Diff of /sml/trunk/HISTORY

Parent Directory | Revision Log | Patch

	revision 975, Wed Oct 31 20:22:44 2001 UTC	revision 1048, Thu Jan 31 06:36:17 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: 2002/01/30
18		Tag: leunga-20020130-nowhere-bug-fix
19		Description:
20
21		MLRISC bug fixes:
22		1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
23		2. MachineInt.fromString was negating its value.
24
25		----------------------------------------------------------------------
26		Name: Matthias Blume
27		Date: 2002/01/29
28		Tag: blume-20020129-INSTALL
29		Description:
30
31		- Added somewhat detailed installation instructions (file INSTALL).
32		- Fixed curl-detection bug in config/install.sh.
33		- It is now possible to select the URL getter using the URLGETTER
34		environment variable:
35
36		not set / "unknown"      --> automatic detection (script tries wget,
37		curl, and lynx)
38		"wget" / "curl" / "lynx" --> use the specified program (script "knows"
39		how to properly invoke them)
40		other                    --> use $URLGETTER directly, it must take
41		precisely two command-line arguments
42		(source URL and destination file name)
43
44		----------------------------------------------------------------------
45		Name: Matthias Blume
46		Date: 2002/01/28
47		Tag: blume-20020128-sparc-ccalls
48		Description:
49
50		- Fixed problem with calculation of "used" registers in sparc-c-calls.
51		- Make use of the allocParam argument in sparc-c-calls.
52
53		----------------------------------------------------------------------
54		Name: Matthias Blume
55		Date: 2002/01/28
56		Tag: blume-20020128-allocParam
57		Description:
58
59		John Reppy:  Changes c-calls API to accept client-callback for
60		allocating extra stack space.
61		me: Corresponding changes to mlriscGen (using a dummy argument that
62		does not change the current behavior).
63
64		----------------------------------------------------------------------
65		Name: Matthias Blume
66		Date: 2002/01/28 12:00:00
67		Tag: Release_110_38
68		Description:
69
70		This time for real!!!
71
72		----------------------------------------------------------------------
73		Name: Matthias Blume
74		Date: 2002/01/28 10:56:00 EST
75		Tag: blume-20020128-retraction
76		Description:
77
78		0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
79		with blume-Release_110_38-retracted.)
80
81		1. Fixed a problem with incorrect rounding modes in real64.sml.
82		(Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
83
84		2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
85		and enums fixed.  The naming of corresponding ML identifiers should
86		now be consistent again.
87
88		----------------------------------------------------------------------
89		Name: Allen Leung
90		Date: 2002/01/27
91		Tag: leunga-20020127-nowhere
92		Description:
93
94		Added a target called nowhere in the configuration scripts.
95		Enabling this will build the MLRISC 'nowhere' tool (for translating
96		programs with where-clauses into legal SML code) during installation.
97
98		----------------------------------------------------------------------
99		Name: Matthias Blume
100		Date: 2002/01/25 21:27:00 EST
101		Tag: blume-Release_110_38-retracted
102		Description:
103
104		Call it a (working) release!  Version is 110.38. Bootfiles are ready.
105
106		README will be added later.
107
108		!!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
109		(Release_110_38) removed.  Reason: Last-minute bug fixes.
110
111		----------------------------------------------------------------------
112		Name: Matthias Blume
113		Date: 2002/01/25
114		Tag: blume-20020125-ffi
115		Description:
116
117		A large number of tweaks and improvements to ml-nlffi-lib and
118		ml-nlffigen:
119
120		- ML represenation types have been streamlined
121		- getter and setter functions work with concrete values, not abstract
122		ones where possible
123		- ml-nlffigen command line more flexible (see README file there)
124		- some bugs have been fixed (hopefully)
125
126		----------------------------------------------------------------------
127		Name: Lal George
128		Date: 2002/01/24
129		Tag: george-20020124-risc-ra-interface
130		Description:
131
132		There is a dramatic simplification in the interface to the
133		register allocator for RISC architectures as a result of making
134		parallel copy instructions explicit.
135
136		----------------------------------------------------------------------
137		Name: Matthias Blume
138		Date: 2002/01/22
139		Tag: blume-20020122-x86-ccalls
