Annotation of /sml/trunk/HISTORY

Parent Directory | Revision Log

---

Revision 646 - (view) (download)

1 :	dbm	570	This is the HISTORY file for the Yale SML/NJ CVS repository.
2 :
3 :			An entry should be made for _every_ commit to the repository.
4 :			The entries in this file will be used when creating the README
5 :			for new versions, so keep that in mind when writing the
6 :			description.
7 :
8 :			The form of an entry should be:
9 :
10 :			Name:
11 :			Date:
12 :			Tag: <post-commit CVS tag>
13 :			Description:
14 :	leunga	591	----------------------------------------------------------------------
15 :	leunga	646	Name: Allen Leung
16 :			Date: 2000/05/15 16:02 EDT
17 :			Tag: leunga-20000515-alpha-x86-ra
18 :			Description:
19 :
20 :			1. Alpha
21 :
22 :			Slight cleanup. Removed the instruction SGNXL
23 :
24 :			2. X86
25 :
26 :			Added the following instructions to the instruction set:
27 :
28 :			ROLx, RORx,
29 :			BTx, BTSx, BTLx, BTRx,
30 :			XCHGx, and variants with the LOCK prefix
31 :
32 :			3. Register Allocation
33 :
34 :			The module ra-rewrite-with-renaming has been improved.
35 :
36 :			----------------------------------------------------------------------
37 :	blume	642	Name: Matthias Blume
38 :	blume	645	Date: 2000/05/15 16:20:00 JST
39 :			Tag: blume-20000515-lightrebuild
40 :			Description:
41 :
42 :			1. I added an alternative to "-rebuild" to "makeml". The difference is
43 :			that prior to calling CMB.make' the CM-variable "LIGHT" will be
44 :			defined. In effect, the command will not build any cross-compiler
45 :			backends and therefore finish more quickly.
46 :
47 :			The "fixpt" script also takes a "-light" switch to be able to use
48 :			this new facility while compiling for a fixpoint.
49 :
50 :			2. I replaced all mentions of anchored paths in group owner specifications
51 :			with simple relative paths (usually starting with "..").
52 :			The rationale is that a library's internal workings should not be
53 :			compromised by the lack of some anchor. (An anchor is necessary
54 :			for someone who wants to refer to the library by an anchored path,
55 :			but it should not be necessary to build the same library in the first
56 :			place.)
57 :
58 :			3. I changed the way CM's tool mechanism determines the shell command
59 :			string used for things like ml-yacc etc. so that it does not break
60 :			when CM.Control.implicit_anchors is turned off.
61 :
62 :			----------------------------------------------------------------------
63 :			Name: Matthias Blume
64 :	blume	644	Date: 2000/05/12 18:20:00 JST
65 :			Tag: blume-20000512-ml-build
66 :			Description:
67 :
68 :			Fixed a bug in config/_ml-build that prevented ml-yacc and ml-lex from
69 :			getting installed properly (by config/install.sh).
70 :
71 :			----------------------------------------------------------------------
72 :			Name: Matthias Blume
73 :	blume	643	Date: 2000/05/12 17:30:00 JST
74 :			Tag: blume-20000512-anchors
75 :			Description:
76 :
77 :			!!! NEW BOOT FILES !!!
78 :
79 :			This change is in preparation of fading out support for "implicitly
80 :			anchored path names". I went through all sources and used the
81 :			explicit (and relatively new) $-notation. See system/README and the
82 :			CM manual for more info on this.
83 :
84 :			I also modified the anchoring scheme for some things such as "smlnj",
85 :			"MLRISC", "cm", etc. to take advantage of the fact that explicit
86 :			anchors are more expressive: anchor name and first arc do not have to
87 :			coincide. This entails the following user-visible change:
88 :
89 :			You have to write $smlnj/foo/bar instead of smlnj/foo/bar. In
90 :			particular, when you fire up sml with a command-line argument, say,
91 :			e.g.:
92 :
93 :			sml '$smlnj/cmb.cm'
94 :
95 :			At the ML toplevel prompt:
96 :
97 :			CM.autoload "$smlnj/cmb.cm";
98 :
99 :			There is also a new controller in CM.Control that can be used to turn
100 :			off all remaining support for implicit anchors by saying:
101 :
102 :			CM.autoload "$smlnj/
103 :			#set CM.Control.implicit_anchors false;
104 :
105 :			This causes CM to reject implicitly anchored paths. This is (for the
106 :			time being) less permissive than the "final" version where there will
107 :			be no more such implicit anchors and relative paths will be just that:
108 :			relative.
109 :
110 :			The next step (version after next version?) will be to make the
111 :			default for CM.Control.implicit_anchors false. After the dust has
112 :			settled, I can then produce the "final" version of this...
113 :
114 :			Note: Since bootstrapping is a bit tricky, I provided new boot files.
115 :
116 :			----------------------------------------------------------------------
117 :			Name: Matthias Blume
118 :	blume	642	Date: 2000/05/11 16:30:00 JST
119 :			Tag: blume-20000511-sources
120 :			Description:
121 :
122 :			The main change is that I added function CM.sources as a generalized
123 :			version of the earlier CM.makedepend. This entails the following
124 :			additional changes:
125 :
126 :			- CM.makedepend has been dropped.
127 :
128 :			- CM manual has been updated.
129 :
130 :			- TOOLS signature and API have been changed.
131 :
132 :			----------------------------------------------------------------------
133 :	leunga	640	Name: Allen Leung
134 :	leunga	641	Date: 2000/05/10 21:17 EDT
135 :			Tag: leunga-20000510-moby-c--ssa
136 :			Description:
137 :
138 :			Various bug fixes and new features for C--, Moby and MLRISC optimizations.
139 :			None of these affect SML/NJ.
140 :
141 :			1. Register Allocation
142 :
143 :			a. A new ra spilling module (ra/ra-spill-with-renaming) is implemented.
144 :			This module tries to remove local (i.e. basic block level) redundancies
145 :			during spilling.
146 :
147 :			b. A new framework for performing region based register allocation.
148 :			Not yet entirely functional.
149 :
150 :			2. X86
151 :
152 :			a. DefUse for POP was missing the stack pointer [found by Lal]
153 :			b. Reload for CALL was incorrect in X86Spill [found by John]
154 :			c. Various fixes in X86Spill so that it can be used correctly for
155 :			the new spilling module.
