Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 656, Fri Jun 9 03:39:04 2000 UTC	revision 1115, Sun Mar 3 18:16:31 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: Allen Leung
17		Date: 2002/03/03 13:20:00 EST
18		Tag: leunga-20020303-mlrisc-tools
19
20		Added #[ ... ] expressions to mlrisc tools
21
22		----------------------------------------------------------------------
23		Name: Matthias Blume
24		Date: 2002/02/27 12:29:00 EST
25		Tag: blume-20020227-cdebug
26		Description:
27
28		- made types in structure C and C_Debug to be equal
29		- got rid of code duplication (c-int.sml vs. c-int-debug.sml)
30		- there no longer is a C_Int_Debug (C_Debug is directly derived from C)
31
32		----------------------------------------------------------------------
33		Name: Matthias Blume
34		Date: 2002/02/26 12:00:00 EST
35		Tag: blume-20020226-ffi
36		Description:
37
38		1. Fixed a minor bug in CM's "noweb" tool:
39		If numbering is turned off, then truly don't number (i.e., do not
40		supply the -L option to noweb).  The previous behavior was to supply
41		-L'' -- which caused noweb to use the "default" line numbering scheme.
42		Thanks to Chris Richards for pointing this out (and supplying the fix).
43
44		2. Once again, I reworked some aspects of the FFI:
45
46		A. The incomplete/complete type business:
47
48		- Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
49		gone!
50		- ML types representing an incomplete type are now *equal* to
51		ML types representing their corresponding complete types (just like
52		in C).  This is still safe because ml-nlffigen will not generate
53		RTTI for incomplete types, nor will it generate functions that
54		require access to such RTTI.   But when ML code generated from both
55		incomplete and complete versions of the C type meet, the ML types
56		are trivially interoperable.
57
58		NOTE:  These changes restore the full generality of the translation
59		(which was previously lost when I eliminated functorization)!
60
61		B. Enum types:
62
63		- Structure C now has a type constructor "enum" that is similar to
64		how the "su" constructor works.  However, "enum" is not a phantom
65		type because each "T enum" has values (and is isomorphic to
66		MLRep.Signed.int).
67		- There are generic access operations for enum objects (using
68		MLRep.Signed.int).
69		- ml-nlffigen will generate a structure E_foo for each "enum foo".
70		* The structure contains the definition of type "mlrep" (the ML-side
71		representation type of the enum).  Normally, mlrep is the same
72		as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
73		then mlrep will be defined as a datatype -- thus facilitating
74		pattern matching on mlrep values.
75		("-ec" will be suppressed if there are duplicate values in an
76		enumeration.)
77		* Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
78		will be generated for each C enum constant xxx.
79		* Conversion functions m2i and i2m convert between mlrep and
80		MLRep.Signed.int.  (Without "-ec", these functions are identities.)
81		* Coversion functions c and ml convert between mlrep and "tag enum".
82		* Access functions (get/set) fetch and store mlrep values.
83		- By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
84		enumerations are merged into one single enumeration represented by
85		structure E_'.
86
87		----------------------------------------------------------------------
88		Name: Allen Leung
89		Date: 2002/02/25 04:45:00 EST
90		Tag: leunga-20020225-cps-spill
91
92		This is a new implementation of the CPS spill phase.
93		The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
94		In case of problems, replace it with the old file spill.sml
95
96		The current compiler runs into some serious performance problems when
97		constructing a large record.  This can happen when we try to compile a
98		structure with many items.  Even a very simple structure like the following
99		makes the compiler slow down.
100
101		structure Foo = struct
102		val x_1 = 0w1 : Word32.int
103		val x_2 = 0w2 : Word32.int
104		val x_3 = 0w3 : Word32.int
105		...
106		val x_N = 0wN : Word32.int
107		end
108
109		The following table shows the compile time, from N=1000 to N=4000,
110		with the old compiler:
111
112		N
113		1000   CPS 100 spill                           0.04u  0.00s  0.00g
114		MLRISC ra                               0.06u  0.00s  0.05g
115		(spills = 0 reloads = 0)
116		TOTAL                                   0.63u  0.07s  0.21g
117
118		1100   CPS 100 spill                           8.25u  0.32s  0.64g
119		MLRISC ra                               5.68u  0.59s  3.93g
120		(spills = 0 reloads = 0)
121		TOTAL                                   14.71u  0.99s  4.81g
122
123		1500   CPS 100 spill                           58.55u  2.34s  1.74g
124		MLRISC ra                               5.54u  0.65s  3.91g
125		(spills = 543 reloads = 1082)
126		TOTAL                                   65.40u  3.13s  6.00g
127
128		2000   CPS 100 spill                           126.69u  4.84s  3.08g
129		MLRISC ra                               0.80u  0.10s  0.55g
130		(spills = 42 reloads = 84)
131		TOTAL                                   129.42u  5.10s  4.13g
132
133		3000   CPS 100 spill                           675.59u  19.03s  11.64g
134		MLRISC ra                               2.69u  0.27s  1.38g
135		(spills = 62 reloads = 124)
136		TOTAL                                   682.48u  19.61s  13.99g
137
138		4000   CPS 100 spill                           2362.82u  56.28s  43.60g
139		MLRISC ra                               4.96u  0.27s  2.72g
140		(spills = 85 reloads = 170)
141		TOTAL                                   2375.26u  57.21s  48.00g
142
143		As you can see the old cps spill module suffers from some serious
144		performance problem.  But since I cannot decipher the old code fully,
145		instead of patching the problems up, I'm reimplementing it
146		with a different algorithm.  The new code is more modular,
147		smaller when compiled, and substantially faster
148		(O(n log n) time and O(n) space).  Timing of the new spill module:
149
150		4000  CPS 100 spill                           0.02u  0.00s  0.00g
151		MLRISC ra                               0.25u  0.02s  0.15g
152		(spills=1 reloads=3)
153		TOTAL                                   7.74u  0.34s  1.62g
154
155		Implementation details:
156
157		As far as I can tell, the purpose of the CPS spill module is to make sure the
158		number of live variables at any program point (the bandwidth)
159		does not exceed a certain limit, which is determined by the
160		size of the spill area.
161
162		When the bandwidth is too large, we decrease the register pressure by
163		packing live variables into spill records.  How we achieve this is
164		completely different than what we did in the old code.
165
166		First, there is something about the MLRiscGen code generator
167		that we should be aware of:
168
169		o MLRiscGen performs code motion!
170
171		In particular, it will move floating point computations and
172		address computations involving only the heap pointer to
173		their use sites (if there is only a single use).
174		What this means is that if we have a CPS record construction
175		statement
176
177		RECORD(k,vl,w,e)
178
179		we should never count the new record address w as live if w
180		has only one use (which is often the case).
181
182		We should do something similar to floating point, but the transformation
183		there is much more complex, so I won't deal with that.
184
185		Secondly, there are now two new cps primops at our disposal:
186
187		1. rawrecord of record_kind option
188		This pure operator allocates some uninitialized storage from the heap.
189		There are two forms:
190
191		rawrecord NONE [INT n]  allocates a tagless record of length n
192		rawrecord (SOME rk) [INT n] allocates a tagged record of length n
193		and initializes the tag.
194
195		2. rawupdate of cty
196		rawupdate cty (v,i,x)
197		Assigns to x to the ith component of record v.
198		The storelist is not updated.
199
200		We use these new primops for both spilling and increment record construction.
201
202		1. Spilling.
203
204		This is implemented with a linear scan algorithm (but generalized
205		to trees).  The algorithm will create a single spill record at the
206		beginning of the cps function and use rawupdate to spill to it,
207		and SELECT or SELp to reload from it.  So both spills and reloads
208		are fine-grain operations.  In contrast, in the old algorithm
209		"spills" have to be bundled together in records.
210
211		Ideally, we should sink the spill record construction to where
212		it is needed.  We can even split the spill record into multiple ones
213		at the places where they are needed.  But CPS is not a good
214		representation for global code motion, so I'll keep it simple and
215		am not attempting this.
216
217		2. Incremental record construction (aka record splitting).
218
219		Long records with many component values which are simulatenously live
220		(recall that single use record addresses are not considered to
221		be live) are constructed with rawrecord and rawupdate.
222		We allocate space on the heap with rawrecord first, then gradually
223		fill it in with rawupdate.  This is the technique suggested to me
224		by Matthias.
225
226		Some restrictions on when this is applicable:
227		1. It is not a VECTOR record.  The code generator currently does not handle
228		this case. VECTOR record uses double indirection like arrays.
229		2. All the record component values are defined in the same "basic block"
230		as the record constructor.  This is to prevent speculative
231		record construction.
232
233		----------------------------------------------------------------------
234		Name: Allen Leung
235		Date: 2002/02/22 01:02:00 EST
236		Tag: leunga-20020222-mlrisc-tools
237
238		Minor bug fixes in the parser and rewriter
239
240		----------------------------------------------------------------------
241		Name: Allen Leung
242		Date: 2002/02/21 20:20:00 EST
243		Tag: leunga-20020221-peephole
244
245		Regenerated the peephole files.  Some contained typos in the specification
246		and some didn't compile because of pretty printing bugs in the old version
247		of 'nowhere'.
248
249		----------------------------------------------------------------------
250		Name: Allen Leung
251		Date: 2002/02/19 20:20:00 EST
252		Tag: leunga-20020219-mlrisc-tools
253		Description:
254
255		Minor bug fixes to the mlrisc-tools library:
256
257		1.  Fixed up parsing colon suffixed keywords
258		2.  Added the ability to shut the error messages up
259		3.  Reimplemented the pretty printer and fixed up/improved
260		the pretty printing of handle and -> types.
261		4.  Fixed up generation of literal symbols in the nowhere tool.
262		5.  Added some SML keywords to to sml.sty
263
264		----------------------------------------------------------------------
265		Name: Matthias Blume
266		Date: 2002/02/19 16:20:00 EST
267		Tag: blume-20020219-cmffi
268		Description:
269
270		A wild mix of changes, some minor, some major:
271
272		* All C FFI-related libraries are now anchored under $c:
273		$/c.cm      --> $c/c.cm
274		$/c-int.cm  --> $c/internals/c-int.cm
275		$/memory.cm --> $c/memory/memory.cm
276
277		* "make" tool (in CM) now treats its argument pathname slightly
278		differently:
279		1. If the native expansion is an absolute name, then before invoking
280		the "make" command on it, CM will apply OS.Path.mkRelative
281		(with relativeTo = OS.FileSys.getDir()) to it.
282		2. The argument will be passed through to subsequent phases of CM
283		processing without "going native".  In particular, if the argument
284		was an anchored path, then "make" will not lose track of that anchor.
285
286		* Compiler backends now "know" their respective C calling conventions
287		instead of having to be told about it by ml-nlffigen.  This relieves
288		ml-nlffigen from one of its burdens.
289
290		* The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
291
292		* Export C_DEBUG and C_Debug from $c/c.cm.
293
294		* C type encoding in ml-nlffi-lib has been improved to model the conceptual
295		subtyping relationship between incomplete pointers and their complete
296		counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
297		with the convention of instantiating 'o with ('t, 'c) obj whenever
298		the pointer target type is complete.  In the incomplete case, 'o
299		will be instantiated with some "'c iobj" -- a type obtained by
300		using one of the functors PointerToIncompleteType or PointerToCompleteType.
301
302		Operations that work on both incomplete and complete pointer types are
303		typed as taking an 'o ptr while operations that require the target to
304		be known are typed as taking some ('t, 'c) obj ptr.
305
306		voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
307		where void is an eqtype without any values.  This makes it possible
308		to work on voidptr values using functions meant to operate on light
309		incomplete pointers.
310
311		* As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
312		been vastly simplified.
313
314		----------------------------------------------------------------------
315		Name: Matthias Blume
316		Date: 2002/02/19 10:48:00 EST
317		Tag: blume-20020219-pqfix
318		Description:
319
320		Applied Chris Okasaki's bug fix for priority queues.
321
322		----------------------------------------------------------------------
323		Name: Matthias Blume
324		Date: 2002/02/15 17:05:00
325		Tag: Release_110_39
326		Description:
327
328		Last-minute retagging is becoming a tradition... :-(
329
330		This is the working release 110.39.
331
332		----------------------------------------------------------------------
333		Name: Matthias Blume
334		Date: 2002/02/15 16:00:00 EST
335		Tag: Release_110_39-orig
336		Description:
337
338		Working release 110.39.  New bootfiles.
339
340		(Update: There was a small bug in the installer so it wouldn't work
341		with all shells.  So I retagged. -Matthias)
342
343		----------------------------------------------------------------------
344		Name: Matthias Blume
345		Date: 2002/02/15 14:17:00 EST
346		Tag: blume-20020215-showbindings
347		Description:
348
349		Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
350		the latter can be useful for exploring what bindings are available at
351		the interactive prompt.  (The first function returns only the list
352		of symbols that are really bound, the second prints those but also the
353		ones that CM's autoloading mechanism knows about.)
354
355		----------------------------------------------------------------------
356		Name: Matthias Blume
357		Date: 2002/02/15 12:08:00 EST
358		Tag: blume-20020215-iptrs
359		Description:
360
361		Two improvements to ml-nlffigen:
362
363		1. Write files only if they do not exist or if their current contents
364		do not coincide with what's being written.  (That is, avoid messing
365		with the time stamps unless absolutely necessary.)
366
367		2. Implement a "repository" mechanism for generated files related
368		to "incomplete pointer types".   See the README file for details.
369
370		----------------------------------------------------------------------
371		Name: Matthias Blume
372		Date: 2002/02/14 11:50:00 EST
373		Tag: blume-20020214-quote
374		Description:
375
376		Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
377		because of the new and improved tag generation scheme.  (Thanks to Allen
378		Leung for pointing it out.)
379
380		----------------------------------------------------------------------
381		Name: Lal George
382		Date: 2002/02/14 09:55:27 EST 2002
383		Tag: george-20020214-isabelle-bug
384		Description:
385
386		Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
387		of Isabelle on the x86.
388
389		From Allen:
390		-----------
391		I've found the problem:
392
393		in ra-core.sml, I use the counter "blocked" to keep track of the
394		true number of elements in the freeze queue.  When the counter goes
395		to zero, I skip examining the queue.  But I've messed up the
396		bookkeeping in combine():
397
398		else ();
399		case !ucol of
400		PSEUDO => (if !cntv > 0 then
401		(if !cntu > 0 then blocked := !blocked - 1 else ();
402		^^^^^^^^^^^^^^^^^^^^^^^
403		moveu := mergeMoveList(!movev, !moveu)
404		)
405		else ();
406
407		combine() is called to coalesce two nodes u and v.
408		I think I was thinking that if the move counts of u and v are both
409		greater than zero then after they are coalesced then one node is
410		removed from the freeze queue.  Apparently I was thinking that
411		both u and v are of low degree, but that's clearly not necessarily true.
412
413
414		02/12/2002:
415		Here's the patch.  HOL now compiles.
416
417		I don't know how this impact on performance (compile
418		time or runtime).  This bug caused the RA (especially on the x86)
419		to go thru the potential spill phase when there are still nodes on the
420		freeze queue.
421
422
423
424
425		----------------------------------------------------------------------
426		Name: Matthias Blume
427		Date: 2002/02/13 22:40:00 EST
428		Tag: blume-20020213-fptr-rtti
429		Description:
430
431		Fixed a bug in ml-nlffigen that was introduced with one of the previous
432		updates.
433
434		----------------------------------------------------------------------
435		Name: Matthias Blume
436		Date: 2002/02/13 16:41:00 EST
437		Tag: blume-20020213-cmlpq
438		Description:
439
440		Added new priority queue export symbols (which have just been added to
441		smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
442		not compile and the installer would choke.)
443
444		----------------------------------------------------------------------
445		Name: Matthias Blume
446		Date: 2002/02/13 16:15:00 EST
447		Tag: blume-20020213-various
448		Description:
449
450		1. More tweaks to ml-nlffigen:
451
452		- better internal datastructures (resulting in slight speedup)
453		- "-match" option requires exact match
454		- "localized" gensym counters (untagged structs/unions nested within
455		other structs/unions or within typedefs get a fresh counter; their
456		tag will be prefixed by a concatenation of their parents' tags)
457		- bug fixes (related to calculation of transitive closure of types
458		to be included in the output)
459
460		2. Minor Basis updates:
461
462		- added implementations for List.collate and Option.app
463
464		----------------------------------------------------------------------
465		Name: Matthias Blume
466		Date: 2002/02/11 15:55:00 EST
467		Tag: blume-20020211-gensym
468		Description:
469
470		Added a "-gensym" option to command line of ml-nlffigen.  This can be
471		used to specify a "stem" -- a string that is inserted in all "gensym'd"
472		names (ML structure names that correspond to unnamed C structs, unions,
473		and enums), so that separate runs of ml-nlffigen do not clash.
