12 |
Tag: <post-commit CVS tag> |
Tag: <post-commit CVS tag> |
13 |
Description: |
Description: |
14 |
---------------------------------------------------------------------- |
---------------------------------------------------------------------- |
15 |
|
Name: Matthias Blume |
16 |
|
Date: 2000/06/06 11:15:00 JST |
17 |
|
Tag: blume-20000606-lazierpickle |
18 |
|
Description: |
19 |
|
|
20 |
|
!!!! NEW BOOT FILES !!!! |
21 |
|
|
22 |
|
* The main purpose of this update is to make library pickles lazier in |
23 |
|
order to reduce the initial space penalty for autoloading a library. |
24 |
|
As a result, it is now possible to have $smlnj/compiler.cm |
25 |
|
pre-registered. This should take care of the many complaints or |
26 |
|
inquiries about missing structure Compiler. This required changes to |
27 |
|
CM's internal data structures and small tweaks to some algorithms. |
28 |
|
|
29 |
|
As a neat additional effect, it is no longer necessary (for the sake |
30 |
|
of lean heap image files) to distinguish between a "minimal" CM and a |
31 |
|
"full" CM. Now, there is only one CM (i.e., the "full" version: |
32 |
|
$smlnj/cm.cm aka $smlnj/cm/full.cm), and it is always available at the |
33 |
|
interactive top level. ($smlnj/cm/minimal.cm is gone.) |
34 |
|
|
35 |
|
To make the life of compiler-hackers easier, "makeml" now also |
36 |
|
pre-registers $smlnj/cmb.cm (aka $smlnj/cmb/current.cm). In other |
37 |
|
words, after you bootstrap a new sml for the first time, you will not |
38 |
|
have to autoload $smlnj/cmb.cm again afterwards. (The first time |
39 |
|
around you will still have to do it, though.) |
40 |
|
|
41 |
|
* A second change consists of major updates to the CM manual. There |
42 |
|
are now several appendices with summary information and also a full |
43 |
|
specification of the CM description file syntax. |
44 |
|
|
45 |
|
* In directory src/system I added the script "allcross". This script |
46 |
|
invokes sml and cross-compiles the compiler for all supported |
47 |
|
architectures. (Useful when providing a new set of boot files.) |
48 |
|
|
49 |
|
* There seems to be a latent bug in my "lazy pickles" mechanism. I |
50 |
|
added a small tweak to pickle-util.sml to work around this problem, |
51 |
|
but it is not a proper fix yet. I will investigate further. (The |
52 |
|
effect of the bug was an inflation of library pickle size.) |
53 |
|
|
54 |
|
* Version number increased to 110.28.1 (to avoid compatibility problems). |
55 |
|
|
56 |
|
---------------------------------------------------------------------- |
57 |
|
Name: Allen Leung |
58 |
|
Date: 2000/05/25 17:28 EDT |
59 |
|
Tag: leunga-20000525-ra |
60 |
|
Description: |
61 |
|
|
62 |
|
Fixed a bug in freezing phase of the register allocator. |
63 |
|
|
64 |
|
---------------------------------------------------------------------- |
65 |
|
Name: Allen Leung |
66 |
|
Date: 2000/05/15 22:53 EDT |
67 |
|
Tag: leunga-20000515-alpha-x86-ra |
68 |
|
Description: |
69 |
|
|
70 |
|
1. Alpha |
71 |
|
|
72 |
|
Slight cleanup. Removed the instruction SGNXL |
73 |
|
|
74 |
|
2. X86 |
75 |
|
|
76 |
|
Added the following instructions to the instruction set: |
77 |
|
|
78 |
|
ROLx, RORx, |
79 |
|
BTx, BTSx, BTLx, BTRx, |
80 |
|
XCHGx, and variants with the LOCK prefix |
81 |
|
|
82 |
|
3. Register Allocation |
83 |
|
|
84 |
|
The module ra-rewrite-with-renaming has been improved. |
85 |
|
|
86 |
|
These have no effect on SML/NJ. |
87 |
|
|
88 |
|
---------------------------------------------------------------------- |
89 |
|
Name: Matthias Blume |
90 |
|
Date: 2000/05/15 16:20:00 JST |
91 |
|
Tag: blume-20000515-lightrebuild |
92 |
|
Description: |
93 |
|
|
94 |
|
1. I added an alternative to "-rebuild" to "makeml". The difference is |
95 |
|
that prior to calling CMB.make' the CM-variable "LIGHT" will be |
96 |
|
defined. In effect, the command will not build any cross-compiler |
97 |
|
backends and therefore finish more quickly. |
98 |
|
|
99 |
|
The "fixpt" script also takes a "-light" switch to be able to use |
100 |
|
this new facility while compiling for a fixpoint. |
101 |
|
|
102 |
|
2. I replaced all mentions of anchored paths in group owner specifications |
103 |
|
with simple relative paths (usually starting with ".."). |
104 |
|
The rationale is that a library's internal workings should not be |
105 |
|
compromised by the lack of some anchor. (An anchor is necessary |
106 |
|
for someone who wants to refer to the library by an anchored path, |
107 |
|
but it should not be necessary to build the same library in the first |
108 |
|
place.) |
109 |
|
|
110 |
|
3. I changed the way CM's tool mechanism determines the shell command |
111 |
|
string used for things like ml-yacc etc. so that it does not break |
112 |
|
when CM.Control.implicit_anchors is turned off. |
113 |
|
|
114 |
|
---------------------------------------------------------------------- |
115 |
|
Name: Matthias Blume |
116 |
|
Date: 2000/05/12 18:20:00 JST |
117 |
|
Tag: blume-20000512-ml-build |
118 |
|
Description: |
119 |
|
|
120 |
|
Fixed a bug in config/_ml-build that prevented ml-yacc and ml-lex from |
121 |
|
getting installed properly (by config/install.sh). |
122 |
|
|
123 |
|
---------------------------------------------------------------------- |
124 |
|