156 :
157 :			3. SSA/IR
158 :
159 :			a. New module ir/dj-dataflow.sml implements elimination based
160 :			data flow analysis.
161 :
162 :			4. MLRiscGen
163 :
164 :			a. Fix for gc type annotation
165 :
166 :			5. MDGen
167 :
168 :			Various fixes for machine description -> ml code translation. For ssa
169 :			only.
170 :
171 :			----------------------------------------------------------------------
172 :			Name: Allen Leung
173 :	leunga	640	Date: 2000/05/08 22:17 EDT
174 :			Tag: leunga-20000508-labexp
175 :			Description:
176 :
177 :			Fermin has found a few assembly problems with constant expressions
178 :			generated in LabelExp. Mostly, the problems involve extra parentheses,
179 :			which choke on dumb assemblers. This is his fix.
180 :
181 :			----------------------------------------------------------------------
182 :	dbm	635	Name: Dave MacQueen
183 :			Date: 2000/04/09 14:00 EDT
184 :			Tag: dbm-20000502-Version_110_28
185 :			Description:
186 :
187 :			1. Updated src/compiler/TopLevel/main/version.sml to version 110.28
188 :
189 :			2. Updated config/version to 110.28
190 :
191 :			3. Updated config/srcarchiveurl
192 :
193 :			3. New boot files!
194 :			ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.28/
195 :
196 :			----------------------------------------------------------------------
197 :	blume	634	Name: Matthias Blume
198 :			Date: 2000/05/01 19:05:00 JST
199 :			Tag: blume-20000501-noweb
200 :			Description:
201 :
202 :			A new noweb tool has been added. The existing system is entirely
203 :			unaffected by this, but some CM users have asked for renewed noweb
204 :			support. Everything is documented in the CM manual.
205 :
206 :			New (plugin) libraries:
207 :
208 :			noweb-tool.cm
209 :			nw-ext.cm
210 :
211 :			----------------------------------------------------------------------
212 :	dbm	633	Name: Dave MacQueen
213 :			Date: 2000/04/30 12:40PM EDT
214 :			Tag: dbm-20000430-bug_fixes
215 :			Description:
216 :
217 :			1. Fix for bug 1498
218 :			smlnj/src/system/Basis/Implementation/Unsafe/object.sig
219 :			smlnj/src/system/Basis/Implementation/Unsafe/object.sml
220 :			added toRealArray function
221 :			smlnj/src/compiler/MiscUtil/print/ppobj.sml
222 :			added check for tag Obj.RealArray to array printing case in ppObj
223 :
224 :			2. Fix for bug 1510
225 :			smlnj/src/compiler/Semant/types/typesutil.sml
226 :			fixed definition of dummyargs (used by equalTycon) so that
227 :			dummy args are distinct types
228 :
229 :			----------------------------------------------------------------------
230 :	blume	629	Name: Matthias Blume
231 :	blume	632	Date: 2000/04/30 01:00:00 JST
232 :			Tag: blume-20000430-versions
233 :			Description:
234 :
235 :			1. CM version numbering added. This is an implementation of Lal's
236 :			proposal for adding version numbers and version checking to .cm
237 :			files. Lal said that his proposal was just that -- a proposal.
238 :			For the time being I went ahead and implemented it so that people
239 :			can comment on it. Everything is completely backward-compatible
240 :			(except for the stable library format, i.e., new bootfiles!).
241 :
242 :			As usual, see the CM manual for details.
243 :
244 :			2. An alternative syntax for anchored paths has been implemented.
245 :			Dave has recently voiced the same concerns that I had when I did
246 :			this, so there should be some support. My take is that eventually
247 :			I will let support for the current syntax (where anchors are
248 :			"implicit") fade out in favor of the new, explicit syntax.
249 :			In order to be backward-compatible, both old and new syntax are
250 :			currently supported.
251 :
252 :			Again, see the CM manual for details.
253 :
254 :			3. Parallel make is trying to be slightly smarter: When the master
255 :			process finds a "bottleneck", i.e., when there is only one
256 :			compilation unit that can be compiled and everybody else is
257 :			waiting on it, then it will simply compile it directly instead
258 :			of clumsily telling one of the slaves to do it.
259 :
260 :			4. Support for "unsharing" added. This is necessary in order to be
261 :			able to have two different versions of the same library running
262 :			at the same time (e.g., for trying out a new MLRISC while still
263 :			having the old MLRISC linked into the current compiler, etc.)
264 :			See the CM manual.
265 :
266 :			5. Simple "makedepend" functionality added for generating Makefile
267 :			dependency information. (This is rather crude at the moment.
268 :			Expect some changes here in the future.)
269 :
270 :			6. ".fun" added as a recognized suffix for ML files. Also documented
271 :			explicitly in the manual that the fallback behavior (unknown suffix
272 :			-> ML file) is not an official feature!
273 :
274 :			7. Small changes to the pickler for stable libraries.
275 :
276 :			8. Several internal changes to CM (for cleanup/improvement).
277 :
278 :
279 :			!!!! NEW BINFILES !!!!
280 :
281 :			----------------------------------------------------------------------
282 :			Name: Matthias Blume
283 :	blume	631	Date: 2000/04/28 17:30:00 JST
284 :			Tag: blume-20000428-pathconfig
285 :			Description:
286 :
287 :			1. I changed config/install.sh to remove duplicate entries from the
288 :			lib/pathconfig file at the end. Moreover, the final version of
289 :			lib/pathconfig is sorted alphabetically. The same (sorting) is done
290 :			in src/system/installml.
291 :
292 :			2. The config/install.sh script now consistently uses relative
293 :			pathnames in lib/pathconfig whenever the anchor is in the lib
294 :			directory. (So far this was true for the libraries that come
295 :			pre-compiled and bundled as part of the bootfiles but not for
296 :			libraries that are compiled by the script itself.)
297 :
298 :			----------------------------------------------------------------------
299 :			Name: Matthias Blume
300 :	blume	629	Date: 2000/04/26 13:10:00 JST
301 :			Tag: blume-20000426-fun_suffix
302 :			Description:
303 :
304 :			Added ".fun" as a recognized file name suffix (for ML code).