474
475		----------------------------------------------------------------------
476		Name: Matthias Blume
477		Date: 2002/02/11 12:05:00 EST
478		Tag: blume-20020211-gensml
479		Description:
480
481		A quick fix for a problem with GenSML (in the pgraph-util library):
482		Make generation of toplevel "local" optional.  (Strictly speaking,
483		signature definitions within "local" are not legal SML.)
484
485		Other than that: updates to INSTALL and cm/TODO.
486
487		----------------------------------------------------------------------
488		Name: Matthias Blume
489		Date: 2002/02/08 15:00:00 EST
490		Tag: blume-20020208-uniquepid
491		Description:
492
493		0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
494
495		1. The installer (config/install.sh) has gotten smarter:
496
497		- Configuration options are a bit easier to specify now
498		(in config/targets).
499		- Bug in recognizing .tar.bz2 files fixed.
500		- Installer automatically resolves dependencies between
501		configuration options (e.g., if you ask for eXene, you will
502		also get cml -- regardless whether you asked for it or not).
503		- Installer can run in "quieter mode" by setting the environment
504		variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
505		"completely silent", though.
506		- Build HashCons library as part of smlnj-lib.
507
508		2. A new scheme for assigning persistent identifiers to compilation
509		units (and, by extension, to types etc.) has been put into place.
510		This fixes a long-standing bug where types and even dynamic values
511		can get internally confused, thereby compromising type safety
512		(abstraction) and dynamic correctness.  See
513
514		http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
515
516		for an example of how things could go wrong until now.
517
518		The downside of the new scheme is that pids are not quite as
519		persistent as they used to be: CM will generate a fresh pid
520		for every compilation unit that it thinks it sees for the first
521		time.  That means that if you compile starting from a clean, fresh
522		source tree at two different times, you end up with different
523		binaries.
524
525		Cutoff recompilation, however, has not been compromised because
526		CM keeps pid information in special caches between runs.
527
528		----------------------------------------------------------------------
529		Name: Lal George
530		Date: 2002/02/07 15:34:13 EST 2002
531		Tag: <none>
532		Description:
533
534		Compilers that generate assembly code may produce  global labels
535		whose value is resolved at link time. The various peephole optimization
536		modules did not take this in account.
537
538		TODO. The Labels.addrOf function should really return an option
539		type so that clients are forced to deal with this issue, rather
540		than an exception being raised.
541
542		----------------------------------------------------------------------
543		Name: Lal George
544		Date: 2002/02/06 13:55:02 EST
545		Tag: george-20020206-ra-breakup
546		Description:
547
548		1. A bug fix from Allen.
549
550		A typo causes extra fstp %st(0)'s to be generated at compensation
551		edges, which might cause stack underflow traps at runtime.  This
552		occurs in fft where there are extraneous fstps right before the 'into'
553		trap instruction (in this case they are harmless since none of the
554		integers overflow.)
555
556		2. Pulled out various utility modules that were embedded in the modules
557		of the register allocator. I need these modules for other purposes, but
558		they are not complete enough to put into a library (just yet).
559		----------------------------------------------------------------------
560		Name: Matthias Blume
561		Date: 2002/01/31 16:05:00 EST
562		Tag: blume-20020131-sparc-ccalls
563		Description:
564
565		1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
566		of extra stack space by mistake.  Fixed.
567
568		2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
569
570		----------------------------------------------------------------------
571		Name: Allen Leung
572		Date: 2002/01/30
573		Tag: leunga-20020130-nowhere-bug-fix
574		Description:
575
576		MLRISC bug fixes:
577		1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
578		2. MachineInt.fromString was negating its value.
579
580		----------------------------------------------------------------------
581		Name: Matthias Blume
582		Date: 2002/01/29
583		Tag: blume-20020129-INSTALL
584		Description:
585
586		- Added somewhat detailed installation instructions (file INSTALL).
587		- Fixed curl-detection bug in config/install.sh.
588		- It is now possible to select the URL getter using the URLGETTER
589		environment variable:
590
591		not set / "unknown"      --> automatic detection (script tries wget,
592		curl, and lynx)
593		"wget" / "curl" / "lynx" --> use the specified program (script "knows"
594		how to properly invoke them)
595		other                    --> use $URLGETTER directly, it must take
596		precisely two command-line arguments
597		(source URL and destination file name)
598
599		----------------------------------------------------------------------
600		Name: Matthias Blume
601		Date: 2002/01/28
602		Tag: blume-20020128-sparc-ccalls
603		Description:
604
605		- Fixed problem with calculation of "used" registers in sparc-c-calls.
606		- Make use of the allocParam argument in sparc-c-calls.
607
608		----------------------------------------------------------------------
609		Name: Matthias Blume
610		Date: 2002/01/28
611		Tag: blume-20020128-allocParam
612		Description:
613
614		John Reppy:  Changes c-calls API to accept client-callback for
615		allocating extra stack space.
616		me: Corresponding changes to mlriscGen (using a dummy argument that
617		does not change the current behavior).
618
619		----------------------------------------------------------------------
620		Name: Matthias Blume
621		Date: 2002/01/28 12:00:00
622		Tag: Release_110_38
623		Description:
624
625		This time for real!!!
626
627		----------------------------------------------------------------------
628		Name: Matthias Blume
629		Date: 2002/01/28 10:56:00 EST
630		Tag: blume-20020128-retraction
631		Description:
632
633		0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
634		with blume-Release_110_38-retracted.)
635
636		1. Fixed a problem with incorrect rounding modes in real64.sml.
637		(Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
638
639		2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
640		and enums fixed.  The naming of corresponding ML identifiers should
641		now be consistent again.
642
643		----------------------------------------------------------------------
644		Name: Allen Leung
645		Date: 2002/01/27
646		Tag: leunga-20020127-nowhere
647		Description:
648
649		Added a target called nowhere in the configuration scripts.
650		Enabling this will build the MLRISC 'nowhere' tool (for translating
651		programs with where-clauses into legal SML code) during installation.
652
653		----------------------------------------------------------------------
654		Name: Matthias Blume
655		Date: 2002/01/25 21:27:00 EST
656		Tag: blume-Release_110_38-retracted
657		Description:
658
659		Call it a (working) release!  Version is 110.38. Bootfiles are ready.
660
661		README will be added later.
662
663		!!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
664		(Release_110_38) removed.  Reason: Last-minute bug fixes.
665
666		----------------------------------------------------------------------
667		Name: Matthias Blume
668		Date: 2002/01/25
669		Tag: blume-20020125-ffi
670		Description:
671
672		A large number of tweaks and improvements to ml-nlffi-lib and
673		ml-nlffigen:
674
675		- ML represenation types have been streamlined
676		- getter and setter functions work with concrete values, not abstract
677		ones where possible
678		- ml-nlffigen command line more flexible (see README file there)
679		- some bugs have been fixed (hopefully)
680
681		----------------------------------------------------------------------
682		Name: Lal George
683		Date: 2002/01/24
684		Tag: george-20020124-risc-ra-interface
685		Description:
686
687		There is a dramatic simplification in the interface to the
688		register allocator for RISC architectures as a result of making
689		parallel copy instructions explicit.
690
691		----------------------------------------------------------------------
692		Name: Matthias Blume
693		Date: 2002/01/22
694		Tag: blume-20020122-x86-ccalls
695		Description:
696
697		Bug fix for c-calls on x86 (having to do with how char- and
698		short-arguments are being handled).
699
700		----------------------------------------------------------------------
701		Name: Matthias Blume
702		Date: 2002/01/21
703		Tag: blume-20020121-ff
704		Description:
705
706		Another day of fiddling with the FFI...
707
708		1. Bug fix/workaround:  CKIT does not complain about negative array
709		dimensions, so ml-nlffigen has to guard itself against this possibility.
710		(Otherwise a negative dimension would send it into an infinite loop.)
711
712		2. Some of the abstract types (light objects, light pointers, most "base"
713		types) in structure C are now eqtypes.
714
715		3. Added constructors and test functions for NULL function pointers.
716
717		----------------------------------------------------------------------
718		Name: Matthias Blume
719		Date: 2002/01/18
720		Tag: blume-20020118-ready-for-new-release
721		Description:
722
723		Made config/srcarchiveurl point to a new place.  (Will provide boot
724		files shortly.)
725
726		Maybe we christen this to be 110.38?
727
728		----------------------------------------------------------------------
729		Name: Matthias Blume
730		Date: 2002/01/18
731		Tag: blume-20020118-more-ffifiddle
732		Description:
733
734		Today's FFI fiddling:
735
736		- Provided a structure CGetSet with "convenient" versions of C.Get.* and
737		C.Set.* that use concrete (MLRep.*) arguments and results instead
738		of abstract ones.
739
740		- Provided word-style bit operations etc. for "int" representation
741		types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
742		and Long.
743
744		----------------------------------------------------------------------
745		Name: Matthias Blume
746		Date: 2002/01/18
747		Tag: blume-20020118-use-x86-fp
748		Description:
749
750		Now that x86-fast-fp seems to be working, I turned it back on again
751		by default.  (Seems to work fine now, even with the FFI.)
752
753		Other than that, I added some documentation about the FFI to
754		src/ml-nlffigen/README and updated the FFI test examples in
755		src/ml-nlffi-lib/Tests/*.
756
757		----------------------------------------------------------------------
758		Name: Allen Leung
759		Date: 2002/01/17
760		Tag: leunga-20020117-x86-fast-fp-call
761		Description:
762
763		1. Fixed a problem with handling return fp values when x86's fast fp
764		mode is turned on.
765
766		2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
767		of %f32.
768
769		3. Added a constructor INT32lit to the ast of MLRISC tools.
770
771		----------------------------------------------------------------------
772		Name: Matthias Blume
773		Date: 2002/01/16
774		Tag: blume-20020116-ffifiddle
775		Description:
776
777		More fiddling with the FFI interface:
778
779		- Make constness 'c instead of rw wherever possible.  This eliminates
780		the need for certain explicit coercions.  (However, due to ML's
781		value polymorphism, there will still be many cases where explicit
782		coercions are necessary.  Phantom types are not the whole answer
783		to modeling a subtyping relationship in ML.)
784
785		- ro/rw coersions for pointers added.  (Avoids the detour through */&.)
786
787		- "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
788		clumsy workaround for varargs problem.)
789
790		----------------------------------------------------------------------
791		Name: Lal George
792		Date: 2002/01/15
793		Tag: <none>
794		Description:
795
796		1. Since COPY instructions are no longer native to the architecture,
797		a generic functor can be used to implement the expandCopies function.
798
799		2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
800		TEXT segment.
801
802		----------------------------------------------------------------------
803		Name: Matthias Blume
804		Date: 2002/01/15
805		Tag: blume-20020115-ffiupdates
806		Description:
807
808		1. Fix for bug resulting in single-precision float values being returned
809		incorrectly from FFI calls.
810
811		2. Small modifications to C FFI API:
812
813		- memory-allocation routines return straight objects (no options)
814		and raise an exception in out-of-memory situations
815		- unsafe extensions to cast between function pointers and pointers
816		from/to ints
817		- added structure C_Debug as an alternative to structure C where
818		pointer-dereferencing (|*| and |*!) always check for null-pointers
819		- added open_lib' to DynLinkage;  open_lib' works like open_lib
820		but also takes a (possibly empty) list of existing library handles
821		that the current library depends on
822
823		----------------------------------------------------------------------
824		Name: Matthias Blume
825		Date: 2002/01/10
826		Tag: blume-20020110-newffigen
827		Description:
828
829		1. Updates to portable graph code.
830
831		2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
832		more scalable now so that even huge interfaces such as the one
833		for GTK compile in finite time and space. :-)
834		See src/ml-nlffigen/README for details on what's new.
835
836		----------------------------------------------------------------------
837		Name: Lal George
838		Date: 2001/01/09 14:31:35 EST 2002
839		Tag: george-20011206-rm-native-copy
840		Description:
841
842		Removed the native COPY and FCOPY instructions
843		from all the architectures and replaced it with the
844		explicit COPY instruction from the previous commit.
845
846		It is now possible to simplify many of the optimizations
847		modules that manipulate copies. This has not been
848		done in this change.
849
850		----------------------------------------------------------------------
851		Name: Lal George
852		Date: 2001/12/06 16:50:13 EST 2001
853		Tag: george-20011206-mlrisc-instruction
854		Description:
855
856		Changed the representation of instructions from being fully abstract
857		to being partially concrete. That is to say:
858
859		from
860		type instruction
861
862		to
863		type instr                              (* machine instruction *)
864
865		datatype instruction =
866		LIVE of {regs: C.cellset, spilled: C.cellset}
867		| KILL of {regs: C.cellset, spilled: C.cellset}
868		| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
869		| ANNOTATION of {i: instruction, a: Annotations.annotation}
870		| INSTR of instr
871
872		This makes the handling of certain special instructions that appear on
873		all architectures easier and uniform.
874
875		LIVE and KILL say that a list of registers are live or killed at the
876		program point where they appear. No spill code is generated when an
877		element of the 'regs' field is spilled, but the register is moved to
878		the 'spilled' (which is present, more for debugging than anything else).
879
880		LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
881		We used to generate:
882
883		DEFFREG f1
884		f1 := f2 + f3
885		trapb
886
887		but now generate:
888
889		f1 := f2 + f3
890		trapb
891		LIVE {regs=[f1,f2,f3], spilled=[]}
892
893		Furthermore, the DEFFREG (hack) required that all floating point instruction
894		use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
895		defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
896		in a cleaner alpha implementation. (Hopefully, intel will not get rid of
897		this architecture).
898
899		COPYXXX is intended to replace the parallel COPY and FCOPY  available on
900		all the architectures. This will result in further simplification of the
901		register allocator that must be aware of them for coalescing purposes, and
902		will also simplify certain aspects of the machine description that provides
903		callbacks related to parallel copies.
904
905		ANNOTATION should be obvious, and now INSTR represents the honest to God
906		machine instruction set!
907
908		The <arch>/instructions/<arch>Instr.sml files define certain utility
909		functions for making porting easier -- essentially converting upper case
910		to lower case. All machine instructions (of type instr) are in upper case,
911		and the lower case form generates an MLRISC instruction. For example on
912		the alpha we have:
913
914		datatype instr =
915		LDA of {r:cell, b:cell, d:operand}
916		| ...
917
918		val lda : {r:cell, b:cell, d:operand} -> instruction
919		...
920
921		where lda is just (INSTR o LDA), etc.
922
923		----------------------------------------------------------------------
924		Name: Matthias Blume
925		Date: 2001/11/22 21:40:00 EST
926		Tag: Release_110_37
927		Description:
928
929		Release 110.37.  This time for real.
930
931		----------------------------------------------------------------------
932		Name: Matthias Blume
933		Date: 2001/11/21 16:35:00 EST
934		Tag: blume-20011121-foot-in-mouth
935		Description:
936
937		Removed the "Release_110_37" tag because of a serious bug.
938		This will be re-tagged once the bug is fixed.
939
940		----------------------------------------------------------------------
941		Name: Matthias Blume
942		Date: 2001/11/21 16:14:00 EST
943		Tag: blume-20011121-forgottenfile
944		Description:
945
946		Forgot to add a file.  (Just a .tex-file -- part of
947		the CM manual source.)
948
949		----------------------------------------------------------------------
950		Name: Matthias Blume
951		Date: 2001/11/21 16:10:00 EST
952		Tag: blume-20011121-invalid_110_37
953		Description:
954
955		Note: I removed the original tag "Release_110_37" from this commit
956		because we found a serious bug in all non-x86 backends.
957		- Matthias
958
959		1. Modifications to the SML/NJ code generator and to the runtime system
960		so that code object name strings are directly inserted into code
961		objects at code generation time.  The only business the runtime system
962		has with this is now to read the name strings on occasions.
963		(The encoding of the name string has also changed somewhat.)
964
965		2. CM now implements a simple "set calculus" for specifying export lists.
966		In particular, it is now possible to refer to the export lists of
967		other libraries/groups/sources and form unions as well as differences.
968		See the latest CM manual for details.
969
970		3. An separate notion of "proxy" libraries has again be eliminated from
971		CM's model.  (Proxy libraries are now simply a special case of using
972		the export list calculus.)
973
974		4. Some of the existing libraries now take advantage of the new set
975		calculus.
976		(Notice that not all libraries have been converted because some
977		of the existing .cm-files are supposed to be backward compatible
978		with 110.0.x.)
979
980		5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
981		"General.exnMessage"!  The former relies on a certain hook to be
982		initialized, and that often does not happen in the stand-alone case.)
