Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 987, Wed Nov 21 21:14:44 2001 UTC	revision 1065, Thu Feb 14 15:05:16 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: Lal George
17		Date: 2002/02/14 09:55:27 EST 2002
18		Tag: george-20020214-isabelle-bug
19		Description:
20
21		Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
22		of Isabelle on the x86.
23
24		From Allen:
25		-----------
26		I've found the problem:
27
28		in ra-core.sml, I use the counter "blocked" to keep track of the
29		true number of elements in the freeze queue.  When the counter goes
30		to zero, I skip examining the queue.  But I've messed up the
31		bookkeeping in combine():
32
33		else ();
34		case !ucol of
35		PSEUDO => (if !cntv > 0 then
36		(if !cntu > 0 then blocked := !blocked - 1 else ();
37		^^^^^^^^^^^^^^^^^^^^^^^
38		moveu := mergeMoveList(!movev, !moveu)
39		)
40		else ();
41
42		combine() is called to coalesce two nodes u and v.
43		I think I was thinking that if the move counts of u and v are both
44		greater than zero then after they are coalesced then one node is
45		removed from the freeze queue.  Apparently I was thinking that
46		both u and v are of low degree, but that's clearly not necessarily true.
47
48
49		02/12/2002:
50		Here's the patch.  HOL now compiles.
51
52		I don't know how this impact on performance (compile
53		time or runtime).  This bug caused the RA (especially on the x86)
54		to go thru the potential spill phase when there are still nodes on the
55		freeze queue.
56
57
58
59
60		----------------------------------------------------------------------
61		Name: Matthias Blume
62		Date: 2002/02/13 22:40:00 EST
63		Tag: blume-20020213-fptr-rtti
64		Description:
65
66		Fixed a bug in ml-nlffigen that was introduced with one of the previous
67		updates.
68
69		----------------------------------------------------------------------
70		Name: Matthias Blume
71		Date: 2002/02/13 16:41:00 EST
72		Tag: blume-20020213-cmlpq
73		Description:
74
75		Added new priority queue export symbols (which have just been added to
76		smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
77		not compile and the installer would choke.)
78
79		----------------------------------------------------------------------
80		Name: Matthias Blume
81		Date: 2002/02/13 16:15:00 EST
82		Tag: blume-20020213-various
83		Description:
84
85		1. More tweaks to ml-nlffigen:
86
87		- better internal datastructures (resulting in slight speedup)
88		- "-match" option requires exact match
89		- "localized" gensym counters (untagged structs/unions nested within
90		other structs/unions or within typedefs get a fresh counter; their
91		tag will be prefixed by a concatenation of their parents' tags)
92		- bug fixes (related to calculation of transitive closure of types
93		to be included in the output)
94
95		2. Minor Basis updates:
96
97		- added implementations for List.collate and Option.app
98
99		----------------------------------------------------------------------
100		Name: Matthias Blume
101		Date: 2002/02/11 15:55:00 EST
102		Tag: blume-20020211-gensym
103		Description:
104
105		Added a "-gensym" option to command line of ml-nlffigen.  This can be
106		used to specify a "stem" -- a string that is inserted in all "gensym'd"
107		names (ML structure names that correspond to unnamed C structs, unions,
108		and enums), so that separate runs of ml-nlffigen do not clash.
109
110		----------------------------------------------------------------------
111		Name: Matthias Blume
112		Date: 2002/02/11 12:05:00 EST
113		Tag: blume-20020211-gensml
114		Description:
115
116		A quick fix for a problem with GenSML (in the pgraph-util library):
117		Make generation of toplevel "local" optional.  (Strictly speaking,
118		signature definitions within "local" are not legal SML.)
119
120		Other than that: updates to INSTALL and cm/TODO.
121
122		----------------------------------------------------------------------
123		Name: Matthias Blume
124		Date: 2002/02/08 15:00:00 EST
125		Tag: blume-20020208-uniquepid
126		Description:
127
128		0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
129
130		1. The installer (config/install.sh) has gotten smarter:
131
132		- Configuration options are a bit easier to specify now
133		(in config/targets).