305 :
306 :			----------------------------------------------------------------------
307 :	leunga	628	Name: Allen Leung
308 :			Date: 2000/04/25 17:00:00 EST
309 :			Tag: leunga-20000425-alpha-ra
310 :			Description:
311 :
312 :			1. Alpha
313 :
314 :			PSEUDOARITH was missing in AlphaRewrite. This causes an endless loop
315 :			in C--.
316 :
317 :			2. RA
318 :
319 :			Added a flag "ra-dump-size" to print out the size of the flowgraph
320 :			and the interference graph.
321 :
322 :			----------------------------------------------------------------------
323 :	dbm	627	Name: Dave MacQueen
324 :			Date: 2000/04/25/
325 :			Tag: dbm-20000425-mlyacc_doc_examples
326 :			Description:
327 :			Updated mlyacc.tex sections 5 and 7 for SML '97 and CM.
328 :			Updated all three examples in src/ml-yacc/examples to run
329 :			under 110.* using CM.make.
330 :
331 :			----------------------------------------------------------------------
332 :	leunga	624	Name: Allen Leung
333 :			Date: 2000/04/20 23:04:00 EST
334 :			Tag: leunga-20000420-ssa-c---stuff
335 :			Description:
336 :
337 :			This update synchronizes my repository with Yale's. Most of these
338 :			changes, however, do not affect SML/NJ at all (the RA is an exception).
339 :
340 :			1. Register Allocator
341 :
342 :			a. An improvement in the interference graph construction:
343 :			Given a copy
344 :
345 :			s <- t
346 :
347 :	leunga	628	no interference edge between s and t is added for this definition of s.
348 :	leunga	624
349 :			b. I've added two new spill heuristic modules that Fermin and I developed
350 :			(in the new library RA.cm). These are unused in SML/NJ but maybe
351 :			useful for others (Moby?)
352 :
353 :			2. X86
354 :
355 :			a. Various fixes in the backend provided by Fermin [C--] and Lal.
356 :
357 :			3. Alpha
358 :
359 :			a. Added the BSR instruction and code generation that goes with it [C--]
360 :			b. Other fixes too numerous to recount provided by Fermin [C--]
361 :
362 :			4. Regmaps
363 :
364 :			a. The regmaps are not initialized with the identity physical bindings
365 :			at creation time. This is unneeded.
366 :
367 :			5. MLRISC Optimizations
368 :
369 :			a. The DJ-Graph module can now compute the iterated dominance frontiers
370 :			intersects with liveness incrementally in linear time! Woohoo!
371 :			This is now used in my new SSA construction algorithm.
372 :
373 :			b. THe branch reorganization module is now smarter about linear chains of
374 :			basic blocks.
375 :
376 :
377 :			----------------------------------------------------------------------
378 :	blume	615	Name: Matthias Blume
379 :			Date: 2000/04/12 13:52:00 JST
380 :			Tag: blume_main_v110p27_1
381 :			Description:
382 :
383 :			Changed install.sh script to handle archive files without version number
384 :			and to use "boot.<arch>-<os>" instead of "sml.boot.<arch>-<os>" for the
385 :			name of the boot file archive.
386 :
387 :			----------------------------------------------------------------------
388 :	dbm	607	Name: Dave MacQueen
389 :			Date: 2000/04/09 14:00 EDT
390 :			Tag: dbm-20000410-Version_110_27
391 :			Description:
392 :
393 :			1. Updated src/compiler/TopLevel/main/version.sml to version 110.27
394 :
395 :			2. Updated src/config/version to 110.27
396 :
397 :			3. New boot files!
398 :
399 :			----------------------------------------------------------------------
400 :	leunga	606	Name: Allen Leung
401 :			Date: 2000/04/09 19:09:00 EST
402 :			Tag: leunga-20000409-misc
403 :			Description:
404 :
405 :			1. Yet another fix for x86 assembly for idivl, imull, mull and friends.
406 :
407 :			2. Miscellaneous improvements to MLRISC (unused in sml/nj)
408 :
409 :			----------------------------------------------------------------------
410 :	monnier	605	Name: Stefan
411 :			Date: 2000/04/07 10:00:00 EDT
412 :			Tag: monnier-20000406-branch-handling
413 :			Description:
414 :
415 :			Improved handling of branches (mostly those generated from
416 :			polymorphic equality), removed switchoff and changed the
417 :			default optimization settings (more cpsopt and less flintopt).
418 :
419 :			----------------------------------------------------------------------
420 :	leunga	601	Name: Allen Leung
421 :	leunga	602	Date: 2000/04/06 01:30:00 EST
422 :			Tag: leunga-20000406-peephole-x86-SSA-2
423 :			Description:
424 :
425 :			Forgot a few files.
426 :
427 :			----------------------------------------------------------------------
428 :			Name: Allen Leung
429 :	leunga	601	Date: 2000/04/06 00:36:00 EST
430 :			Tag: leunga-20000406-peephole-x86-SSA
431 :			Description:
432 :
433 :			1. New Peephole code
434 :
435 :			2. Minor improvement to X86 instruction selection
436 :
437 :			3. Various fixes to SSA and machine description -> code translator
438 :
439 :			----------------------------------------------------------------------
440 :	blume	595	Name: Matthias Blume
441 :			Date: 2000/04/05 12:30:00 JST
442 :			Tag: blume_main_v110p26p2_3
443 :			Description:
444 :
445 :			This update just merges three minor cosmetic updates to CM's sources
446 :			to get ready for the 110.27 code freeze on Friday. No functionality
447 :			has changed.
448 :
449 :			----------------------------------------------------------------------
450 :	leunga	593	Name: Allen Leung
451 :	leunga	594	Date: 2000/04/04 19:39:00 EST
452 :			Tag: leunga-20000404-x86-asm
453 :			Description:
454 :
455 :			1. Fixed a problem in X86 assembly.
456 :
457 :			Things like
458 :
459 :			jmp %eax
460 :			jmp (%eax)
461 :
462 :			should be output as
463 :
464 :			jmp *%eax
465 :			jmp *(%eax)
466 :
467 :			2. Assembly output
468 :
469 :			Added a new flag
470 :
471 :			"asm-indent-copies" (default to false)
472 :
473 :			When this flag is on, parallel copies will be indented an extra level.
