8 |
The form of an entry should be: |
The form of an entry should be: |
9 |
|
|
10 |
Name: |
Name: |
11 |
Date: |
Date: yyyy/mm/dd |
12 |
Tag: <post-commit CVS tag> |
Tag: <post-commit CVS tag> |
13 |
Description: |
Description: |
14 |
|
|
15 |
|
---------------------------------------------------------------------- |
16 |
|
Name: Allen Leung |
17 |
|
Date: 2002/02/21 20:20:00 EST |
18 |
|
Tag: leunga-20020221-peephole |
19 |
|
|
20 |
|
Regenerated the peephole files. Some contained typos in the specification |
21 |
|
and some didn't compile because of pretty printing bugs in the old version |
22 |
|
of 'nowhere'. |
23 |
|
|
24 |
|
---------------------------------------------------------------------- |
25 |
|
Name: Allen Leung |
26 |
|
Date: 2002/02/19 20:20:00 EST |
27 |
|
Tag: leunga-20020219-mlrisc-tools |
28 |
|
Description: |
29 |
|
|
30 |
|
Minor bug fixes to the mlrisc-tools library: |
31 |
|
|
32 |
|
1. Fixed up parsing colon suffixed keywords |
33 |
|
2. Added the ability to shut the error messages up |
34 |
|
3. Reimplemented the pretty printer and fixed up/improved |
35 |
|
the pretty printing of handle and -> types. |
36 |
|
4. Fixed up generation of literal symbols in the nowhere tool. |
37 |
|
5. Added some SML keywords to to sml.sty |
38 |
|
|
39 |
|
---------------------------------------------------------------------- |
40 |
|
Name: Matthias Blume |
41 |
|
Date: 2002/02/19 16:20:00 EST |
42 |
|
Tag: blume-20020219-cmffi |
43 |
|
Description: |
44 |
|
|
45 |
|
A wild mix of changes, some minor, some major: |
46 |
|
|
47 |
|
* All C FFI-related libraries are now anchored under $c: |
48 |
|
$/c.cm --> $c/c.cm |
49 |
|
$/c-int.cm --> $c/internals/c-int.cm |
50 |
|
$/memory.cm --> $c/memory/memory.cm |
51 |
|
|
52 |
|
* "make" tool (in CM) now treats its argument pathname slightly |
53 |
|
differently: |
54 |
|
1. If the native expansion is an absolute name, then before invoking |
55 |
|
the "make" command on it, CM will apply OS.Path.mkRelative |
56 |
|
(with relativeTo = OS.FileSys.getDir()) to it. |
57 |
|
2. The argument will be passed through to subsequent phases of CM |
58 |
|
processing without "going native". In particular, if the argument |
59 |
|
was an anchored path, then "make" will not lose track of that anchor. |
60 |
|
|
61 |
|
* Compiler backends now "know" their respective C calling conventions |
62 |
|
instead of having to be told about it by ml-nlffigen. This relieves |
63 |
|
ml-nlffigen from one of its burdens. |
64 |
|
|
65 |
|
* The X86Backend has been split into X86CCallBackend and X86StdCallBackend. |
66 |
|
|
67 |
|
* Export C_DEBUG and C_Debug from $c/c.cm. |
68 |
|
|
69 |
|
* C type encoding in ml-nlffi-lib has been improved to model the conceptual |
70 |
|
subtyping relationship between incomplete pointers and their complete |
71 |
|
counterparts. For this, ('t, 'c) ptr has been changed to 'o ptr -- |
72 |
|
with the convention of instantiating 'o with ('t, 'c) obj whenever |
73 |
|
the pointer target type is complete. In the incomplete case, 'o |
74 |
|
will be instantiated with some "'c iobj" -- a type obtained by |
75 |
|
using one of the functors PointerToIncompleteType or PointerToCompleteType. |
76 |
|
|
77 |
|
Operations that work on both incomplete and complete pointer types are |
78 |
|
typed as taking an 'o ptr while operations that require the target to |
79 |
|
be known are typed as taking some ('t, 'c) obj ptr. |
80 |
|
|
81 |
|
voidptr is now a bit "more concrete", namely "type voidptr = void ptr'" |
