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

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