983
984		----------------------------------------------------------------------
985		Name: Lal George
986		Date: 2001/11/21  13:56:18 EST
987		Tag: george-2001121-pseudo-ops
988		Description:
989
990		Implemented a complete redesign of MLRISC pseudo-ops. Now there
991		ought to never be any question of incompatabilities with
992		pseudo-op syntax expected by host assemblers.
993
994		For now, only modules supporting GAS syntax are implemented
995		but more should follow, such as MASM, and vendor assembler
996		syntax, e.g. IBM as, Sun as, etc.
997
998		----------------------------------------------------------------------
999		Name: Matthias Blume
1000		Date: 2001/11/14 11:52:00 EST
1001		Tag: blume-20011114-srcname
1002		Description:
1003
1004		1. Routed the name of the current source file to mlriscgen where it
1005		should be directly emitted into the code object.  (This last part
1006		is yet to be done.)
1007
1008		2. Some cleanup of the pgraph code to make it match the proposal that
1009		I put out the other day.  (The proposal notwithstanding, things are
1010		still in flux here.)
1011
1012		----------------------------------------------------------------------
1013		Name: Lal George
1014		Date: 2001/11/14 09:44:04 EST
1015		Tag:
1016		Description:
1017
1018		Fix for a backpatching bug reported by Allen.
1019
1020		Because the boundary between short and long span-dependent
1021		instructions is +/- 128, there are an astounding number of
1022		span-dependent instructions whose size is over estimated.
1023
1024		Allen came up with the idea of letting the size of span
1025		dependent instructions be non-monotonic, for a maxIter
1026		number of times, after which the size must be monotonically
1027		increasing.
1028
1029		This table shows the number of span-dependent instructions
1030		whose size was over-estimated as a function of maxIter, for the
1031		file Parse/parse/ml.grm.sml:
1032
1033		maxIter            # of instructions:
1034		10                      687
1035		20                      438
1036		30                      198
1037		40                        0
1038
1039		In compiling the compiler, there is no significant difference in
1040		compilation speed between maxIter=10 and maxIter=40. Actually,
1041		my measurements showed that maxIter=40 was a tad faster than
1042		maxIter=10! Also 96% of the  files in the compiler reach a fix
1043		point within 13 iterations, so fixing maxIter at 40, while high,
1044		is okay.
1045
1046		----------------------------------------------------------------------
1047		Name: Matthias Blume
1048		Date: 2001/10/31 15:25:00 EST
1049		Tag: blume-20011031-pgraph
1050		Description:
1051
1052		CKIT:
1053		* Changed the "Function" constructor of type Ast.ctype to carry optional
1054		argument identifiers.
1055		* Changed the return type of TypeUtil.getFunction accordingly.
1056		* Type equality ignores the argument names.
1057		* TypeUtil.composite tries to preserve argument names but gives up quickly
1058		if there is a mismatch.
1059
1060		installation script:
1061		* attempts to use "curl" if available (unless "wget" is available as well)
1062
1063		CM:
1064		* has an experimental implementation of "portable graphs" which I will
1065		soon propose as an implementation-independent library format
1066		* there are also new libraries $/pgraph.cm and $/pgraph-util.cm
1067
1068		NLFFI-LIB:
1069		* some cleanup (all cosmetic)
1070
1071		NLFFIGEN:
1072		* temporarily disabled the mechanism that suppresses ML output for
1073		C definitions whose identifiers start with an underscore character
1074		* generate val bindings for enum constants
1075		* user can request that only one style (light or heavy) is being used;
1076		default is to use both (command-line arguments: -heavy and -light)
1077		* fixed bug in handling of function types involving incomplete pointers
1078		* generate ML entry points that take record arguments (i.e., using
1079		named arguments) for C functions that have a prototype with named
1080		arguments
1081		(see changes to CKIT)
1082
1083		----------------------------------------------------------------------
1084		Name: Allen Leung
1085		Date: 2001/10/27 20:34:00 EDT
1086		Tag: leunga-20011027-x86-fast-fp-call
1087		Description:
1088
1089		Fixed the bug described in blume-20010920-slowfp.
1090
1091		The fix involves
1092		1. generating FCOPYs in FSTP in ia32-svid
1093		2. marking a CALL with the appropriate annotation
1094
1095		----------------------------------------------------------------------
1096		Name: Matthias Blume
1097		Date: 2001/10/16 11:32:00 EDT
1098		Tag: blume-20011016-netbsd
1099		Description:
1100
1101		Underscore patch from Chris Richards (fixing problem with compiling
1102		runtime system under recent NetBSD).
1103
1104		----------------------------------------------------------------------
1105		Name: Allen Leung
1106		Date: 2001/10/12 17:18:32 EDT 2001
1107		Tag: leung-20011012-x86-printflowgraph
1108		Description:
1109
1110		X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
1111
1112		----------------------------------------------------------------------
1113		Name: Lal George
1114		Date: 2001/10/11 23:51:34 EDT
1115		Tag: george-20011011-too-many-instrs
1116		Description:
1117
1118		The representation of a program point never expected to see more
1119		than 65536 instructions in a basic block!
1120
1121		----------------------------------------------------------------------
1122		Name: Lal George
1123		Date: 2001/10/09 09:41:37 EDT
1124		Tag: george-20011008-mlrisc-labels
1125		Description:
1126
1127		Changed the machine description files to support printing of
1128		local and global labels in assembly code, based on host assembler
1129		conventions.
1130
1131		----------------------------------------------------------------------
1132		Name: Matthias Blume
1133		Date: 2001/09/25 15:25:00 EDT
1134		Tag: blume-20010925-exninfo
1135		Description:
1136
1137		I provided a non-hook implementation of exnName (at the toplevel) and
1138		made the "dummy" implementation of exnMessage (at the toplevel) more
1139		useful: if nothing gets "hooked in", then at least you are going to
1140		see the exception name and a message indicating why you don't see more.
1141
1142		[For the time being, programs that need exnMessage and want to use
1143		ml-build should either use General.exnMessage (strongly recommended) or
1144		refer to structure General at some other point so that CM sees a
1145		static dependency.]
1146
1147		[Similar remarks go for "print" and "use":  If you want to use their
1148		functionality in stand-alone programs generated by ml-build, then use
1149		TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
1150
1151		----------------------------------------------------------------------
1152		Name: Matthias Blume
1153		Date: 2001/09/20 17:28:00 EDT
1154		Tag: blume-20010920-slowfp
1155		Description:
1156
1157		Allen says that x86-fast-fp is not safe yet, so I turned it off again...
1158
1159		----------------------------------------------------------------------
1160		Name: Matthias Blume
1161		Date: 2001/09/20 17:20:00 EDT
1162		Tag: blume-20010920-canonicalpaths
1163		Description:
1164
1165		0. Updated the BOOT file (something that I forgot to do earlier).
1166
1167		1. Small internal change to CM so that it avoids "/../" in filenames
1168		as much as possible (but only where it is safe).
1169
1170		2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
1171		that arguments that contain delimiters are passed through correctly.
1172		This change also means that all "special" arguments of the form
1173		@SMLxxx... must come first.
1174
1175		3. Changed install script to put relative anchor names for tool commands
1176		into pathconfig.
1177
1178		----------------------------------------------------------------------
1179		Name: Matthias Blume
1180		Date: 2001/09/18 15:35:00 EDT
1181		Tag: blume-20010918-readme11036
1182		Description:
1183
1184		Added README files.
1185
1186		----------------------------------------------------------------------
1187		Name: Matthias Blume
1188		Date: 2001/09/18 11:45:00 EDT
1189		Tag: Release_110_36 (retag)
1190		Description:
1191
1192		Fixed mistake in config/preloads. Retagged as 110.36.
1193
1194		----------------------------------------------------------------------
1195		Name: Matthias Blume
1196		Date: 2001/09/18 09:40:00 EDT
1197		Tag: Release_110_36_orig (tag changed)
1198		Description:
1199
1200		New version (110.36).  New bootfiles.
1201
1202		----------------------------------------------------------------------
1203		Name: Matthias Blume
1204		Date: 2001/09/14 16:15:00 EDT
1205		Tag: blume-20010914-x86fastfp
1206		Description:
1207
1208		John committed some changes that Allen made, in particular a (hopefully)
1209		correctly working version of the x86-fp module.
1210
1211		I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
1212		flag to "true".  Everything seems to compile to a fixpoint ok, and
1213		"mandelbrot" speeds up by about 15%.
1214
1215		----------------------------------------------------------------------
1216		Name: Matthias Blume
1217		Date: 2001/09/13 11:20:00 EDT
1218		Tag: blume-20010913-minimal
1219		Description:
1220
1221		1. Stefan Monnier's patch to fix a miscompilation problem that
1222		was brought to light by John Reppy's work on Moby.
1223
1224		2. Implemented a minimal "structure Compiler" that contains just
1225		"version" and "architecture".  The minimal version will be
1226		available when the full version is not.  This is for backward-
1227		compatibility with code that wants to test Compiler.version.
1228
1229		----------------------------------------------------------------------
1230		Name: Matthias Blume
1231		Date: 2001/08/28 14:03:00 EDT
1232		Tag: blume-20010828-ml-lex
1233		Description:
1234
1235		Fix for bug 1581, received from Neophytos Michael.
1236
1237		----------------------------------------------------------------------
1238		Name: Matthias Blume
1239		Date: 2001/08/27 11:20:00 EDT
1240		Tag: blume-20010827-readme11035
1241		Description:
1242
1243		Fleshed out the README file for 110.35.
1244
1245		----------------------------------------------------------------------
1246		Name: Matthias Blume
1247		Date: 2001/08/24 17:10:00 EDT
1248		Tag: Release_110_35
1249		Description:
1250
1251		New version number (110.35).  New bootfiles.
1252
1253		----------------------------------------------------------------------
1254		Name: Lal George
1255		Date: 2001/08/24 13:47:18 EDT 2001
1256		Tag: george-20010824-MLRISC-graphs
1257		Description:
1258
1259		removed clusters from MLRISC completely and replaced with graphs.
1260
1261		----------------------------------------------------------------------
1262		Name: Matthias Blume
1263		Date: 2001/08/23 17:50:00 EDT
1264		Tag: blume-20010823-toplevel
1265		Description:
1266
1267		- some reorganization of the code that implements various kinds of
1268		environments in the compiler (static, dynamic, symbolic, combined)
1269		- re-implemented the EnvRef module so that evalStream works properly
1270		(if the stream contains references to "use", "CM.make", etc.)
1271		- cleaned up evalloop.sml and interact.sml (but they need more cleaning)
1272
1273		----------------------------------------------------------------------
1274		Name: Matthias Blume
1275		Date: 2001/08/20 15:50 EDT
1276		Tag: blume20010820-slipup
1277		Description:
1278
1279		I forgot to commit a few files.  Here they are...
1280
1281		----------------------------------------------------------------------
1282		Name: Matthias Blume
1283		Date: 2001/08/20 15:35:00 EDT
1284		Tag: blume-20010820-debugprof
1285		Description:
1286
1287		!!!! NEW BOOTFILES !!!!
1288
1289		This is another round of reorganizing the compiler sources.  This
1290		time the main goal was to factor out all the "instrumentation"
1291		passes (for profiling and backtracing) into their own library.
1292		The difficulty was to do it in such a way that it does not depend
1293		on elaborate.cm but only on elabdata.cm.
1294
1295		Therefore there have been further changes to both elaborate.cm and
1296		elabdata.cm -- more "generic" things have been moved from the former
1297		to the latter.  As a result, I was forced to split the assignment
1298		of numbers indicating "primtyc"s into two portions: SML-generic and
1299		SML/NJ-specific.  Since it would have been awkward to maintain,
1300		I bit the bullet and actually _changed_ the mapping between these
1301		numbers and primtycs.  The bottom line of this is that you need
1302		a new set of bin- and bootfiles.
1303
1304		I have built new bootfiles for all architectures, so doing a fresh
1305		checkout and config/install.sh should be all you need.
1306
1307		The newly created library's name is
1308
1309		$smlnj/viscomp/debugprof.cm
1310
1311		and its sources live under
1312
1313		src/compiler/DebugProf
1314
1315		----------------------------------------------------------------------
1316		Name: Matthias Blume
1317		Date: 2001/08/15 17:15:00 EDT
1318		Tag: blume-20010815-compreorg
1319		Description:
1320
1321		This is a first cut at reorganizing the CM libraries that make up the
1322		core of the compiler.  The idea is to separate out pieces that could
1323		be used independently by tools, e.g., the parser, the typechecker, etc.
1324
1325		The current status is a step in this direction, but it is not quite
1326		satisfactory yet.  Expect more changes in the future.
1327
1328		Here is the current (new) organization...
1329
1330		What used to be $smlnj/viscomp/core.cm is now divided into
1331		six CM libraries:
1332
1333		$smlnj/viscomp/basics.cm
1334		/parser.cm
1335		/elabdata.cm
1336		/elaborate.cm
1337		/execute.cm
1338		/core.cm
1339
1340		The CM files for these libraries live under src/system/smlnj/viscomp.
1341		All these libraries are proxy libraries that contain precisely
1342		one CM library component.  Here are the locations of the components
1343		(all within the src/compiler tree):
1344
1345		Basics/basics.cm
1346		Parse/parser.cm
1347		ElabData/elabdata.cm
1348		Elaborator/elaborate.cm
1349		Execution/execute.cm
1350		core.cm
1351
1352		[This organization is the same that has been used already
1353		for a while for the architecture-specific parts of the visible
1354		compiler and for the old version of core.cm.]
1355
1356		As you will notice, many source files have been moved from their
1357		respective original locations to a new home in one of the above
1358		subtrees.
1359
1360		The division of labor between the new libraries is the following:
1361
1362		basics.cm:
1363		- Simple, basic definitions that pertain to many (or all) of
1364		the other libraries.
1365		parser.cm:
1366		- The SML parser, producing output of type Ast.dec.
1367		- The type family for Ast is also defined and exported here.
1368		elabdata.cm:
1369		- The datatypes that describe input and output of the elaborator.
1370		This includes types, absyn, and static environments.
1371		elaborator.cm:
1372		- The SML/NJ type checker and elaborator.
1373		This maps an Ast.dec (with a given static environment) to
1374		an Absyn.dec (with a new static environment).
1375		- This libraries implements certain modules that used to be
1376		structures as functors (to remove dependencies on FLINT).
1377		execute.cm:
1378		- Everything having to do with executing binary code objects.
1379		- Dynamic environments.
1380		core.cm:
1381		- SML/NJ-specific instantiations of the elaborator and MLRISC.
1382		- Top-level modules.
1383		- FLINT (this should eventually become its own library)
1384
1385		Notes:
1386
1387		I am not 100% happy with the way I separated the elaborator (and its
1388		data structures) from FLINT.  Two instances of the same problem:
1389
1390		1. Data structures contain certain fields that carry FLINT-specific
1391		information.  I hacked around this using exn and the property list
1392		module from smlnj-lib.  But the fact that there are middle-end
1393		specific fields around at all is a bit annoying.
1394
1395		2. The elaborator calculates certain FLINT-related information.  I tried
1396		to make this as abstract as I could using functorization, but, again,
1397		the fact that the elaborator has to perform calculations on behalf
1398		of the middle-end at all is not nice.
1399
1400		3. Having to used exn and property lists is unfortunate because it
1401		weakens type checking.  The other alternative (parameterizing
1402		nearly *everything*) is not appealing, though.
1403
1404		I removed the "rebinding =" warning hack because due to the new organization
1405		it was awkward to maintain it.  As a result, the compiler now issues some of
1406		these warnings when compiling init.cmi during bootstrap compilation. On
1407		the plus side, you also get a warning when you do, for example:
1408		val op = = Int32.+
1409		which was not the case up to now.
1410
1411		I placed "assign" and "deref" into the _Core structure so that the
1412		code that deals with the "lazy" keyword can find them there.  This
1413		removes the need for having access to the primitive environment
1414		during elaboration.
1415
1416		----------------------------------------------------------------------
1417		Name: Matthias Blume
1418		Date: 2001/08/13
1419		Tag: blume-20010813-closures
1420		Description:
1421
1422		This fix was sent to us by Zhong Shao.  It is supposed to improve the
1423		performance of certain loops by avoiding needless closure allocation.
1424
1425		----------------------------------------------------------------------
1426		Name: Lal George
1427		Date: 2001/07/31 10:03:23 EDT 2001
1428		Tag: george-20010731-x86-fmalloc
1429		Description: Fixed bug in x86 calls
1430
1431		There was a bug where call instructions would mysteriously
1432		vanish. The call instruction had to be one that returned
1433		a floating point value.
