Annotation of /sml/trunk/HISTORY

Parent Directory | Revision Log

---

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

---

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