82 |
|
where void is an eqtype without any values. This makes it possible |
83 |
|
to work on voidptr values using functions meant to operate on light |
84 |
|
incomplete pointers. |
85 |
|
|
86 |
|
* As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has |
87 |
|
been vastly simplified. |
88 |
|
|
89 |
|
---------------------------------------------------------------------- |
90 |
|
Name: Matthias Blume |
91 |
|
Date: 2002/02/19 10:48:00 EST |
92 |
|
Tag: blume-20020219-pqfix |
93 |
|
Description: |
94 |
|
|
95 |
|
Applied Chris Okasaki's bug fix for priority queues. |
96 |
|
|
97 |
|
---------------------------------------------------------------------- |
98 |
|
Name: Matthias Blume |
99 |
|
Date: 2002/02/15 17:05:00 |
100 |
|
Tag: Release_110_39 |
101 |
|
Description: |
102 |
|
|
103 |
|
Last-minute retagging is becoming a tradition... :-( |
104 |
|
|
105 |
|
This is the working release 110.39. |
106 |
|
|
107 |
|
---------------------------------------------------------------------- |
108 |
|
Name: Matthias Blume |
109 |
|
Date: 2002/02/15 16:00:00 EST |
110 |
|
Tag: Release_110_39-orig |
111 |
|
Description: |
112 |
|
|
113 |
|
Working release 110.39. New bootfiles. |
114 |
|
|
115 |
|
(Update: There was a small bug in the installer so it wouldn't work |
116 |
|
with all shells. So I retagged. -Matthias) |
117 |
|
|
118 |
|
---------------------------------------------------------------------- |
119 |
|
Name: Matthias Blume |
120 |
|
Date: 2002/02/15 14:17:00 EST |
121 |
|
Tag: blume-20020215-showbindings |
122 |
|
Description: |
123 |
|
|
124 |
|
Added EnvRef.listBoundSymbols and CM.State.showBindings. Especially |
125 |
|
the latter can be useful for exploring what bindings are available at |
126 |
|
the interactive prompt. (The first function returns only the list |
127 |
|
of symbols that are really bound, the second prints those but also the |
128 |
|
ones that CM's autoloading mechanism knows about.) |
129 |
|
|
130 |
|
---------------------------------------------------------------------- |
131 |
|
Name: Matthias Blume |
132 |
|
Date: 2002/02/15 12:08:00 EST |
133 |
|
Tag: blume-20020215-iptrs |
134 |
|
Description: |
135 |
|
|
136 |
|
Two improvements to ml-nlffigen: |
137 |
|
|
138 |
|
1. Write files only if they do not exist or if their current contents |
139 |
|
do not coincide with what's being written. (That is, avoid messing |
140 |
|
with the time stamps unless absolutely necessary.) |
141 |
|
|
142 |
|
2. Implement a "repository" mechanism for generated files related |
143 |
|
to "incomplete pointer types". See the README file for details. |
144 |
|
|
145 |
|
---------------------------------------------------------------------- |
146 |
|
Name: Matthias Blume |
147 |
|
Date: 2002/02/14 11:50:00 EST |
148 |
|
Tag: blume-20020214-quote |
149 |
|
Description: |
150 |
|
|
151 |
|
Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm). This is required |
152 |
|
because of the new and improved tag generation scheme. (Thanks to Allen |
153 |
|
Leung for pointing it out.) |
154 |
|
|
155 |
|
---------------------------------------------------------------------- |
156 |
|
Name: Lal George |
157 |
|
Date: 2002/02/14 09:55:27 EST 2002 |
158 |
|
Tag: george-20020214-isabelle-bug |
159 |
|
Description: |
160 |
|
|
161 |
|
Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation |
162 |
|
of Isabelle on the x86. |
163 |
|
|
164 |
|
From Allen: |
165 |
|
----------- |
166 |
|
I've found the problem: |