140		Description:
141
142		Bug fix for c-calls on x86 (having to do with how char- and
143		short-arguments are being handled).
144
145		----------------------------------------------------------------------
146		Name: Matthias Blume
147		Date: 2002/01/21
148		Tag: blume-20020121-ff
149		Description:
150
151		Another day of fiddling with the FFI...
152
153		1. Bug fix/workaround:  CKIT does not complain about negative array
154		dimensions, so ml-nlffigen has to guard itself against this possibility.
155		(Otherwise a negative dimension would send it into an infinite loop.)
156
157		2. Some of the abstract types (light objects, light pointers, most "base"
158		types) in structure C are now eqtypes.
159
160		3. Added constructors and test functions for NULL function pointers.
161
162		----------------------------------------------------------------------
163		Name: Matthias Blume
164		Date: 2002/01/18
165		Tag: blume-20020118-ready-for-new-release
166		Description:
167
168		Made config/srcarchiveurl point to a new place.  (Will provide boot
169		files shortly.)
170
171		Maybe we christen this to be 110.38?
172
173		----------------------------------------------------------------------
174		Name: Matthias Blume
175		Date: 2002/01/18
176		Tag: blume-20020118-more-ffifiddle
177		Description:
178
179		Today's FFI fiddling:
180
181		- Provided a structure CGetSet with "convenient" versions of C.Get.* and
182		C.Set.* that use concrete (MLRep.*) arguments and results instead
183		of abstract ones.
184
185		- Provided word-style bit operations etc. for "int" representation
186		types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
187		and Long.
188
189		----------------------------------------------------------------------
190		Name: Matthias Blume
191		Date: 2002/01/18
192		Tag: blume-20020118-use-x86-fp
193		Description:
194
195		Now that x86-fast-fp seems to be working, I turned it back on again
196		by default.  (Seems to work fine now, even with the FFI.)
197
198		Other than that, I added some documentation about the FFI to
199		src/ml-nlffigen/README and updated the FFI test examples in
200		src/ml-nlffi-lib/Tests/*.
201
202		----------------------------------------------------------------------
203		Name: Allen Leung
204		Date: 2002/01/17
205		Tag: leunga-20020117-x86-fast-fp-call
206		Description:
207
208		1. Fixed a problem with handling return fp values when x86's fast fp
209		mode is turned on.
210
211		2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
212		of %f32.
213
214		3. Added a constructor INT32lit to the ast of MLRISC tools.
215
216		----------------------------------------------------------------------
217		Name: Matthias Blume
218		Date: 2002/01/16
219		Tag: blume-20020116-ffifiddle
220		Description:
221
222		More fiddling with the FFI interface:
223
224		- Make constness 'c instead of rw wherever possible.  This eliminates
225		the need for certain explicit coercions.  (However, due to ML's
226		value polymorphism, there will still be many cases where explicit
227		coercions are necessary.  Phantom types are not the whole answer
228		to modeling a subtyping relationship in ML.)
229
230		- ro/rw coersions for pointers added.  (Avoids the detour through */&.)
231
232		- "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
233		clumsy workaround for varargs problem.)
234
235		----------------------------------------------------------------------
236		Name: Lal George
237		Date: 2002/01/15
238		Tag: <none>
239		Description:
240
241		1. Since COPY instructions are no longer native to the architecture,
242		a generic functor can be used to implement the expandCopies function.
243
244		2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
245		TEXT segment.
246
247		----------------------------------------------------------------------
248		Name: Matthias Blume
249		Date: 2002/01/15
250		Tag: blume-20020115-ffiupdates
251		Description:
252
253		1. Fix for bug resulting in single-precision float values being returned
254		incorrectly from FFI calls.
255
256		2. Small modifications to C FFI API:
257
258		- memory-allocation routines return straight objects (no options)
259		and raise an exception in out-of-memory situations
260		- unsafe extensions to cast between function pointers and pointers
261		from/to ints
262		- added structure C_Debug as an alternative to structure C where
263		pointer-dereferencing (|*| and |*!) always check for null-pointers
264		- added open_lib' to DynLinkage;  open_lib' works like open_lib
265		but also takes a (possibly empty) list of existing library handles