474 :
475 :			----------------------------------------------------------------------
476 :			Name: Allen Leung
477 :	leunga	593	Date: 2000/04/04 03:18:00 EST
478 :			Tag: leunga-20000404-C--Moby
479 :			Description:
480 :
481 :			All of these fixes are related to C--, Moby, and my own optimization
482 :			stuff; so they shouldn't affect SML/NJ.
483 :
484 :			1. X86
485 :
486 :			Various fixes related floating point, and extensions.
487 :
488 :			2. Alpha
489 :
490 :			Some extra patterns related to loads with signed/zero extension
491 :			provided by Fermin.
492 :
493 :			3. Assembly
494 :
495 :			When generating assemby, resolve the value of client defined constants,
496 :			instead of generating symbolic values. This is controlled by the
497 :			new flag "asm-resolve-constants", which is default to true.
498 :
499 :			4. Machine Descriptions
500 :
501 :			a. The precedence parser was slightly broken when parsing infixr symbols.
502 :			b. The type generalizing code had the bound variables reversed, resulting
503 :			in a problem during arity raising.
504 :			c. Various fixes in machine descriptions.
505 :
506 :			----------------------------------------------------------------------
507 :	blume	592	Name: Matthias Blume
508 :			Date: 2000/04/03 16:05:00 JST
509 :			Tag: blume_main_v110p26p2_2
510 :			Description:
511 :
512 :			I eliminated coreEnv from compInfo. Access to the "Core" structure is
513 :			now done via the ordinary static environment that is context to each
514 :			compilation unit.
515 :
516 :			To this end, I arranged that instead of "structure Core" as "structure
517 :			_Core" is bound in the pervasive environment. Core access is done via
518 :			_Core (which can never be accidentially rebound because _Core is not a
519 :			legal surface-syntax symbol).
520 :
521 :			The current solution is much cleaner because the core environment is
522 :			now simply part of the pervasive environment which is part of every
523 :			compilation unit's context anyway. In particular, this eliminates all
524 :			special-case handling that was necessary until now in order to deal
525 :			with dynamic and symbolic parts of the core environment.
526 :
527 :			Remaining hackery (to bind the "magic" symbol _Core) is localized in the
528 :			compilation mananger's bootstrap compiler (actually: in the "init group"
529 :			handling). See the comments in src/system/smlnj/init/init.cmi for
530 :			more details.
531 :
532 :			I also tried to track down all mentions of "Core" (as string argument
533 :			to Symbol.strSymbol) in the compiler and replaced them with a
534 :			reference to the new CoreSym.coreSym. Seems cleaner since the actual
535 :			name appears in one place only.
536 :
537 :			Binfile and bootfile format have not changed, but the switchover from
538 :			the old "init.cmi" to the new one is a bit tricky, so I supplied new
539 :			bootfiles anyway.
540 :
541 :			----------------------------------------------------------------------
542 :	leunga	591	Name: Allen Leung
543 :			Date: 2000/04/02 21:17:00 EST
544 :			Tag: leunga-20000402-mltree
545 :			Description:
546 :	leunga	585
547 :	leunga	591	1. Renamed the constructor CALL in MLTREE by popular demand.
548 :			2. Added a bunch of files from my repository. These are currently
549 :			used by other non-SMLNJ backends.
550 :
551 :	leunga	576	----------------------------------------------------------------------
552 :	leunga	590	Name: Allen Leung
553 :			Date: 2000/03/31 21:15:00 EST
554 :			Tag: leunga-20000331-aliasing
555 :			Description:
556 :
557 :			This update contains a rewritten (and hopefully more correct) module
558 :			for extracting aliasing information from CPS.
559 :
560 :			To turn on this feature:
561 :
562 :			Compiler.Control.CG.memDisambiguate := true
563 :
564 :			To pretty print the region information with assembly
565 :
566 :			Compiler.Control.MLRISC.getFlag "asm-show-region" := true;
567 :
568 :			To control how many levels of aliasing information are printed, use:
569 :
570 :			Compiler.Control.MLRISC.getInt "points-to-show-level" := n
571 :
572 :			The default of n is 3.
573 :
574 :			----------------------------------------------------------------------
575 :	dbm	589	Name: David MacQueen
576 :			Date: 2000/03/31 11:15:00 EST
577 :			Tag: dbm-20000331-runtime_fix
578 :			Description:
579 :
580 :			This update contains:
581 :
582 :			1. runtime/c-lib/c-libraries.c
583 :			includes added in revision 1.2 caused compilation errors on hppa-hpux
584 :
585 :			2. fix for bug 1556
586 :			system/Basis/Implementation/NJ/internal-signals.sml
587 :
588 :			----------------------------------------------------------------------
589 :	blume	587	Name: Matthias Blume
590 :	blume	588	Date: 2000/03/31 18:00:00 JST
591 :			Tag: blume_main_v110p26p2_1
592 :			Description:
593 :
594 :			This update contains:
595 :
596 :			1. A small change to CM's handling of stable libraries:
597 :			CM now maintains one "global" modmap that is used for all stable
598 :			libraries. The use of such a global modmap maximizes sharing and
599 :			minimizes the need for re-traversing parts of environments during
600 :			modmap construction. (However, this has minor impact since modmap
601 :			construction seems to account for just one percent or less of total
602 :			compile time.)
603 :
604 :			2. I added a "genmap" phase to the statistics. This is where I got the
605 :			"one percent" number (see above).
606 :
607 :			3. CM's new tool parameter mechanism just became _even_ better. :)
608 :			- The parser understands named parameters and recursive options.
609 :			- The "make" and "shell" tools use these new features.
610 :			(This makes it a lot easier to cascade these tools.)
611 :			- There is a small syntax change: named parameters use a
612 :
613 :			<name> : ( <option> ... ) or
614 :			<name> : <string>
615 :
616 :			syntax. Previously, named parameters were implemented in an
617 :			ad-hoc fashion by each tool individually (by parsing strings)
618 :			and had the form
619 :
620 :			<name>=<string>
621 :
622 :			See the CM manual for a full description of these issues.
623 :
624 :			----------------------------------------------------------------------
625 :			Name: Matthias Blume
626 :	blume	587	Date: 2000/03/30 18:00:00 JST
627 :			Tag: blume_main_v110p26p2_0
628 :			Description:
629 :
630 :			!!!!! WARNING !!!!!!