134		- Bug in recognizing .tar.bz2 files fixed.
135		- Installer automatically resolves dependencies between
136		configuration options (e.g., if you ask for eXene, you will
137		also get cml -- regardless whether you asked for it or not).
138		- Installer can run in "quieter mode" by setting the environment
139		variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
140		"completely silent", though.
141		- Build HashCons library as part of smlnj-lib.
142
143		2. A new scheme for assigning persistent identifiers to compilation
144		units (and, by extension, to types etc.) has been put into place.
145		This fixes a long-standing bug where types and even dynamic values
146		can get internally confused, thereby compromising type safety
147		(abstraction) and dynamic correctness.  See
148
149		http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
150
151		for an example of how things could go wrong until now.
152
153		The downside of the new scheme is that pids are not quite as
154		persistent as they used to be: CM will generate a fresh pid
155		for every compilation unit that it thinks it sees for the first
156		time.  That means that if you compile starting from a clean, fresh
157		source tree at two different times, you end up with different
158		binaries.
159
160		Cutoff recompilation, however, has not been compromised because
161		CM keeps pid information in special caches between runs.
162
163		----------------------------------------------------------------------
164		Name: Lal George
165		Date: 2002/02/07 15:34:13 EST 2002
166		Tag: <none>
167		Description:
168
169		Compilers that generate assembly code may produce  global labels
170		whose value is resolved at link time. The various peephole optimization
171		modules did not take this in account.
172
173		TODO. The Labels.addrOf function should really return an option
174		type so that clients are forced to deal with this issue, rather
175		than an exception being raised.
176
177		----------------------------------------------------------------------
178		Name: Lal George
179		Date: 2002/02/06 13:55:02 EST
180		Tag: george-20020206-ra-breakup
181		Description:
182
183		1. A bug fix from Allen.
184
185		A typo causes extra fstp %st(0)'s to be generated at compensation
186		edges, which might cause stack underflow traps at runtime.  This
187		occurs in fft where there are extraneous fstps right before the 'into'
188		trap instruction (in this case they are harmless since none of the
189		integers overflow.)
190
191		2. Pulled out various utility modules that were embedded in the modules
192		of the register allocator. I need these modules for other purposes, but
193		they are not complete enough to put into a library (just yet).
194		----------------------------------------------------------------------
195		Name: Matthias Blume
196		Date: 2002/01/31 16:05:00 EST
197		Tag: blume-20020131-sparc-ccalls
198		Description:
199
200		1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
201		of extra stack space by mistake.  Fixed.
202
203		2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
204
205		----------------------------------------------------------------------
206		Name: Allen Leung
207		Date: 2002/01/30
208		Tag: leunga-20020130-nowhere-bug-fix
209		Description:
210
211		MLRISC bug fixes:
212		1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
213		2. MachineInt.fromString was negating its value.
214
215		----------------------------------------------------------------------
216		Name: Matthias Blume
217		Date: 2002/01/29
218		Tag: blume-20020129-INSTALL
219		Description:
220
221		- Added somewhat detailed installation instructions (file INSTALL).
222		- Fixed curl-detection bug in config/install.sh.
223		- It is now possible to select the URL getter using the URLGETTER
224		environment variable:
225
226		not set / "unknown"      --> automatic detection (script tries wget,
227		curl, and lynx)
228		"wget" / "curl" / "lynx" --> use the specified program (script "knows"
229		how to properly invoke them)
230		other                    --> use $URLGETTER directly, it must take
231		precisely two command-line arguments
232		(source URL and destination file name)
233
234		----------------------------------------------------------------------
235		Name: Matthias Blume
236		Date: 2002/01/28
237		Tag: blume-20020128-sparc-ccalls
238		Description:
239
240		- Fixed problem with calculation of "used" registers in sparc-c-calls.
241		- Make use of the allocParam argument in sparc-c-calls.
242
243		----------------------------------------------------------------------
244		Name: Matthias Blume
245		Date: 2002/01/28
246		Tag: blume-20020128-allocParam
247		Description:
248
249		John Reppy:  Changes c-calls API to accept client-callback for
250		allocating extra stack space.
