumentation"
1042		passes (for profiling and backtracing) into their own library.
1043		The difficulty was to do it in such a way that it does not depend
1044		on elaborate.cm but only on elabdata.cm.
1045
1046		Therefore there have been further changes to both elaborate.cm and
1047		elabdata.cm -- more "generic" things have been moved from the former
1048		to the latter.  As a result, I was forced to split the assignment
1049		of numbers indicating "primtyc"s into two portions: SML-generic and
1050		SML/NJ-specific.  Since it would have been awkward to maintain,
1051		I bit the bullet and actually _changed_ the mapping between these
1052		numbers and primtycs.  The bottom line of this is that you need
1053		a new set of bin- and bootfiles.
1054
1055		I have built new bootfiles for all architectures, so doing a fresh
1056		checkout and config/install.sh should be all you need.
1057
1058		The newly created library's name is
1059
1060		$smlnj/viscomp/debugprof.cm
1061
1062		and its sources live under
1063
1064		src/compiler/DebugProf
1065
1066		----------------------------------------------------------------------
1067		Name: Matthias Blume
1068		Date: 2001/08/15 17:15:00 EDT
1069		Tag: blume-20010815-compreorg
1070		Description:
1071
1072		This is a first cut at reorganizing the CM libraries that make up the
1073		core of the compiler.  The idea is to separate out pieces that could
1074		be used independently by tools, e.g., the parser, the typechecker, etc.
1075
1076		The current status is a step in this direction, but it is not quite
1077		satisfactory yet.  Expect more changes in the future.
1078
1079		Here is the current (new) organization...
1080
1081		What used to be $smlnj/viscomp/core.cm is now divided into
1082		six CM libraries:
1083
1084		$smlnj/viscomp/basics.cm
1085		/parser.cm
1086		/elabdata.cm
1087		/elaborate.cm
1088		/execute.cm
1089		/core.cm
1090
1091		The CM files for these libraries live under src/system/smlnj/viscomp.
1092		All these libraries are proxy libraries that contain precisely
1093		one CM library component.  Here are the locations of the components
1094		(all within the src/compiler tree):
1095
1096		Basics/basics.cm
1097		Parse/parser.cm
1098		ElabData/elabdata.cm
1099		Elaborator/elaborate.cm
1100		Execution/execute.cm
1101		core.cm
1102
1103		[This organization is the same that has been used already
1104		for a while for the architecture-specific parts of the visible
1105		compiler and for the old version of core.cm.]
1106
1107		As you will notice, many source files have been moved from their
1108		respective original locations to a new home in one of the above
1109		subtrees.
1110
1111		The division of labor between the new libraries is the following:
1112
1113		basics.cm:
1114		- Simple, basic definitions that pertain to many (or all) of
1115		the other libraries.
1116		parser.cm:
1117		- The SML parser, producing output of type Ast.dec.
1118		- The type family for Ast is also defined and exported here.
1119		elabdata.cm:
1120		- The datatypes that describe input and output of the elaborator.
1121		This includes types, absyn, and static environments.
1122		elaborator.cm:
1123		- The SML/NJ type checker and elaborator.
1124		This maps an Ast.dec (with a given static environment) to
1125		an Absyn.dec (with a new static environment).
1126		- This libraries implements certain modules that used to be
1127		structures as functors (to remove dependencies on FLINT).
1128		execute.cm:
1129		- Everything having to do with executing binary code objects.
1130		- Dynamic environments.
1131		core.cm:
1132		- SML/NJ-specific instantiations of the elaborator and MLRISC.
1133		- Top-level modules.
1134		- FLINT (this should eventually become its own library)
1135
1136		Notes:
1137
1138		I am not 100% happy with the way I separated the elaborator (and its
1139		data structures) from FLINT.  Two instances of the same problem:
1140
1141		1. Data structures contain certain fields that carry FLINT-specific
1142		information.  I hacked around this using exn and the property list
1143		module from smlnj-lib.  But the fact that there are middle-end
1144		specific fields around at all is a bit annoying.
1145
1146		2. The elaborator calculates certain FLINT-related information.  I tried
1147		to make this as abstract as I could using functorization, but, again,
1148		the fact that the elaborator has to perform calculations on behalf
1149		of the middle-end at all is not nice.
1150
1151		3. Having to used exn and property lists is unfortunate because it
1152		weakens type checking.  The other alternative (parameterizing
1153		nearly *everything*) is not appealing, though.
1154
1155		I removed the "rebinding =" warning hack because due to the new organization
1156		it was awkward to maintain it.  As a result, the compiler now issues some of
1157		these warnings when compiling init.cmi during bootstrap compilation. On
1158		the plus side, you also get a warning when you do, for example:
1159		val op = = Int32.+
1160		which was not the case up to now.
1161
1162		I placed "assign" and "deref" into the _Core structure so that the
1163		code that deals with the "lazy" keyword can find them there.  This
1164		removes the need for having access to the primitive environment
1165		during elaboration.
1166
1167		----------------------------------------------------------------------
1168		Name: Matthias Blume
1169		Date: 2001/08/13
1170		Tag: blume-20010813-closures
1171		Description:
1172
1173		This fix was sent to us by Zhong Shao.  It is supposed to improve the
1174		performance of certain loops by avoiding needless closure allocation.
1175
1176		----------------------------------------------------------------------
1177		Name: Lal George
1178		Date: 2001/07/31 10:03:23 EDT 2001
1179		Tag: george-20010731-x86-fmalloc
1180		Description: Fixed bug in x86 calls
1181
1182		There was a bug where call instructions would mysteriously
1183		vanish. The call instruction had to be one that returned
1184		a floating point value.
1185
1186		----------------------------------------------------------------------
1187		Name: Lal George
1188		Date: 2001/07/19 16:36:29 EDT 2001
1189		Tag: george-20010719-simple-cells
1190		Description:
1191
1192		I have dramatically simplified the interface for CELLS in MLRISC.
1193
1194		In summary, the cells interface is broken up into three parts:
1195
1196		1. CellsBasis : CELLS_BASIS
1197
1198		CellsBasis is a top level structure and common for all
1199		architectures.  it contains the definitions of basic datatypes
1200		and utility  functions over these types.
1201
1202		2. functor Cells() : CELLS
1203
1204		Cells generates an interface for CELLS that incorporates the
1205		specific resources on the target architecture, such as the
1206		presence of special register classes, their number and size,
1207		and various useful substructures.
1208
1209		3. <ARCH>CELLS
1210
1211		e.g. SparcCells: SPARCCELLS
1212
1213		<ARCH>CELLS usually contains additional bindings for special
1214		registers  on the architecture, such as:
1215
1216		val r0 : cell           (* register zero *)
1217		val y : cell            (* Y register *)
1218		val psr : cell          (* processor status register *)
1219		...
1220
1221		The structure returned by applying the Cells functor is opened
1222		in this interface.
1223
1224		The main implication of all this is that the datatypes for cells is
1225		split between CellsBasis and CELLS -- a fairly simple change for user
1226		code.
1227
1228		In the old scheme the CELLS interface had a definitional binding of
1229		the form:
1230
1231		signature CELLS = sig
1232
1233		structure CellsBasis = CellsBasis
1234
1235		...
1236
1237		end
1238
1239		With all the sharing constraints that goes on in MLRISC, this old
1240		design  quickly leads to errors such as:
1241
1242		"structure definition spec inside of sharing ... "
1243
1244
1245		and appears to require an unacceptable amount of sharing and where
1246		constraint hackery.
1247
1248		I think this error message (the interaction of definitional specs and
1249		sharing) requires more explanation on our web page.
1250
1251		----------------------------------------------------------------------
1252		Name: Matthias Blume
1253		Date: 2001/07/19 15:00:00 EDT
1254		Tag: blume-20010719-libreorg
1255		Description:
1256
1257		This update puts together a fairly extensive but straightforward change
1258		to the way the libraries that implement the interactive system are
1259		organized:
1260
1261		The biggest change is the elimination of structure Compiler.  As a
1262		replacement for this structure, there is now a CM library
1263		(known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
1264		that exports all the substructures of the original structure Compiler
1265		directly.  So instead of saying Compiler.Foo.bar one now simply
1266		says Foo.bar.  (The CM libraries actually export a collection of
1267		structures that is richer than the collection of substructures of
1268		structure Compiler.)
1269
1270		To make the transition smooth, there is a separate library called
1271		$smlnj/compiler/compiler.cm which puts together and exports the
1272		original structure Compiler (or at least something very close to it).
1273
1274		There are five members of the original structure Compiler
1275		that are not exported directly but which instead became members
1276		of a new structure Backend (described by signature BACKEND).  These are:
1277		structure Profile (: PROFILE), structure Compile (: COMPILE), structure
1278		Interact (: INTERACT), structure Machine (: MACHINE), and val
1279		architecture (: string).
1280
1281		Structure Compiler.Version has become structure CompilerVersion.
1282
1283		Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
1284		by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
1285		or x86, respectively.
1286		Each of these exports the same frontend structures that
1287		$smlnj/compiler.cm exports.  But they do not have a structure Backend
1288		and instead export some structure <Arch>Backend where <Arch> is Alpha32,
1289		Hppa, PPC, Sparc, or X86, respectively.
1290
1291		Library $smlnj/compiler/all.cm exports the union of the exports of
1292		$smlnj/compiler/<arch>.cm
1293
1294		There are no structures <Arch>Compiler anymore, use
1295		$smlnj/compiler/<arch>.cm instead.
