Home My Page Projects Code Snippets Project Openings SML/NJ
Summary Activity Forums Tracker Lists Tasks Docs Surveys News SCM Files

SCM Repository

[smlnj] Annotation of /sml/trunk/HISTORY
ViewVC logotype

Annotation of /sml/trunk/HISTORY

Parent Directory Parent Directory | Revision Log Revision Log


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

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