Name: Matthias Blume |
125 |
|
Date: 2000/05/12 17:30:00 JST |
126 |
|
Tag: blume-20000512-anchors |
127 |
|
Description: |
128 |
|
|
129 |
|
!!! NEW BOOT FILES !!! |
130 |
|
|
131 |
|
This change is in preparation of fading out support for "implicitly |
132 |
|
anchored path names". I went through all sources and used the |
133 |
|
explicit (and relatively new) $-notation. See system/README and the |
134 |
|
CM manual for more info on this. |
135 |
|
|
136 |
|
I also modified the anchoring scheme for some things such as "smlnj", |
137 |
|
"MLRISC", "cm", etc. to take advantage of the fact that explicit |
138 |
|
anchors are more expressive: anchor name and first arc do not have to |
139 |
|
coincide. This entails the following user-visible change: |
140 |
|
|
141 |
|
You have to write $smlnj/foo/bar instead of smlnj/foo/bar. In |
142 |
|
particular, when you fire up sml with a command-line argument, say, |
143 |
|
e.g.: |
144 |
|
|
145 |
|
sml '$smlnj/cmb.cm' |
146 |
|
|
147 |
|
At the ML toplevel prompt: |
148 |
|
|
149 |
|
CM.autoload "$smlnj/cmb.cm"; |
150 |
|
|
151 |
|
There is also a new controller in CM.Control that can be used to turn |
152 |
|
off all remaining support for implicit anchors by saying: |
153 |
|
|
154 |
|
CM.autoload "$smlnj/ |
155 |
|
#set CM.Control.implicit_anchors false; |
156 |
|
|
157 |
|
This causes CM to reject implicitly anchored paths. This is (for the |
158 |
|
time being) less permissive than the "final" version where there will |
159 |
|
be no more such implicit anchors and relative paths will be just that: |
160 |
|
relative. |
161 |
|
|
162 |
|
The next step (version after next version?) will be to make the |
163 |
|
default for CM.Control.implicit_anchors false. After the dust has |
164 |
|
settled, I can then produce the "final" version of this... |
165 |
|
|
166 |
|
Note: Since bootstrapping is a bit tricky, I provided new boot files. |
167 |
|
|
168 |
|
---------------------------------------------------------------------- |
169 |
|
Name: Matthias Blume |
170 |
|
Date: 2000/05/11 16:30:00 JST |
171 |
|
Tag: blume-20000511-sources |
172 |
|
Description: |
173 |
|
|
174 |
|
The main change is that I added function CM.sources as a generalized |
175 |
|
version of the earlier CM.makedepend. This entails the following |
176 |
|
additional changes: |
177 |
|
|
178 |
|
- CM.makedepend has been dropped. |
179 |
|
|
180 |
|
- CM manual has been updated. |
181 |
|
|
182 |
|
- TOOLS signature and API have been changed. |
183 |
|
|
184 |
|
---------------------------------------------------------------------- |
185 |
|
Name: Allen Leung |
186 |
|
Date: 2000/05/10 21:17 EDT |
187 |
|
Tag: leunga-20000510-moby-c--ssa |
188 |
|
Description: |
189 |
|
|
190 |
|
Various bug fixes and new features for C--, Moby and MLRISC optimizations. |
191 |
|
None of these affect SML/NJ. |
192 |
|
|
193 |
|
1. Register Allocation |
194 |
|
|
195 |
|
a. A new ra spilling module (ra/ra-spill-with-renaming) is implemented. |
196 |
|
This module tries to remove local (i.e. basic block level) redundancies |
197 |
|
during spilling. |
198 |
|
|
199 |
|
b. A new framework for performing region based register allocation. |
200 |
|
Not yet entirely functional. |
201 |
|
|
202 |
|
2. X86 |
203 |
|
|
204 |
|
a. DefUse for POP was missing the stack pointer [found by Lal] |
205 |
|
b. Reload for CALL was incorrect in X86Spill [found by John] |
206 |
|
c. Various fixes in X86Spill so that it can be used correctly for |
207 |
|
the new spilling module. |
208 |
|
|
209 |
|
3. SSA/IR |
210 |
|
|
211 |
|
a. New module ir/dj-dataflow.sml implements elimination based |
212 |
|
data flow analysis. |
213 |
|
|
214 |
|
4. MLRiscGen |
215 |
|
|
216 |
|
a. Fix for gc type annotation |
217 |
|
|
218 |
|
5. MDGen |
219 |
|
|
220 |
|
Various fixes for machine description -> ml code translation. For ssa |
221 |
|
only. |
222 |
|
|
223 |
|
---------------------------------------------------------------------- |
224 |
|
Name: Allen Leung |
225 |
|
Date: 2000/05/08 22:17 EDT |
226 |
|
Tag: leunga-20000508-labexp |
227 |
|
Description: |
228 |
|
|
229 |
|
Fermin has found a few assembly problems with constant expressions |
230 |
|
generated in LabelExp. Mostly, the problems involve extra parentheses, |
231 |
|
which choke on dumb assemblers. This is his fix. |
232 |
|
|
233 |
|
---------------------------------------------------------------------- |
234 |
|
Name: Dave MacQueen |
235 |
|
Date: 2000/04/09 14:00 EDT |
236 |
|
Tag: dbm-20000502-Version_110_28 |
237 |
|
Description: |
238 |
|
|
239 |
|
1. Updated src/compiler/TopLevel/main/version.sml to version 110.28 |
240 |
|
|
241 |
|
2. Updated config/version to 110.28 |
242 |
|
|
243 |
|
3. Updated config/srcarchiveurl |
244 |
|
|
245 |
|
3. New boot files! |
246 |
|
ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.28/ |
247 |
|
|
248 |
|
---------------------------------------------------------------------- |
249 |
|
Name: Matthias Blume |
250 |
|
Date: 2000/05/01 19:05:00 JST |
251 |
|
Tag: blume-20000501-noweb |
252 |
|
Description: |
253 |
|
|
254 |
|
A new noweb tool has been added. The existing system is entirely |
255 |
|
unaffected by this, but some CM users have asked for renewed noweb |
256 |
|
support. Everything is documented in the CM manual. |