1296
1297		Library host-compiler-0.cm is gone.  Instead, the internal library
1298		that instantiates CM is now called cm0.cm.  Selection of the host
1299		compiler (backend) is no longer done here but. (Responsibility for it
1300		now lies with $smlnj/compiler/current.cm.  This seems to be more
1301		logical.)
1302
1303		Many individual files have been moved or renamed.  Some files have
1304		been split into multiple files, and some "dead" files have been deleted.
1305
1306		Aside from these changes to library organization, there are also changes
1307		to the way the code itself is organized:
1308
1309		Structure Binfile has been re-implemented in such a way that it no
1310		longer needs any knowledge of the compiler.  It exclusively deals
1311		with the details of binfile layout.  It no longer invokes the
1312		compiler (for the purpose of creating new prospective binfile
1313		content), and it no longer has any knowledge of how to interpret
1314		pickles.
1315
1316		Structure Compile (: COMPILE) has been stripped down to the bare
1317		essentials of compilation.  It no longer deals with linking/execution.
1318		The interface has been cleaned up considerably.
1319
1320		Utility routines for dealing with linking and execution have been
1321		moved into their own substructures.
1322
1323		(The ultimate goal of these changes is to provide a light-weight
1324		binfile loader/linker (at least for, e.g., stable libraries) that
1325		does not require CM or the compiler to be present.)
1326
1327		CM documentation has been updated to reflect the changes to library
1328		organization.
1329
1330		----------------------------------------------------------------------
1331		Name: Matthias Blume
1332		Date: 2001/07/10 17:30:00 EDT
1333		Tag: Release_110_34
1334		Description:
1335
1336		Minor tweak to 110.34 (re-tagged):
1337
1338		- README.html file added to CVS repository
1339		- runtime compiles properly under FreeBSD 3.X and 4.X
1340
1341		----------------------------------------------------------------------
1342		Name: Matthias Blume
1343		Date: 2001/07/10 17:30:00 EDT
1344		Tag: Release_110_34
1345		Description:
1346
1347		New version number (110.34). New bootfiles.
1348
1349		----------------------------------------------------------------------
1350		Name: Matthias Blume
1351		Date: 2001/07/09 16:00:00 EDT
1352		Tag: blume-20010709-more-varargs
1353		Description:
1354
1355		I changed the handling of varargs in ml-nlffigen again:
1356		The ellipsis ... will now simply be ignored (with an accompanying warning).
1357
1358		The immediate effect is that you can actually call a varargs function
1359		from ML -- but you can't actually supply any arguments beyond the ones
1360		specified explicitly.  (For example, you can call printf with its format
1361		string, but you cannot pass additional arguments.)
1362
1363		This behavior is only marginally more useful than the one before, but
1364		it has the advantage that a function or, more importantly, a function
1365		type never gets dropped on the floor, thus avoiding follow-up problems with
1366		other types that refer to the offending one.
1367
1368		----------------------------------------------------------------------
1369		Name: Matthias Blume
1370		Date: 2001/07/09 11:25:00 EDT
1371		Tag: blume-20010709-varargs
1372		Description:
1373
1374		1. ckit-lib.cm now exports structure Error
1375		2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
1376		with a warning accompanied by a source location.  Moreover, it
1377		merely skips the offending function or type and proceeds with the
1378		rest of its work.u  As a result, one can safely feed C code containing
1379		"..." to ml-nlffigen.
1380		3. There are some internal improvements to CM, providing slightly
1381		more general string substitutions in the tools subsystem.
1382
1383		----------------------------------------------------------------------
1384		Name: Matthias Blume
1385		Date: 2001/06/27 15:10:00 EDT
1386		Tag: blume-20010627-concur
1387		Description:
1388
1389		Fixed a small bug in CM's handling of parallel compilation.
1390		(You could observe the bug by Control-C-interrupting an ordinary
1391		CMB.make or CM.stabilize and then attaching some compile servers.
1392		The result was that all of a sudden the previously interrupted
1393		compilation would continue on its own.  This was because of
1394		an over-optimization: CM did not bother to clean out certain queues
1395		when no servers were attached "anyway", resulting in the contents
1396		of these queues to grab control when new servers did get attached.)
1397
1398		There is also another minor update to the CM manual.
1399
1400		----------------------------------------------------------------------
1401		Name: Matthias Blume
1402		Date: 2001/06/26 16:15:00 EDT
1403		Tag: blume-20010626-cmdoc
1404		Description:
1405
1406		Minor typo fixed in CM manual (syntax diagram for libraries).
1407
1408		----------------------------------------------------------------------
1409		Name: Matthias Blume
1410		Date: 2001/06/25 22:55:00 EDT
1411		Tag: blume-20010625-x86pc
1412		Description:
1413
1414		Fixed a nasty bug in the X86 assembly code that caused signal
1415		handlers to fail (crash) randomly.
1416
1417		----------------------------------------------------------------------
1418		Name: Matthias Blume
1419		Date: 2001/06/25 12:05:00 EDT
1420		Tag: blume-20010625-nlffigen
1421		Description:
1422
1423		This update fixes a number of minor bugs in ml-nlffigen as reported by
1424		Nick Carter <nbc@andrew.cmu.edu>.
1425
1426		1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
1427		2. Default names for generated files are now derived from the name of
1428		the C file *without its directory*.  In particular, this causes generated
1429		files to be placed locally even if the C file is in some system directory.
1430		3. Default names for generated signatures and structures are also derived
1431		from the C file name without its directory.  This avoids silly things
1432		like "structure GL/GL".
1433		(Other silly names are still possible because ml-nlffigen does not do
1434		a thorough check of whether generated names are legal ML identifiers.
1435		When in doubt, use command line arguments to force particular names.)
1436
1437		----------------------------------------------------------------------
1438		Name: Matthias Blume
1439		Date: 2001/06/21 12:25:00 EDT
1440		Tag: blume-20010621-eXene
1441		Description:
1442
1443		eXene now compiles and (sort of) works again.
1444
1445		The library name (for version > 110.33) is $/eXene.cm.
1446
1447		I also added an new example in src/eXene/examples/nbody.  See the
1448		README file there for details.
1449
1450		----------------------------------------------------------------------
1451		Name: Matthias Blume
1452		Date: 2001/06/20 16:40:00 EDT
1453		Tag: blume-20010620-cml
1454		Description:
1455
1456		CML now compiles and works again.
1457
1458		Libraries (for version > 110.33):
1459
1460		$cml/cml.cm            Main CML library.
1461		$cml/basis.cm          CML's version of $/basis.cm.
1462		$cml/cml-internal.cm   Internal helper library.
1463		$cml/core-cml.cm       Internal helper library.
1464		$cml-lib/trace-cml.cm  Tracing facility.
1465		$cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
1466
1467		The installer (config/install.sh) has been taught how to properly
1468		install this stuff.
1469
1470		----------------------------------------------------------------------
1471		Name: Matthias Blume
1472		Date: 2001/06/19 17:55:00 EDT
1473		Tag: blume-20010619-instantiate
1474		Description:
1475
1476		This un-breaks the fix for bug 1432.
1477		(The bug was originally fixed in 110.9 but I broke it again some
1478		time after that.)
1479
1480		----------------------------------------------------------------------
1481		Name: Matthias Blume
1482		Date: 2001/06/19 17:25:00 EDT
1483		Tag: blume-20010619-signals
1484		Description:
1485
1486		This should (hopefully) fix the long-standing signal handling bug.
1487		(The runtime system was constructing a continuation record with an
1488		incorrect descriptor which would cause the GC to drop data on the floor...)
1489
1490		----------------------------------------------------------------------
1491		Name: Matthias Blume
1492		Date: 2001/06/15 15:05:00 EDT
1493		Tag: blume-20010615-moresparc
1494		Description:
1495
1496		Here is a short late-hour update related to Sparc c-calls:
1497
1498		-- made handling of double-word arguments a bit smarter
1499
1500		-- instruction selection phase tries to collapse certain clumsily
1501		constructed ML-Trees; typical example:
1502
1503		ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
1504
1505		This currently has no further impact on SML/NJ since mlriscGen does
1506		not seem to generate such patterns in the first place, and c-calls
1507		(which did generate them in the beginning) has meanwhile been fixed
1508		so as to avoid them as well.
1509
1510		----------------------------------------------------------------------
1511		Name: Matthias Blume
1512		Date: 2001/06/15 15:05:00 EDT
1513		Tag: blume-20010615-sparc
1514		Description:
1515
1516		The purpose of this update is to provide an implementation of NLFFI
1517		on Sparc machines.
1518
1519		Here are the changes in detail:
1520
1521		* src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
1522		the Sparc implementation of the c-calls API.
1523		* The Sparc backend of SML/NJ has been modified to uniformely use %fp
1524		for accessing the ML frame.  Thus, we have a real frame pointer and
1525		can freely modify %sp without need for an omit-frame-ptr phase.
1526		The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
1527		this case.
1528		* ml-nlffigen has been taught to produce code for different architectures
1529		and calling conventions.
1530		* In a way similar to what was done in the x86 case, the Sparc
1531		backend uses its own specific extension to mltree.  (For example,
1532		it needs to be able to generate UNIMP instructions which are part
1533		of the calling convention.)
1534		* ml-nlffi-lib was reorganized to make it more modular (in particular,
1535		to make it easier to plug in new machine- and os-dependent parts).
1536
1537		There are some other fairly unrelated bug fixes and cleanups as well:
1538
1539		* I further hacked the .cm files for MLRISC tools (like MDLGen) so
1540		that they properly share their libraries with existing SML/NJ libraries.
1541		* I fixed a minor cosmetic bug in CM, supressing certain spurious
1542		follow-up error messages.