631 :			!! New binfiles !!
632 :			!!!!!!!!!!!!!!!!!!!!
633 :
634 :			This update contains:
635 :
636 :			1. Moderate changes to CM:
637 :
638 :			- Changes to CM's tools mechanism. In particular, it is now possible
639 :			to have tools that accept additional "command line" parameters
640 :			(specified in the .cm file at each instance where the tool's class is
641 :			used).
642 :
643 :			This was done to accomodate the new "make" and "shell" tools which
644 :			facilitate fairly seemless hookup to portions of code managed using
645 :			Makefiles or Shell scripts.
646 :
647 :			There are no classes "shared" or "private" anymore. Instead, the
648 :			sharing annotation is now a parameter to the "sml" class.
649 :
650 :			There is a bit of generic machinery for implementing one's own
651 :			tools that accept command-line parameters. However, I am not yet fully
652 :			satisfied with that part, so expect changes here in the future.
653 :
654 :			All existing tools are described in the CM manual.
655 :
656 :			- Slightly better error handling. (CM now surpresses many followup
657 :			error messages that tended to be more annoying than helpful.)
658 :
659 :			2. Major changes to the compiler's static environment data structures.
660 :
661 :			- no CMStaticEnv anymore.
662 :			- no CMEnv, no "BareEnvironment" (actually, _only_ BareEnvironment,
663 :			but it is called Environment), no conversions between different
664 :			kinds of static environments
665 :
666 :			- There is still a notion of a "modmap", but such modmaps are generated
667 :			on demand at the time when they are needed. This sounds slow, but I
668 :			sped up the code that generates modmaps enough for this not to lead to
669 :			a slowdown of the compiler (at least I didn't detect any).
670 :
671 :			- To facilitate rapid modmap generation, static environments now
672 :			contain an (optional) "modtree" structure. Modtree annotations are
673 :			constructed by the unpickler during unpickling. (This means that
674 :			the elaborator does not have to worry about modtrees at all.)
675 :			Modtrees have the advantage that they are compositional in the same
676 :			way as the environment data structure itself is compositional.
677 :			As a result, modtrees never hang on to parts of an environment that
678 :			has already been rendered "stale" by filtering or rebinding.
679 :
680 :			- I went through many, many trials and errors before arriving at the
681 :			current solution. (The initial idea of "linkpaths" did not work.)
682 :			But the result of all this is that I have touched a lot of files that
683 :			depend on the "modules" and "types" data structures (most of the
684 :			elaborator). There were a lot of changes during my "linkpath" trials
685 :			that could have been reverted to their original state but weren't.
686 :			Please, don't be too harsh on me for messing with this code a bit more
687 :			than what was strictly necessary... (I _did_ resist the tempation
688 :			of doing any "global reformatting" to avoid an untimely death at
689 :			Dave's hands. :)
690 :
691 :			- One positive aspect of the previous point: At least I made sure that
692 :			all files that I touched now compile without warnings (other than
693 :			"polyEqual").
694 :
695 :			- compiler now tends to run "leaner" (i.e., ties up less memory in
696 :			redundant modmaps)
697 :
698 :			----------------------------------------------------------------------
699 :	leunga	580	Name: Allen Leung
700 :	leunga	585	Date: 2000/03/29 18:00:00
701 :			Tag: leunga-20000327-mlriscGen_hppa_alpha_x86
702 :	leunga	586	Boot files (optional): ftp://react-ilp.cs.nyu.edu/leunga/110.26.1-sml.boot.x86-unix-20000330.tar.gz
703 :	leunga	585	Description:
704 :
705 :			This update contains *MAJOR* changes to the way code is generated from CPS
706 :			in the module mlriscGen, and in various backend modules.
707 :
708 :			CHANGES
709 :			=======
710 :
711 :			1. MLRiscGen: forward propagation fix.
712 :
713 :			There was a bug in forward propagation introduced at about the same time
714 :			as the MLRISC x86 backend, which prohibits coalescing to be
715 :			performed effectively in loops.
716 :
717 :			Effect: speed up of loops in RISC architectures.
718 :			By itself, this actually slowed down certain benchmarks on the x86.
719 :
720 :			2. MLRiscGen: forward propagating addresses from consing.
721 :
722 :			I've changed the way consing code is generated. Basically I separated
723 :			out the initialization part:
724 :
725 :			store tag, offset(allocptr)
726 :			store elem1, offset+4(allocptr)
727 :			store elem2, offset+8(allocptr)
728 :			...
729 :			store elemn, offset+4n(allocptr)
730 :
731 :			and the address computation part:
732 :
733 :			celladdr <- offset+4+alloctpr
734 :
735 :			and move the address computation part
736 :
737 :			Effect: register pressure is generally lower as a result. This
738 :			makes compilation of certain expressions much faster, such as
739 :			long lists with non-trivial elements.
740 :
741 :			[(0,0), (0,0), .... (0,0)]
742 :
743 :			3. MLRiscGen: base pointer elimination.
744 :
745 :			As part of the linkage mechanism, we generate the sequence:
746 :
747 :			L: ... <- start of the code fragment
748 :
749 :			L1:
750 :			base pointer <- linkreg - L1 + L
751 :
752 :			The base pointer was then used for computing relocatable addresses
753 :			in the code fragment. Frequently (such as in lots of continuations)
754 :			this is not needed. We now eliminate this sequence whenever possible.
755 :
756 :			For compile time efficiency, I'm using a very stupid local heuristic.
757 :			But in general, this should be done as a control flow analysis.
758 :
759 :			Effect: Smaller code size. Speed up of most programs.
760 :
761 :			4. Hppa back end
762 :
763 :			Long jumps in span dependence resolution used to depend on the existence
764 :			of the base pointer.
765 :
766 :			A jump to a long label L was expanded into the following sequence:
767 :
768 :			LDIL %hi(L-8192), %r29
769 :			LDO %lo(L-8192)(%r29), %r29
770 :			ADD %r29, baseptr, %r29
771 :			BV,n %r0(%r29)
772 :
773 :			In the presence of change (3) above, this will not work. I've changed
774 :			it so that the following sequence of instructions are generated, which
775 :			doesn't mention the base pointer at all:
776 :
777 :			BL,n L', %r29 /* branch and link, L' + 4 -> %r29 */
778 :			L': ADDIL L-(L'+4), %r29 /* Compute address of L */
779 :			BV,n %r0(%r29) /* Jump */
780 :
781 :			5. Alpha back end
782 :
783 :			New alpha instructions LDB/LDW have been added, as per Fermin's
784 :			suggestions. This is unrelated to all other changes.