257 |
|
|
258 |
|
New (plugin) libraries: |
259 |
|
|
260 |
|
noweb-tool.cm |
261 |
|
nw-ext.cm |
262 |
|
|
263 |
|
---------------------------------------------------------------------- |
264 |
|
Name: Dave MacQueen |
265 |
|
Date: 2000/04/30 12:40PM EDT |
266 |
|
Tag: dbm-20000430-bug_fixes |
267 |
|
Description: |
268 |
|
|
269 |
|
1. Fix for bug 1498 |
270 |
|
smlnj/src/system/Basis/Implementation/Unsafe/object.sig |
271 |
|
smlnj/src/system/Basis/Implementation/Unsafe/object.sml |
272 |
|
added toRealArray function |
273 |
|
smlnj/src/compiler/MiscUtil/print/ppobj.sml |
274 |
|
added check for tag Obj.RealArray to array printing case in ppObj |
275 |
|
|
276 |
|
2. Fix for bug 1510 |
277 |
|
smlnj/src/compiler/Semant/types/typesutil.sml |
278 |
|
fixed definition of dummyargs (used by equalTycon) so that |
279 |
|
dummy args are distinct types |
280 |
|
|
281 |
|
---------------------------------------------------------------------- |
282 |
|
Name: Matthias Blume |
283 |
|
Date: 2000/04/30 01:00:00 JST |
284 |
|
Tag: blume-20000430-versions |
285 |
|
Description: |
286 |
|
|
287 |
|
1. CM version numbering added. This is an implementation of Lal's |
288 |
|
proposal for adding version numbers and version checking to .cm |
289 |
|
files. Lal said that his proposal was just that -- a proposal. |
290 |
|
For the time being I went ahead and implemented it so that people |
291 |
|
can comment on it. Everything is completely backward-compatible |
292 |
|
(except for the stable library format, i.e., new bootfiles!). |
293 |
|
|
294 |
|
As usual, see the CM manual for details. |
295 |
|
|
296 |
|
2. An alternative syntax for anchored paths has been implemented. |
297 |
|
Dave has recently voiced the same concerns that I had when I did |
298 |
|
this, so there should be some support. My take is that eventually |
299 |
|
I will let support for the current syntax (where anchors are |
300 |
|
"implicit") fade out in favor of the new, explicit syntax. |
301 |
|
In order to be backward-compatible, both old and new syntax are |
302 |
|
currently supported. |
303 |
|
|
304 |
|
Again, see the CM manual for details. |
305 |
|
|
306 |
|
3. Parallel make is trying to be slightly smarter: When the master |
307 |
|
process finds a "bottleneck", i.e., when there is only one |
308 |
|
compilation unit that can be compiled and everybody else is |
309 |
|
waiting on it, then it will simply compile it directly instead |
310 |
|
of clumsily telling one of the slaves to do it. |
311 |
|
|
312 |
|
4. Support for "unsharing" added. This is necessary in order to be |
313 |
|
able to have two different versions of the same library running |
314 |
|
at the same time (e.g., for trying out a new MLRISC while still |
315 |
|
having the old MLRISC linked into the current compiler, etc.) |
316 |
|
See the CM manual. |
317 |
|
|
318 |
|
5. Simple "makedepend" functionality added for generating Makefile |
319 |
|
dependency information. (This is rather crude at the moment. |
320 |
|
Expect some changes here in the future.) |
321 |
|
|
322 |
|
6. ".fun" added as a recognized suffix for ML files. Also documented |
323 |
|
explicitly in the manual that the fallback behavior (unknown suffix |
324 |
|
-> ML file) is not an official feature! |
325 |
|
|
326 |
|
7. Small changes to the pickler for stable libraries. |
327 |
|
|
328 |
|
8. Several internal changes to CM (for cleanup/improvement). |
329 |
|
|
330 |
|
|
331 |
|
!!!! NEW BINFILES !!!! |
332 |
|
|
333 |
|
---------------------------------------------------------------------- |
334 |
|
Name: Matthias Blume |
335 |
|
Date: 2000/04/28 17:30:00 JST |
336 |
|
Tag: blume-20000428-pathconfig |
337 |
|
Description: |
338 |
|
|
339 |
|
1. I changed config/install.sh to remove duplicate entries from the |
340 |
|
lib/pathconfig file at the end. Moreover, the final version of |
341 |
|
lib/pathconfig is sorted alphabetically. The same (sorting) is done |
342 |
|
in src/system/installml. |
343 |
|
|
344 |
|
2. The config/install.sh script now consistently uses relative |
345 |
|
pathnames in lib/pathconfig whenever the anchor is in the lib |
346 |
|
directory. (So far this was true for the libraries that come |
347 |
|
pre-compiled and bundled as part of the bootfiles but not for |
348 |
|
libraries that are compiled by the script itself.) |
349 |
|
|
350 |
|
---------------------------------------------------------------------- |
351 |
|
Name: Matthias Blume |
352 |
|
Date: 2000/04/26 13:10:00 JST |
353 |
|
Tag: blume-20000426-fun_suffix |
354 |
|
Description: |
355 |
|
|
356 |
|
Added ".fun" as a recognized file name suffix (for ML code). |
357 |
|
|
358 |
|
---------------------------------------------------------------------- |
359 |
|
Name: Allen Leung |
360 |
|
Date: 2000/04/25 17:00:00 EST |
361 |
|
Tag: leunga-20000425-alpha-ra |
362 |
|
Description: |
363 |
|
|
364 |
|
1. Alpha |
365 |
|
|
366 |
|
PSEUDOARITH was missing in AlphaRewrite. This causes an endless loop |
367 |
|
in C--. |
368 |
|
|
369 |
|
2. RA |
370 |
|
|
371 |
|
Added a flag "ra-dump-size" to print out the size of the flowgraph |
372 |
|
and the interference graph. |
373 |
|
|
374 |
|
---------------------------------------------------------------------- |
375 |
|
Name: Dave MacQueen |
376 |
|
Date: 2000/04/25/ |