1543		* Updates to CM/CMB documentation.
1544
1545		TODO items:
1546
1547		* MLRISC should use a different register as its asmTemp on the Sparc.
1548		(The current %o2 is a really bad choice because it is part of the
1549		calling conventions, so things might interfere in unexpected ways.)
1550
1551		----------------------------------------------------------------------
1552		Name: Matthias Blume
1553		Date: 2001/06/07
1554		Tag: blume-20010607-calls
1555		Description:
1556
1557		A number of internal changes related to C calls and calling conventions:
1558
1559		1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
1560		number of bytes popped implicitly (by the callee).  In most cases
1561		this field is 0 but on x86/win32 it is some non-zero value.  This
1562		is information provided for the benefit of the "omit-frameptr" pass.
1563		2. The CALL instruction on the x86 carries a similar "pops" field.
1564		The instruction selection phase copies its value from the ML-Tree
1565		CALL statement.
1566		3. On all other architectures, the instruction selection phase checks
1567		whether "pops=0" and complains if not.
1568		4. The c-calls implementation for x86 now accepts two calling conventions:
1569		"ccall" and "stdcall".  When "ccall" is selected, the caller cleans
1570		up after the call and pops is set to 0.  For "stdcall", the caller
1571		does nothing, leaving the cleanup to the callee; pops is set to
1572		the number of bytes that were pushed onto the stack.
1573		5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
1574		distinguish between "ccall" and "stdcall".
1575		6. The UNIMP instruction has been added to the supported Sparc instruction
1576		set. (This is needed for implementing the official C calling convention
1577		on this architecture.)
1578		7. I fixed some of the .cm files under src/MLRISC/Tools to make them
1579		work with the latest CM.
1580
1581		----------------------------------------------------------------------
1582		Name: Matthias Blume
1583		Date: 2001/06/05 15:10:00 EDT
1584		Tag: blume-20010605-cm-index
1585		Description:
1586
1587		0. The "lambdasplit" parameter for class "sml" in CM has been documented.
1588
1589		1. CM can now generate "index files".  These are human-readable files
1590		that list on a per-.cm-file basis each toplevel symbol defined or
1591		imported.  The location of the index file for
1592		<p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
1593		To enable index-file generation, set CM.Control.generate_index to true
1594		or export an environment-symbol: export CM_GENERATE_INDEX=true.
1595
1596		The CM manual has been updated accordingly.
1597
1598		2. I made some slight modifications to the c-calls API in MLRISC.
1599
1600		a) There is now a callback to support saving/restoring of
1601		dedicated but caller-save registers around the actual call
1602		instruction.
1603		b) One can optionally specify a comment-annotation for the
1604		call instruction.
1605
1606		3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
1607		(For example, the comment annotation shows the C prototype of
1608		the function being called.)
1609
1610		----------------------------------------------------------------------
1611		Name: Matthias Blume
1612		Date: 2001/06/01 13:30:00 EDT
1613		Tag: blume-20010601-nlffi-cleanup
1614		Description:
1615
1616		This is mostly a cleanup of MLFFI stuff:
1617
1618		- some signature files have been put into a more exposed place
1619		- the ugly 'f type parameter is gone (simplifies types tremendously!)
1620		- ml-nlffigen changed accordingly
1621		- tutorial updated
1622
1623		Other changes:
1624
1625		- author's affiliation in CM manual(s) updated
1626		- some more recognized keywords added to Allen's sml.sty
1627
1628		----------------------------------------------------------------------
1629		Name: Matthias Blume
1630		Date: 2001/05/25 15:30:00 EDT
1631		Tag: blume-20010525-iptr
1632		Description:
1633
1634		- put the official 110.33-README (as it appears on the ftp server) under
1635		CVS
1636		- fixed a small bug related to incomplete pointer types in
1637		ml-nlffigen
1638		- small cosmetic change to the ml-nlffi-lib's "arr" type constructor
1639		(it does not need the 'f type parameter)
1640
1641		----------------------------------------------------------------------
1642		Name: Matthias Blume
1643		Date: 2001/05/23 14:30:00 EDT
1644		Tag: Release_110_33
1645		Description:
1646
1647		New version number (110.33).  New bootfiles.
1648
1649		----------------------------------------------------------------------
1650		Name: Matthias Blume
1651		Date: 2001/05/22 18:06:00 EDT
1652		Tag: blume-20010522-targets
1653		Description:
1654
1655		Made install.sh use file config/targets.customized if it exists, falling
1656		back to config/targets if it doesn't.  This way one can have a customized
1657		version of the targets file without touching the "real thing", thus
1658		eliminating the constant fear of accidentally checking something bogus
1659		back into the CVS repository...  (File config/targets.customized must
1660		not be added to the repository!)
1661
1662		----------------------------------------------------------------------
1663		Name: Matthias Blume
1664		Date: 2001/05/22 16:30:00 EDT
1665		Tag: blume-20010522-minitut
1666		Description:
1667
1668		1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
1669		struct returns.
1670		2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
1671		incomplete, preliminary documentation for NLFFI.
1672
1673		----------------------------------------------------------------------
1674		Name: Matthias Blume
1675		Date: 2001/05/14 11:30:00 EDT
1676		Tag: blume-20010514-script
1677		Description:
1678
1679		Some bugs in install script fixed.
1680
1681		In addition to that I also made a slight change to the NLFFI API:
1682		Functors generated by ml-nlffigen now take the dynamic library as a
1683		straight functor argument, not as a suspended one.  (The original
1684		functor code used to force the suspension right away anyway, so there
1685		was nothing gained by this complication of the interface.)
1686
1687		----------------------------------------------------------------------
1688		Name: Matthias Blume
1689		Date: 2001/05/11 14:35:00 EDT
1690		Tag: blume-20010511-ml-nlffi
1691		Description:
1692
1693		I finally took the plunge and added my new FFI code to the main
1694		repository.  For x86-linux it is now ready for prime-time.
1695
1696		There are two new subdirectories of "src":
1697
1698		- ml-nlffi-lib:
1699		The utility library for programs using the FFI interface.
1700		Here is the implementation of $/c.cm and its associated low-level
1701		partners $/c-int.cm and $/memory.cm.
1702		- ml-nlffigen:
1703		A stand-alone program for generating ML glue code from C source
1704		code.
1705
1706		Building ml-nlffigen requires $/ckit-lib.cm.
1707
1708		The config/install.sh script has been updates to do the Right Thing
1709		(hopefully).
1710
1711		Notice that the source tree for the C-Kit will not be put under "src"
1712		but directly under the installation root directory.  (This is the
1713		structure that currently exists on the CVS server when you check out
1714		module "sml".)  Fortunately, config/install.sh knows about this oddity.
1715
1716		Bugs: No documentation yet.
1717
1718		----------------------------------------------------------------------
1719		Name: Matthias Blume
1720		Date: 2001/05/09 16:35:00 EDT
1721		Tag: blume-20010509-cpscontract
1722		Description:
1723
1724		Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
1725		wrapper/unwrapper elimination did not decrement usage counts and some
1726		dead variables got overlooked by the dead-up logic.)
1727
1728		----------------------------------------------------------------------
1729		Name: Lal George
1730		Date: 2001/05/08  17:26:09 EDT
1731		Tag: george-20010508-omit-frameptr
1732		Description:
1733
1734		Changes to implement the omit-frame-pointer optimization to support
1735		raw C calls. For now, there is only support on the Intel x86, but
1736		other architectures will follow as more experience is gained with this.
1737
1738
1739		----------------------------------------------------------------------
1740		Name: Matthias Blume
1741		Date: 2001/05/07 14:40:00 EDT
1742		Tag: blume-20010507-proxies
1743		Description:
1744
1745		I made into "proxy libraries" all libraries that qualify for such a
1746		change.  (A qualifying library is a library that has another library or
1747		groups as its sole member and repeats that member's export list
1748		verbatim.  A proxy library avoids this repetition by omitting its export
1749		list, effectively inheriting the list that its (only) member exports.
1750		See the CM manual for more explanation.)
1751		The main effect is that explicit export lists for these libraries
1752		do not have to be kepts in sync, making maintenance a bit easier.
1753
1754		I also added copyright notices to many .cm-files.
1755
1756		Last but not least, I made a new set of bootfiles.
1757
1758		----------------------------------------------------------------------
1759		Name: Matthias Blume
1760		Date: 2001/05/04 17:00:00 EDT
1761		Tag: blume-20010504-cm-lsplit
1762		Description:
1763
1764		0. John merged pending changes to $/smlnj-lib.cm
1765
1766		1. Allen's previous change accidentally backed out of one of Lal's
1767		earlier changes.  I undid this mistake (re-introducing Lal's change).
1768
1769		2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
1770		within the compiler where applicable.  There is some code simplification
1771		because of that.
1772
1773		3. The "split" phase (in FLINT) is now part of the default list of phases.
1774		Compiler.Control.LambdaSplitting.* can be used to globally control the
1775		lambda-splitting (cross-module-inlining) engine.  In addition to that,
1776		it can now also be controlled on a per-source basis: CM has been taught
1777		a new tool parameter applicable to ML source files.
