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/NOTES/HISTORY
ViewVC logotype

Annotation of /sml/trunk/NOTES/HISTORY

Parent Directory Parent Directory | Revision Log Revision Log


Revision 633 - (view) (download)
Original Path: sml/trunk/HISTORY

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

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