785 :
786 :			6. X86 back end
787 :
788 :			I've changed andl to testl in the floating point test sequence
789 :			whenever appropriate. The Intel optimization guide states that
790 :			testl is perferable to andl.
791 :
792 :			7. RA (x86 only)
793 :
794 :			I've improved the spill propagation algorithm, using an approximation
795 :			of maximal weighted independent sets. This seems to be necessary to
796 :			alleviate the negative effect in light of the slow down in (1).
797 :
798 :			I'll write down the algorithm one of these days.
799 :
800 :			8. MLRiscGen: frequencies
801 :
802 :			I've added an annotation that states that all call gc blocks have zero
803 :			execution frequencies. This improves register allocation on the x86.
804 :
805 :			BENCHMARKS
806 :			==========
807 :
808 :			I've only perform the comparison on 110.25.
809 :
810 :			The platforms are:
811 :
812 :			HPPA A four processor HP machine (E9000) with 5G of memory.
813 :			X86 A 300Hhz Pentium II with 128M of memory, and
814 :			SPARC An Ultra sparc 2 with 512M of memory.
815 :
816 :			I used the following parameters for the SML benchmarks:
817 :
818 :			@SMLalloc
819 :			HPPA 256k
820 :			SPARC 512k
821 :			X86 256k
822 :
823 :			COMPILATION TIME
824 :			----------------
825 :			Here are the numbers comparing the compilation times of the compilers.
826 :			I've only compared 110.25 compiling the new sources versus
827 :			a fixpoint version of the new compiler compiling the same.
828 :
829 :			110.25 New
830 :			Total Time in RA Spill+Reload Total Time In RA Spill+Reload
831 :			HPPA 627s 116s 2684+3584 599s 95s 1003+1879
832 :			SPARC 892s 173s 2891+3870 708s 116s 1004+1880
833 :			X86 999s 315s 94006+130691 987s 296s 108877+141957
834 :
835 :			110.25 New
836 :			Code Size Code Size
837 :			HPPA 8596736 8561421
838 :			SPARC 8974299 8785143
839 :			X86 9029180 8716783
840 :
841 :			So in summary, things are at least as good as before. Dramatic
842 :			reduction in compilation is obtained on the Sparc; I can't explain it,
843 :			but it is reproducible. Perhaps someone should try to reproduce this
844 :			on their own machines.
845 :
846 :			SML BENCHMARKS
847 :			--------------
848 :
849 :			On the average, all benchmarks perform at least as well as before.
850 :
851 :			HPPA Compilation Time Spill+Reload Run Time
852 :			110.25 New 110.25 New 110.25 New
853 :
854 :			barnesHut 3.158 3.015 4.75% 1+1 0+0 2.980 2.922 2.00%
855 :			boyer 6.152 5.708 7.77% 0+0 0+0 0.218 0.213 2.34%
856 :			count-graphs 1.168 1.120 4.32% 0+0 0+0 22.705 23.073 -1.60%
857 :			fft 0.877 0.792 10.74% 1+3 1+3 0.602 0.587 2.56%
858 :			knuthBendix 3.180 2.857 11.32% 0+0 0+0 0.675 0.662 2.02%
859 :			lexgen 6.190 5.290 17.01% 0+0 0+0 0.913 0.788 15.86%
860 :			life 0.803 0.703 14.22% 25+25 0+0 0.153 0.140 9.52%
861 :			logic 2.048 2.007 2.08% 6+6 1+1 4.133 4.008 3.12%
862 :			mandelbrot 0.077 0.080 -4.17% 0+0 0+0 0.765 0.712 7.49%
863 :			mlyacc 22.932 20.937 9.53% 154+181 32+57 0.468 0.430 8.91%
864 :			nucleic 5.183 5.060 2.44% 2+2 0+0 0.125 0.120 4.17%
865 :			ratio-regions 3.357 3.142 6.84% 0+0 0+0 116.225 113.173 2.70%
866 :			ray 1.283 1.290 -0.52% 0+0 0+0 2.887 2.855 1.11%
867 :			simple 6.307 6.032 4.56% 28+30 5+7 3.705 3.658 1.28%
868 :			tsp 0.888 0.862 3.09% 0+0 0+0 7.040 6.893 2.13%
869 :			vliw 24.378 23.455 3.94% 106+127 25+45 2.758 2.707 1.91%
870 :			--------------------------------------------------------------------------
871 :			Average 6.12% 4.09%
872 :
873 :			SPARC Compilation Time Spill+Reload Run Time
874 :			110.25 New 110.25 New 110.25 New
875 :
876 :			barnesHut 3.778 3.592 5.20% 2+2 0+0 3.648 3.453 5.65%
877 :			boyer 6.632 6.110 8.54% 0+0 0+0 0.258 0.242 6.90%
878 :			count-graphs 1.435 1.325 8.30% 0+0 0+0 33.672 34.737 -3.07%
879 :			fft 0.980 0.940 4.26% 3+9 2+6 0.838 0.827 1.41%
880 :			knuthBendix 3.590 3.138 14.39% 0+0 0+0 0.962 0.967 -0.52%
881 :			lexgen 6.593 6.072 8.59% 1+1 0+0 1.077 1.078 -0.15%
882 :			life 0.972 0.868 11.90% 26+26 0+0 0.143 0.140 2.38%
883 :			logic 2.525 2.387 5.80% 7+7 1+1 5.625 5.158 9.05%
884 :			mandelbrot 0.090 0.093 -3.57% 0+0 0+0 0.855 0.728 17.39%
885 :			mlyacc 26.732 23.827 12.19% 162+189 32+57 0.550 0.560 -1.79%
886 :			nucleic 6.233 6.197 0.59% 3+3 0+0 0.163 0.173 -5.77%
887 :			ratio-regions 3.780 3.507 7.79% 0+0 0+0 133.993 131.035 2.26%
888 :			ray 1.595 1.550 2.90% 1+1 0+0 3.440 3.418 0.63%
889 :			simple 6.972 6.487 7.48% 29+32 5+7 3.523 3.525 -0.05%
890 :			tsp 1.115 1.063 4.86% 0+0 0+0 7.393 7.265 1.77%
891 :			vliw 27.765 24.818 11.87% 110+135 25+45 2.265 2.135 6.09%