1778
1779		- To turn lambda-splitting off completely:
1780		local open Compiler.Control.LambdaSplitting in
1781		val _ = set Off
1782		end
1783		- To make "no lambda-splitting" the global default (but allow per-source
1784		overriding); this is the initial setting:
1785		local open Compiler.Control.LambdaSplitting in
1786		val _ = set (Default NONE)
1787		end
1788		- To make "lambda-splitting with aggressiveness a" the global default
1789		(and allow per-source overriding):
1790		local open Compiler.Control.LambdaSplitting in
1791		val _ = set (Default (SOME a))
1792		end
1793
1794		- To turn lambda-splitting off for a given ML souce file (say: a.sml)
1795		write (in the respective .cm-file):
1796		a.sml (lambdasplitting:off)
1797		- To turn lambda-splitting for a.sml on with minimal aggressiveness:
1798		a.sml (lambdasplitting:on)
1799		- To turn lambda-splitting for a.sml on with aggressiveness <a> (where
1800		<a> is a decimal non-negative integer):
1801		a.sml (lambdasplitting:<a>)
1802		- To turn lambda-splitting for a.sml on with maximal aggressiveness:
1803		a.sml (lambdasplitting:infinity)
1804		- To use the global default for a.sml:
1805		a.sml (lambdasplitting:default)
1806		or simply
1807		a.sml
1808
1809		----------------------------------------------------------------------
1810		Name: Allen Leung
1811		Date: 2001/05/04 01:57:00 EDT
1812		Tag: leunga-20010504-sync
1813		Description:
1814
1815		MLRISC features.
1816
1817		1. Fix to CMPXCHG instructions.
1818		2. Changed RA interface to allow annotations in callbacks.
1819		3. Added a new method to the stream interface to allow annotations updates.
1820
1821		----------------------------------------------------------------------
1822		Name: Matthias Blume
1823		Date: 2001/05/01 11:45:00 EDT
1824		Tag: blume-20010501-pcedittmp
1825		Description:
1826
1827		Changed install.sh to use the current working directory instead of
1828		/usr/tmp for a temporary file (pcedittmp).  The previous choice
1829		of /usr/tmp caused trouble with MacOS X because of file premission
1830		problems.
1831
1832		----------------------------------------------------------------------
1833		Name: Matthias Blume
1834		Date: 2001/04/20 11:10:00 EDT
1835		Tag: blume-20010420-inMLflag
1836		Description:
1837
1838		- added vp_limitPtrMask to vproc-state.h
1839		(for use by the raw-C-calls mechanism to implement proper interrupt
1840		handling)
1841		- made the ML compiler aware of various data-structure offsets so it
1842		can generate code for accessing the vp_inML flag and vp_limitPtrMask
1843		- tweaked mlriscGen.sml to have it emit interrupt-handling code for
1844		raw C-calls
1845
1846		----------------------------------------------------------------------
1847		Name: Lal George
1848		Date: 2001/04/20 09:15:28 EDT
1849		Tag: george-20010420-macosX
1850		Description:
1851
1852		- Changes to port to Mac OS X; Darwin.
1853
1854		- In the process I found that sqrt was broken on the PPC, because the
1855		fsqrt instruction is not implemented.
1856
1857		----------------------------------------------------------------------
1858		Name: Matthias Blume
1859		Date: 2001/04/18 12:45:00 EDT
1860		Tag: blume-20010418-ccalls
1861		Description:
1862
1863		- fixed two off-by-4 errors in the x86-specific c-calls implementation
1864		(this bug prevented structure arguments containing pointers from being
1865		passed correctly)
1866		- changed the raw-C-call code in mlriscGen.sml in such a way that
1867		structure arguments are represented as a pointer to the beginning
1868		of the structure (instead of having a series of synthesized arguments,
1869		one for each structure member)
1870
1871		- made makeml script's verbosity level configurable via environment
1872		variable (MAKEML_VERBOSITY)
1873
1874		- eliminated placeholder implementations for f32l, w16s, i16s, and f32s
1875		in rawmem-x86.sml; we are now using the real thing
1876
1877		----------------------------------------------------------------------
1878		Name: Matthias Blume
1879		Date: 2001/03/22 16:25:00 EST
1880		Tag: blume-20010322-bootfiles
1881		Description:
1882
1883		Created a new set of bootfiles (for your automatic installation convenience).
1884
1885		----------------------------------------------------------------------
1886		Name: Matthias Blume
1887		Date: 2001/03/22 15:10:00 EST
1888		Tag: blume-20010322-rawmem-parcm
1889		Description:
1890
1891		1. All "raw memory access" primitives for the new FFI are implemented now
1892		(at least on the x86).
1893		2. Some further cleanup of CM's parallel make mechanism.
1894
1895		----------------------------------------------------------------------
1896		Name: Matthias Blume
1897		Date: 2001/03/19 17:53:00 EST
1898		Tag: blume-20010319-parallel
1899		Description:
1900
1901		Parallel make (using compile servers) now works again.
1902
1903		To this end, CM.stabilize and CMB.make have been modified to work in
1904		two passes when compile servers are attached:
1905		1. Compile everything, do not perform stabilization; this pass
1906		uses compile servers
1907		2. Stabilize everything; this pass does not use compile servers
1908		If there are no compile servers, the two passes are combined into one
1909		(as before).  Splitting the passes increases the inherent parallelism
1910		in the dependency graph because the entire graph including all
1911		libraries is available at the same time.  This, in turn, improves
1912		server utilization.  The downside is that the master process will
1913		have to do some extra work after compilation is done (because for
1914		technical reasons it must re-read all the binfiles during stabilization).
1915
1916		----------------------------------------------------------------------
1917		Name: Matthias Blume
1918		Date: 2001/03/16 12:22:00 EST
1919		Tag: blume-20010316-bootfiles
1920		Description:
1921
1922		Created a new set of bootfiles (for your automatic installation convenience).
1923
1924		----------------------------------------------------------------------
1925		Name: Matthias Blume
1926		Date: 2001/03/16 11:00:00 EST
1927		Tag: blume-20010316-MLTREE-fixup
1928		Description:
1929
1930		This is a minor fixup for an (untagged) earlier commit by Allen.
1931		(A file was missing).
1932
1933		----------------------------------------------------------------------
1934		Name: Allen Leung
1935		Date: Mon Mar  5 18:54:57 EST 2001
1936		Tag: leunga-20010305-cut-support
1937
1938		1. New support for alternative control-flow in MLTREE.
1939		Currently we support
1940
1941		FLOW_TO(CALL ...., [k1,...,kn])
1942
1943		This is needed for 'cuts to' in C-- and try/handle-like constructs
1944		in Moby
1945
1946		New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
1947
1948		2. Register Allocator
1949
1950		Changes in interface [from Fermin, John]
1951
1952		3. Alpha 8-bit SLL support [Fermin]
1953
1954		4. All architectures
1955
1956		A new module (ClusterExpandCopies) for expanding parallel copies.
1957
1958		----------------------------------------------------------------------
1959		Name: Allen Leung
1960		Date: 2001/02/27 23:07:00 EST
1961		Tag: leunga-20010227-minor-stuff
1962
1963		1. Alpha bug fix for CMOVNE
1964		2. Handle mltree COND(..,FCMP ...,...)
1965		3. Bug fix in simplifier
1966
1967		----------------------------------------------------------------------
1968		Name: Matthias Blume
1969		Date: 2001/01/30 17:50:00 EST
1970		Tag: blume-20010130-sync
1971		Description:
1972
1973		This is just a minor update to sync my devel branch with the main brach.
1974		The only visible change is the addition of some README files.
1975
1976		----------------------------------------------------------------------
1977		Name: Matthias Blume
1978		Date: 2001/01/12 23:30:00 JST
1979		Tag: blume-20010112-bootfiles
1980		Description:
1981
1982		Made a new set of bootfiles that goes with the current state of the
1983		repository.
1984
1985		----------------------------------------------------------------------
1986		Name: Matthias Blume
1987		Date: 2001/01/12 21:20:00 JST
1988		Tag: blume-20010112-sync
1989		Description:
1990
1991		I am just flushing out some minor changes that had accumulated in
1992		my private branch in order to sync with the main tree.  (This is
1993		mainly because I had CVS trouble when trying to merge _into_ my
1994		private branch.)
1995
1996		Most people should be completely unaffected by this.
1997
1998		----------------------------------------------------------------------
1999		Name: Allen Leung
2000		Date: Thu Jan 11 21:03:00 EST 2001
2001		Tag: leunga-20010111-labexp=mltree
2002		Description:
2003
2004		1.  Removed the type LabelExp and replace it by MLTree.
2005		2.  Rewritten mltree-simplify with the pattern matcher tool.
2006		3.  There were some bugs in alpha code generator which would break
2007		64-bit code generation.
2008		4.  Redo the tools to generate code with the
2009		5.  The CM files in MLRISC (and in src/system/smlnj/MLRISC)
2010		are now generated by perl scripts.
2011
2012		----------------------------------------------------------------------
2013		Name: Matthias Blume
2014		Date: 2001/01/10 21:55:00 JST
2015		Tag: blume-20010110-rcc
2016		Description:
2017
2018		The RCC stuff now seems to work (but only on the x86).
2019		This required hacking of the c-calls interface (and -implementation) in
2020		MLRISC.
2021
2022		Normal compiler users should be unaffected.
2023
2024		----------------------------------------------------------------------
2025		Name: Matthias Blume
2026		Date: 2001/01/09 01:20:00 JST
2027		Tag: blume-20010109-rcc
2028		Description:
2029
2030		This is a fairly big patch, flushing out a large number of pending
2031		changes that I made to my development copy over the last couple of days.
2032
2033		Of practical relevance at this moment is a workaround for a pickling
2034		bug that Allen ran into the other day.  The cause of the bug itself is
2035		still unknown and it might be hard to fix it properly, but the
2036		workaround has some merits of its own (namely somewhat reducing pickling
2037		overhead for certain libraries).  Therefore, I think this solution should
2038		be satisfactory at this time.