377 |
|
Tag: dbm-20000425-mlyacc_doc_examples |
378 |
|
Description: |
379 |
|
Updated mlyacc.tex sections 5 and 7 for SML '97 and CM. |
380 |
|
Updated all three examples in src/ml-yacc/examples to run |
381 |
|
under 110.* using CM.make. |
382 |
|
|
383 |
|
---------------------------------------------------------------------- |
384 |
|
Name: Allen Leung |
385 |
|
Date: 2000/04/20 23:04:00 EST |
386 |
|
Tag: leunga-20000420-ssa-c---stuff |
387 |
|
Description: |
388 |
|
|
389 |
|
This update synchronizes my repository with Yale's. Most of these |
390 |
|
changes, however, do not affect SML/NJ at all (the RA is an exception). |
391 |
|
|
392 |
|
1. Register Allocator |
393 |
|
|
394 |
|
a. An improvement in the interference graph construction: |
395 |
|
Given a copy |
396 |
|
|
397 |
|
s <- t |
398 |
|
|
399 |
|
no interference edge between s and t is added for this definition of s. |
400 |
|
|
401 |
|
b. I've added two new spill heuristic modules that Fermin and I developed |
402 |
|
(in the new library RA.cm). These are unused in SML/NJ but maybe |
403 |
|
useful for others (Moby?) |
404 |
|
|
405 |
|
2. X86 |
406 |
|
|
407 |
|
a. Various fixes in the backend provided by Fermin [C--] and Lal. |
408 |
|
|
409 |
|
3. Alpha |
410 |
|
|
411 |
|
a. Added the BSR instruction and code generation that goes with it [C--] |
412 |
|
b. Other fixes too numerous to recount provided by Fermin [C--] |
413 |
|
|
414 |
|
4. Regmaps |
415 |
|
|
416 |
|
a. The regmaps are not initialized with the identity physical bindings |
417 |
|
at creation time. This is unneeded. |
418 |
|
|
419 |
|
5. MLRISC Optimizations |
420 |
|
|
421 |
|
a. The DJ-Graph module can now compute the iterated dominance frontiers |
422 |
|
intersects with liveness incrementally in linear time! Woohoo! |
423 |
|
This is now used in my new SSA construction algorithm. |
424 |
|
|
425 |
|
b. THe branch reorganization module is now smarter about linear chains of |
426 |
|
basic blocks. |
427 |
|
|
428 |
|
|
429 |
|
---------------------------------------------------------------------- |
430 |
|
Name: Matthias Blume |
431 |
|
Date: 2000/04/12 13:52:00 JST |
432 |
|
Tag: blume_main_v110p27_1 |
433 |
|
Description: |
434 |
|
|
435 |
|
Changed install.sh script to handle archive files without version number |
436 |
|
and to use "boot.<arch>-<os>" instead of "sml.boot.<arch>-<os>" for the |
437 |
|
name of the boot file archive. |
438 |
|
|
439 |
|
---------------------------------------------------------------------- |
440 |
|
Name: Dave MacQueen |
441 |
|
Date: 2000/04/09 14:00 EDT |
442 |
|
Tag: dbm-20000410-Version_110_27 |
443 |
|
Description: |
444 |
|
|
445 |
|
1. Updated src/compiler/TopLevel/main/version.sml to version 110.27 |
446 |
|
|
447 |
|
2. Updated src/config/version to 110.27 |
448 |
|
|
449 |
|
3. New boot files! |
450 |
|
|
451 |
|
---------------------------------------------------------------------- |
452 |
|
Name: Allen Leung |
453 |
|
Date: 2000/04/09 19:09:00 EST |
454 |
|
Tag: leunga-20000409-misc |
455 |
|
Description: |
456 |
|
|
457 |
|
1. Yet another fix for x86 assembly for idivl, imull, mull and friends. |
458 |
|
|
459 |
|
2. Miscellaneous improvements to MLRISC (unused in sml/nj) |
460 |
|
|
461 |
|
---------------------------------------------------------------------- |
462 |
|
Name: Stefan |
463 |
|
Date: 2000/04/07 10:00:00 EDT |
464 |
|
Tag: monnier-20000406-branch-handling |
465 |
|
Description: |
466 |
|
|
467 |
|
Improved handling of branches (mostly those generated from |
468 |
|
polymorphic equality), removed switchoff and changed the |
469 |
|
default optimization settings (more cpsopt and less flintopt). |
470 |
|
|
471 |
|
---------------------------------------------------------------------- |
472 |
|
Name: Allen Leung |
473 |
|
Date: 2000/04/06 01:30:00 EST |
474 |
|
Tag: leunga-20000406-peephole-x86-SSA-2 |
475 |
|
Description: |
476 |
|
|
477 |
|
Forgot a few files. |
478 |
|
|
479 |
|
---------------------------------------------------------------------- |
480 |
|
Name: Allen Leung |
481 |
|
Date: 2000/04/06 00:36:00 EST |
482 |
|
Tag: leunga-20000406-peephole-x86-SSA |
483 |
|
Description: |
484 |
|
|
485 |
|
1. New Peephole code |
486 |
|
|
487 |
|
2. Minor improvement to X86 instruction selection |
488 |
|
|
489 |
|
3. Various fixes to SSA and machine description -> code translator |
490 |
|
|
491 |
|
---------------------------------------------------------------------- |
492 |
|
Name: Matthias Blume |
493 |
|
Date: 2000/04/05 12:30:00 JST |
494 |
|
Tag: blume_main_v110p26p2_3 |
495 |
|
Description: |
496 |
|
|
497 |
|
This update just merges three minor cosmetic updates to CM's sources |
498 |
|
to get ready for the 110.27 code freeze on Friday. No functionality |
499 |
|
has changed. |
500 |
|
|
501 |
|
---------------------------------------------------------------------- |
502 |
|
Name: Allen Leung |
503 |
|
Date: 2000/04/04 19:39:00 EST |
504 |
|
Tag: leunga-20000404-x86-asm |
505 |
|
Description: |
506 |
|
|
507 |
|
1. Fixed a problem in X86 assembly. |
508 |
|
|
509 |
|
Things like |
510 |
|
|
511 |
|
jmp %eax |
512 |
|
jmp (%eax) |
513 |
|
|
514 |
|
should be output as |
515 |
|
|
516 |
|
jmp *%eax |
517 |
|
jmp *(%eax) |
518 |
|
|
519 |
|
2. Assembly output |
520 |
|
|
521 |
|