167 |
|
|
168 |
|
in ra-core.sml, I use the counter "blocked" to keep track of the |
169 |
|
true number of elements in the freeze queue. When the counter goes |
170 |
|
to zero, I skip examining the queue. But I've messed up the |
171 |
|
bookkeeping in combine(): |
172 |
|
|
173 |
|
else (); |
174 |
|
case !ucol of |
175 |
|
PSEUDO => (if !cntv > 0 then |
176 |
|
(if !cntu > 0 then blocked := !blocked - 1 else (); |
177 |
|
^^^^^^^^^^^^^^^^^^^^^^^ |
178 |
|
moveu := mergeMoveList(!movev, !moveu) |
179 |
|
) |
180 |
|
else (); |
181 |
|
|
182 |
|
combine() is called to coalesce two nodes u and v. |
183 |
|
I think I was thinking that if the move counts of u and v are both |
184 |
|
greater than zero then after they are coalesced then one node is |
185 |
|
removed from the freeze queue. Apparently I was thinking that |
186 |
|
both u and v are of low degree, but that's clearly not necessarily true. |
187 |
|
|
188 |
|
|
189 |
|
02/12/2002: |
190 |
|
Here's the patch. HOL now compiles. |
191 |
|
|
192 |
|
I don't know how this impact on performance (compile |
193 |
|
time or runtime). This bug caused the RA (especially on the x86) |
194 |
|
to go thru the potential spill phase when there are still nodes on the |
195 |
|
freeze queue. |
196 |
|
|
197 |
|
|
198 |
|
|
199 |
|
|
200 |
|
---------------------------------------------------------------------- |
201 |
|
Name: Matthias Blume |
202 |
|
Date: 2002/02/13 22:40:00 EST |
203 |
|
Tag: blume-20020213-fptr-rtti |
204 |
|
Description: |
205 |
|
|
206 |
|
Fixed a bug in ml-nlffigen that was introduced with one of the previous |
207 |
|
updates. |
208 |
|
|
209 |
|
---------------------------------------------------------------------- |
210 |
|
Name: Matthias Blume |
211 |
|
Date: 2002/02/13 16:41:00 EST |
212 |
|
Tag: blume-20020213-cmlpq |
213 |
|
Description: |
214 |
|
|
215 |
|
Added new priority queue export symbols (which have just been added to |
216 |
|
smlnj-lib.cm) to CML's version of smlnj-lib.cm. (Otherwise CML would |
217 |
|
not compile and the installer would choke.) |
218 |
|
|
219 |
|
---------------------------------------------------------------------- |
220 |
|
Name: Matthias Blume |
221 |
|
Date: 2002/02/13 16:15:00 EST |
222 |
|
Tag: blume-20020213-various |
223 |
|
Description: |
224 |
|
|
225 |
|
1. More tweaks to ml-nlffigen: |
226 |
|
|
227 |
|
- better internal datastructures (resulting in slight speedup) |
228 |
|
- "-match" option requires exact match |
229 |
|
- "localized" gensym counters (untagged structs/unions nested within |
230 |
|
other structs/unions or within typedefs get a fresh counter; their |
231 |
|
tag will be prefixed by a concatenation of their parents' tags) |
232 |
|
- bug fixes (related to calculation of transitive closure of types |
233 |
|
to be included in the output) |
234 |
|
|
235 |
|
2. Minor Basis updates: |
236 |
|
|
237 |
|
- added implementations for List.collate and Option.app |
238 |
|
|
239 |
|
---------------------------------------------------------------------- |
240 |
|
Name: Matthias Blume |
241 |
|
Date: 2002/02/11 15:55:00 EST |
242 |
|
Tag: blume-20020211-gensym |
243 |
|
Description: |
244 |
|
|
245 |
|
Added a "-gensym" option to command line of ml-nlffigen. This can be |
246 |
|
used to specify a "stem" -- a string that is inserted in all "gensym'd" |
247 |
|
names (ML structure names that correspond to unnamed C structs, unions, |
248 |
|
and enums), so that separate runs of ml-nlffigen do not clash. |