2039
2040		The rest of the changes (i.e., the vast majority) has to do with my
2041		ongoing efforts of providing direct support for C function calls from
2042		ML.  At the moment there is a new primop "RAW_CCALL", typing magic
2043		in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new
2044		case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to
2045		translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to
2046		handle RCC.
2047
2048		The last part (the changes to mlriscGen.sml) are still known to be
2049		wrong on the x86 and not implemented on all other architectures.  But
2050		the infrastructure is in place. I had to change a few functor
2051		signatures in the backend to be able to route the CCalls interface
2052		from MLRISC there, and I had to specialize the mltree type (on the
2053		x86) to include the necessary extensions. (The extensions themselves
2054		were already there and redy to go in MLRISC/x86).
2055
2056		Everything should be very happy as soon as someone helps me with
2057		mlriscGen.sml...
2058
2059		In any case, nothing of this should matter to anyone as long as the
2060		new primop is not being used (which is going to be the case unless you
2061		find it where I hid it :). The rest of the compiler is completely
2062		unaffected.
2063
2064		----------------------------------------------------------------------
2065		Name: Matthias Blume
2066		Date: 2001/01/05 00:30:00 JST
2067		Tag: blume-20010105-primops
2068		Description:
2069
2070		Added some experimental support for work that I am doing right now.
2071		These changes mostly concern added primops, but there is also a new
2072		experimental C library in the runtime system (but currently not enabled
2073		anywhere except on Linux/X86).
2074
2075		In the course of adding primops (and playing with them), I discovered that
2076		Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly
2077		broken.  (Zhong was very right he labeled this stuff as "major gross hack".)
2078		To recover, I made type information in INL_PRIM mandatory and changed
2079		prim.sml as well as built-in.sml accordingly.  The InLine structure now
2080		has complete, correct type information (i.e., no bottom types).
2081
2082		Since all these changes mean that we need new binfiles, I also bumped the
2083		version number to 110.32.1.
2084
2085		----------------------------------------------------------------------
2086		Name: Matthias Blume
2087		Date: 2000/12/30 22:10:00 JST
2088		Tag: blume-20001230-various
2089		Description:
2090
2091		Added proxy libraries for MLRISC and let MLRISC libraries refer
2092		to each other using path anchors.  (See CM manual for explanation.)
2093
2094		Updated CM documentation.
2095
2096		Fixed some bugs in CM.
2097
2098		Implemented "proxy" libraries (= syntactic sugar for CM).
2099
2100		Added "-quiet" option to makeml and changed runtime system accordingly.
2101
2102		Added cleanup handler for exportML to reset timers and compiler stats.
2103
2104		----------------------------------------------------------------------
2105		Name: Lal George
2106		Date: 2000/12/22 22:22:58 EST 2000
2107		Tag: Release_110_32
2108		Description:
2109
2110		Infinite precision used throughout MLRISC.
2111		see MLRISC/mltree/machine-int.sig
2112
2113		----------------------------------------------------------------------
2114		Name: Matthias Blume
2115		Date: 2000/12/22 23:16:00 JST
2116		Tag: blume-20001222-warn
2117		Description:
2118
2119		Corrected wording and formatting of some CM warning message which I
2120		broke in my previous patch.
2121
2122		----------------------------------------------------------------------
2123		Name: Matthias Blume
2124		Date: 2000/12/22 21:20:00 JST
2125		Tag: blume-20001222-anchorenv
2126		Description:
2127
2128		Fixed CM's handling of anchor environments in connection with CMB.make.
2129
2130		----------------------------------------------------------------------
2131		Name: Matthias Blume
2132		Date: 2000/12/22 13:15:00 JST
2133		Tag: blume-20001222-cleanup
2134		Description:
2135
2136		Removed src/cm/ffi which does not (and did not) belong here.
2137
2138		----------------------------------------------------------------------
2139		Name: Matthias Blume
2140		Date: 2000/12/21 23:55:00 JST
2141		Tag: blume-20001221-exn
2142		Description:
2143
2144		Probably most important: CM no longer silently swallows all exceptions
2145		in the compiler.
2146		Plus: some other minor CM changes.  For example, CM now reports some
2147		sizes for generated binfiles (code, data, envpickle, lambdapickle).
2148
2149		----------------------------------------------------------------------
2150		Name: Matthias Blume
2151		Date: 2000/12/15 00:01:05 JST
2152		Tag: blume-20001215-dirtool
2153		Description:
2154
2155		- "dir" tool added.
2156		- improvements and cleanup to Tools structure
2157		- documentation updates
2158
2159		----------------------------------------------------------------------
2160		Name: Allen Leung
2161		Date: Thu Dec 14 03:45:24 EST 2000
2162		Description:
2163		Tag:  leunga-20001214-int-inf
2164		Description:
2165
2166		In IntInf, added these standard functions, which are missing from our
2167		implementation:
2168
2169		andb : int * int -> int
2170		xorb : int * int -> int
2171		orb  : int * int -> int
2172		notb : int -> int
2173		<<   : int * word -> int
2174		~>>  : int * word -> int
2175
2176		Not tested, I hope they are correct.
2177
2178		----------------------------------------------------------------------
2179		Name: Allen Leung
2180		Date: Fri Dec  8 19:23:26 EST 2000
2181		Description:
2182		Tag:  leunga-20001208-nowhere
2183		Description:
2184
2185		Slight improvements to the 'nowhere' tool to handle OR-patterns,
2186		to generate better error messages etc.  Plus a brief manual.
2187
2188		----------------------------------------------------------------------
2189		Name: Lal George
2190		Date: 2000/12/08 09:54:02 EST 2000
2191		Tag: Release_110_31
2192		Description:
2193
2194		- Version 110.31
2195		----------------------------------------------------------------------
2196		Name: Allen Leung
2197		Date: Thu Dec  7 22:01:04 EST 2000
2198		Tag:  leunga-20001207-cell-monster-hack
2199		Description:
2200
2201		Major MLRISC internal changes.  Affect all clients.
2202		Summary:
2203
2204		1.  Type CELLS.cell = int is now replaced by a datatype.
2205		As a result, the old regmap is now gone.  Almost all interfaces
2206		in MLRISC change as a consequence.
2207
2208		2.  A new brand version of machine description tool (v3.0) that generates
2209		modules expecting the new interface.  The old version is removed.
2210
2211		3.  The RA interface has been further abstracted into two new functors.
2212		RISC_RA and X86RA.  These functors have much simpler interfaces.
2213		[See also directory MLRISC/demo.]
2214
2215		4.  Some other new source->source code generation tools are available:
2216
2217		a. MLRISC/Tools/RewriteGen -- generate rewriters from rules.
2218		b. MLRISC/Tools/WhereGen -- expands conditional pattern matching rules.
2219		I use this tool to generate the peephole optimizers---with the new
2220		cell type changes, peephole rules are becoming difficult to write
2221		without conditional pattern matching.
2222
2223		5.  More Intmap -> IntHashTable change.  Previous changes by Matthias didn't
2224		cover the entire MLRISC source tree so many things broke.
2225
2226		6.  CM files have been moved to the subdirectory MLRISC/cm.
2227		They are moved because there are a lot of them and they clutter up the
2228		root dir.
2229
2230		7.  More detailed documentation to come...
2231
2232		NOTE: To rebuild from 110.30 (ftp distribution), you'll have to do
2233		a makeml -rebuild first.  This is because of other other
2234		changes that Matthias has made (see below).
2235
2236
2237		----------------------------------------------------------------------
2238		Name: Matthias Blume
2239		Date: 2000/11/30 23:12:00 JST
2240		Tag: blume-20001130-filereorg
2241		Description:
2242
2243		Some manual updates and some file reorganizations in CM.
2244
2245		----------------------------------------------------------------------
2246		Name: Matthias Blume
2247		Date: 2000/11/24 17:45:00 JST
2248		Tag: blume-20001124-link
2249		Description:
2250
2251		Drastically improved link traversal code for the case that the dynamic
2252		value was already loaded at bootstrap time.  As a result, CM and CMB
2253		now both load blazingly fast -- even on a very slow machine.  Also,
2254		memory consumption has been further reduced by this.
2255
2256		Warning: The format of the PIDMAP file has changed.  THerefore, to
2257		bootstrap you have to do this:
2258
2259		1. Run CMB.make
2260		2. Make a symbolic link for the boot directory:
2261		ln -s sml.boot.ARCH-OS xxx
2262		3. "Rebuild" the boot directory:
2263		./makeml -boot xxx -rebuild sml ; rm xxx
2264		4. Boot normally:
2265		./makeml
2266
2267		----------------------------------------------------------------------
2268		Name: Matthias Blume
2269		Date: 2000/11/21 21:20:00 JST
2270		Tag: blume-20001121-tools
2271		Description:
2272
2273		Continued hacking on autoloading problem -- with success this time.
2274		Also changed tool-plugin mechanism.  See new CM manual.
2275
2276		----------------------------------------------------------------------
2277		Name: Matthias Blume
2278		Date: 2000/11/19 14:30:00 JST
2279		Tag:  blume-20001119-autoload
2280		Description:
2281
2282		Some hacking to make autoloading faster.  Success for CMB, no success
2283		so far for CM.  There is a reduced structure CM' that autoloads faster.
2284		(This is a temporary, non-documented hack to be eliminated again when
2285		the general problem is solved.)