251		me: Corresponding changes to mlriscGen (using a dummy argument that
252		does not change the current behavior).
253
254		----------------------------------------------------------------------
255		Name: Matthias Blume
256		Date: 2002/01/28 12:00:00
257		Tag: Release_110_38
258		Description:
259
260		This time for real!!!
261
262		----------------------------------------------------------------------
263		Name: Matthias Blume
264		Date: 2002/01/28 10:56:00 EST
265		Tag: blume-20020128-retraction
266		Description:
267
268		0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
269		with blume-Release_110_38-retracted.)
270
271		1. Fixed a problem with incorrect rounding modes in real64.sml.
272		(Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
273
274		2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
275		and enums fixed.  The naming of corresponding ML identifiers should
276		now be consistent again.
277
278		----------------------------------------------------------------------
279		Name: Allen Leung
280		Date: 2002/01/27
281		Tag: leunga-20020127-nowhere
282		Description:
283
284		Added a target called nowhere in the configuration scripts.
285		Enabling this will build the MLRISC 'nowhere' tool (for translating
286		programs with where-clauses into legal SML code) during installation.
287
288		----------------------------------------------------------------------
289		Name: Matthias Blume
290		Date: 2002/01/25 21:27:00 EST
291		Tag: blume-Release_110_38-retracted
292		Description:
293
294		Call it a (working) release!  Version is 110.38. Bootfiles are ready.
295
296		README will be added later.
297
298		!!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
299		(Release_110_38) removed.  Reason: Last-minute bug fixes.
300
301		----------------------------------------------------------------------
302		Name: Matthias Blume
303		Date: 2002/01/25
304		Tag: blume-20020125-ffi
305		Description:
306
307		A large number of tweaks and improvements to ml-nlffi-lib and
308		ml-nlffigen:
309
310		- ML represenation types have been streamlined
311		- getter and setter functions work with concrete values, not abstract
312		ones where possible
313		- ml-nlffigen command line more flexible (see README file there)
314		- some bugs have been fixed (hopefully)
315
316		----------------------------------------------------------------------
317		Name: Lal George
318		Date: 2002/01/24
319		Tag: george-20020124-risc-ra-interface
320		Description:
321
322		There is a dramatic simplification in the interface to the
323		register allocator for RISC architectures as a result of making
324		parallel copy instructions explicit.
325
326		----------------------------------------------------------------------
327		Name: Matthias Blume
328		Date: 2002/01/22
329		Tag: blume-20020122-x86-ccalls
330		Description:
331
332		Bug fix for c-calls on x86 (having to do with how char- and
333		short-arguments are being handled).
334
335		----------------------------------------------------------------------
336		Name: Matthias Blume
337		Date: 2002/01/21
338		Tag: blume-20020121-ff
339		Description:
340
341		Another day of fiddling with the FFI...
342
343		1. Bug fix/workaround:  CKIT does not complain about negative array
344		dimensions, so ml-nlffigen has to guard itself against this possibility.
345		(Otherwise a negative dimension would send it into an infinite loop.)
346
347		2. Some of the abstract types (light objects, light pointers, most "base"
348		types) in structure C are now eqtypes.
349
350		3. Added constructors and test functions for NULL function pointers.
351
352		----------------------------------------------------------------------
353		Name: Matthias Blume
354		Date: 2002/01/18
355		Tag: blume-20020118-ready-for-new-release
356		Description:
357
358		Made config/srcarchiveurl point to a new place.  (Will provide boot
359		files shortly.)
360
361		Maybe we christen this to be 110.38?
362
363		----------------------------------------------------------------------
364		Name: Matthias Blume
365		Date: 2002/01/18
366		Tag: blume-20020118-more-ffifiddle
367		Description:
368
369		Today's FFI fiddling:
370
371		- Provided a structure CGetSet with "convenient" versions of C.Get.* and
372		C.Set.* that use concrete (MLRep.*) arguments and results instead
373		of abstract ones.