892 :			----------------------------------------------------------------------------
893 :			Average 6.94% 2.64%
894 :
895 :			X86 Compilation Time Spill+Reload Run Time
896 :			110.25 New 110.25 New 110.25 New
897 :
898 :			barnesHut 5.530 5.420 2.03% 593+893 597+915 3.532 3.440 2.66%
899 :			boyer 8.768 7.747 13.19% 493+199 301+289 0.327 0.297 10.11%
900 :			count-graphs 2.040 2.010 1.49% 298+394 315+457 26.578 28.660 -7.26%
901 :			fft 1.327 1.302 1.92% 112+209 115+210 1.055 0.962 9.71%
902 :			knuthBendix 5.218 5.475 -4.69% 451+598 510+650 0.928 0.932 -0.36%
903 :			lexgen 9.970 9.623 3.60% 1014+841 1157+885 0.947 0.928 1.97%
904 :			life 1.183 1.183 0.00% 162+182 145+148 0.127 0.103 22.58%
905 :			logic 3.285 3.512 -6.45% 514+684 591+836 5.682 5.577 1.88%
906 :			mandelbrot 0.147 0.143 2.33% 38+41 33+54 0.703 0.690 1.93%
907 :			mlyacc 35.457 32.763 8.22% 3496+4564 3611+4860 0.552 0.550 0.30%
908 :			nucleic 7.100 6.888 3.07% 239+168 201+158 0.175 0.173 0.96%
909 :			ratio-regions 6.388 6.843 -6.65% 1182+257 981+300 120.142 120.345 -0.17%
910 :			ray 2.332 2.338 -0.29% 346+398 402+494 3.593 3.540 1.51%
911 :			simple 9.912 9.903 0.08% 1475+941 1579+1168 3.057 3.178 -3.83%
912 :			tsp 1.623 1.532 5.98% 266+200 250+211 8.045 7.878 2.12%
913 :			vliw 33.947 35.470 -4.29% 2629+2774 2877+3171 2.072 1.890 9.61%
914 :			----------------------------------------------------------------------------
915 :			Average 1.22% 3.36%
916 :
917 :			----------------------------------------------------------------------
918 :			Name: Allen Leung
919 :	leunga	583	Date: 2000/03/23 16:25:00
920 :			Tag: leunga-20000323-fix_x86_alpha
921 :			Description:
922 :
923 :			1. X86 fixes/changes
924 :
925 :			a. The old code generated for SETcc was completely wrong.
926 :			The Intel optimization guide is VERY misleading.
927 :
928 :			2. ALPHA fixes/changes
929 :
930 :			a. Added the instructions LDBU, LDWU, STB, STW as per Fermin's suggestion.
931 :			b. Added a new mode byteWordLoadStores to the functor parameter to Alpha()
932 :			c. Added reassociation code for address computation.
933 :
934 :			----------------------------------------------------------------------
935 :			Name: Allen Leung
936 :	leunga	580	Date: 2000/03/22 01:23:00
937 :			Tag: leunga-20000322-fix_x86_hppa_ra
938 :			Description:
939 :
940 :			1. X86 fixes/changes
941 :
942 :			a. x86Rewrite bug with MUL3 (found by Lal)
943 :			b. Added the instructions FSTS, FSTL
944 :
945 :			2. PA-RISC fixes/changes
946 :
947 :			a. B label should not be a delay slot candidate! Why did this work?
948 :			b. ADDT(32, REG(32, r), LI n) now generates one instruction instead of two,
949 :			as it should be.
950 :			c. The assembly syntax for fstds and fstdd was wrong.
951 :			d. Added the composite instruction COMICLR/LDO, which is the immediate
952 :			operand variant of COMCLR/LDO.
953 :
954 :			3. Generic MLRISC
955 :
956 :			a. shuffle.sml rewritten to be slightly more efficient
957 :			b. DIV bug in mltree-simplify fixed (found by Fermin)
958 :
959 :			4. Register Allocator
960 :
961 :			a. I now release the interference graph earlier during spilling.
962 :			May improve memory usage.
963 :
964 :			----------------------------------------------------------------------
965 :	blume	577	Name: Matthias Blume
966 :	blume	578	Date: 2000/03/14 14:15:32
967 :			Tag: blume_main_v110p26p1_2
968 :			Description:
969 :
970 :			1. Tools.registerStdShellCmdTool (from smlnj/cm/tool.cm) takes an
971 :			additional argument called "template" which is an optional string that
972 :			specifiel the layout of the tool command line. See the CM manual for
973 :			explanation.
974 :
975 :			2. A special-purpose tool can be "regisitered" by simply dropping the
976 :			corresponding <...>-tool.cm (and/or <...>-ext.cm) into the same
977 :			directory where the .cm file lives that uses this tool. (The
978 :			behavior/misfeature until now was to look for the tool description
979 :			files in the current working directory.) As before, tool description
980 :			files could also be anchored -- in which case they can live anywhere
981 :			they like. Following the recent e-mail discussion, this change should
982 :			make it easier to have special-purpose tools that are shipped together
983 :			with the sources of the program that uses them.
984 :
985 :			----------------------------------------------------------------------
986 :			Name: Matthias Blume
987 :	blume	577	Date: 2000/03/10 07:48:34
988 :			Tag: blume_main_v110p26p1_1
989 :			Description:
990 :
991 :			I added a re-written version of Dave's fixpt script to src/system.