2286
2287		----------------------------------------------------------------------
2288		Name: Matthias Blume
2289		Date: 2000/11/17 14:10:00 JST
2290		Tag: blume-20001117-pickle-lib
2291		Description:
2292
2293		1. Eliminated comp-lib.cm
2294		2. Made pickle-lib.cm
2295		3. Eliminated all uses of intset.sml (from comp-lib.cm)
2296		4. Replaced all uses of intmap.{sig,sml} (from comp-lib.cm) with
2297		equivalent constructs from smlnj-lib.cm (INtHashTable).
2298		5. Point 4. also goes for those uses of intmap.* in MLRISC.
2299		Duplicated intmap modules thrown out.
2300		6. Hunted down all duplicated SCC code and replaced it with
2301		equivalent stuff (GraphSCCFn from smlnj-lib.cm).
2302		7. Rewrote Feedback module.
2303		8. Moved sortedlist.sml into viscomp-lib.cm.  Eventually it
2304		should be thrown out and equivalent modules from smlnj-lib.cm
2305		should be used (IntRedBlackSet, IntListSet, ...).
2306
2307		Confirmed that compiler compiles to fixpoint.
2308
2309		----------------------------------------------------------------------
2310		Name: Allen Leung
2311		Date: 2000/11/10 18:00:00
2312		Tag: leunga-20001110-new-x86-fp
2313
2314		A new x86 floating point code generator has been added.
2315		By default this is turned off.  To turn this on, do:
2316
2317		CM.autoload "$smlnj/compiler.cm";
2318		Compiler.Control.MLRISC.getFlag "x86-fast-fp" := true;
2319
2320		Changes:
2321
2322		1.  Changed FTAN to FPTAN so that the assembly output is correct.
2323		2.  Changed the extension callback for FTANGENT to generate:
2324
2325		fptan
2326		fstp  %st(0)
2327		instead of
2328		fptan
2329		fstpl ftempmem
2330
2331		3.  Numerous assembly fixes for x86.
2332
2333		5.  Cleaned up the machine code output module x86/x86MC.sml and added
2334		support for a whole bunch of instructions and addressing modes:
2335
2336		fadd/fsub/fsubr/fmul/fdiv/fdivr  %st, %st(n)
2337		faddp/fsubp/fsubrp/fmulp/fdivp/fdivrp  %st, %st(n)
2338		fadd/fsub/fsubr/fmul/fdiv/fdivr  %st(n), %st
2339		fiadd/fisub/fisubr/fimul/fidiv/fidivr mem
2340		fxch %st(n)
2341		fld %st(n)
2342		fst %st(n)
2343		fst mem
2344		fstp %st(n)
2345		fucom %st(n)
2346		fucomp %st(n)
2347
2348		All these are now generated when the fast fp mode is turned on.
2349
2350		6.  Removed the dedicated registers %st(0), ..., %st(7) from X86CpsRegs
2351
2352		----------------------------------------------------------------------
2353		Name: Matthias Blume
2354		Date: 2000/11/09 11:20:00 JST
2355		Tag: blume-20001109-scc
2356		Description:
2357
2358		Eliminated some code duplication:
2359
2360		1. Added "where" clause to GraphSCCFn in SML/NJ Library.
2361		(Otherwise the functor is useless.)
2362		2. Used GraphSCCFn where SCCUtilFun was used previously.
2363		3. Got rid of SCCUtilFun (in comp-lib.cm).
2364
2365		----------------------------------------------------------------------
2366		Name: Lal George
2367		Date: 2000/11/06 09:02:21 EST 2000
2368		Tag: Release_110_30
2369		Description:
2370
2371		- Version 110.30
2372		----------------------------------------------------------------------
2373		Name: Matthias Blume
2374		Date: 2000/11/04 14:45:00
2375		Tag: blume-20001104-mlbuild
2376		Description:
2377
2378		- Made ml-build faster on startup.
2379		- Documentation fixes.
2380
2381		----------------------------------------------------------------------
2382		Name: Matthias Blume
2383		Date: 2000/11/02 17:00:00 JST
2384		Tag: blume-20001102-condcomp
2385		Description:
2386
2387		- Small tweaks to pickler -- new BOOTFILES!
2388		- Version bumped to 110.29.2.
2389		- Added conditional compilation facility to init.cmi (see comment there).
2390		----------------------------------------------------------------------
2391		Name: Allen Leung
2392		Date: 2000/10/23 19:31:00
2393		Tag: leunga-20001023-demo-ra
2394
2395		1. Minor RA changes that improves spilling on x86 (affects Moby and C-- only)
2396		2. Test programs for the graph library updated
2397		3. Some new MLRISC demo programs added
2398
2399		----------------------------------------------------------------------
2400		Name: Matthias Blume
2401		Date: 2000/08/31 22:15:00 JST
2402		Tag: blume-20001017-errmsg
2403		Description:
2404
2405		More error message grief: Where there used to be no messages, there
2406		now were some that had bogus error regions.  Fixed.
2407
2408		----------------------------------------------------------------------
2409		Name: Matthias Blume
2410		Date: 2000/08/31 17:30:00 JST
2411		Tag: blume-20001017-v110p29p1
2412		Description:
2413
2414		I made a version 110.29.1 with new bootfiles.
2415
2416		Changes:  Modified pickler/unpickler for faster and leaner unpickling.
2417		CM documentation changes and a small bugfix in CM's error reporting.
2418
2419		----------------------------------------------------------------------
2420		Name: Lal George
2421		Date: 2000/09/27 14:42:35 EDT
2422		Tag: george-20000927-nodestatus
2423		Description:
2424
2425		Changed the type of the nodestatus, so that:
2426
2427		SPILLED(~1)             is now SPILLED
2428		SPILLED(m) where m>=0   is now MEMREG(m)
2429		SPILLED(s) where s<~1   is now SPILL_LOC(~s)
2430
2431		----------------------------------------------------------------------
2432		Name: Matthias Blume
2433		Date: 2000/09/07 14:45:00 JST
2434		Tag: blume-20000907-cmerrmsg
2435		Description:
2436
2437		Small tweak to CM to avoid getting ML syntax error messages twice.
2438
2439		----------------------------------------------------------------------
2440		Name: Matthias Blume
2441		Date: 2000/08/31 18:00:00 JST
2442		Tag: blume-20000831-cvsbootfiles
2443		Description:
2444
2445		New URL for boot files (because the 110.29 files on the BL server do
2446		now work correctly with my updated install scripts for yacc and lex).
2447
2448		----------------------------------------------------------------------
2449		Name: Matthias Blume
2450		Date: 2000/08/08 12:33:00 JST
2451		Tag: blume-20000808-manual
2452		Description:
2453
2454		Tiny update to CM manual.
2455
2456		----------------------------------------------------------------------
2457		Name: Allen Leung
2458		Date: 2000/08/7 19:31:00
2459		Tag: leunga-20000807-a-whole-bunch-of-stuff
2460
2461		Moby, C--, SSA, x86, machine descriptions etc.  Should only affect C--
2462		and Mobdy.
2463
2464		1.  x86
2465
2466		a.  Fixes to peephole module by John and Dan.
2467		b.  Assembly fix to SETcc by Allen.
2468		c.  Fix to c-call by John.
2469		d.  Fix to spilling by John.  (This one deals with the missing FSTPT case)
2470		e.  Instruction selection optimization to SETcc as suggested by John.
2471
2472		For example,
2473
2474		MV(32, x, COND(32, CMP(32, LT, a, b), LI 1, LI 0))
2475
2476		should generate:
2477
2478		MOVL a, x
2479		SUBL b, x
2480		SHRL 31, x
2481
2482		2.  IR stuff
2483
2484		A bunch of new DJ-graph related algorithms added.  These
2485		speed up SSA construction.
2486
2487		3.  SSA + Scheduling
2488
2489		Added code for SSA and scheduling to the repository
2490
2491		----------------------------------------------------------------------
2492		Name: Lal George
2493		Date: 2000/07/27 11:53:14 EDT
2494
2495		Tag: lal-20000727-linux-ppc
2496		Description:
2497
2498		Made changes to support Linux PPC.
2499		p.s. I have confirmation that the 110.29 boot files work fine.
2500
2501		----------------------------------------------------------------------
2502		Name: Matthias Blume
2503		Date: 2000/07/27 17:40:00 JST
2504		Tag: blume-20000727-scripts
2505		Description:
2506
2507		!!!! WARNING !!!!
2508		You must recompile the runtime system!
2509		!!!! WARNING !!!!
2510
2511		This is basically another round of script-enhancements:
2512
2513		1. sml, ml-build, and ml-makedepend accept options -D and -U to define
2514		and undefine CM preprocessor symbols.
2515
2516		2. ml-build avoids generating a new heap image if it finds that the
2517		existing one is still ok.  (The condition is that no ML file had to
2518		be recompiled and all ML files are found to be older that the heap
2519		file.)
2520
2521		To make this work smoothly, I also hacked the runtime system as
2522		well as SMLofNJ.SysInfo to get access to the heap image suffix
2523		(.sparc-solaris, ...) that is currently being used.
2524
2525		Moreover, the signature of CM.mk_standalone has changed.  See the
2526		CM manual.
2527
2528		3. ml-makedepend accepts additional options -n, -a, and -o.  (See the
2529		CM manual for details.)
2530
2531		4. More CM manual updates:
2532		- all of the above has been documented.