374
375		- Provided word-style bit operations etc. for "int" representation
376		types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
377		and Long.
378
379		----------------------------------------------------------------------
380		Name: Matthias Blume
381		Date: 2002/01/18
382		Tag: blume-20020118-use-x86-fp
383		Description:
384
385		Now that x86-fast-fp seems to be working, I turned it back on again
386		by default.  (Seems to work fine now, even with the FFI.)
387
388		Other than that, I added some documentation about the FFI to
389		src/ml-nlffigen/README and updated the FFI test examples in
390		src/ml-nlffi-lib/Tests/*.
391
392		----------------------------------------------------------------------
393		Name: Allen Leung
394		Date: 2002/01/17
395		Tag: leunga-20020117-x86-fast-fp-call
396		Description:
397
398		1. Fixed a problem with handling return fp values when x86's fast fp
399		mode is turned on.
400
401		2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
402		of %f32.
403
404		3. Added a constructor INT32lit to the ast of MLRISC tools.
405
406		----------------------------------------------------------------------
407		Name: Matthias Blume
408		Date: 2002/01/16
409		Tag: blume-20020116-ffifiddle
410		Description:
411
412		More fiddling with the FFI interface:
413
414		- Make constness 'c instead of rw wherever possible.  This eliminates
415		the need for certain explicit coercions.  (However, due to ML's
416		value polymorphism, there will still be many cases where explicit
417		coercions are necessary.  Phantom types are not the whole answer
418		to modeling a subtyping relationship in ML.)
419
420		- ro/rw coersions for pointers added.  (Avoids the detour through */&.)
421
422		- "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
423		clumsy workaround for varargs problem.)
424
425		----------------------------------------------------------------------
426		Name: Lal George
427		Date: 2002/01/15
428		Tag: <none>
429		Description:
430
431		1. Since COPY instructions are no longer native to the architecture,
432		a generic functor can be used to implement the expandCopies function.
433
434		2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
435		TEXT segment.
436
437		----------------------------------------------------------------------
438		Name: Matthias Blume
439		Date: 2002/01/15
440		Tag: blume-20020115-ffiupdates
441		Description:
442
443		1. Fix for bug resulting in single-precision float values being returned
444		incorrectly from FFI calls.
445
446		2. Small modifications to C FFI API:
447
448		- memory-allocation routines return straight objects (no options)
449		and raise an exception in out-of-memory situations
450		- unsafe extensions to cast between function pointers and pointers
451		from/to ints
452		- added structure C_Debug as an alternative to structure C where
453		pointer-dereferencing (|*| and |*!) always check for null-pointers
454		- added open_lib' to DynLinkage;  open_lib' works like open_lib
455		but also takes a (possibly empty) list of existing library handles
456		that the current library depends on
457
458		----------------------------------------------------------------------
459		Name: Matthias Blume
460		Date: 2002/01/10
461		Tag: blume-20020110-newffigen
462		Description:
463
464		1. Updates to portable graph code.
465
466		2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
467		more scalable now so that even huge interfaces such as the one
468		for GTK compile in finite time and space. :-)
469		See src/ml-nlffigen/README for details on what's new.
470
471		----------------------------------------------------------------------
472		Name: Lal George
473		Date: 2001/01/09 14:31:35 EST 2002
474		Tag: george-20011206-rm-native-copy
475		Description:
476
477		Removed the native COPY and FCOPY instructions
478		from all the architectures and replaced it with the
479		explicit COPY instruction from the previous commit.
480
481		It is now possible to simplify many of the optimizations
482		modules that manipulate copies. This has not been
483		done in this change.
484
485		----------------------------------------------------------------------
486		Name: Lal George
487		Date: 2001/12/06 16:50:13 EST 2001
488		Tag: george-20011206-mlrisc-instruction
489		Description:
490
491		Changed the representation of instructions from being fully abstract
492		to being partially concrete. That is to say:
493
494		from
495		type instruction
496
497		to
498		type instr                              (* machine instruction *)
499
500		datatype instruction =
501		LIVE of {regs: C.cellset, spilled: C.cellset}
502		| KILL of {regs: C.cellset, spilled: C.cellset}
503		| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
504		| ANNOTATION of {i: instruction, a: Annotations.annotation}
505		| INSTR of instr
506
507		This makes the handling of certain special instructions that appear on
508		all architectures easier and uniform.