1434
1435		----------------------------------------------------------------------
1436		Name: Lal George
1437		Date: 2001/07/19 16:36:29 EDT 2001
1438		Tag: george-20010719-simple-cells
1439		Description:
1440
1441		I have dramatically simplified the interface for CELLS in MLRISC.
1442
1443		In summary, the cells interface is broken up into three parts:
1444
1445		1. CellsBasis : CELLS_BASIS
1446
1447		CellsBasis is a top level structure and common for all
1448		architectures.  it contains the definitions of basic datatypes
1449		and utility  functions over these types.
1450
1451		2. functor Cells() : CELLS
1452
1453		Cells generates an interface for CELLS that incorporates the
1454		specific resources on the target architecture, such as the
1455		presence of special register classes, their number and size,
1456		and various useful substructures.
1457
1458		3. <ARCH>CELLS
1459
1460		e.g. SparcCells: SPARCCELLS
1461
1462		<ARCH>CELLS usually contains additional bindings for special
1463		registers  on the architecture, such as:
1464
1465		val r0 : cell           (* register zero *)
1466		val y : cell            (* Y register *)
1467		val psr : cell          (* processor status register *)
1468		...
1469
1470		The structure returned by applying the Cells functor is opened
1471		in this interface.
1472
1473		The main implication of all this is that the datatypes for cells is
1474		split between CellsBasis and CELLS -- a fairly simple change for user
1475		code.
1476
1477		In the old scheme the CELLS interface had a definitional binding of
1478		the form:
1479
1480		signature CELLS = sig
1481
1482		structure CellsBasis = CellsBasis
1483
1484		...
1485
1486		end
1487
1488		With all the sharing constraints that goes on in MLRISC, this old
1489		design  quickly leads to errors such as:
1490
1491		"structure definition spec inside of sharing ... "
1492
1493
1494		and appears to require an unacceptable amount of sharing and where
1495		constraint hackery.
1496
1497		I think this error message (the interaction of definitional specs and
1498		sharing) requires more explanation on our web page.
1499
1500		----------------------------------------------------------------------
1501		Name: Matthias Blume
1502		Date: 2001/07/19 15:00:00 EDT
1503		Tag: blume-20010719-libreorg
1504		Description:
1505
1506		This update puts together a fairly extensive but straightforward change
1507		to the way the libraries that implement the interactive system are
1508		organized:
1509
1510		The biggest change is the elimination of structure Compiler.  As a
1511		replacement for this structure, there is now a CM library
1512		(known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
1513		that exports all the substructures of the original structure Compiler
1514		directly.  So instead of saying Compiler.Foo.bar one now simply
1515		says Foo.bar.  (The CM libraries actually export a collection of
1516		structures that is richer than the collection of substructures of
1517		structure Compiler.)
1518
1519		To make the transition smooth, there is a separate library called
1520		$smlnj/compiler/compiler.cm which puts together and exports the
1521		original structure Compiler (or at least something very close to it).
1522
1523		There are five members of the original structure Compiler
1524		that are not exported directly but which instead became members
1525		of a new structure Backend (described by signature BACKEND).  These are:
1526		structure Profile (: PROFILE), structure Compile (: COMPILE), structure
1527		Interact (: INTERACT), structure Machine (: MACHINE), and val
1528		architecture (: string).
1529
1530		Structure Compiler.Version has become structure CompilerVersion.
1531
1532		Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
1533		by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
1534		or x86, respectively.
1535		Each of these exports the same frontend structures that
1536		$smlnj/compiler.cm exports.  But they do not have a structure Backend
1537		and instead export some structure <Arch>Backend where <Arch> is Alpha32,
1538		Hppa, PPC, Sparc, or X86, respectively.
1539
1540		Library $smlnj/compiler/all.cm exports the union of the exports of
1541		$smlnj/compiler/<arch>.cm
1542
1543		There are no structures <Arch>Compiler anymore, use
1544		$smlnj/compiler/<arch>.cm instead.
1545
1546		Library host-compiler-0.cm is gone.  Instead, the internal library
1547		that instantiates CM is now called cm0.cm.  Selection of the host
1548		compiler (backend) is no longer done here but. (Responsibility for it
1549		now lies with $smlnj/compiler/current.cm.  This seems to be more
1550		logical.)
1551
1552		Many individual files have been moved or renamed.  Some files have
1553		been split into multiple files, and some "dead" files have been deleted.
1554
1555		Aside from these changes to library organization, there are also changes
1556		to the way the code itself is organized:
1557
1558		Structure Binfile has been re-implemented in such a way that it no
1559		longer needs any knowledge of the compiler.  It exclusively deals
1560		with the details of binfile layout.  It no longer invokes the
1561		compiler (for the purpose of creating new prospective binfile
1562		content), and it no longer has any knowledge of how to interpret
1563		pickles.
1564
1565		Structure Compile (: COMPILE) has been stripped down to the bare
1566		essentials of compilation.  It no longer deals with linking/execution.
1567		The interface has been cleaned up considerably.
1568
1569		Utility routines for dealing with linking and execution have been
1570		moved into their own substructures.
1571
1572		(The ultimate goal of these changes is to provide a light-weight
1573		binfile loader/linker (at least for, e.g., stable libraries) that
1574		does not require CM or the compiler to be present.)
1575
1576		CM documentation has been updated to reflect the changes to library
1577		organization.
1578
1579		----------------------------------------------------------------------
1580		Name: Matthias Blume
1581		Date: 2001/07/10 17:30:00 EDT
1582		Tag: Release_110_34
1583		Description:
1584
1585		Minor tweak to 110.34 (re-tagged):
1586
1587		- README.html file added to CVS repository
1588		- runtime compiles properly under FreeBSD 3.X and 4.X
1589
1590		----------------------------------------------------------------------
1591		Name: Matthias Blume
1592		Date: 2001/07/10 17:30:00 EDT
1593		Tag: Release_110_34
1594		Description:
1595
1596		New version number (110.34). New bootfiles.
1597
1598		----------------------------------------------------------------------
1599		Name: Matthias Blume
1600		Date: 2001/07/09 16:00:00 EDT
1601		Tag: blume-20010709-more-varargs
1602		Description:
1603
1604		I changed the handling of varargs in ml-nlffigen again:
1605		The ellipsis ... will now simply be ignored (with an accompanying warning).
1606
1607		The immediate effect is that you can actually call a varargs function
1608		from ML -- but you can't actually supply any arguments beyond the ones
1609		specified explicitly.  (For example, you can call printf with its format
1610		string, but you cannot pass additional arguments.)
1611
1612		This behavior is only marginally more useful than the one before, but
1613		it has the advantage that a function or, more importantly, a function
1614		type never gets dropped on the floor, thus avoiding follow-up problems with
1615		other types that refer to the offending one.
1616
1617		----------------------------------------------------------------------
1618		Name: Matthias Blume
1619		Date: 2001/07/09 11:25:00 EDT
1620		Tag: blume-20010709-varargs
1621		Description:
1622
1623		1. ckit-lib.cm now exports structure Error
1624		2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
1625		with a warning accompanied by a source location.  Moreover, it
1626		merely skips the offending function or type and proceeds with the
1627		rest of its work.u  As a result, one can safely feed C code containing
1628		"..." to ml-nlffigen.
1629		3. There are some internal improvements to CM, providing slightly
1630		more general string substitutions in the tools subsystem.
1631
1632		----------------------------------------------------------------------
1633		Name: Matthias Blume
1634		Date: 2001/06/27 15:10:00 EDT
1635		Tag: blume-20010627-concur
1636		Description:
1637
1638		Fixed a small bug in CM's handling of parallel compilation.
1639		(You could observe the bug by Control-C-interrupting an ordinary
1640		CMB.make or CM.stabilize and then attaching some compile servers.
1641		The result was that all of a sudden the previously interrupted
1642		compilation would continue on its own.  This was because of
1643		an over-optimization: CM did not bother to clean out certain queues
1644		when no servers were attached "anyway", resulting in the contents
1645		of these queues to grab control when new servers did get attached.)
1646
1647		There is also another minor update to the CM manual.
1648
1649		----------------------------------------------------------------------
1650		Name: Matthias Blume
1651		Date: 2001/06/26 16:15:00 EDT
1652		Tag: blume-20010626-cmdoc
1653		Description:
1654
1655		Minor typo fixed in CM manual (syntax diagram for libraries).
1656
1657		----------------------------------------------------------------------
1658		Name: Matthias Blume
1659		Date: 2001/06/25 22:55:00 EDT
1660		Tag: blume-20010625-x86pc
1661		Description:
1662
1663		Fixed a nasty bug in the X86 assembly code that caused signal
1664		handlers to fail (crash) randomly.
1665
1666		----------------------------------------------------------------------
1667		Name: Matthias Blume
1668		Date: 2001/06/25 12:05:00 EDT
1669		Tag: blume-20010625-nlffigen
1670		Description:
1671
1672		This update fixes a number of minor bugs in ml-nlffigen as reported by
1673		Nick Carter <nbc@andrew.cmu.edu>.
1674
1675		1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
1676		2. Default names for generated files are now derived from the name of
1677		the C file *without its directory*.  In particular, this causes generated
1678		files to be placed locally even if the C file is in some system directory.
1679		3. Default names for generated signatures and structures are also derived
1680		from the C file name without its directory.  This avoids silly things
1681		like "structure GL/GL".
1682		(Other silly names are still possible because ml-nlffigen does not do
1683		a thorough check of whether generated names are legal ML identifiers.
1684		When in doubt, use command line arguments to force particular names.)
1685
1686		----------------------------------------------------------------------
1687		Name: Matthias Blume
1688		Date: 2001/06/21 12:25:00 EDT
1689		Tag: blume-20010621-eXene
1690		Description:
1691
1692		eXene now compiles and (sort of) works again.
1693
1694		The library name (for version > 110.33) is $/eXene.cm.
1695
1696		I also added an new example in src/eXene/examples/nbody.  See the
1697		README file there for details.
1698
1699		----------------------------------------------------------------------
1700		Name: Matthias Blume
1701		Date: 2001/06/20 16:40:00 EDT
1702		Tag: blume-20010620-cml
1703		Description:
1704
1705		CML now compiles and works again.
1706
1707		Libraries (for version > 110.33):
1708
1709		$cml/cml.cm            Main CML library.
1710		$cml/basis.cm          CML's version of $/basis.cm.
1711		$cml/cml-internal.cm   Internal helper library.
1712		$cml/core-cml.cm       Internal helper library.
1713		$cml-lib/trace-cml.cm  Tracing facility.
1714		$cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
1715
1716		The installer (config/install.sh) has been taught how to properly
1717		install this stuff.
1718
1719		----------------------------------------------------------------------
1720		Name: Matthias Blume
1721		Date: 2001/06/19 17:55:00 EDT
1722		Tag: blume-20010619-instantiate
1723		Description:
1724
1725		This un-breaks the fix for bug 1432.
1726		(The bug was originally fixed in 110.9 but I broke it again some
1727		time after that.)
1728
1729		----------------------------------------------------------------------
1730		Name: Matthias Blume
1731		Date: 2001/06/19 17:25:00 EDT
1732		Tag: blume-20010619-signals
1733		Description:
1734
1735		This should (hopefully) fix the long-standing signal handling bug.
1736		(The runtime system was constructing a continuation record with an
1737		incorrect descriptor which would cause the GC to drop data on the floor...)
1738
1739		----------------------------------------------------------------------
1740		Name: Matthias Blume
1741		Date: 2001/06/15 15:05:00 EDT
1742		Tag: blume-20010615-moresparc
1743		Description:
1744
1745		Here is a short late-hour update related to Sparc c-calls:
1746
1747		-- made handling of double-word arguments a bit smarter
1748
1749		-- instruction selection phase tries to collapse certain clumsily
1750		constructed ML-Trees; typical example:
1751
1752		ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
1753
1754		This currently has no further impact on SML/NJ since mlriscGen does
1755		not seem to generate such patterns in the first place, and c-calls
1756		(which did generate them in the beginning) has meanwhile been fixed
1757		so as to avoid them as well.
1758
1759		----------------------------------------------------------------------
1760		Name: Matthias Blume
1761		Date: 2001/06/15 15:05:00 EDT
1762		Tag: blume-20010615-sparc
1763		Description:
1764
1765		The purpose of this update is to provide an implementation of NLFFI
1766		on Sparc machines.
1767
1768		Here are the changes in detail:
1769
1770		* src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
1771		the Sparc implementation of the c-calls API.
1772		* The Sparc backend of SML/NJ has been modified to uniformely use %fp
1773		for accessing the ML frame.  Thus, we have a real frame pointer and
1774		can freely modify %sp without need for an omit-frame-ptr phase.
1775		The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
1776		this case.
1777		* ml-nlffigen has been taught to produce code for different architectures
1778		and calling conventions.
1779		* In a way similar to what was done in the x86 case, the Sparc
1780		backend uses its own specific extension to mltree.  (For example,
1781		it needs to be able to generate UNIMP instructions which are part
1782		of the calling convention.)
1783		* ml-nlffi-lib was reorganized to make it more modular (in particular,
1784		to make it easier to plug in new machine- and os-dependent parts).
1785
1786		There are some other fairly unrelated bug fixes and cleanups as well:
1787
1788		* I further hacked the .cm files for MLRISC tools (like MDLGen) so
1789		that they properly share their libraries with existing SML/NJ libraries.
1790		* I fixed a minor cosmetic bug in CM, supressing certain spurious
1791		follow-up error messages.
1792		* Updates to CM/CMB documentation.
1793
1794		TODO items:
1795
1796		* MLRISC should use a different register as its asmTemp on the Sparc.
1797		(The current %o2 is a really bad choice because it is part of the
1798		calling conventions, so things might interfere in unexpected ways.)
1799
1800		----------------------------------------------------------------------
1801		Name: Matthias Blume
1802		Date: 2001/06/07
1803		Tag: blume-20010607-calls
1804		Description:
1805
1806		A number of internal changes related to C calls and calling conventions:
1807
1808		1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
1809		number of bytes popped implicitly (by the callee).  In most cases
1810		this field is 0 but on x86/win32 it is some non-zero value.  This
1811		is information provided for the benefit of the "omit-frameptr" pass.
1812		2. The CALL instruction on the x86 carries a similar "pops" field.
1813		The instruction selection phase copies its value from the ML-Tree
1814		CALL statement.
1815		3. On all other architectures, the instruction selection phase checks
1816		whether "pops=0" and complains if not.
1817		4. The c-calls implementation for x86 now accepts two calling conventions:
1818		"ccall" and "stdcall".  When "ccall" is selected, the caller cleans
1819		up after the call and pops is set to 0.  For "stdcall", the caller
1820		does nothing, leaving the cleanup to the callee; pops is set to
1821		the number of bytes that were pushed onto the stack.
1822		5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
1823		distinguish between "ccall" and "stdcall".
1824		6. The UNIMP instruction has been added to the supported Sparc instruction
1825		set. (This is needed for implementing the official C calling convention
1826		on this architecture.)
1827		7. I fixed some of the .cm files under src/MLRISC/Tools to make them
1828		work with the latest CM.
1829
1830		----------------------------------------------------------------------
1831		Name: Matthias Blume
1832		Date: 2001/06/05 15:10:00 EDT
1833		Tag: blume-20010605-cm-index
1834		Description:
1835
1836		0. The "lambdasplit" parameter for class "sml" in CM has been documented.
1837
1838		1. CM can now generate "index files".  These are human-readable files
1839		that list on a per-.cm-file basis each toplevel symbol defined or
1840		imported.  The location of the index file for
1841		<p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
1842		To enable index-file generation, set CM.Control.generate_index to true
1843		or export an environment-symbol: export CM_GENERATE_INDEX=true.
1844
1845		The CM manual has been updated accordingly.
1846
1847		2. I made some slight modifications to the c-calls API in MLRISC.
1848
1849		a) There is now a callback to support saving/restoring of
1850		dedicated but caller-save registers around the actual call
1851		instruction.
1852		b) One can optionally specify a comment-annotation for the
1853		call instruction.
1854
1855		3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
1856		(For example, the comment annotation shows the C prototype of
1857		the function being called.)
1858
1859		----------------------------------------------------------------------
1860		Name: Matthias Blume
1861		Date: 2001/06/01 13:30:00 EDT
1862		Tag: blume-20010601-nlffi-cleanup
1863		Description:
1864
1865		This is mostly a cleanup of MLFFI stuff:
1866
1867		- some signature files have been put into a more exposed place
1868		- the ugly 'f type parameter is gone (simplifies types tremendously!)