2533		- there is now a section describing the (CM-related) command line
2534		arguments that are accepted by the "sml" command
2535
2536		----------------------------------------------------------------------
2537		Name: Matthias Blume
2538		Date: 2000/07/25 16:20:00 JST
2539		Tag: blume-20000725-makedepend
2540		Description:
2541
2542		Added a script called ml-makedepend.  This can be used in makefiles
2543		for Unix' make in a way very similar to the "makedepend" command for
2544		C.
2545
2546		The script internally uses function CM.sources.
2547
2548		Synopsis:
2549
2550		ml-makedepend [-f makefile] cmfile targetname
2551
2552		The default for the makefile is "makefile" (or "Makefile" should
2553		"makefile" not exist).
2554
2555		ml-makedepend adds a cmfile/targetname-specific section to this
2556		makefile (after removing the previous version of this section).  The
2557		section contains a single dependency specification with targetname on
2558		the LHS (targetname is an arbitrary name), and a list of files derived
2559		from the cmfile on the RHS.  Some of the files on the RHS are
2560		ARCH/OPSYS-specific.  Therefore, ml-makedepend inserts references to
2561		"make" variables $(ARCH) and $(OPSYS) in place of the corresponding
2562		path names.  The makefile writer is responsible for making sure that
2563		these variables have correct at the time "make" is invoked.
2564
2565		----------------------------------------------------------------------
2566		Name: Matthias Blume
2567		Date: 2000/07/22 23:30:00 JST
2568		Tag: blume-20000722-urlupdate
2569		Description:
2570
2571		Changed BOOT and config/srcarchiveurl to point to BL server:
2572
2573		ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.29/
2574
2575		----------------------------------------------------------------------
2576		Name: Matthias Blume
2577		Date: 2000/07/18 18:00:00 JST
2578		Tag: blume-20000718-Version_110_29
2579		Description:
2580
2581		1. Updated src/compiler/TopLevel/main/version.sml to version 110.29
2582
2583		2. Updated config/version to 110.29
2584
2585		3. Updated config/srcarchiveurl
2586
2587		3. New boot files!
2588		ftp://ftp.cs.princeton.edu/pub/people/blume/sml/110.29-autofetch
2589
2590		----------------------------------------------------------------------
2591		Name: Matthias Blume
2592		Date: 2000/07/11 13:58:00 JST
2593		Tag: blume-20000711-doctypo
2594		Description:
2595
2596		Fixed a few typos in CM manual.
2597
2598		----------------------------------------------------------------------
2599		Name: Allen Leung
2600		Date: 2000/06/15 00:38:00
2601		Tag: leunga-20000704-sparc-x86
2602
2603		1. x86 peephole improvement sp += k; sp -= k => nop  [from John]
2604		2. fix to x86 RET bug [found by Dan Grossman]
2605		3. sparc assembly bug fix for ticc instructions [found by Fermin]
2606
2607		Affects c-- and moby only
2608
2609		----------------------------------------------------------------------
2610		Name: Matthias Blume
2611		Date: 2000/07/04 15:26:00
2612		Tag: blume-20000704-trigger
2613		Description:
2614
2615		1. Improvements to CM manual.
2616		2. SMLofNJ.Internals.BTrace.trigger reinstated as an alternative way
2617		of getting a back-trace.  The function, when called, raises an
2618		internal exception which explicitly carries the full back-trace history,
2619		so it is unaffected by any intervening handle-raise pairs ("trivial"
2620		or not).  The interactive loop will print that history once it arrives
2621		at top level.
2622		Short of having all exceptions implicitly carry the full history, the
2623		recommended way of using this facility is:
2624		- compile your program with instrumentation "on"
2625		- run it, when it raises an exception, look at the history
2626		- if the history is "cut off" because of some handler, go and modify
2627		your program so that it explicitly calls BTrace.trigger
2628		- recompile (still instrumented), and rerun; look at the full history
2629
2630		----------------------------------------------------------------------
2631		Name: Matthias Blume
2632		Date: 2000/07/03 15:36:00 JST
2633		Tag: blume-20000702-manual
2634		Description:
2635
2636		Small corrections and updates to CM manual.
2637
2638		----------------------------------------------------------------------
2639		Name: Matthias Blume
2640		Date: 2000/06/29 16:04:00 JST
2641		Tag: blume-20000629-yacctool
2642		Description:
2643
2644		Changes:
2645
2646		1. Class "mlyacc" now takes separate arguments to pass options to
2647		generated .sml- and .sig-files independently.
2648		2. Corresponding CM manual updates.
2649		3. BTrace module now also reports call sites.  (However, for loop clusters
2650		it only shows from where the cluster was entered.)  There are associated
2651		modifications to core.sml, internals.{sig,sml}, btrace.sml, and btimp.sml.
2652
2653		----------------------------------------------------------------------
2654		Name: Matthias Blume
2655		Date: 2000/06/27 16:51:00 JST
2656		Tag: blume-20000627-noweb
2657		Description:
2658
2659		Changes:
2660
2661		1. Implemented "subdir" and "witness" options for noweb tool.
2662		This caused some slight internal changes in CM's tool implementation.
2663		2. Fixed bug in "tool plugin" mechanism.  This is essentially cleaning
2664		some remaining issues from earlier path anchor changes.
2665		3. Updated CM manual accordingly.
2666
2667		4. Changed implementation of back-tracing so that I now consider it
2668		ready for prime-time.
2669
2670		In particular, you don't have to explicitly trigger the back-trace
2671		anymore.  Instead, if you are running BTrace-instrumented code and
2672		there is an uncaught exception (regardless of whether or not it was
2673		raised in instrumented code), the top-level evalloop will print
2674		the back-trace.
2675
2676		Features:
2677
2678		- Instrumented and uninstrumented code work together seemlessly.
2679		(Of course, uninstrumented code is never mentioned in actual
2680		back-traces.)
2681
2682		- Asymptotic time- and space-complexity of instrumented code is
2683		equal to that of uninstrumented code.  (This means that
2684		tail-recursion is preserved by the instrumentation phase.)
2685
2686		- Modules whose code has been instrumented in different sessions
2687		work together without problem.
2688
2689		- There is no penalty whatsoever on uninstrumented code.
2690
2691		- There is no penalty on "raise" expressions, even in
2692		instrumented code.
2693
2694		A potential bug (or perhaps it is a feature, too):
2695
2696		A back-trace reaches no further than the outermost instrumented
2697		non-trivial "raise".  Here, a "trivial" raise is one that is the
2698		sole RHS of a "handle" rule.  Thus, back-traces reach trough
2699
2700		<exp> handle e => raise e
2701
2702		and even
2703
2704		<exp> handle Foo => raise Bar
2705
2706		and, of course, through
2707
2708		<exp> handle Foo => ...
2709
2710		if the exception was not Foo.
2711
2712		Back-traces always reach right through any un-instrumented code
2713		including any of its "handle" expressions, trivial or not.
2714
2715		To try this out, do the following:
2716
2717		- Erase all existing binfiles for your program.
2718		(You may keep binfiles for those modules where you think you
2719		definitely don't need back-tracing.)
2720		- Turn on back-trace instrumentation:
2721		SMLofNJ.Internals.BTrace.mode (SOME true);
2722		- Recompile your program.  (I.e., run "CM.make" or "use".)
2723		- You may now turn instrumentation off again (if you want):
2724		SMLofNJ.Internals.BTrace.mode (SOME false);
2725		- Run your program as usual.  If it raises an exception that
2726		reaches the interactive toplevel, then a back-trace will
2727		automatically be printed.  After that, the toplevel loop
2728		will print the exception history as usual.
2729
2730		----------------------------------------------------------------------
2731		Name: Matthias Blume
2732		Date: 2000/06/26 09:56:46 JST
2733		Tag: blume-20000626-setup
2734		Description:
2735
2736		CM: - setup-parameter to "sml" added; this can be used to run arbitrary
2737		ML code before and after compiling a file (e.g., to set compiler
2738		flags)
2739
2740		Compiler: - improved btrace API (in core.sml, internals.{sig,sml})
2741		- associated changes to btrace.sml (BTrace instrumentation pass)
2742		- cleaner implementation of btimp.sml (BTrace tracing and report
2743		module)
2744
2745		CM manual: * new path encoding documented
2746		* description of setup-parameter to "sml" added
2747
2748		The biggest user-visible change to back-tracing is that it is no
2749		longer necessary to compile all traced modules within the same
2750		session.  (This was a real limitation.)
2751
2752		----------------------------------------------------------------------
2753		Name: Matthias Blume
2754		Date: 2000/06/24 12:40:00 JST
2755		Tag: blume-20000624-startup
2756		Description:
2757
2758		Fixes startup slowdown problem.  (I was calling SrcPath.sync a _tad_
2759		bit too often -- to put it mildly. :)
2760
2761		----------------------------------------------------------------------
2762		Name: Matthias Blume
2763		Date: 2000/06/23 18:20:00 JST
2764		Tag: blume-20000623-btrace
2765		Description:
2766
2767		This updates adds a backtrace facility to aid programmers in debugging
2768		their programs.  This involves the following changes:
2769
2770		1. Module system/smlnj/init/core.sml (structure _Core) now has hooks for
2771		keeping track of the current call stack.  When programs are compiled
2772		in a special mode, the compiler will insert calls to these hooks
2773		into the user program.
2774		"Hook" means that it is possible for different implementations of
2775		back-tracing to register themselves (at different times).
2776
2777		2. compiler/MiscUtil/profile/btrace.sml implements the annotation phase
2778		as an Absyn.dec->Absyn.dec rewrite.  Normally this phase is turned off.