Added a new flag |
522 |
|
|
523 |
|
"asm-indent-copies" (default to false) |
524 |
|
|
525 |
|
When this flag is on, parallel copies will be indented an extra level. |
526 |
|
|
527 |
|
---------------------------------------------------------------------- |
528 |
|
Name: Allen Leung |
529 |
|
Date: 2000/04/04 03:18:00 EST |
530 |
|
Tag: leunga-20000404-C--Moby |
531 |
|
Description: |
532 |
|
|
533 |
|
All of these fixes are related to C--, Moby, and my own optimization |
534 |
|
stuff; so they shouldn't affect SML/NJ. |
535 |
|
|
536 |
|
1. X86 |
537 |
|
|
538 |
|
Various fixes related floating point, and extensions. |
539 |
|
|
540 |
|
2. Alpha |
541 |
|
|
542 |
|
Some extra patterns related to loads with signed/zero extension |
543 |
|
provided by Fermin. |
544 |
|
|
545 |
|
3. Assembly |
546 |
|
|
547 |
|
When generating assemby, resolve the value of client defined constants, |
548 |
|
instead of generating symbolic values. This is controlled by the |
549 |
|
new flag "asm-resolve-constants", which is default to true. |
550 |
|
|
551 |
|
4. Machine Descriptions |
552 |
|
|
553 |
|
a. The precedence parser was slightly broken when parsing infixr symbols. |
554 |
|
b. The type generalizing code had the bound variables reversed, resulting |
555 |
|
in a problem during arity raising. |
556 |
|
c. Various fixes in machine descriptions. |
557 |
|
|
558 |
|
---------------------------------------------------------------------- |
559 |
|
Name: Matthias Blume |
560 |
|
Date: 2000/04/03 16:05:00 JST |
561 |
|
Tag: blume_main_v110p26p2_2 |
562 |
|
Description: |
563 |
|
|
564 |
|
I eliminated coreEnv from compInfo. Access to the "Core" structure is |
565 |
|
now done via the ordinary static environment that is context to each |
566 |
|
compilation unit. |
567 |
|
|
568 |
|
To this end, I arranged that instead of "structure Core" as "structure |
569 |
|
_Core" is bound in the pervasive environment. Core access is done via |
570 |
|
_Core (which can never be accidentially rebound because _Core is not a |
571 |
|
legal surface-syntax symbol). |
572 |
|
|
573 |
|
The current solution is much cleaner because the core environment is |
574 |
|
now simply part of the pervasive environment which is part of every |
575 |
|
compilation unit's context anyway. In particular, this eliminates all |
576 |
|
special-case handling that was necessary until now in order to deal |
577 |
|
with dynamic and symbolic parts of the core environment. |
578 |
|
|
579 |
|
Remaining hackery (to bind the "magic" symbol _Core) is localized in the |
580 |
|
compilation mananger's bootstrap compiler (actually: in the "init group" |
581 |
|
handling). See the comments in src/system/smlnj/init/init.cmi for |
582 |
|
more details. |
583 |
|
|
584 |
|
I also tried to track down all mentions of "Core" (as string argument |
585 |
|
to Symbol.strSymbol) in the compiler and replaced them with a |
586 |
|
reference to the new CoreSym.coreSym. Seems cleaner since the actual |
587 |
|
name appears in one place only. |
588 |
|
|
589 |
|
Binfile and bootfile format have not changed, but the switchover from |
590 |
|
the old "init.cmi" to the new one is a bit tricky, so I supplied new |
591 |
|
bootfiles anyway. |
592 |
|
|
593 |
|
---------------------------------------------------------------------- |
594 |
|
Name: Allen Leung |
595 |
|
Date: 2000/04/02 21:17:00 EST |
596 |
|
Tag: leunga-20000402-mltree |
597 |
|
Description: |
598 |
|
|
599 |
|
1. Renamed the constructor CALL in MLTREE by popular demand. |
600 |
|
2. Added a bunch of files from my repository. These are currently |
601 |
|
used by other non-SMLNJ backends. |
602 |
|
|
603 |
|
---------------------------------------------------------------------- |
604 |
|
Name: Allen Leung |
605 |
|
Date: 2000/03/31 21:15:00 EST |
606 |
|
Tag: leunga-20000331-aliasing |
607 |
|
Description: |
608 |
|
|
609 |
|
This update contains a rewritten (and hopefully more correct) module |
610 |
|
for extracting aliasing information from CPS. |
611 |
|
|
612 |
|
To turn on this feature: |
613 |
|
|
614 |
|
Compiler.Control.CG.memDisambiguate := true |
615 |
|
|
616 |
|
To pretty print the region information with assembly |
617 |
|
|
618 |
|
Compiler.Control.MLRISC.getFlag "asm-show-region" := true; |
619 |
|
|
620 |
|
To control how many levels of aliasing information are printed, use: |
621 |
|
|
622 |
|
Compiler.Control.MLRISC.getInt "points-to-show-level" := n |
623 |
|
|
624 |
|
The default of n is 3. |
625 |
|
|
626 |
|
---------------------------------------------------------------------- |
627 |
|
Name: David MacQueen |
628 |
|
Date: 2000/03/31 11:15:00 EST |
629 |
|
Tag: dbm-20000331-runtime_fix |
630 |
|
Description: |
631 |
|
|
632 |
|
This update contains: |
633 |
|
|
634 |
|
1. runtime/c-lib/c-libraries.c |
635 |
|
includes added in revision 1.2 caused compilation errors on hppa-hpux |
636 |
|
|
637 |
|
2. fix for bug 1556 |
638 |
|
system/Basis/Implementation/NJ/internal-signals.sml |
639 |
|
|
640 |
|
---------------------------------------------------------------------- |
641 |
|
Name: Matthias Blume |
642 |
|
Date: 2000/03/31 18:00:00 JST |
643 |
|
Tag: blume_main_v110p26p2_1 |
644 |
|
Description: |
645 |
|
|
646 |
|
This update contains: |
647 |
|
|
648 |
|
1. A small change to CM's handling of stable libraries: |