1869		- ml-nlffigen changed accordingly
1870		- tutorial updated
1871
1872		Other changes:
1873
1874		- author's affiliation in CM manual(s) updated
1875		- some more recognized keywords added to Allen's sml.sty
1876
1877		----------------------------------------------------------------------
1878		Name: Matthias Blume
1879		Date: 2001/05/25 15:30:00 EDT
1880		Tag: blume-20010525-iptr
1881		Description:
1882
1883		- put the official 110.33-README (as it appears on the ftp server) under
1884		CVS
1885		- fixed a small bug related to incomplete pointer types in
1886		ml-nlffigen
1887		- small cosmetic change to the ml-nlffi-lib's "arr" type constructor
1888		(it does not need the 'f type parameter)
1889
1890		----------------------------------------------------------------------
1891		Name: Matthias Blume
1892		Date: 2001/05/23 14:30:00 EDT
1893		Tag: Release_110_33
1894		Description:
1895
1896		New version number (110.33).  New bootfiles.
1897
1898		----------------------------------------------------------------------
1899		Name: Matthias Blume
1900		Date: 2001/05/22 18:06:00 EDT
1901		Tag: blume-20010522-targets
1902		Description:
1903
1904		Made install.sh use file config/targets.customized if it exists, falling
1905		back to config/targets if it doesn't.  This way one can have a customized
1906		version of the targets file without touching the "real thing", thus
1907		eliminating the constant fear of accidentally checking something bogus
1908		back into the CVS repository...  (File config/targets.customized must
1909		not be added to the repository!)
1910
1911		----------------------------------------------------------------------
1912		Name: Matthias Blume
1913		Date: 2001/05/22 16:30:00 EDT
1914		Tag: blume-20010522-minitut
1915		Description:
1916
1917		1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
1918		struct returns.
1919		2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
1920		incomplete, preliminary documentation for NLFFI.
1921
1922		----------------------------------------------------------------------
1923		Name: Matthias Blume
1924		Date: 2001/05/14 11:30:00 EDT
1925		Tag: blume-20010514-script
1926		Description:
1927
1928		Some bugs in install script fixed.
1929
1930		In addition to that I also made a slight change to the NLFFI API:
1931		Functors generated by ml-nlffigen now take the dynamic library as a
1932		straight functor argument, not as a suspended one.  (The original
1933		functor code used to force the suspension right away anyway, so there
1934		was nothing gained by this complication of the interface.)
1935
1936		----------------------------------------------------------------------
1937		Name: Matthias Blume
1938		Date: 2001/05/11 14:35:00 EDT
1939		Tag: blume-20010511-ml-nlffi
1940		Description:
1941
1942		I finally took the plunge and added my new FFI code to the main
1943		repository.  For x86-linux it is now ready for prime-time.
1944
1945		There are two new subdirectories of "src":
1946
1947		- ml-nlffi-lib:
1948		The utility library for programs using the FFI interface.
1949		Here is the implementation of $/c.cm and its associated low-level
1950		partners $/c-int.cm and $/memory.cm.
1951		- ml-nlffigen:
1952		A stand-alone program for generating ML glue code from C source
1953		code.
1954
1955		Building ml-nlffigen requires $/ckit-lib.cm.
1956
1957		The config/install.sh script has been updates to do the Right Thing
1958		(hopefully).
1959
1960		Notice that the source tree for the C-Kit will not be put under "src"
1961		but directly under the installation root directory.  (This is the
1962		structure that currently exists on the CVS server when you check out
1963		module "sml".)  Fortunately, config/install.sh knows about this oddity.
1964
1965		Bugs: No documentation yet.
1966
1967		----------------------------------------------------------------------
1968		Name: Matthias Blume
1969		Date: 2001/05/09 16:35:00 EDT
1970		Tag: blume-20010509-cpscontract
1971		Description:
1972
1973		Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
1974		wrapper/unwrapper elimination did not decrement usage counts and some
1975		dead variables got overlooked by the dead-up logic.)
1976
1977		----------------------------------------------------------------------
1978		Name: Lal George
1979		Date: 2001/05/08  17:26:09 EDT
1980		Tag: george-20010508-omit-frameptr
1981		Description:
1982
1983		Changes to implement the omit-frame-pointer optimization to support
1984		raw C calls. For now, there is only support on the Intel x86, but
1985		other architectures will follow as more experience is gained with this.
1986
1987
1988		----------------------------------------------------------------------
1989		Name: Matthias Blume
1990		Date: 2001/05/07 14:40:00 EDT
1991		Tag: blume-20010507-proxies
1992		Description:
1993
1994		I made into "proxy libraries" all libraries that qualify for such a
1995		change.  (A qualifying library is a library that has another library or
1996		groups as its sole member and repeats that member's export list
1997		verbatim.  A proxy library avoids this repetition by omitting its export
1998		list, effectively inheriting the list that its (only) member exports.
1999		See the CM manual for more explanation.)
2000		The main effect is that explicit export lists for these libraries
2001		do not have to be kepts in sync, making maintenance a bit easier.
2002
2003		I also added copyright notices to many .cm-files.
2004
2005		Last but not least, I made a new set of bootfiles.
2006
2007		----------------------------------------------------------------------
2008		Name: Matthias Blume
2009		Date: 2001/05/04 17:00:00 EDT
2010		Tag: blume-20010504-cm-lsplit
2011		Description:
2012
2013		0. John merged pending changes to $/smlnj-lib.cm
2014
2015		1. Allen's previous change accidentally backed out of one of Lal's
2016		earlier changes.  I undid this mistake (re-introducing Lal's change).
2017
2018		2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
2019		within the compiler where applicable.  There is some code simplification
2020		because of that.
2021
2022		3. The "split" phase (in FLINT) is now part of the default list of phases.
2023		Compiler.Control.LambdaSplitting.* can be used to globally control the
2024		lambda-splitting (cross-module-inlining) engine.  In addition to that,
2025		it can now also be controlled on a per-source basis: CM has been taught
2026		a new tool parameter applicable to ML source files.
2027
2028		- To turn lambda-splitting off completely:
2029		local open Compiler.Control.LambdaSplitting in
2030		val _ = set Off
2031		end
2032		- To make "no lambda-splitting" the global default (but allow per-source
2033		overriding); this is the initial setting:
2034		local open Compiler.Control.LambdaSplitting in
2035		val _ = set (Default NONE)
2036		end
2037		- To make "lambda-splitting with aggressiveness a" the global default
2038		(and allow per-source overriding):
2039		local open Compiler.Control.LambdaSplitting in
2040		val _ = set (Default (SOME a))
2041		end
2042
2043		- To turn lambda-splitting off for a given ML souce file (say: a.sml)
2044		write (in the respective .cm-file):
2045		a.sml (lambdasplitting:off)
2046		- To turn lambda-splitting for a.sml on with minimal aggressiveness:
2047		a.sml (lambdasplitting:on)
2048		- To turn lambda-splitting for a.sml on with aggressiveness <a> (where
2049		<a> is a decimal non-negative integer):
2050		a.sml (lambdasplitting:<a>)
2051		- To turn lambda-splitting for a.sml on with maximal aggressiveness:
2052		a.sml (lambdasplitting:infinity)
2053		- To use the global default for a.sml:
2054		a.sml (lambdasplitting:default)
2055		or simply
2056		a.sml
2057
2058		----------------------------------------------------------------------
2059		Name: Allen Leung
2060		Date: 2001/05/04 01:57:00 EDT
2061		Tag: leunga-20010504-sync
2062		Description:
2063
2064		MLRISC features.
2065
2066		1. Fix to CMPXCHG instructions.
2067		2. Changed RA interface to allow annotations in callbacks.
2068		3. Added a new method to the stream interface to allow annotations updates.
2069
2070		----------------------------------------------------------------------
2071		Name: Matthias Blume
2072		Date: 2001/05/01 11:45:00 EDT
2073		Tag: blume-20010501-pcedittmp
2074		Description:
2075
2076		Changed install.sh to use the current working directory instead of
2077		/usr/tmp for a temporary file (pcedittmp).  The previous choice
2078		of /usr/tmp caused trouble with MacOS X because of file premission
2079		problems.
2080
2081		----------------------------------------------------------------------
2082		Name: Matthias Blume
2083		Date: 2001/04/20 11:10:00 EDT
2084		Tag: blume-20010420-inMLflag
2085		Description:
2086
2087		- added vp_limitPtrMask to vproc-state.h
2088		(for use by the raw-C-calls mechanism to implement proper interrupt
2089		handling)
2090		- made the ML compiler aware of various data-structure offsets so it
2091		can generate code for accessing the vp_inML flag and vp_limitPtrMask
2092		- tweaked mlriscGen.sml to have it emit interrupt-handling code for
2093		raw C-calls
2094
2095		----------------------------------------------------------------------
2096		Name: Lal George
2097		Date: 2001/04/20 09:15:28 EDT
2098		Tag: george-20010420-macosX
2099		Description:
2100
2101		- Changes to port to Mac OS X; Darwin.
2102
2103		- In the process I found that sqrt was broken on the PPC, because the
2104		fsqrt instruction is not implemented.
2105
2106		----------------------------------------------------------------------
2107		Name: Matthias Blume
2108		Date: 2001/04/18 12:45:00 EDT
2109		Tag: blume-20010418-ccalls
2110		Description:
2111
2112		- fixed two off-by-4 errors in the x86-specific c-calls implementation
2113		(this bug prevented structure arguments containing pointers from being
2114		passed correctly)
2115		- changed the raw-C-call code in mlriscGen.sml in such a way that
2116		structure arguments are represented as a pointer to the beginning
2117		of the structure (instead of having a series of synthesized arguments,
2118		one for each structure member)
2119
2120		- made makeml script's verbosity level configurable via environment
2121		variable (MAKEML_VERBOSITY)
2122
2123		- eliminated placeholder implementations for f32l, w16s, i16s, and f32s
2124		in rawmem-x86.sml; we are now using the real thing
2125
2126		----------------------------------------------------------------------
2127		Name: Matthias Blume
2128		Date: 2001/03/22 16:25:00 EST
2129		Tag: blume-20010322-bootfiles
2130		Description:
2131
2132		Created a new set of bootfiles (for your automatic installation convenience).
2133
2134		----------------------------------------------------------------------
2135		Name: Matthias Blume
2136		Date: 2001/03/22 15:10:00 EST
2137		Tag: blume-20010322-rawmem-parcm
2138		Description:
2139
2140		1. All "raw memory access" primitives for the new FFI are implemented now
2141		(at least on the x86).
2142		2. Some further cleanup of CM's parallel make mechanism.
2143
2144		----------------------------------------------------------------------
2145		Name: Matthias Blume
2146		Date: 2001/03/19 17:53:00 EST
2147		Tag: blume-20010319-parallel
2148		Description:
2149
2150		Parallel make (using compile servers) now works again.
2151
2152		To this end, CM.stabilize and CMB.make have been modified to work in
2153		two passes when compile servers are attached:
2154		1. Compile everything, do not perform stabilization; this pass
2155		uses compile servers
2156		2. Stabilize everything; this pass does not use compile servers
2157		If there are no compile servers, the two passes are combined into one
2158		(as before).  Splitting the passes increases the inherent parallelism
2159		in the dependency graph because the entire graph including all
2160		libraries is available at the same time.  This, in turn, improves
2161		server utilization.  The downside is that the master process will
2162		have to do some extra work after compilation is done (because for
2163		technical reasons it must re-read all the binfiles during stabilization).
2164
2165		----------------------------------------------------------------------
2166		Name: Matthias Blume
2167		Date: 2001/03/16 12:22:00 EST
2168		Tag: blume-20010316-bootfiles
2169		Description:
2170
2171		Created a new set of bootfiles (for your automatic installation convenience).
2172
2173		----------------------------------------------------------------------
2174		Name: Matthias Blume
2175		Date: 2001/03/16 11:00:00 EST
2176		Tag: blume-20010316-MLTREE-fixup
2177		Description:
2178
2179		This is a minor fixup for an (untagged) earlier commit by Allen.
2180		(A file was missing).
2181
2182		----------------------------------------------------------------------
2183		Name: Allen Leung
2184		Date: Mon Mar  5 18:54:57 EST 2001
2185		Tag: leunga-20010305-cut-support
2186
2187		1. New support for alternative control-flow in MLTREE.
2188		Currently we support
2189
2190		FLOW_TO(CALL ...., [k1,...,kn])
2191
2192		This is needed for 'cuts to' in C-- and try/handle-like constructs
2193		in Moby
2194
2195		New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
2196
2197		2. Register Allocator
2198
2199		Changes in interface [from Fermin, John]
2200
2201		3. Alpha 8-bit SLL support [Fermin]
2202
2203		4. All architectures
2204
2205		A new module (ClusterExpandCopies) for expanding parallel copies.
2206
2207		----------------------------------------------------------------------
2208		Name: Allen Leung
2209		Date: 2001/02/27 23:07:00 EST
2210		Tag: leunga-20010227-minor-stuff
2211
2212		1. Alpha bug fix for CMOVNE
2213		2. Handle mltree COND(..,FCMP ...,...)
2214		3. Bug fix in simplifier
2215
2216		----------------------------------------------------------------------
2217		Name: Matthias Blume
2218		Date: 2001/01/30 17:50:00 EST
2219		Tag: blume-20010130-sync
2220		Description:
2221
2222		This is just a minor update to sync my devel branch with the main brach.
2223		The only visible change is the addition of some README files.
2224
2225		----------------------------------------------------------------------
2226		Name: Matthias Blume
2227		Date: 2001/01/12 23:30:00 JST
2228		Tag: blume-20010112-bootfiles
2229		Description:
2230
2231		Made a new set of bootfiles that goes with the current state of the
2232		repository.
2233
2234		----------------------------------------------------------------------
2235		Name: Matthias Blume
2236		Date: 2001/01/12 21:20:00 JST
2237		Tag: blume-20010112-sync
2238		Description:
2239
2240		I am just flushing out some minor changes that had accumulated in
2241		my private branch in order to sync with the main tree.  (This is
2242		mainly because I had CVS trouble when trying to merge _into_ my
2243		private branch.)
2244
2245		Most people should be completely unaffected by this.
2246
2247		----------------------------------------------------------------------
2248		Name: Allen Leung
2249		Date: Thu Jan 11 21:03:00 EST 2001
2250		Tag: leunga-20010111-labexp=mltree
2251		Description:
2252
2253		1.  Removed the type LabelExp and replace it by MLTree.
2254		2.  Rewritten mltree-simplify with the pattern matcher tool.
2255		3.  There were some bugs in alpha code generator which would break
2256		64-bit code generation.
2257		4.  Redo the tools to generate code with the
2258		5.  The CM files in MLRISC (and in src/system/smlnj/MLRISC)
2259		are now generated by perl scripts.
2260
2261		----------------------------------------------------------------------
2262		Name: Matthias Blume
2263		Date: 2001/01/10 21:55:00 JST
2264		Tag: blume-20010110-rcc
2265		Description:
2266
2267		The RCC stuff now seems to work (but only on the x86).
2268		This required hacking of the c-calls interface (and -implementation) in
2269		MLRISC.
2270
2271		Normal compiler users should be unaffected.
2272
2273		----------------------------------------------------------------------
2274		Name: Matthias Blume
2275		Date: 2001/01/09 01:20:00 JST
2276		Tag: blume-20010109-rcc
2277		Description:
2278
2279		This is a fairly big patch, flushing out a large number of pending
2280		changes that I made to my development copy over the last couple of days.
2281
2282		Of practical relevance at this moment is a workaround for a pickling
2283		bug that Allen ran into the other day.  The cause of the bug itself is
2284		still unknown and it might be hard to fix it properly, but the
2285		workaround has some merits of its own (namely somewhat reducing pickling
2286		overhead for certain libraries).  Therefore, I think this solution should
2287		be satisfactory at this time.
2288
2289		The rest of the changes (i.e., the vast majority) has to do with my
2290		ongoing efforts of providing direct support for C function calls from
2291		ML.  At the moment there is a new primop "RAW_CCALL", typing magic
2292		in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new
2293		case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to
2294		translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to
2295		handle RCC.
2296
2297		The last part (the changes to mlriscGen.sml) are still known to be
2298		wrong on the x86 and not implemented on all other architectures.  But
2299		the infrastructure is in place. I had to change a few functor
2300		signatures in the backend to be able to route the CCalls interface
2301		from MLRISC there, and I had to specialize the mltree type (on the
2302		x86) to include the necessary extensions. (The extensions themselves
2303		were already there and redy to go in MLRISC/x86).
2304
2305		Everything should be very happy as soon as someone helps me with
2306		mlriscGen.sml...
2307
2308		In any case, nothing of this should matter to anyone as long as the
2309		new primop is not being used (which is going to be the case unless you
2310		find it where I hid it :). The rest of the compiler is completely
2311		unaffected.
