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 641 - (view) (download)

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

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