2779		It can be turned on using this call:
2780		SMLofNJ.Internals.BTrace.mode (SOME true);
2781		Turning it off again:
2782		SMLofNJ.Internals.BTrace.mode (SOME false);
2783		Querying the current status:
2784		SMLofNJ.Internals.BTrace.mode NONE;
2785		Annotated programs are about twice as big as normal ones, and they
2786		run a factor of 2 to 4 slower with a dummy back-trace plugin (one
2787		where all hooks do nothing).  The slowdown with a plugin that is
2788		actually useful (such as the one supplied by default) is even greater,
2789		but in the case of the default plugin it is still only an constant
2790		factor (amortized).
2791
2792		3. system/Basis/Implementation/NJ/internals.{sig,sml} have been augmented
2793		with a sub-structure BTrace for controlling back-tracing.  In particular,
2794		the above-mentioned function "mode" controls whether the annotation
2795		phase is invoked by the compiler.  Another important function is
2796		"trigger": when called it aborts the current execution and causes
2797		the top-level loop to print a full back-trace.
2798
2799		4. compiler/MiscUtil/profile/btimp.sml is the current default plugin
2800		for back-tracing.  It keeps track of the dynamic call stack and in
2801		addition to that it keeps a partial history at each "level" of that
2802		stack.  For example, if a tail-calls b, b tail-calls c, and c tail-calls
2803		d and b (at separate times, dynamically), then the report will show:
2804
2805		GOTO   d
2806		/c
2807		GOTO  \b
2808		CALL   a
2809
2810		This shows that there was an initial non-tail call of a, then a
2811		tail-call to b or c, looping behavior in a cluster of functions that
2812		consist of b and c, and then a goto from that cluster (i.e., either from
2813		b or from c) to d.
2814
2815		Note that (depending on the user program) the amount of information
2816		that the back-trace module has to keep track of at each level is bounded
2817		by a constant.  Thus, the whole implementation has the same asymptotical
2818		complexity as the original program (both in space and in time).
2819
2820		5. compiler/TopLevel/interact/evalloop.sml has been modified to
2821		handle the special exception SMLofNJ.Internals.BTrace.BTrace
2822		which is raised by the "trigger" function mentioned above.
2823
2824		Notes on usage:
2825
2826		- Annotated code works well together with unannotated code:
2827		Unannotated calls simply do not show up at all in the backtrace.
2828
2829		- It is not a good idea to let modules that were annotated during
2830		different sessions run at the same time.  This is because the compiler
2831		chooses small integers to identify individual functions, and there
2832		will be clashes if different modules were compiled in separate sessions.
2833		(Nothing will crash, and you will even be told about the clashes, but
2834		back-trace information will in general not be useful.)
2835
2836		- Back-tracing can be confused by callcc and capture.
2837
2838		- The only way of getting a back-trace right now is to explicitly
2839		invoke the "trigger" function from your user program.  Eventually, we
2840		should make every exception carry back-trace information (if
2841		available).  But since this creates more overhead at "raise"-time
2842		(similar to the current exnHistory overhead), I have not yet
2843		implemented this.  (The implementation will be rather easy.)  With
2844		exceptions carrying back-trace information, this facility will be even
2845		more useful because users don't need to modify their programs...
2846
2847		- While it is possible to compile the compiler with back-trace
2848		annotations turned on (I did it to get some confidence in
2849		correctness), you must make absolutely sure that core.sml and
2850		btimp.sml are compiled WITHOUT annotation!  (core.sml cannot actually
2851		be compiled with annotation because there is no core access yet, but
2852		if you compile btimp.sml with annotation, then the system will go into
2853		an infinite recursion and crash.)
2854		Since CM currently does not know about BTrace, the only way to turn
2855		annotations on and off for different modules of the compiler is to
2856		interrupt CMB.make, change the settings, and re-invoke it.  Of course,
2857		this is awkward and clumsy.
2858
2859		Sample sessions:
2860
2861		Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
2862		- SMLofNJ.Internals.BTrace.mode (SOME true);
2863		[autoloading]
2864		[autoloading done]
2865		val it = false : bool
2866		- structure X = struct
2867		-     fun main n = let
2868		-         fun a (x, 0) = d x
2869		-           | a (x, n) = b (x, n - 1)
2870		-         and b (x, n) = c (x, n)
2871		-         and c (x, n) = a (x, n)
2872		-         and d x = e (x, 3)
2873		-         and e (x, 0) = f x
2874		-           | e (x, n) = e (x, n - 1)
2875		-         and f 0 = SMLofNJ.Internals.BTrace.trigger ()
2876		-           | f n = n * g (n - 1)
2877		-         and g n = a (n, 3)
2878		-     in
2879		-         f n
2880		-     end
2881		- end;
2882		structure X : sig val main : int -> int end
2883		- X.main 3;
2884		*** BACK-TRACE ***
2885		GOTO   stdIn:4.2-13.20: X.main[2].f
2886		GOTO-( stdIn:4.2-13.20: X.main[2].e
2887		GOTO   stdIn:4.2-13.20: X.main[2].d
2888		/ stdIn:4.2-13.20: X.main[2].a
2889		| stdIn:4.2-13.20: X.main[2].b
2890		GOTO-\ stdIn:4.2-13.20: X.main[2].c
2891		CALL   stdIn:4.2-13.20: X.main[2].g
2892		GOTO   stdIn:4.2-13.20: X.main[2].f
2893		GOTO-( stdIn:4.2-13.20: X.main[2].e
2894		GOTO   stdIn:4.2-13.20: X.main[2].d
2895		/ stdIn:4.2-13.20: X.main[2].a
2896		| stdIn:4.2-13.20: X.main[2].b
2897		GOTO-\ stdIn:4.2-13.20: X.main[2].c
2898		CALL   stdIn:4.2-13.20: X.main[2].g
2899		GOTO   stdIn:4.2-13.20: X.main[2].f
2900		GOTO-( stdIn:4.2-13.20: X.main[2].e
2901		GOTO   stdIn:4.2-13.20: X.main[2].d
2902		/ stdIn:4.2-13.20: X.main[2].a
2903		| stdIn:4.2-13.20: X.main[2].b
2904		GOTO-\ stdIn:4.2-13.20: X.main[2].c
2905		CALL   stdIn:4.2-13.20: X.main[2].g
2906		GOTO   stdIn:4.2-13.20: X.main[2].f
2907		CALL   stdIn:2.15-17.4: X.main[2]
2908		-
2909
2910		(Note that because of a FLINt bug the above code currently does not
2911		compile without BTrace turned on.)
2912
2913		Here is another example, using my modified Tiger compiler:
2914
2915		Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
2916		- SMLofNJ.Internals.BTrace.mode (SOME true);
2917		[autoloading]
2918		[autoloading done]
2919		val it = false : bool
2920		- CM.make "sources.cm";
2921		[autoloading]
2922		...
2923		[autoloading done]
2924		[scanning sources.cm]
2925		[parsing (sources.cm):parse.sml]
2926		[creating directory CM/SKEL ...]
2927		[parsing (sources.cm):tiger.lex.sml]
2928		...
2929		[wrote CM/sparc-unix/semant.sml]
2930		[compiling (sources.cm):main.sml]
2931		[wrote CM/sparc-unix/main.sml]
2932		[New bindings added.]
2933		val it = true : bool
2934		- Main.compile ("../testcases/merge.tig", "foo.out");
2935		*** BACK-TRACE ***
2936		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trvar
2937		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
2938		CALL   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
2939		GOTO   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
2940		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
2941		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
2942		CALL   lib/semant.sml:488.3-505.6: SemantFun[2].transDec.trdec[2].transBody[2]
2943		/ lib/semant.sml:411.65-543.8: SemantFun[2].transDec
2944		CALL-\ lib/semant.sml:413.2-540.9: SemantFun[2].transDec.trdec[2]
2945		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
2946		CALL   lib/semant.sml:8.52-558.4: SemantFun[2].transProg[2]
2947		CALL   main.sml:1.18-118.4: Main.compile[2]
2948		-
2949
2950		----------------------------------------------------------------------
2951		Name: Matthias Blumen
2952		Date: 2000/06/21 18:00:00 JST
2953		Tag: blume-20000621-manual
2954		Description:
2955
2956		CM manual update: Path environments documented.
2957
2958		----------------------------------------------------------------------
2959		Name: Matthias Blume
2960		Date: 2000/06/19 13:40:00
2961		Tag: blume-20000619-manual
2962		Description:
2963
2964		CM manual and system/README update.  This only covers the fact that
2965		there are no more implicit anchors.  (Path environments and the "bind"
2966		option to "cm" have yet to be documented.)
2967
2968		----------------------------------------------------------------------
2969		Name: Matthias Blume
2970		Date: 2000/06/19 11:05:00 JST
2971		Tag: blume-20000619-chdir-bugfix
2972		Description:
2973
2974		Fixed a bug in new SrcPath module that sometimes led to a bad chDir call.
2975
2976	----------------------------------------------------------------------	----------------------------------------------------------------------
2977	Name: Matthias Blume	Name: Matthias Blume
2978	Date: 2000/06/18 22:00:10 JST	Date: 2000/06/18 22:00:10 JST
#	Line 942	Line 3904
3904	elaborator). There were a lot of changes during my "linkpath" trials	elaborator). There were a lot of changes during my "linkpath" trials
3905	that could have been reverted to their original state but weren't.	that could have been reverted to their original state but weren't.
3906	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
3907	than what was strictly necessary...  (I _did_ resist the tempation	than what was strictly necessary...  (I _did_ resist the temptation
3908	of doing any "global reformatting" to avoid an untimely death at	of doing any "global reformatting" to avoid an untimely death at
3909	Dave's hands. :)	Dave's hands. :)
3910

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.670   changed lines   Added in v.1078

---

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