2312
2313		----------------------------------------------------------------------
2314		Name: Matthias Blume
2315		Date: 2001/01/05 00:30:00 JST
2316		Tag: blume-20010105-primops
2317		Description:
2318
2319		Added some experimental support for work that I am doing right now.
2320		These changes mostly concern added primops, but there is also a new
2321		experimental C library in the runtime system (but currently not enabled
2322		anywhere except on Linux/X86).
2323
2324		In the course of adding primops (and playing with them), I discovered that
2325		Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly
2326		broken.  (Zhong was very right he labeled this stuff as "major gross hack".)
2327		To recover, I made type information in INL_PRIM mandatory and changed
2328		prim.sml as well as built-in.sml accordingly.  The InLine structure now
2329		has complete, correct type information (i.e., no bottom types).
2330
2331		Since all these changes mean that we need new binfiles, I also bumped the
2332		version number to 110.32.1.
2333
2334		----------------------------------------------------------------------
2335		Name: Matthias Blume
2336		Date: 2000/12/30 22:10:00 JST
2337		Tag: blume-20001230-various
2338		Description:
2339
2340		Added proxy libraries for MLRISC and let MLRISC libraries refer
2341		to each other using path anchors.  (See CM manual for explanation.)
2342
2343		Updated CM documentation.
2344
2345		Fixed some bugs in CM.
2346
2347		Implemented "proxy" libraries (= syntactic sugar for CM).
2348
2349		Added "-quiet" option to makeml and changed runtime system accordingly.
2350
2351		Added cleanup handler for exportML to reset timers and compiler stats.
2352
2353		----------------------------------------------------------------------
2354		Name: Lal George
2355		Date: 2000/12/22 22:22:58 EST 2000
2356		Tag: Release_110_32
2357		Description:
2358
2359		Infinite precision used throughout MLRISC.
2360		see MLRISC/mltree/machine-int.sig
2361
2362		----------------------------------------------------------------------
2363		Name: Matthias Blume
2364		Date: 2000/12/22 23:16:00 JST
2365		Tag: blume-20001222-warn
2366		Description:
2367
2368		Corrected wording and formatting of some CM warning message which I
2369		broke in my previous patch.
2370
2371		----------------------------------------------------------------------
2372		Name: Matthias Blume
2373		Date: 2000/12/22 21:20:00 JST
2374		Tag: blume-20001222-anchorenv
2375		Description:
2376
2377		Fixed CM's handling of anchor environments in connection with CMB.make.
2378
2379		----------------------------------------------------------------------
2380		Name: Matthias Blume
2381		Date: 2000/12/22 13:15:00 JST
2382		Tag: blume-20001222-cleanup
2383		Description:
2384
2385		Removed src/cm/ffi which does not (and did not) belong here.
2386
2387		----------------------------------------------------------------------
2388		Name: Matthias Blume
2389		Date: 2000/12/21 23:55:00 JST
2390		Tag: blume-20001221-exn
2391		Description:
2392
2393		Probably most important: CM no longer silently swallows all exceptions
2394		in the compiler.
2395		Plus: some other minor CM changes.  For example, CM now reports some
2396		sizes for generated binfiles (code, data, envpickle, lambdapickle).
2397
2398		----------------------------------------------------------------------
2399		Name: Matthias Blume
2400		Date: 2000/12/15 00:01:05 JST
2401		Tag: blume-20001215-dirtool
2402		Description:
2403
2404		- "dir" tool added.
2405		- improvements and cleanup to Tools structure
2406		- documentation updates
2407
2408		----------------------------------------------------------------------
2409		Name: Allen Leung
2410		Date: Thu Dec 14 03:45:24 EST 2000
2411		Description:
2412		Tag:  leunga-20001214-int-inf
2413		Description:
2414
2415		In IntInf, added these standard functions, which are missing from our
2416		implementation:
2417
2418		andb : int * int -> int
2419		xorb : int * int -> int
2420		orb  : int * int -> int
2421		notb : int -> int
2422		<<   : int * word -> int
2423		~>>  : int * word -> int
2424
2425		Not tested, I hope they are correct.
2426
2427		----------------------------------------------------------------------
2428		Name: Allen Leung
2429		Date: Fri Dec  8 19:23:26 EST 2000
2430		Description:
2431		Tag:  leunga-20001208-nowhere
2432		Description:
2433
2434		Slight improvements to the 'nowhere' tool to handle OR-patterns,
2435		to generate better error messages etc.  Plus a brief manual.
2436
2437		----------------------------------------------------------------------
2438		Name: Lal George
2439		Date: 2000/12/08 09:54:02 EST 2000
2440		Tag: Release_110_31
2441		Description:
2442
2443		- Version 110.31
2444		----------------------------------------------------------------------
2445		Name: Allen Leung
2446		Date: Thu Dec  7 22:01:04 EST 2000
2447		Tag:  leunga-20001207-cell-monster-hack
2448		Description:
2449
2450		Major MLRISC internal changes.  Affect all clients.
2451		Summary:
2452
2453		1.  Type CELLS.cell = int is now replaced by a datatype.
2454		As a result, the old regmap is now gone.  Almost all interfaces
2455		in MLRISC change as a consequence.
2456
2457		2.  A new brand version of machine description tool (v3.0) that generates
2458		modules expecting the new interface.  The old version is removed.
2459
2460		3.  The RA interface has been further abstracted into two new functors.
2461		RISC_RA and X86RA.  These functors have much simpler interfaces.
2462		[See also directory MLRISC/demo.]
2463
2464		4.  Some other new source->source code generation tools are available:
2465
2466		a. MLRISC/Tools/RewriteGen -- generate rewriters from rules.
2467		b. MLRISC/Tools/WhereGen -- expands conditional pattern matching rules.
2468		I use this tool to generate the peephole optimizers---with the new
2469		cell type changes, peephole rules are becoming difficult to write
2470		without conditional pattern matching.
2471
2472		5.  More Intmap -> IntHashTable change.  Previous changes by Matthias didn't
2473		cover the entire MLRISC source tree so many things broke.
2474
2475		6.  CM files have been moved to the subdirectory MLRISC/cm.
2476		They are moved because there are a lot of them and they clutter up the
2477		root dir.
2478
2479		7.  More detailed documentation to come...
2480
2481		NOTE: To rebuild from 110.30 (ftp distribution), you'll have to do
2482		a makeml -rebuild first.  This is because of other other
2483		changes that Matthias has made (see below).
2484
2485
2486		----------------------------------------------------------------------
2487		Name: Matthias Blume
2488		Date: 2000/11/30 23:12:00 JST
2489		Tag: blume-20001130-filereorg
2490		Description:
2491
2492		Some manual updates and some file reorganizations in CM.
2493
2494		----------------------------------------------------------------------
2495		Name: Matthias Blume
2496		Date: 2000/11/24 17:45:00 JST
2497		Tag: blume-20001124-link
2498		Description:
2499
2500		Drastically improved link traversal code for the case that the dynamic
2501		value was already loaded at bootstrap time.  As a result, CM and CMB
2502		now both load blazingly fast -- even on a very slow machine.  Also,
2503		memory consumption has been further reduced by this.
2504
2505		Warning: The format of the PIDMAP file has changed.  THerefore, to
2506		bootstrap you have to do this:
2507
2508		1. Run CMB.make
2509		2. Make a symbolic link for the boot directory:
2510		ln -s sml.boot.ARCH-OS xxx
2511		3. "Rebuild" the boot directory:
2512		./makeml -boot xxx -rebuild sml ; rm xxx
2513		4. Boot normally:
2514		./makeml
2515
2516		----------------------------------------------------------------------
2517		Name: Matthias Blume
2518		Date: 2000/11/21 21:20:00 JST
2519		Tag: blume-20001121-tools
2520		Description:
2521
2522		Continued hacking on autoloading problem -- with success this time.
2523		Also changed tool-plugin mechanism.  See new CM manual.
2524
2525		----------------------------------------------------------------------
2526		Name: Matthias Blume
2527		Date: 2000/11/19 14:30:00 JST
2528		Tag:  blume-20001119-autoload
2529		Description:
2530
2531		Some hacking to make autoloading faster.  Success for CMB, no success
2532		so far for CM.  There is a reduced structure CM' that autoloads faster.
2533		(This is a temporary, non-documented hack to be eliminated again when
2534		the general problem is solved.)
2535
2536		----------------------------------------------------------------------
2537		Name: Matthias Blume
2538		Date: 2000/11/17 14:10:00 JST
2539		Tag: blume-20001117-pickle-lib
2540		Description:
2541
2542		1. Eliminated comp-lib.cm
2543		2. Made pickle-lib.cm
2544		3. Eliminated all uses of intset.sml (from comp-lib.cm)
2545		4. Replaced all uses of intmap.{sig,sml} (from comp-lib.cm) with
2546		equivalent constructs from smlnj-lib.cm (INtHashTable).
2547		5. Point 4. also goes for those uses of intmap.* in MLRISC.
2548		Duplicated intmap modules thrown out.
2549		6. Hunted down all duplicated SCC code and replaced it with
2550		equivalent stuff (GraphSCCFn from smlnj-lib.cm).
2551		7. Rewrote Feedback module.
2552		8. Moved sortedlist.sml into viscomp-lib.cm.  Eventually it
2553		should be thrown out and equivalent modules from smlnj-lib.cm
2554		should be used (IntRedBlackSet, IntListSet, ...).
2555
2556		Confirmed that compiler compiles to fixpoint.
2557
2558		----------------------------------------------------------------------
2559		Name: Allen Leung
2560		Date: 2000/11/10 18:00:00
2561		Tag: leunga-20001110-new-x86-fp
2562
2563		A new x86 floating point code generator has been added.
2564		By default this is turned off.  To turn this on, do:
2565
2566		CM.autoload "$smlnj/compiler.cm";
2567		Compiler.Control.MLRISC.getFlag "x86-fast-fp" := true;
2568
2569		Changes:
2570
2571		1.  Changed FTAN to FPTAN so that the assembly output is correct.
2572		2.  Changed the extension callback for FTANGENT to generate:
2573
2574		fptan
2575		fstp  %st(0)
2576		instead of
2577		fptan
2578		fstpl ftempmem
2579
2580		3.  Numerous assembly fixes for x86.
2581
2582		5.  Cleaned up the machine code output module x86/x86MC.sml and added
2583		support for a whole bunch of instructions and addressing modes:
2584
2585		fadd/fsub/fsubr/fmul/fdiv/fdivr  %st, %st(n)
2586		faddp/fsubp/fsubrp/fmulp/fdivp/fdivrp  %st, %st(n)
2587		fadd/fsub/fsubr/fmul/fdiv/fdivr  %st(n), %st
2588		fiadd/fisub/fisubr/fimul/fidiv/fidivr mem
2589		fxch %st(n)
2590		fld %st(n)
2591		fst %st(n)
2592		fst mem
2593		fstp %st(n)
2594		fucom %st(n)
2595		fucomp %st(n)
2596
2597		All these are now generated when the fast fp mode is turned on.
2598
2599		6.  Removed the dedicated registers %st(0), ..., %st(7) from X86CpsRegs
2600
2601		----------------------------------------------------------------------
2602		Name: Matthias Blume
2603		Date: 2000/11/09 11:20:00 JST
2604		Tag: blume-20001109-scc
2605		Description:
2606
2607		Eliminated some code duplication:
2608
2609		1. Added "where" clause to GraphSCCFn in SML/NJ Library.
2610		(Otherwise the functor is useless.)
2611		2. Used GraphSCCFn where SCCUtilFun was used previously.
2612		3. Got rid of SCCUtilFun (in comp-lib.cm).
2613
2614		----------------------------------------------------------------------
2615		Name: Lal George
2616		Date: 2000/11/06 09:02:21 EST 2000
2617		Tag: Release_110_30
2618		Description:
2619
2620		- Version 110.30
2621		----------------------------------------------------------------------
2622		Name: Matthias Blume
2623		Date: 2000/11/04 14:45:00
2624		Tag: blume-20001104-mlbuild
2625		Description:
2626
2627		- Made ml-build faster on startup.
2628		- Documentation fixes.
2629
2630		----------------------------------------------------------------------
2631		Name: Matthias Blume
2632		Date: 2000/11/02 17:00:00 JST
2633		Tag: blume-20001102-condcomp
2634		Description:
2635
2636		- Small tweaks to pickler -- new BOOTFILES!
2637		- Version bumped to 110.29.2.
2638		- Added conditional compilation facility to init.cmi (see comment there).
2639		----------------------------------------------------------------------
2640		Name: Allen Leung
2641		Date: 2000/10/23 19:31:00
2642		Tag: leunga-20001023-demo-ra
2643
2644		1. Minor RA changes that improves spilling on x86 (affects Moby and C-- only)
2645		2. Test programs for the graph library updated
2646		3. Some new MLRISC demo programs added
2647
2648		----------------------------------------------------------------------
2649		Name: Matthias Blume
2650		Date: 2000/08/31 22:15:00 JST
2651		Tag: blume-20001017-errmsg
2652		Description:
2653
2654		More error message grief: Where there used to be no messages, there
2655		now were some that had bogus error regions.  Fixed.
2656
2657		----------------------------------------------------------------------
2658		Name: Matthias Blume
2659		Date: 2000/08/31 17:30:00 JST
2660		Tag: blume-20001017-v110p29p1
2661		Description:
2662
2663		I made a version 110.29.1 with new bootfiles.
2664
2665		Changes:  Modified pickler/unpickler for faster and leaner unpickling.
2666		CM documentation changes and a small bugfix in CM's error reporting.
2667
2668		----------------------------------------------------------------------
2669		Name: Lal George
2670		Date: 2000/09/27 14:42:35 EDT
2671		Tag: george-20000927-nodestatus
2672		Description:
2673
2674		Changed the type of the nodestatus, so that:
2675
2676		SPILLED(~1)             is now SPILLED
2677		SPILLED(m) where m>=0   is now MEMREG(m)
2678		SPILLED(s) where s<~1   is now SPILL_LOC(~s)
2679
2680		----------------------------------------------------------------------
2681		Name: Matthias Blume
2682		Date: 2000/09/07 14:45:00 JST
2683		Tag: blume-20000907-cmerrmsg
2684		Description:
2685
2686		Small tweak to CM to avoid getting ML syntax error messages twice.
2687
2688		----------------------------------------------------------------------
2689		Name: Matthias Blume
2690		Date: 2000/08/31 18:00:00 JST
2691		Tag: blume-20000831-cvsbootfiles
2692		Description:
2693
2694		New URL for boot files (because the 110.29 files on the BL server do
2695		now work correctly with my updated install scripts for yacc and lex).
2696
2697		----------------------------------------------------------------------
2698		Name: Matthias Blume
2699		Date: 2000/08/08 12:33:00 JST
2700		Tag: blume-20000808-manual
2701		Description:
2702
2703		Tiny update to CM manual.
2704
2705		----------------------------------------------------------------------
2706		Name: Allen Leung
2707		Date: 2000/08/7 19:31:00
2708		Tag: leunga-20000807-a-whole-bunch-of-stuff
2709
2710		Moby, C--, SSA, x86, machine descriptions etc.  Should only affect C--
2711		and Mobdy.
2712
2713		1.  x86
2714
2715		a.  Fixes to peephole module by John and Dan.
2716		b.  Assembly fix to SETcc by Allen.
2717		c.  Fix to c-call by John.
2718		d.  Fix to spilling by John.  (This one deals with the missing FSTPT case)
2719		e.  Instruction selection optimization to SETcc as suggested by John.
2720
2721		For example,
2722
2723		MV(32, x, COND(32, CMP(32, LT, a, b), LI 1, LI 0))
2724
2725		should generate:
2726
2727		MOVL a, x
2728		SUBL b, x
2729		SHRL 31, x
2730
2731		2.  IR stuff
2732
2733		A bunch of new DJ-graph related algorithms added.  These
2734		speed up SSA construction.
2735
2736		3.  SSA + Scheduling
2737
2738		Added code for SSA and scheduling to the repository
2739
2740		----------------------------------------------------------------------
2741		Name: Lal George
2742		Date: 2000/07/27 11:53:14 EDT
2743
2744		Tag: lal-20000727-linux-ppc
2745		Description:
2746
2747		Made changes to support Linux PPC.
2748		p.s. I have confirmation that the 110.29 boot files work fine.
2749
2750		----------------------------------------------------------------------
2751		Name: Matthias Blume
2752		Date: 2000/07/27 17:40:00 JST
2753		Tag: blume-20000727-scripts
2754		Description:
2755
2756		!!!! WARNING !!!!
2757		You must recompile the runtime system!
2758		!!!! WARNING !!!!