509
510		LIVE and KILL say that a list of registers are live or killed at the
511		program point where they appear. No spill code is generated when an
512		element of the 'regs' field is spilled, but the register is moved to
513		the 'spilled' (which is present, more for debugging than anything else).
514
515		LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
516		We used to generate:
517
518		DEFFREG f1
519		f1 := f2 + f3
520		trapb
521
522		but now generate:
523
524		f1 := f2 + f3
525		trapb
526		LIVE {regs=[f1,f2,f3], spilled=[]}
527
528		Furthermore, the DEFFREG (hack) required that all floating point instruction
529		use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
530		defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
531		in a cleaner alpha implementation. (Hopefully, intel will not get rid of
532		this architecture).
533
534		COPYXXX is intended to replace the parallel COPY and FCOPY  available on
535		all the architectures. This will result in further simplification of the
536		register allocator that must be aware of them for coalescing purposes, and
537		will also simplify certain aspects of the machine description that provides
538		callbacks related to parallel copies.
539
540		ANNOTATION should be obvious, and now INSTR represents the honest to God
541		machine instruction set!
542
543		The <arch>/instructions/<arch>Instr.sml files define certain utility
544		functions for making porting easier -- essentially converting upper case
545		to lower case. All machine instructions (of type instr) are in upper case,
546		and the lower case form generates an MLRISC instruction. For example on
547		the alpha we have:
548
549		datatype instr =
550		LDA of {r:cell, b:cell, d:operand}
551		| ...
552
553		val lda : {r:cell, b:cell, d:operand} -> instruction
554		...
555
556		where lda is just (INSTR o LDA), etc.
557
558		----------------------------------------------------------------------
559		Name: Matthias Blume
560		Date: 2001/11/22 21:40:00 EST
561		Tag: Release_110_37
562		Description:
563
564		Release 110.37.  This time for real.
565
566		----------------------------------------------------------------------
567		Name: Matthias Blume
568		Date: 2001/11/21 16:35:00 EST
569		Tag: blume-20011121-foot-in-mouth
570		Description:
571
572		Removed the "Release_110_37" tag because of a serious bug.
573		This will be re-tagged once the bug is fixed.
574
575		----------------------------------------------------------------------
576	Name: Matthias Blume	Name: Matthias Blume
577	Date: 2001/11/21 16:14:00 EST	Date: 2001/11/21 16:14:00 EST
578	Tag: blume-20011121-forgottenfile	Tag: blume-20011121-forgottenfile
#	Line 24	Line 584
584	----------------------------------------------------------------------	----------------------------------------------------------------------
585	Name: Matthias Blume	Name: Matthias Blume
586	Date: 2001/11/21 16:10:00 EST	Date: 2001/11/21 16:10:00 EST
587	Tag: Release_110_37	Tag: blume-20011121-invalid_110_37
588	Description:	Description:
589
590		Note: I removed the original tag "Release_110_37" from this commit
591		because we found a serious bug in all non-x86 backends.
592		- Matthias
593
594	1. Modifications to the SML/NJ code generator and to the runtime system	1. Modifications to the SML/NJ code generator and to the runtime system
595	so that code object name strings are directly inserted into code	so that code object name strings are directly inserted into code
596	objects at code generation time.  The only business the runtime system	objects at code generation time.  The only business the runtime system
#	Line 248	Line 812
812
813	----------------------------------------------------------------------	----------------------------------------------------------------------
814	Name: Matthias Blume	Name: Matthias Blume
	>>>>>>> 1.169
815	Date: 2001/09/18 15:35:00 EDT	Date: 2001/09/18 15:35:00 EDT
816	Tag: blume-20010918-readme11036	Tag: blume-20010918-readme11036
817	Description:	Description:

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.987   changed lines   Added in v.1065

---

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