649 |
|
CM now maintains one "global" modmap that is used for all stable |
650 |
|
libraries. The use of such a global modmap maximizes sharing and |
651 |
|
minimizes the need for re-traversing parts of environments during |
652 |
|
modmap construction. (However, this has minor impact since modmap |
653 |
|
construction seems to account for just one percent or less of total |
654 |
|
compile time.) |
655 |
|
|
656 |
|
2. I added a "genmap" phase to the statistics. This is where I got the |
657 |
|
"one percent" number (see above). |
658 |
|
|
659 |
|
3. CM's new tool parameter mechanism just became _even_ better. :) |
660 |
|
- The parser understands named parameters and recursive options. |
661 |
|
- The "make" and "shell" tools use these new features. |
662 |
|
(This makes it a lot easier to cascade these tools.) |
663 |
|
- There is a small syntax change: named parameters use a |
664 |
|
|
665 |
|
<name> : ( <option> ... ) or |
666 |
|
<name> : <string> |
667 |
|
|
668 |
|
syntax. Previously, named parameters were implemented in an |
669 |
|
ad-hoc fashion by each tool individually (by parsing strings) |
670 |
|
and had the form |
671 |
|
|
672 |
|
<name>=<string> |
673 |
|
|
674 |
|
See the CM manual for a full description of these issues. |
675 |
|
|
676 |
|
---------------------------------------------------------------------- |
677 |
|
Name: Matthias Blume |
678 |
|
Date: 2000/03/30 18:00:00 JST |
679 |
|
Tag: blume_main_v110p26p2_0 |
680 |
|
Description: |
681 |
|
|
682 |
|
!!!!! WARNING !!!!!! |
683 |
|
!! New binfiles !! |
684 |
|
!!!!!!!!!!!!!!!!!!!! |
685 |
|
|
686 |
|
This update contains: |
687 |
|
|
688 |
|
1. Moderate changes to CM: |
689 |
|
|
690 |
|
- Changes to CM's tools mechanism. In particular, it is now possible |
691 |
|
to have tools that accept additional "command line" parameters |
692 |
|
(specified in the .cm file at each instance where the tool's class is |
693 |
|
used). |
694 |
|
|
695 |
|
This was done to accomodate the new "make" and "shell" tools which |
696 |
|
facilitate fairly seemless hookup to portions of code managed using |
697 |
|
Makefiles or Shell scripts. |
698 |
|
|
699 |
|
There are no classes "shared" or "private" anymore. Instead, the |
700 |
|
sharing annotation is now a parameter to the "sml" class. |
701 |
|
|
702 |
|
There is a bit of generic machinery for implementing one's own |
703 |
|
tools that accept command-line parameters. However, I am not yet fully |
704 |
|
satisfied with that part, so expect changes here in the future. |
705 |
|
|
706 |
|
All existing tools are described in the CM manual. |
707 |
|
|
708 |
|
- Slightly better error handling. (CM now surpresses many followup |
709 |
|
error messages that tended to be more annoying than helpful.) |
710 |
|
|
711 |
|
2. Major changes to the compiler's static environment data structures. |
712 |
|
|
713 |
|
- no CMStaticEnv anymore. |
714 |
|
- no CMEnv, no "BareEnvironment" (actually, _only_ BareEnvironment, |
715 |
|
but it is called Environment), no conversions between different |
716 |
|
kinds of static environments |
717 |
|
|
718 |
|
- There is still a notion of a "modmap", but such modmaps are generated |
719 |
|
on demand at the time when they are needed. This sounds slow, but I |
720 |
|
sped up the code that generates modmaps enough for this not to lead to |
721 |
|
a slowdown of the compiler (at least I didn't detect any). |
722 |
|
|
723 |
|
- To facilitate rapid modmap generation, static environments now |
724 |
|
contain an (optional) "modtree" structure. Modtree annotations are |
725 |
|
constructed by the unpickler during unpickling. (This means that |
726 |
|
the elaborator does not have to worry about modtrees at all.) |
727 |
|
Modtrees have the advantage that they are compositional in the same |
728 |
|
way as the environment data structure itself is compositional. |
729 |
|
As a result, modtrees never hang on to parts of an environment that |
730 |
|
has already been rendered "stale" by filtering or rebinding. |
731 |
|
|
732 |
|
- I went through many, many trials and errors before arriving at the |
733 |
|
current solution. (The initial idea of "linkpaths" did not work.) |
734 |
|
But the result of all this is that I have touched a lot of files that |
735 |
|
depend on the "modules" and "types" data structures (most of the |
736 |
|
elaborator). There were a lot of changes during my "linkpath" trials |
737 |
|
that could have been reverted to their original state but weren't. |
738 |
|
Please, don't be too harsh on me for messing with this code a bit more |
739 |
|
than what was strictly necessary... (I _did_ resist the tempation |
740 |
|
of doing any "global reformatting" to avoid an untimely death at |
741 |
|
Dave's hands. :) |
742 |
|
|
743 |
|
- One positive aspect of the previous point: At least I made sure that |
744 |
|
all files that I touched now compile without warnings (other than |
745 |
|
"polyEqual"). |
746 |
|
|
747 |
|
- compiler now tends to run "leaner" (i.e., ties up less memory in |
748 |
|
redundant modmaps) |
749 |
|
|
750 |
|
---------------------------------------------------------------------- |
751 |
|
Name: Allen Leung |
752 |
|
Date: 2000/03/29 18:00:00 |
753 |
|
Tag: leunga-20000327-mlriscGen_hppa_alpha_x86 |