2759
2760		This is basically another round of script-enhancements:
2761
2762		1. sml, ml-build, and ml-makedepend accept options -D and -U to define
2763		and undefine CM preprocessor symbols.
2764
2765		2. ml-build avoids generating a new heap image if it finds that the
2766		existing one is still ok.  (The condition is that no ML file had to
2767		be recompiled and all ML files are found to be older that the heap
2768		file.)
2769
2770		To make this work smoothly, I also hacked the runtime system as
2771		well as SMLofNJ.SysInfo to get access to the heap image suffix
2772		(.sparc-solaris, ...) that is currently being used.
2773
2774		Moreover, the signature of CM.mk_standalone has changed.  See the
2775		CM manual.
2776
2777		3. ml-makedepend accepts additional options -n, -a, and -o.  (See the
2778		CM manual for details.)
2779
2780		4. More CM manual updates:
2781		- all of the above has been documented.
2782		- there is now a section describing the (CM-related) command line
2783		arguments that are accepted by the "sml" command
2784
2785		----------------------------------------------------------------------
2786		Name: Matthias Blume
2787		Date: 2000/07/25 16:20:00 JST
2788		Tag: blume-20000725-makedepend
2789		Description:
2790
2791		Added a script called ml-makedepend.  This can be used in makefiles
2792		for Unix' make in a way very similar to the "makedepend" command for
2793		C.
2794
2795		The script internally uses function CM.sources.
2796
2797		Synopsis:
2798
2799		ml-makedepend [-f makefile] cmfile targetname
2800
2801		The default for the makefile is "makefile" (or "Makefile" should
2802		"makefile" not exist).
2803
2804		ml-makedepend adds a cmfile/targetname-specific section to this
2805		makefile (after removing the previous version of this section).  The
2806		section contains a single dependency specification with targetname on
2807		the LHS (targetname is an arbitrary name), and a list of files derived
2808		from the cmfile on the RHS.  Some of the files on the RHS are
2809		ARCH/OPSYS-specific.  Therefore, ml-makedepend inserts references to
2810		"make" variables $(ARCH) and $(OPSYS) in place of the corresponding
2811		path names.  The makefile writer is responsible for making sure that
2812		these variables have correct at the time "make" is invoked.
2813
2814		----------------------------------------------------------------------
2815		Name: Matthias Blume
2816		Date: 2000/07/22 23:30:00 JST
2817		Tag: blume-20000722-urlupdate
2818		Description:
2819
2820		Changed BOOT and config/srcarchiveurl to point to BL server:
2821
2822		ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.29/
2823
2824		----------------------------------------------------------------------
2825		Name: Matthias Blume
2826		Date: 2000/07/18 18:00:00 JST
2827		Tag: blume-20000718-Version_110_29
2828		Description:
2829
2830		1. Updated src/compiler/TopLevel/main/version.sml to version 110.29
2831
2832		2. Updated config/version to 110.29
2833
2834		3. Updated config/srcarchiveurl
2835
2836		3. New boot files!
2837		ftp://ftp.cs.princeton.edu/pub/people/blume/sml/110.29-autofetch
2838
2839		----------------------------------------------------------------------
2840		Name: Matthias Blume
2841		Date: 2000/07/11 13:58:00 JST
2842		Tag: blume-20000711-doctypo
2843		Description:
2844
2845		Fixed a few typos in CM manual.
2846
2847		----------------------------------------------------------------------
2848		Name: Allen Leung
2849		Date: 2000/06/15 00:38:00
2850		Tag: leunga-20000704-sparc-x86
2851
2852		1. x86 peephole improvement sp += k; sp -= k => nop  [from John]
2853		2. fix to x86 RET bug [found by Dan Grossman]
2854		3. sparc assembly bug fix for ticc instructions [found by Fermin]
2855
2856		Affects c-- and moby only
2857
2858		----------------------------------------------------------------------
2859		Name: Matthias Blume
2860		Date: 2000/07/04 15:26:00
2861		Tag: blume-20000704-trigger
2862		Description:
2863
2864		1. Improvements to CM manual.
2865		2. SMLofNJ.Internals.BTrace.trigger reinstated as an alternative way
2866		of getting a back-trace.  The function, when called, raises an
2867		internal exception which explicitly carries the full back-trace history,
2868		so it is unaffected by any intervening handle-raise pairs ("trivial"
2869		or not).  The interactive loop will print that history once it arrives
2870		at top level.
2871		Short of having all exceptions implicitly carry the full history, the
2872		recommended way of using this facility is:
2873		- compile your program with instrumentation "on"
2874		- run it, when it raises an exception, look at the history
2875		- if the history is "cut off" because of some handler, go and modify
2876		your program so that it explicitly calls BTrace.trigger
2877		- recompile (still instrumented), and rerun; look at the full history
2878
2879		----------------------------------------------------------------------
2880		Name: Matthias Blume
2881		Date: 2000/07/03 15:36:00 JST
2882		Tag: blume-20000702-manual
2883		Description:
2884
2885		Small corrections and updates to CM manual.
2886
2887		----------------------------------------------------------------------
2888		Name: Matthias Blume
2889		Date: 2000/06/29 16:04:00 JST
2890		Tag: blume-20000629-yacctool
2891		Description:
2892
2893		Changes:
2894
2895		1. Class "mlyacc" now takes separate arguments to pass options to
2896		generated .sml- and .sig-files independently.
2897		2. Corresponding CM manual updates.
2898		3. BTrace module now also reports call sites.  (However, for loop clusters
2899		it only shows from where the cluster was entered.)  There are associated
2900		modifications to core.sml, internals.{sig,sml}, btrace.sml, and btimp.sml.
2901
2902		----------------------------------------------------------------------
2903		Name: Matthias Blume
2904		Date: 2000/06/27 16:51:00 JST
2905		Tag: blume-20000627-noweb
2906		Description:
2907
2908		Changes:
2909
2910		1. Implemented "subdir" and "witness" options for noweb tool.
2911		This caused some slight internal changes in CM's tool implementation.
2912		2. Fixed bug in "tool plugin" mechanism.  This is essentially cleaning
2913		some remaining issues from earlier path anchor changes.
2914		3. Updated CM manual accordingly.
2915
2916		4. Changed implementation of back-tracing so that I now consider it
2917		ready for prime-time.
2918
2919		In particular, you don't have to explicitly trigger the back-trace
2920		anymore.  Instead, if you are running BTrace-instrumented code and
2921		there is an uncaught exception (regardless of whether or not it was
2922		raised in instrumented code), the top-level evalloop will print
2923		the back-trace.
2924
2925		Features:
2926
2927		- Instrumented and uninstrumented code work together seemlessly.
2928		(Of course, uninstrumented code is never mentioned in actual
2929		back-traces.)
2930
2931		- Asymptotic time- and space-complexity of instrumented code is
2932		equal to that of uninstrumented code.  (This means that
2933		tail-recursion is preserved by the instrumentation phase.)
2934
2935		- Modules whose code has been instrumented in different sessions
2936		work together without problem.
2937
2938		- There is no penalty whatsoever on uninstrumented code.
2939
2940		- There is no penalty on "raise" expressions, even in
2941		instrumented code.
2942
2943		A potential bug (or perhaps it is a feature, too):
2944
2945		A back-trace reaches no further than the outermost instrumented
2946		non-trivial "raise".  Here, a "trivial" raise is one that is the
2947		sole RHS of a "handle" rule.  Thus, back-traces reach trough
2948
2949		<exp> handle e => raise e
2950
2951		and even
2952
2953		<exp> handle Foo => raise Bar
2954
2955		and, of course, through
2956
2957		<exp> handle Foo => ...
2958
2959		if the exception was not Foo.
2960
2961		Back-traces always reach right through any un-instrumented code
2962		including any of its "handle" expressions, trivial or not.
2963
2964		To try this out, do the following:
2965
2966		- Erase all existing binfiles for your program.
2967		(You may keep binfiles for those modules where you think you
2968		definitely don't need back-tracing.)
2969		- Turn on back-trace instrumentation:
2970		SMLofNJ.Internals.BTrace.mode (SOME true);
2971		- Recompile your program.  (I.e., run "CM.make" or "use".)
2972		- You may now turn instrumentation off again (if you want):
2973		SMLofNJ.Internals.BTrace.mode (SOME false);
2974		- Run your program as usual.  If it raises an exception that
2975		reaches the interactive toplevel, then a back-trace will
2976		automatically be printed.  After that, the toplevel loop
2977		will print the exception history as usual.
2978
2979		----------------------------------------------------------------------
2980		Name: Matthias Blume
2981		Date: 2000/06/26 09:56:46 JST
2982		Tag: blume-20000626-setup
2983		Description:
2984
2985		CM: - setup-parameter to "sml" added; this can be used to run arbitrary
2986		ML code before and after compiling a file (e.g., to set compiler
2987		flags)
2988
2989		Compiler: - improved btrace API (in core.sml, internals.{sig,sml})
2990		- associated changes to btrace.sml (BTrace instrumentation pass)
2991		- cleaner implementation of btimp.sml (BTrace tracing and report
2992		module)
2993
2994		CM manual: * new path encoding documented
2995		* description of setup-parameter to "sml" added
2996
2997		The biggest user-visible change to back-tracing is that it is no
2998		longer necessary to compile all traced modules within the same
2999		session.  (This was a real limitation.)
3000
3001		----------------------------------------------------------------------
3002		Name: Matthias Blume
3003		Date: 2000/06/24 12:40:00 JST
3004		Tag: blume-20000624-startup
3005		Description:
3006
3007		Fixes startup slowdown problem.  (I was calling SrcPath.sync a _tad_
3008		bit too often -- to put it mildly. :)
3009
3010		----------------------------------------------------------------------
3011		Name: Matthias Blume
3012		Date: 2000/06/23 18:20:00 JST
3013		Tag: blume-20000623-btrace
3014		Description:
3015
3016		This updates adds a backtrace facility to aid programmers in debugging
3017		their programs.  This involves the following changes:
3018
3019		1. Module system/smlnj/init/core.sml (structure _Core) now has hooks for
3020		keeping track of the current call stack.  When programs are compiled
3021		in a special mode, the compiler will insert calls to these hooks
3022		into the user program.
3023		"Hook" means that it is possible for different implementations of
3024		back-tracing to register themselves (at different times).
3025
3026		2. compiler/MiscUtil/profile/btrace.sml implements the annotation phase
3027		as an Absyn.dec->Absyn.dec rewrite.  Normally this phase is turned off.
3028		It can be turned on using this call:
3029		SMLofNJ.Internals.BTrace.mode (SOME true);
3030		Turning it off again:
3031		SMLofNJ.Internals.BTrace.mode (SOME false);
3032		Querying the current status:
3033		SMLofNJ.Internals.BTrace.mode NONE;
3034		Annotated programs are about twice as big as normal ones, and they
3035		run a factor of 2 to 4 slower with a dummy back-trace plugin (one
3036		where all hooks do nothing).  The slowdown with a plugin that is
3037		actually useful (such as the one supplied by default) is even greater,
3038		but in the case of the default plugin it is still only an constant
3039		factor (amortized).
3040
3041		3. system/Basis/Implementation/NJ/internals.{sig,sml} have been augmented
3042		with a sub-structure BTrace for controlling back-tracing.  In particular,
3043		the above-mentioned function "mode" controls whether the annotation
3044		phase is invoked by the compiler.  Another important function is
3045		"trigger": when called it aborts the current execution and causes
3046		the top-level loop to print a full back-trace.
3047
3048		4. compiler/MiscUtil/profile/btimp.sml is the current default plugin
3049		for back-tracing.  It keeps track of the dynamic call stack and in
3050		addition to that it keeps a partial history at each "level" of that
3051		stack.  For example, if a tail-calls b, b tail-calls c, and c tail-calls
3052		d and b (at separate times, dynamically), then the report will show:
3053
3054		GOTO   d
3055		/c
3056		GOTO  \b
3057		CALL   a
3058
3059		This shows that there was an initial non-tail call of a, then a
3060		tail-call to b or c, looping behavior in a cluster of functions that
3061		consist of b and c, and then a goto from that cluster (i.e., either from
3062		b or from c) to d.
3063
3064		Note that (depending on the user program) the amount of information
3065		that the back-trace module has to keep track of at each level is bounded
3066		by a constant.  Thus, the whole implementation has the same asymptotical
3067		complexity as the original program (both in space and in time).
3068
3069		5. compiler/TopLevel/interact/evalloop.sml has been modified to
3070		handle the special exception SMLofNJ.Internals.BTrace.BTrace
3071		which is raised by the "trigger" function mentioned above.
3072
3073		Notes on usage:
3074
3075		- Annotated code works well together with unannotated code:
3076		Unannotated calls simply do not show up at all in the backtrace.
3077
3078		- It is not a good idea to let modules that were annotated during
3079		different sessions run at the same time.  This is because the compiler
3080		chooses small integers to identify individual functions, and there
3081		will be clashes if different modules were compiled in separate sessions.
3082		(Nothing will crash, and you will even be told about the clashes, but
3083		back-trace information will in general not be useful.)
3084
3085		- Back-tracing can be confused by callcc and capture.
3086
3087		- The only way of getting a back-trace right now is to explicitly
3088		invoke the "trigger" function from your user program.  Eventually, we
3089		should make every exception carry back-trace information (if
3090		available).  But since this creates more overhead at "raise"-time
3091		(similar to the current exnHistory overhead), I have not yet
3092		implemented this.  (The implementation will be rather easy.)  With
3093		exceptions carrying back-trace information, this facility will be even
3094		more useful because users don't need to modify their programs...
3095
3096		- While it is possible to compile the compiler with back-trace
3097		annotations turned on (I did it to get some confidence in
3098		correctness), you must make absolutely sure that core.sml and
3099		btimp.sml are compiled WITHOUT annotation!  (core.sml cannot actually
3100		be compiled with annotation because there is no core access yet, but
3101		if you compile btimp.sml with annotation, then the system will go into
3102		an infinite recursion and crash.)
3103		Since CM currently does not know about BTrace, the only way to turn
3104		annotations on and off for different modules of the compiler is to
3105		interrupt CMB.make, change the settings, and re-invoke it.  Of course,
3106		this is awkward and clumsy.
3107
3108		Sample sessions:
3109
3110		Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
3111		- SMLofNJ.Internals.BTrace.mode (SOME true);
3112		[autoloading]
3113		[autoloading done]
3114		val it = false : bool
3115		- structure X = struct
3116		-     fun main n = let
3117		-         fun a (x, 0) = d x
3118		-           | a (x, n) = b (x, n - 1)
3119		-         and b (x, n) = c (x, n)
3120		-         and c (x, n) = a (x, n)
3121		-         and d x = e (x, 3)
3122		-         and e (x, 0) = f x
3123		-           | e (x, n) = e (x, n - 1)
3124		-         and f 0 = SMLofNJ.Internals.BTrace.trigger ()
3125		-           | f n = n * g (n - 1)
3126		-         and g n = a (n, 3)
3127		-     in
3128		-         f n
3129		-     end
3130		- end;
3131		structure X : sig val main : int -> int end
3132		- X.main 3;
3133		*** BACK-TRACE ***
3134		GOTO   stdIn:4.2-13.20: X.main[2].f
3135		GOTO-( stdIn:4.2-13.20: X.main[2].e
3136		GOTO   stdIn:4.2-13.20: X.main[2].d
3137		/ stdIn:4.2-13.20: X.main[2].a
3138		| stdIn:4.2-13.20: X.main[2].b
3139		GOTO-\ stdIn:4.2-13.20: X.main[2].c
3140		CALL   stdIn:4.2-13.20: X.main[2].g
3141		GOTO   stdIn:4.2-13.20: X.main[2].f
3142		GOTO-( stdIn:4.2-13.20: X.main[2].e
3143		GOTO   stdIn:4.2-13.20: X.main[2].d
3144		/ stdIn:4.2-13.20: X.main[2].a
3145		| stdIn:4.2-13.20: X.main[2].b
3146		GOTO-\ stdIn:4.2-13.20: X.main[2].c
3147		CALL   stdIn:4.2-13.20: X.main[2].g
3148		GOTO   stdIn:4.2-13.20: X.main[2].f
3149		GOTO-( stdIn:4.2-13.20: X.main[2].e
3150		GOTO   stdIn:4.2-13.20: X.main[2].d
3151		/ stdIn:4.2-13.20: X.main[2].a
3152		| stdIn:4.2-13.20: X.main[2].b
3153		GOTO-\ stdIn:4.2-13.20: X.main[2].c
3154		CALL   stdIn:4.2-13.20: X.main[2].g
3155		GOTO   stdIn:4.2-13.20: X.main[2].f
3156		CALL   stdIn:2.15-17.4: X.main[2]
3157		-
3158
3159		(Note that because of a FLINt bug the above code currently does not
3160		compile without BTrace turned on.)