249 |
|
|
250 |
|
---------------------------------------------------------------------- |
251 |
|
Name: Matthias Blume |
252 |
|
Date: 2002/02/11 12:05:00 EST |
253 |
|
Tag: blume-20020211-gensml |
254 |
|
Description: |
255 |
|
|
256 |
|
A quick fix for a problem with GenSML (in the pgraph-util library): |
257 |
|
Make generation of toplevel "local" optional. (Strictly speaking, |
258 |
|
signature definitions within "local" are not legal SML.) |
259 |
|
|
260 |
|
Other than that: updates to INSTALL and cm/TODO. |
261 |
|
|
262 |
|
---------------------------------------------------------------------- |
263 |
|
Name: Matthias Blume |
264 |
|
Date: 2002/02/08 15:00:00 EST |
265 |
|
Tag: blume-20020208-uniquepid |
266 |
|
Description: |
267 |
|
|
268 |
|
0. Version number has been bumped to 110.38.1. NEW BOOTFILES!!! |
269 |
|
|
270 |
|
1. The installer (config/install.sh) has gotten smarter: |
271 |
|
|
272 |
|
- Configuration options are a bit easier to specify now |
273 |
|
(in config/targets). |
274 |
|
- Bug in recognizing .tar.bz2 files fixed. |
275 |
|
- Installer automatically resolves dependencies between |
276 |
|
configuration options (e.g., if you ask for eXene, you will |
277 |
|
also get cml -- regardless whether you asked for it or not). |
278 |
|
- Installer can run in "quieter mode" by setting the environment |
279 |
|
variable INSTALL_QUIETLY to "true". "Quieter" does not mean |
280 |
|
"completely silent", though. |
281 |
|
- Build HashCons library as part of smlnj-lib. |
282 |
|
|
283 |
|
2. A new scheme for assigning persistent identifiers to compilation |
284 |
|
units (and, by extension, to types etc.) has been put into place. |
285 |
|
This fixes a long-standing bug where types and even dynamic values |
286 |
|
can get internally confused, thereby compromising type safety |
287 |
|
(abstraction) and dynamic correctness. See |
288 |
|
|
289 |
|
http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz |
290 |
|
|
291 |
|
for an example of how things could go wrong until now. |
292 |
|
|
293 |
|
The downside of the new scheme is that pids are not quite as |
294 |
|
persistent as they used to be: CM will generate a fresh pid |
295 |
|
for every compilation unit that it thinks it sees for the first |
296 |
|
time. That means that if you compile starting from a clean, fresh |
297 |
|
source tree at two different times, you end up with different |
298 |
|
binaries. |
299 |
|
|
300 |
|
Cutoff recompilation, however, has not been compromised because |
301 |
|
CM keeps pid information in special caches between runs. |
302 |
|
|
303 |
|
---------------------------------------------------------------------- |
304 |
|
Name: Lal George |
305 |
|
Date: 2002/02/07 15:34:13 EST 2002 |
306 |
|
Tag: <none> |
307 |
|
Description: |
308 |
|
|
309 |
|
Compilers that generate assembly code may produce global labels |
310 |
|
whose value is resolved at link time. The various peephole optimization |
311 |
|
modules did not take this in account. |
312 |
|
|
313 |
|
TODO. The Labels.addrOf function should really return an option |
314 |
|
type so that clients are forced to deal with this issue, rather |
315 |
|
than an exception being raised. |
316 |
|
|
317 |
|
---------------------------------------------------------------------- |
318 |
|
Name: Lal George |
319 |
|
Date: 2002/02/06 13:55:02 EST |
320 |
|
Tag: george-20020206-ra-breakup |
321 |
|
Description: |
322 |
|
|
323 |
|
1. A bug fix from Allen. |
324 |
|
|
325 |
|