754 |
|
Boot files (optional): ftp://react-ilp.cs.nyu.edu/leunga/110.26.1-sml.boot.x86-unix-20000330.tar.gz |
755 |
|
Description: |
756 |
|
|
757 |
|
This update contains *MAJOR* changes to the way code is generated from CPS |
758 |
|
in the module mlriscGen, and in various backend modules. |
759 |
|
|
760 |
|
CHANGES |
761 |
|
======= |
762 |
|
|
763 |
|
1. MLRiscGen: forward propagation fix. |
764 |
|
|
765 |
|
There was a bug in forward propagation introduced at about the same time |
766 |
|
as the MLRISC x86 backend, which prohibits coalescing to be |
767 |
|
performed effectively in loops. |
768 |
|
|
769 |
|
Effect: speed up of loops in RISC architectures. |
770 |
|
By itself, this actually slowed down certain benchmarks on the x86. |
771 |
|
|
772 |
|
2. MLRiscGen: forward propagating addresses from consing. |
773 |
|
|
774 |
|
I've changed the way consing code is generated. Basically I separated |
775 |
|
out the initialization part: |
776 |
|
|
777 |
|
store tag, offset(allocptr) |
778 |
|
store elem1, offset+4(allocptr) |
779 |
|
store elem2, offset+8(allocptr) |
780 |
|
... |
781 |
|
store elemn, offset+4n(allocptr) |
782 |
|
|
783 |
|
and the address computation part: |
784 |
|
|
785 |
|
celladdr <- offset+4+alloctpr |
786 |
|
|
787 |
|
and move the address computation part |
788 |
|
|
789 |
|
Effect: register pressure is generally lower as a result. This |
790 |
|
makes compilation of certain expressions much faster, such as |
791 |
|
long lists with non-trivial elements. |
792 |
|
|
793 |
|
[(0,0), (0,0), .... (0,0)] |
794 |
|
|
795 |
|
3. MLRiscGen: base pointer elimination. |
796 |
|
|
797 |
|
As part of the linkage mechanism, we generate the sequence: |
798 |
|
|
799 |
|
L: ... <- start of the code fragment |
800 |
|
|
801 |
|
L1: |
802 |
|
base pointer <- linkreg - L1 + L |
803 |
|
|
804 |
|
The base pointer was then used for computing relocatable addresses |
805 |
|
in the code fragment. Frequently (such as in lots of continuations) |
806 |
|
this is not needed. We now eliminate this sequence whenever possible. |
807 |
|
|
808 |
|
For compile time efficiency, I'm using a very stupid local heuristic. |
809 |
|
But in general, this should be done as a control flow analysis. |
810 |
|
|
811 |
|
Effect: Smaller code size. Speed up of most programs. |
812 |
|
|
813 |
|
4. Hppa back end |
814 |
|
|
815 |
|
Long jumps in span dependence resolution used to depend on the existence |
816 |
|
of the base pointer. |
817 |
|
|
818 |
|
A jump to a long label L was expanded into the following sequence: |
819 |
|
|
820 |
|
LDIL %hi(L-8192), %r29 |
821 |
|
LDO %lo(L-8192)(%r29), %r29 |
822 |
|
ADD %r29, baseptr, %r29 |
823 |
|
BV,n %r0(%r29) |
824 |
|
|
825 |
|
In the presence of change (3) above, this will not work. I've changed |
826 |
|
it so that the following sequence of instructions are generated, which |
827 |
|
doesn't mention the base pointer at all: |
828 |
|
|
829 |
|
BL,n L', %r29 /* branch and link, L' + 4 -> %r29 */ |
830 |
|
L': ADDIL L-(L'+4), %r29 /* Compute address of L */ |
831 |
|
BV,n %r0(%r29) /* Jump */ |
832 |
|
|
833 |
|
5. Alpha back end |
834 |
|
|
835 |
|
New alpha instructions LDB/LDW have been added, as per Fermin's |
836 |
|
suggestions. This is unrelated to all other changes. |
837 |
|
|
838 |
|
6. X86 back end |
839 |
|
|
840 |
|
I've changed andl to testl in the floating point test sequence |
841 |
|
whenever appropriate. The Intel optimization guide states that |
842 |
|
testl is perferable to andl. |
843 |
|
|
844 |
|
7. RA (x86 only) |
845 |
|
|
846 |
|
I've improved the spill propagation algorithm, using an approximation |
847 |
|
of maximal weighted independent sets. This seems to be necessary to |
848 |
|
alleviate the negative effect in light of the slow down in (1). |
849 |
|
|
850 |
|
I'll write down the algorithm one of these days. |
851 |
|
|
852 |
|
8. MLRiscGen: frequencies |
853 |
|
|
854 |
|
I've added an annotation that states that all call gc blocks have zero |
855 |
|
execution frequencies. This improves register allocation on the x86. |
856 |
|
|
857 |
|
BENCHMARKS |
858 |
|
========== |
859 |
|
|
860 |
|
I've only perform the comparison on 110.25. |
861 |
|
|
862 |
|
The platforms are: |
863 |
|
|
864 |
|
HPPA A four processor HP machine (E9000) with 5G of memory. |
865 |
|
X86 A 300Hhz Pentium II with 128M of memory, and |
866 |
|
SPARC An Ultra sparc 2 with 512M of memory. |
867 |
|
|
868 |
|
I used the following parameters for the SML benchmarks: |
869 |
|
|
870 |
|
@SMLalloc |
871 |
|
HPPA 256k |
872 |
|
SPARC 512k |
873 |
|
X86 256k |
874 |
|
|
875 |
|
COMPILATION TIME |
876 |
|
---------------- |
877 |
|
Here are the numbers comparing the compilation times of the compilers. |
878 |
|
I've only compared 110.25 compiling the new sources versus |
879 |
|
a fixpoint version of the new compiler compiling the same. |
880 |
|
|
881 |
|
110.25 New |
882 |
|
Total Time in RA Spill+Reload Total Time In RA Spill+Reload |
883 |
|
HPPA 627s 116s 2684+3584 599s 95s 1003+1879 |
884 |
|
SPARC 892s 173s 2891+3870 708s 116s 1004+1880 |
885 |
|
X86 999s 315s 94006+130691 987s 296s 108877+141957 |
886 |
|
|
887 |
|
110.25 New |
888 |
|
Code Size Code Size |
889 |
|
HPPA 8596736 8561421 |
890 |
|
SPARC 8974299 8785143 |
891 |
|
X86 9029180 8716783 |