992 :			Changes relative to the original version:
993 :			- sh-ified (not everybody has ksh)
994 :			- automatically figures out which architecture it runs on
995 :			- uses ./makeml a bit more cleverly
996 :			- never invokes ./installml (and, thus, does not clobber your
997 :			good and working installation of sml in case something goes wrong)
998 :			- accepts max iteration count using option "-iter <n>"
999 :			- accepts a "base" name using option "-base <base>"
1000 :
1001 :			It does not build any extraneous heap images but directly rebuilds
1002 :			bin- and boot-hierarchies using makeml's "-rebuild" switch. Finally,
1003 :			it can incorporate existing bin- and boot- hierarchies. For example,
1004 :			suppose the base is set to "sml" (which is the default). Then it
1005 :			successively builds
1006 :
1007 :			sml.bin.<arch>-unix and sml.boot.<arch>-unix
1008 :			then sml1.bin.<arch>-unix and sml1.boot.<arch>-unix
1009 :			then sml2.bin.<arch>-unix and sml2.boot.<arch>-unix
1010 :			...
1011 :			then sml<n>.bin.<arch>-unix and sml<n>.boot.<arch>-unix
1012 :
1013 :			and so on. If any of these already exist, it will just use what's
1014 :			there. In particular, many people will have the initial set of bin
1015 :			and boot files around, so this saves time for at least one full
1016 :			rebuild. Having sets of the form <base><k>.{bin,boot}.<arch>-unix for
1017 :			<k>=1,2,... is normally not a good idea when invoking fixpt. However,
1018 :			they might be the result of an earlier partial run of fixpt (which
1019 :			perhaps got accidentially killed). In this case, fixpt will quickly
1020 :			move through what exists before continuing where it left off earlier,
1021 :			and, thus, saves a lot of time.
1022 :
1023 :			----------------------------------------------------------------------
1024 :	leunga	576	Name: Allen Leung
1025 :			Date: 00/03/10 02:20:00
1026 :			Tag: leunga-20000310-fix_x86_asm_ra
1027 :			Description:
1028 :	dbm	570
1029 :	leunga	576	More assembly output problems involving the indexed addressing mode
1030 :			on the x86 have been found and corrected. Thanks to Fermin Reig for the
1031 :			fix.
1032 :
1033 :			The interface and implementation of the register allocator have been changed
1034 :			slightly to accommodate the possibility to skip the register allocation
1035 :			phases completely and go directly to memory allocation. This is needed
1036 :			for C-- use.
1037 :
1038 :	dbm	570	----------------------------------------------------------------------
1039 :	blume	572	Name: Matthias Blume
1040 :	blume	575	Date: 00/03/09 10:23:53
1041 :			Tag: blume_main_v110p26p1_0
1042 :			Description:
1043 :
1044 :			* Complete re-organization of library names. Many libraries have been
1045 :			consolidated so that they share the same path anchor. For example,
1046 :			all MLRISC-related libraries are anchored at MLRISC, most libraries that
1047 :			are SML/NJ-specific are under "smlnj". Notice that names like
1048 :			host-cmb.cm or host-compiler.cm no longer exist. See system/README
1049 :			for a complete description of the new naming scheme. Quick reference:
1050 :
1051 :			host-cmb.cm -> smlnj/cmb.cm
1052 :			host-compiler.cm -> smlnj/compiler.cm
1053 :			full-cm.cm -> smlnj/cm.cm
1054 :			<arch>-<os>.cm -> smlnj/cmb/<arch>-<os>.cm
1055 :			<arch>-compiler.cm -> smlnj/compiler/<arch>.cm
1056 :
1057 :			* Bug fixes in CM.
1058 :			- exceptions in user code are being passed through (i.e., reach top level)
1059 :			- more bugs in paranoia mode fixed
1060 :			- bug related to checking group owners fixed
1061 :
1062 :			* New install.sh script that automagically fetches archive files:
1063 :			The new file config/srcarchiveurl must contain the URL of the
1064 :			(remote) directory that contains bin files (or other source archives).
1065 :			If install.sh does not find the archive locally, it tries to get
1066 :			it from that remote directory.
1067 :			This should simplify installation further: For machines that have
1068 :			access to the internet, just fetch <version>-config.tgz, unpack it,
1069 :			edit config/targets, and go (run config/install.sh). The scipt will
1070 :			fetch everything else that it might need all by itself.
1071 :
1072 :			For CVS users, this mechanism is not relevant for source archives, but
1073 :			it is convenient for getting new sets of binfiles.
1074 :
1075 :			Archives should be tar files compressed with either gzip, compress, or
1076 :			bzip2. The script recognizes .tgz, .tar, tar.gz, tz, .tar.Z, and .tar.bz2.
1077 :
1078 :			----------------------------------------------------------------------
1079 :			Name: Matthias Blume
1080 :	blume	572	Date: 2000/03/07 04:01:04
1081 :			Tag: blume_main_v110_26_2
1082 :	dbm	570	Description:
1083 :	blume	572	- size info in BOOTLIST
1084 :			* no fixed upper limits for number of bootfiles or length of
1085 :			bootfile names in runtime
1086 :			* falling back to old behavior if no BOOTLIST size info found
1087 :			- allocation size heuristics in .run-sml
1088 :			* tries to read cache size from /proc/cpuinfo (this is important for
1089 :			small-cache Celeron systems!)
1090 :			- install.sh robustified
1091 :			- CM manual updates
1092 :			- paranoid mode
1093 :			* no more CMB.deliver() (i.e., all done by CMB.make())
1094 :			* can re-use existing sml.boot.* files
1095 :			* init.cmi now treated as library
1096 :			* library stamps for consistency checks
1097 :			- sml.boot.<arch>-<os>/PIDMAP file
1098 :			* This file is read by the CM startup code. This is used to minimize
1099 :			the amount of dynamic state that needs to be stowed away for the
1100 :			purpose of sharing between interactive system and user code.
1101 :			- CM.Anchor.anchor instead of CM.Anchor.{set,cancel}
1102 :			* Upon request by Elsa. Anchors now controlled by get-set-pair
1103 :			like most other CM state variables.
1104 :			- Compiler.CMSA eliminated
1105 :			* No longer supported by CM anyway.
1106 :			- fixed bugs in pickler that kept biting Stefan
1107 :			* past refs to past refs (was caused by the possibility that
1108 :			ad-hoc sharing is more discriminating than hash-cons sharing)
1109 :			* integer overflow on LargeInt.minInt
1110 :			- ml-{lex,yacc} build scripts now use new mechanism
1111 :			for building standalone programs
1112 :			- fixed several gcc -Wall warnings that were caused by missing header
1113 :			files, missing initializations, etc., in runtime (not all warnings
1114 :			eliminated, though)

---

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