266		that the current library depends on
267
268		----------------------------------------------------------------------
269		Name: Matthias Blume
270		Date: 2002/01/10
271		Tag: blume-20020110-newffigen
272		Description:
273
274		1. Updates to portable graph code.
275
276		2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
277		more scalable now so that even huge interfaces such as the one
278		for GTK compile in finite time and space. :-)
279		See src/ml-nlffigen/README for details on what's new.
280
281		----------------------------------------------------------------------
282		Name: Lal George
283		Date: 2001/01/09 14:31:35 EST 2002
284		Tag: george-20011206-rm-native-copy
285		Description:
286
287		Removed the native COPY and FCOPY instructions
288		from all the architectures and replaced it with the
289		explicit COPY instruction from the previous commit.
290
291		It is now possible to simplify many of the optimizations
292		modules that manipulate copies. This has not been
293		done in this change.
294
295		----------------------------------------------------------------------
296		Name: Lal George
297		Date: 2001/12/06 16:50:13 EST 2001
298		Tag: george-20011206-mlrisc-instruction
299		Description:
300
301		Changed the representation of instructions from being fully abstract
302		to being partially concrete. That is to say:
303
304		from
305		type instruction
306
307		to
308		type instr                              (* machine instruction *)
309
310		datatype instruction =
311		LIVE of {regs: C.cellset, spilled: C.cellset}
312		| KILL of {regs: C.cellset, spilled: C.cellset}
313		| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
314		| ANNOTATION of {i: instruction, a: Annotations.annotation}
315		| INSTR of instr
316
317		This makes the handling of certain special instructions that appear on
318		all architectures easier and uniform.
319
320		LIVE and KILL say that a list of registers are live or killed at the
321		program point where they appear. No spill code is generated when an
322		element of the 'regs' field is spilled, but the register is moved to
323		the 'spilled' (which is present, more for debugging than anything else).
324
325		LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
326		We used to generate:
327
328		DEFFREG f1
329		f1 := f2 + f3
330		trapb
331
332		but now generate:
333
334		f1 := f2 + f3
335		trapb
336		LIVE {regs=[f1,f2,f3], spilled=[]}
337
338		Furthermore, the DEFFREG (hack) required that all floating point instruction
339		use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
340		defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
341		in a cleaner alpha implementation. (Hopefully, intel will not get rid of
342		this architecture).
343
344		COPYXXX is intended to replace the parallel COPY and FCOPY  available on
345		all the architectures. This will result in further simplification of the
346		register allocator that must be aware of them for coalescing purposes, and
347		will also simplify certain aspects of the machine description that provides
348		callbacks related to parallel copies.
349
350		ANNOTATION should be obvious, and now INSTR represents the honest to God
351		machine instruction set!
352
353		The <arch>/instructions/<arch>Instr.sml files define certain utility
354		functions for making porting easier -- essentially converting upper case
355		to lower case. All machine instructions (of type instr) are in upper case,
356		and the lower case form generates an MLRISC instruction. For example on
357		the alpha we have:
358
359		datatype instr =
360		LDA of {r:cell, b:cell, d:operand}
361		| ...
362
363		val lda : {r:cell, b:cell, d:operand} -> instruction
364		...
365
366		where lda is just (INSTR o LDA), etc.
367
368		----------------------------------------------------------------------
369		Name: Matthias Blume
370		Date: 2001/11/22 21:40:00 EST
371		Tag: Release_110_37
372		Description:
373
374		Release 110.37.  This time for real.
375
376		----------------------------------------------------------------------
377		Name: Matthias Blume
378		Date: 2001/11/21 16:35:00 EST
379		Tag: blume-20011121-foot-in-mouth
380		Description:
381
382		Removed the "Release_110_37" tag because of a serious bug.
383		This will be re-tagged once the bug is fixed.
384
385		----------------------------------------------------------------------
386		Name: Matthias Blume
387		Date: 2001/11/21 16:14:00 EST
388		Tag: blume-20011121-forgottenfile
389		Description:
390