892 |
|
|
893 |
|
So in summary, things are at least as good as before. Dramatic |
894 |
|
reduction in compilation is obtained on the Sparc; I can't explain it, |
895 |
|
but it is reproducible. Perhaps someone should try to reproduce this |
896 |
|
on their own machines. |
897 |
|
|
898 |
|
SML BENCHMARKS |
899 |
|
-------------- |
900 |
|
|
901 |
|
On the average, all benchmarks perform at least as well as before. |
902 |
|
|
903 |
|
HPPA Compilation Time Spill+Reload Run Time |
904 |
|
110.25 New 110.25 New 110.25 New |
905 |
|
|
906 |
|
barnesHut 3.158 3.015 4.75% 1+1 0+0 2.980 2.922 2.00% |
907 |
|
boyer 6.152 5.708 7.77% 0+0 0+0 0.218 0.213 2.34% |
908 |
|
count-graphs 1.168 1.120 4.32% 0+0 0+0 22.705 23.073 -1.60% |
909 |
|
fft 0.877 0.792 10.74% 1+3 1+3 0.602 0.587 2.56% |
910 |
|
knuthBendix 3.180 2.857 11.32% 0+0 0+0 0.675 0.662 2.02% |
911 |
|
lexgen 6.190 5.290 17.01% 0+0 0+0 0.913 0.788 15.86% |
912 |
|
life 0.803 0.703 14.22% 25+25 0+0 0.153 0.140 9.52% |
913 |
|
logic 2.048 2.007 2.08% 6+6 1+1 4.133 4.008 3.12% |
914 |
|
mandelbrot 0.077 0.080 -4.17% 0+0 0+0 0.765 0.712 7.49% |
915 |
|
mlyacc 22.932 20.937 9.53% 154+181 32+57 0.468 0.430 8.91% |
916 |
|
nucleic 5.183 5.060 2.44% 2+2 0+0 0.125 0.120 4.17% |
917 |
|
ratio-regions 3.357 3.142 6.84% 0+0 0+0 116.225 113.173 2.70% |
918 |
|
ray 1.283 1.290 -0.52% 0+0 0+0 2.887 2.855 1.11% |
919 |
|
simple 6.307 6.032 4.56% 28+30 5+7 3.705 3.658 1.28% |
920 |
|
tsp 0.888 0.862 3.09% 0+0 0+0 7.040 6.893 2.13% |
921 |
|
vliw 24.378 23.455 3.94% 106+127 25+45 2.758 2.707 1.91% |
922 |
|
-------------------------------------------------------------------------- |
923 |
|
Average 6.12% 4.09% |
924 |
|
|
925 |
|
SPARC Compilation Time Spill+Reload Run Time |
926 |
|
110.25 New 110.25 New 110.25 New |
927 |
|
|
928 |
|
barnesHut 3.778 3.592 5.20% 2+2 0+0 3.648 3.453 5.65% |
929 |
|
boyer 6.632 6.110 8.54% 0+0 0+0 0.258 0.242 6.90% |
930 |
|
count-graphs 1.435 1.325 8.30% 0+0 0+0 33.672 34.737 -3.07% |
931 |
|
fft 0.980 0.940 4.26% 3+9 2+6 0.838 0.827 1.41% |
932 |
|
knuthBendix 3.590 3.138 14.39% 0+0 0+0 0.962 0.967 -0.52% |
933 |
|
lexgen 6.593 6.072 8.59% 1+1 0+0 1.077 1.078 -0.15% |
934 |
|
life 0.972 0.868 11.90% 26+26 0+0 0.143 0.140 2.38% |
935 |
|
logic 2.525 2.387 5.80% 7+7 1+1 5.625 5.158 9.05% |
936 |
|
mandelbrot 0.090 0.093 -3.57% 0+0 0+0 0.855 0.728 17.39% |
937 |
|
mlyacc 26.732 23.827 12.19% 162+189 32+57 0.550 0.560 -1.79% |
938 |
|
nucleic 6.233 6.197 0.59% 3+3 0+0 0.163 0.173 -5.77% |
939 |
|
ratio-regions 3.780 3.507 7.79% 0+0 0+0 133.993 131.035 2.26% |
940 |
|
ray 1.595 1.550 2.90% 1+1 0+0 3.440 3.418 0.63% |
941 |
|
simple 6.972 6.487 7.48% 29+32 5+7 3.523 3.525 -0.05% |
942 |
|
tsp 1.115 1.063 4.86% 0+0 0+0 7.393 7.265 1.77% |
943 |
|
vliw 27.765 24.818 11.87% 110+135 25+45 2.265 2.135 6.09% |
944 |
|
---------------------------------------------------------------------------- |
945 |
|
Average 6.94% 2.64% |
946 |
|
|
947 |
|
X86 Compilation Time Spill+Reload Run Time |
948 |
|
110.25 New 110.25 New 110.25 New |
949 |
|
|
950 |
|
barnesHut 5.530 5.420 2.03% 593+893 597+915 3.532 3.440 2.66% |
951 |
|
boyer 8.768 7.747 13.19% 493+199 301+289 0.327 0.297 10.11% |
952 |
|
count-graphs 2.040 2.010 1.49% 298+394 315+457 26.578 28.660 -7.26% |
953 |
|
fft 1.327 1.302 1.92% 112+209 115+210 1.055 0.962 9.71% |
954 |
|
knuthBendix 5.218 5.475 -4.69% 451+598 510+650 0.928 0.932 -0.36% |
955 |
|
lexgen 9.970 9.623 3.60% 1014+841 1157+885 0.947 0.928 1.97% |
956 |
|
life 1.183 1.183 0.00% 162+182 145+148 0.127 0.103 22.58% |
957 |
|
logic 3.285 3.512 -6.45% 514+684 591+836 5.682 5.577 1.88% |
958 |
|
mandelbrot 0.147 0.143 2.33% 38+41 33+54 0.703 0.690 1.93% |
959 |
|
mlyacc 35.457 32.763 8.22% 3496+4564 3611+4860 0.552 0.550 0.30% |
960 |
|
nucleic 7.100 6.888 3.07% 239+168 201+158 0.175 0.173 0.96% |
961 |
|
ratio-regions 6.388 6.843 -6.65% 1182+257 981+300 120.142 120.345 -0.17% |
962 |
|
ray 2.332 2.338 -0.29% 346+398 402+494 3.593 3.540 1.51% |
963 |
|
simple 9.912 9.903 0.08% 1475+941 1579+1168 3.057 3.178 -3.83% |
964 |
|
tsp 1.623 1.532 5.98% 266+200 250+211 8.045 7.878 2.12% |
965 |
|
vliw 33.947 35.470 -4.29% 2629+2774 2877+3171 2.072 1.890 9.61% |
966 |
|
---------------------------------------------------------------------------- |
967 |
|
Average 1.22% 3.36% |
968 |
|
|
969 |
|
---------------------------------------------------------------------- |
970 |
|
Name: Allen Leung |
971 |
|
Date: 2000/03/23 16:25:00 |
972 |
|
Tag: leunga-20000323-fix_x86_alpha |
973 |
|
Description: |
974 |
|
|
975 |
|
1. X86 fixes/changes |
976 |
|
|
977 |
|
a. The old code generated for SETcc was completely wrong. |
978 |
|
The Intel optimization guide is VERY misleading. |
979 |
|
|
980 |
|
2. ALPHA fixes/changes |
981 |
|
|
982 |
|
a. Added the instructions LDBU, LDWU, STB, STW as per Fermin's suggestion. |
983 |
|
b. Added a new mode byteWordLoadStores to the functor parameter to Alpha() |
984 |
|
c. Added reassociation code for address computation. |
985 |
|
|
986 |
|
---------------------------------------------------------------------- |
987 |
Name: Allen Leung |
Name: Allen Leung |
988 |
Date: 2000/03/22 01:23:00 |
Date: 2000/03/22 01:23:00 |
989 |
Tag: leunga-20000322-fix_x86_hppa_ra |
Tag: leunga-20000322-fix_x86_hppa_ra |