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

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