3161
3162		Here is another example, using my modified Tiger compiler:
3163
3164		Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
3165		- SMLofNJ.Internals.BTrace.mode (SOME true);
3166		[autoloading]
3167		[autoloading done]
3168		val it = false : bool
3169		- CM.make "sources.cm";
3170		[autoloading]
3171		...
3172		[autoloading done]
3173		[scanning sources.cm]
3174		[parsing (sources.cm):parse.sml]
3175		[creating directory CM/SKEL ...]
3176		[parsing (sources.cm):tiger.lex.sml]
3177		...
3178		[wrote CM/sparc-unix/semant.sml]
3179		[compiling (sources.cm):main.sml]
3180		[wrote CM/sparc-unix/main.sml]
3181		[New bindings added.]
3182		val it = true : bool
3183		- Main.compile ("../testcases/merge.tig", "foo.out");
3184		*** BACK-TRACE ***
3185		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trvar
3186		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
3187		CALL   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
3188		GOTO   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
3189		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
3190		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
3191		CALL   lib/semant.sml:488.3-505.6: SemantFun[2].transDec.trdec[2].transBody[2]
3192		/ lib/semant.sml:411.65-543.8: SemantFun[2].transDec
3193		CALL-\ lib/semant.sml:413.2-540.9: SemantFun[2].transDec.trdec[2]
3194		CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
3195		CALL   lib/semant.sml:8.52-558.4: SemantFun[2].transProg[2]
3196		CALL   main.sml:1.18-118.4: Main.compile[2]
3197		-
3198
3199		----------------------------------------------------------------------
3200		Name: Matthias Blumen
3201		Date: 2000/06/21 18:00:00 JST
3202		Tag: blume-20000621-manual
3203		Description:
3204
3205		CM manual update: Path environments documented.
3206
3207		----------------------------------------------------------------------
3208		Name: Matthias Blume
3209		Date: 2000/06/19 13:40:00
3210		Tag: blume-20000619-manual
3211		Description:
3212
3213		CM manual and system/README update.  This only covers the fact that
3214		there are no more implicit anchors.  (Path environments and the "bind"
3215		option to "cm" have yet to be documented.)
3216
3217		----------------------------------------------------------------------
3218		Name: Matthias Blume
3219		Date: 2000/06/19 11:05:00 JST
3220		Tag: blume-20000619-chdir-bugfix
3221		Description:
3222
3223		Fixed a bug in new SrcPath module that sometimes led to a bad chDir call.
3224
3225		----------------------------------------------------------------------
3226		Name: Matthias Blume
3227		Date: 2000/06/18 22:00:10 JST
3228		Tag: blume-20000618-implicit-anchors-really-gone
3229		Description:
3230
3231		I updates the previous HISTORY entry where I forgot to mention that
3232		implicit anchors are no longer with us.
3233
3234		The current update also gets rid of the (now useless) controller
3235		CM.Control.implicit_anchors.
3236
3237		----------------------------------------------------------------------
3238		Name: Matthias Blume
3239		Date: 2000/06/16 17:30:00 JST
3240		Tag: blume-20000616-anchorenv
3241		Description:
3242
3243		This patch implements the long anticipated (just kidding :) "anchor
3244		environment" mechanism.  In the course of doing this, I also
3245		re-implemented CM's internal "SrcPath" module from scratch.  The new
3246		one should be more robust in certain boundary cases.  In any case, it
3247		is a lot cleaner than its predecessor (IMHO).
3248
3249		This time, although there is yet another boot file format change, I
3250		kept the unpickler backward-compatible.  As a result, no new bootfiles
3251		are necessary and bootstrapping is straightforward.  (You cannot read
3252		new bootfiles into an old system, but the other way around is no
3253		problem.)
3254
3255		Visible changes:
3256
3257		** 0. Implicit path anchors (without the leading $-symbol) are no
3258		longer recognized at all. This means that such path names are not
3259		illegal either.  For example, the name basis.cm simply refers to a
3260		local file called "basis.cm" (i.e, the name is an ordinary path
3261		relative to .cm-files directory).  Or, to put it differently, only
3262		names that start with $ are anchored paths.
3263
3264		** 1. The $<singlearc> abbreviation for $/<singlearc> has finally
3265		vanished.
3266
3267		John (Reppy) had critizised this as soon as I originally proposed and
3268		implemented it, but at that time I did not really deeply believe
3269		him. :) Now I came full-circle because I need the $<singlearc> syntax
3270		in another place where it cannot be seen as an abbreviation for
3271		$/<singlearc>.  To avoid the confusion, $<singlearc> now means what it
3272		seems to mean (i.e., it "expands" into the corresponding anchor
3273		value).
3274
3275		However, when paths are used as members in CM description files, it
3276		continues to be true that there must be at least another arc after the
3277		anchor.  This is now enforced separately during semantic analysis
3278		(i.e., from a lexical/syntactical point of view, the notation is ok.)
3279
3280		** 2. The "cm" class now accepts an option "bind".  The option's value
3281		is a sub-option list of precisely two items -- one labeled "anchor"
3282		and the other one labeled "value".  As you might expect, "anchor" is
3283		used to specify an anchor name to be bound, and "value" specifies what
3284		the anchor is being bound to.
3285
3286		The value must be a directory name and can be given in either standard
3287		syntax (including the possibility that it is itself an anchored path)
3288		or native syntax.
3289
3290		Examples:
3291
3292		foo.cm (bind:(anchor:bar value:$mystuff/bar))
3293		lib.cm (bind:(anchor:a value:"H:\\x\\y\\z"))  (* only works under windows *)
3294
3295		and so on.
3296
3297		The meaning of this is that the .cm-file will be processed with an
3298		augmented anchor environment where the given anchor(s) is/are bound to
3299		the given values(s).
3300
3301		The rationale for having this feature is this: Suppose you are trying
3302		to use two different (already stable) libraries a.cm and b.cm (that
3303		you perhaps didn't write yourself).  Further, suppose each of these
3304		two libraries internally uses its own auxiliary library $aux/lib.cm.
3305		Normally you would now have a problem because the anchor "lib" can not
3306		be bound to more than one value globally.  Therefore, the project that
3307		uses both a.cm and b.cm must locally redirect the anchor to some other
3308		place:
3309
3310		a.cm (bind:(anchor:lib value:/usr/lib/smlnj/a-stuff))
3311		b.cm (bind:(anchor:lib value:/usr/lib/smlnj/b-stuff))
3312
3313		This hard-wires $lib/aux.cm to /usr/lib/smlnj/a-stuff/aux.cm or
3314		/usr/lib/smlnj/b-stuff/aux.cm, respectively.
3315
3316		Hard-wiring path names is a bit inflexible (and CM will verbosely warn
3317		you when you do so at the time of CM.stabilize).  Therefore, you can
3318		also use an anchored path as the value:
3319
3320		a.cm (bind:(anchor:lib value:$a-lib))
3321		b.cm (bind:(anchor:lib value:$b-lib))
3322
3323		Now you can globally configure (using the usual CM.Anchor.anchor or
3324		pathconfig machinery) bindings for "a-lib" and "b-lib".  Since "lib"
3325		itself is always locally bound, setting it globally is no longer
3326		meaningful or necessary (but it does not hurt either).  In fact, "lib"
3327		can still be used as a global anchor for separate purposes.  As a
3328		matter of fact, one can locally define "lib" in terms of a global
3329		"lib":
3330
3331		a.cm (bind:(anchor:lib value:$lib/a))
3332		b.cm (bind:(anchor:lib value:$lib/b))
3333
3334		** 3: The encoding of path names has changed.  This affects the way
3335		path names are shown in CM's progress report and also the internal
3336		protocol encoding used for parallel make.
3337
3338		The encoding now uses one or more ':'-separated segments.  Each
3339		segments corresponds to a file that has been specified relative to the
3340		file given by its preceding segment.  The first segment is either
3341		relative to the CWD, absolute, or anchored.  Each segment itself is
3342		basically a Unix pathname; all segments but the first are relative.
3343
3344		Example:
3345
3346		$foo/bar/baz.cm:a/b/c.sml
3347
3348		This path denotes the file bar/a/b/c.sml relative to the directory
3349		denoted by anchor "foo".  Notice that the encoding also includes
3350		baz.cm which is the .cm-file that listed a/b/c.sml.  As usual, such
3351		paths are resolved relative to the .cm-files directory, so baz.cm must
3352		be ignored to get the "real" pathname.
3353
3354		To make this fact more obvious, CM puts the names of such "virtual
3355		arcs" into parentheses when they appear in progress reports. (No
3356		parentheses will appear in the internal protocol encoding.)  Thus,
3357		what you really see is:
3358
3359		$foo/bar/(baz.cm):a/b/c.sml
3360
3361		I find this notation to be much more informative than before.
3362
3363		Another new feature of the encoding is that special characters
3364		including parentheses, colons, (back)slashes, and white space are
3365		written as \ddd (where ddd is the decimal encoding of the character).
3366
3367		*** The CM manual still needs to be updated.
3368
3369		----------------------------------------------------------------------
3370		Name: Allen Leung
3371		Date: 2000/06/15 00:38:00
3372		Tag: leunga-20000615-x86-peephole
3373
3374		x86 Peephole fix by Fermin.  Affects c-- and moby only.
3375
3376		----------------------------------------------------------------------
3377		Name: Matthias Blume
3378		Date: 2000/06/12 11:40:00
3379		Tag: blume-20000612-parmakefix
3380		Description:
3381
3382		More cleanup after changing the file naming scheme: This time I
3383		repaired the parallel make mechanism for CMB.make which I broke earlier.
3384
3385		----------------------------------------------------------------------
3386		Name: Allen Leung
3387		Date: 2000/06/09 01:25:00
3388		Tag: leunga-20000609-various
3389
3390		None of these things should affect normal SML/NJ operations
3391
3392		1. Peephole improvements provided by Fermin (c--)
3393		2. New annotation DEFUSE for adding extra dependence (moby)
3394		3. New X86 LOCK instructions (moby)
3395		4. New machine description language for reservation tables (scheduling)
3396		5. Fixes to various optimization/analysis modules (branch chaining, dominator
3397		trees etc.)
3398		6. I've changed the CM files so that they can work with versions
3399		110.0.6, 110.25 and 110.28
3400
3401	----------------------------------------------------------------------	----------------------------------------------------------------------
3402	Name: Matthias Blume	Name: Matthias Blume
3403	Date: 2000/06/09 12:40:00	Date: 2000/06/09 12:40:00
#	Line 574	Line 3961
3961
3962	3.  Assembly	3.  Assembly
3963
3964	When generating assemby, resolve the value of client defined constants,	When generating assembly, resolve the value of client defined constants,
3965	instead of generating symbolic values.  This is controlled by the	instead of generating symbolic values.  This is controlled by the
3966	new flag "asm-resolve-constants", which is default to true.	new flag "asm-resolve-constants", which is default to true.
3967
#	Line 597	Line 3984
3984
3985	To this end, I arranged that instead of "structure Core" as "structure	To this end, I arranged that instead of "structure Core" as "structure
3986	_Core" is bound in the pervasive environment.  Core access is done via	_Core" is bound in the pervasive environment.  Core access is done via
3987	_Core (which can never be accidentially rebound because _Core is not a	_Core (which can never be accidentally rebound because _Core is not a
3988	legal surface-syntax symbol).	legal surface-syntax symbol).
3989
3990	The current solution is much cleaner because the core environment is	The current solution is much cleaner because the core environment is
#	Line 607	Line 3994
3994	with dynamic and symbolic parts of the core environment.	with dynamic and symbolic parts of the core environment.
3995
3996	Remaining hackery (to bind the "magic" symbol _Core) is localized in the	Remaining hackery (to bind the "magic" symbol _Core) is localized in the
3997	compilation mananger's bootstrap compiler (actually: in the "init group"	compilation manager's bootstrap compiler (actually: in the "init group"
3998	handling).  See the comments in src/system/smlnj/init/init.cmi for	handling).  See the comments in src/system/smlnj/init/init.cmi for
3999	more details.	more details.
4000
#	Line 722	Line 4109
4109	(specified in the .cm file at each instance where the tool's class is	(specified in the .cm file at each instance where the tool's class is
4110	used).	used).
4111
4112	This was done to accomodate the new "make" and "shell" tools which	This was done to accommodate the new "make" and "shell" tools which
4113	facilitate fairly seemless hookup to portions of code managed using	facilitate fairly seamless hookup to portions of code managed using
4114	Makefiles or Shell scripts.	Makefiles or Shell scripts.
4115
4116	There are no classes "shared" or "private" anymore.  Instead, the	There are no classes "shared" or "private" anymore.  Instead, the
#	Line 735	Line 4122
4122
4123	All existing tools are described in the CM manual.	All existing tools are described in the CM manual.
4124
4125	- Slightly better error handling.  (CM now surpresses many followup	- Slightly better error handling.  (CM now suppresses many followup
4126	error messages that tended to be more annoying than helpful.)	error messages that tended to be more annoying than helpful.)
4127
4128	2. Major changes to the compiler's static environment data structures.	2. Major changes to the compiler's static environment data structures.
#	Line 766	Line 4153
4153	elaborator). There were a lot of changes during my "linkpath" trials	elaborator). There were a lot of changes during my "linkpath" trials
4154	that could have been reverted to their original state but weren't.	that could have been reverted to their original state but weren't.
4155	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
4156	than what was strictly necessary...  (I _did_ resist the tempation	than what was strictly necessary...  (I _did_ resist the temptation
4157	of doing any "global reformatting" to avoid an untimely death at	of doing any "global reformatting" to avoid an untimely death at
4158	Dave's hands. :)	Dave's hands. :)
4159
#	Line 869	Line 4256
4256
4257	I've changed andl to testl in the floating point test sequence	I've changed andl to testl in the floating point test sequence
4258	whenever appropriate.  The Intel optimization guide states that	whenever appropriate.  The Intel optimization guide states that
4259	testl is perferable to andl.	testl is preferable to andl.
4260
4261	7. RA (x86 only)	7. RA (x86 only)
4262
#	Line 1051	Line 4438
4438
4439	1. Tools.registerStdShellCmdTool (from smlnj/cm/tool.cm) takes an	1. Tools.registerStdShellCmdTool (from smlnj/cm/tool.cm) takes an
4440	additional argument called "template" which is an optional string that	additional argument called "template" which is an optional string that
4441	specifiel the layout of the tool command line.  See the CM manual for	specifies the layout of the tool command line.  See the CM manual for
4442	explanation.	explanation.
4443
4444	2. A special-purpose tool can be "regisitered" by simply dropping the	2. A special-purpose tool can be "registered" by simply dropping the
4445	corresponding <...>-tool.cm (and/or <...>-ext.cm) into the same	corresponding <...>-tool.cm (and/or <...>-ext.cm) into the same
4446	directory where the .cm file lives that uses this tool.  (The	directory where the .cm file lives that uses this tool.  (The
4447	behavior/misfeature until now was to look for the tool description	behavior/misfeature until now was to look for the tool description
#	Line 1098	Line 4485
4485	rebuild.  Having sets of the form <base><k>.{bin,boot}.<arch>-unix for	rebuild.  Having sets of the form <base><k>.{bin,boot}.<arch>-unix for
4486	<k>=1,2,... is normally not a good idea when invoking fixpt.  However,	<k>=1,2,... is normally not a good idea when invoking fixpt.  However,
4487	they might be the result of an earlier partial run of fixpt (which	they might be the result of an earlier partial run of fixpt (which
4488	perhaps got accidentially killed).  In this case, fixpt will quickly	perhaps got accidentally killed).  In this case, fixpt will quickly
4489	move through what exists before continuing where it left off earlier,	move through what exists before continuing where it left off earlier,
4490	and, thus, saves a lot of time.	and, thus, saves a lot of time.
4491
#	Line 1148	Line 4535
4535	it from that remote directory.	it from that remote directory.
4536	This should simplify installation further:  For machines that have	This should simplify installation further:  For machines that have
4537	access to the internet, just fetch <version>-config.tgz, unpack it,	access to the internet, just fetch <version>-config.tgz, unpack it,
4538	edit config/targets, and go (run config/install.sh).  The scipt will	edit config/targets, and go (run config/install.sh).  The script will
4539	fetch everything else that it might need all by itself.	fetch everything else that it might need all by itself.
4540
4541	For CVS users, this mechanism is not relevant for source archives, but	For CVS users, this mechanism is not relevant for source archives, but

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.656   changed lines   Added in v.1115

---

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