391		Forgot to add a file.  (Just a .tex-file -- part of
392		the CM manual source.)
393
394		----------------------------------------------------------------------
395		Name: Matthias Blume
396		Date: 2001/11/21 16:10:00 EST
397		Tag: blume-20011121-invalid_110_37
398		Description:
399
400		Note: I removed the original tag "Release_110_37" from this commit
401		because we found a serious bug in all non-x86 backends.
402		- Matthias
403
404		1. Modifications to the SML/NJ code generator and to the runtime system
405		so that code object name strings are directly inserted into code
406		objects at code generation time.  The only business the runtime system
407		has with this is now to read the name strings on occasions.
408		(The encoding of the name string has also changed somewhat.)
409
410		2. CM now implements a simple "set calculus" for specifying export lists.
411		In particular, it is now possible to refer to the export lists of
412		other libraries/groups/sources and form unions as well as differences.
413		See the latest CM manual for details.
414
415		3. An separate notion of "proxy" libraries has again be eliminated from
416		CM's model.  (Proxy libraries are now simply a special case of using
417		the export list calculus.)
418
419		4. Some of the existing libraries now take advantage of the new set
420		calculus.
421		(Notice that not all libraries have been converted because some
422		of the existing .cm-files are supposed to be backward compatible
423		with 110.0.x.)
424
425		5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
426		"General.exnMessage"!  The former relies on a certain hook to be
427		initialized, and that often does not happen in the stand-alone case.)
428
429		----------------------------------------------------------------------
430		Name: Lal George
431		Date: 2001/11/21  13:56:18 EST
432		Tag: george-2001121-pseudo-ops
433		Description:
434
435		Implemented a complete redesign of MLRISC pseudo-ops. Now there
436		ought to never be any question of incompatabilities with
437		pseudo-op syntax expected by host assemblers.
438
439		For now, only modules supporting GAS syntax are implemented
440		but more should follow, such as MASM, and vendor assembler
441		syntax, e.g. IBM as, Sun as, etc.
442
443		----------------------------------------------------------------------
444		Name: Matthias Blume
445		Date: 2001/11/14 11:52:00 EST
446		Tag: blume-20011114-srcname
447		Description:
448
449		1. Routed the name of the current source file to mlriscgen where it
450		should be directly emitted into the code object.  (This last part
451		is yet to be done.)
452
453		2. Some cleanup of the pgraph code to make it match the proposal that
454		I put out the other day.  (The proposal notwithstanding, things are
455		still in flux here.)
456
457		----------------------------------------------------------------------
458		Name: Lal George
459		Date: 2001/11/14 09:44:04 EST
460		Tag:
461		Description:
462
463		Fix for a backpatching bug reported by Allen.
464
465		Because the boundary between short and long span-dependent
466		instructions is +/- 128, there are an astounding number of
467		span-dependent instructions whose size is over estimated.
468
469		Allen came up with the idea of letting the size of span
470		dependent instructions be non-monotonic, for a maxIter
471		number of times, after which the size must be monotonically
472		increasing.
473
474		This table shows the number of span-dependent instructions
475		whose size was over-estimated as a function of maxIter, for the
476		file Parse/parse/ml.grm.sml:
477
478		maxIter            # of instructions:
479		10                      687
480		20                      438
481		30                      198
482		40                        0
483
484		In compiling the compiler, there is no significant difference in
485		compilation speed between maxIter=10 and maxIter=40. Actually,
486		my measurements showed that maxIter=40 was a tad faster than
487		maxIter=10! Also 96% of the  files in the compiler reach a fix
488		point within 13 iterations, so fixing maxIter at 40, while high,
489		is okay.
490
491		----------------------------------------------------------------------
492	Name: Matthias Blume	Name: Matthias Blume
493	Date: 2001/10/31 15:25:00 EST	Date: 2001/10/31 15:25:00 EST
494	Tag: blume-20011031-pgraph	Tag: blume-20011031-pgraph

---

Colored Diff Long Colored Diff Unidiff Context Diff Side by Side

Legend: Removed from v.975   changed lines   Added in v.1048

---

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