A typo causes extra fstp %st(0)'s to be generated at compensation |
326 |
|
edges, which might cause stack underflow traps at runtime. This |
327 |
|
occurs in fft where there are extraneous fstps right before the 'into' |
328 |
|
trap instruction (in this case they are harmless since none of the |
329 |
|
integers overflow.) |
330 |
|
|
331 |
|
2. Pulled out various utility modules that were embedded in the modules |
332 |
|
of the register allocator. I need these modules for other purposes, but |
333 |
|
they are not complete enough to put into a library (just yet). |
334 |
|
---------------------------------------------------------------------- |
335 |
|
Name: Matthias Blume |
336 |
|
Date: 2002/01/31 16:05:00 EST |
337 |
|
Tag: blume-20020131-sparc-ccalls |
338 |
|
Description: |
339 |
|
|
340 |
|
1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes) |
341 |
|
of extra stack space by mistake. Fixed. |
342 |
|
|
343 |
|
2. Bug in logic of handling of command-line options in ml-nlffigen fixed. |
344 |
|
|
345 |
|
---------------------------------------------------------------------- |
346 |
|
Name: Allen Leung |
347 |
|
Date: 2002/01/30 |
348 |
|
Tag: leunga-20020130-nowhere-bug-fix |
349 |
|
Description: |
350 |
|
|
351 |
|
MLRISC bug fixes: |
352 |
|
1. Fixed a bindings computation bug in the 'nowhere' program generator tool. |
353 |
|
2. MachineInt.fromString was negating its value. |
354 |
|
|
355 |
|
---------------------------------------------------------------------- |
356 |
|
Name: Matthias Blume |
357 |
|
Date: 2002/01/29 |
358 |
|
Tag: blume-20020129-INSTALL |
359 |
|
Description: |
360 |
|
|
361 |
|
- Added somewhat detailed installation instructions (file INSTALL). |
362 |
|
- Fixed curl-detection bug in config/install.sh. |
363 |
|
- It is now possible to select the URL getter using the URLGETTER |
364 |
|
environment variable: |
365 |
|
|
366 |
|
not set / "unknown" --> automatic detection (script tries wget, |
367 |
|
curl, and lynx) |
368 |
|
"wget" / "curl" / "lynx" --> use the specified program (script "knows" |
369 |
|
how to properly invoke them) |
370 |
|
other --> use $URLGETTER directly, it must take |
371 |
|
precisely two command-line arguments |
372 |
|
(source URL and destination file name) |
373 |
|
|
374 |
|
---------------------------------------------------------------------- |
375 |
|
Name: Matthias Blume |
376 |
|
Date: 2002/01/28 |
377 |
|
Tag: blume-20020128-sparc-ccalls |
378 |
|
Description: |
379 |
|
|
380 |
|
- Fixed problem with calculation of "used" registers in sparc-c-calls. |
381 |
|
- Make use of the allocParam argument in sparc-c-calls. |
382 |
|
|
383 |
|
---------------------------------------------------------------------- |
384 |
|
Name: Matthias Blume |
385 |
|
Date: 2002/01/28 |
386 |
|
Tag: blume-20020128-allocParam |
387 |
|
Description: |
388 |
|
|
389 |
|
John Reppy: Changes c-calls API to accept client-callback for |
390 |
|
allocating extra stack space. |
391 |
|
me: Corresponding changes to mlriscGen (using a dummy argument that |
392 |
|
does not change the current behavior). |
393 |
|
|
394 |
|
---------------------------------------------------------------------- |
395 |
|
Name: Matthias Blume |
396 |
|
Date: 2002/01/28 12:00:00 |
397 |
|
Tag: Release_110_38 |
398 |
|
Description: |
399 |
|
|
400 |
|
This time for real!!! |
401 |
|
|
402 |
|
---------------------------------------------------------------------- |
403 |
|
Name: Matthias Blume |
404 |
|
Date: 2002/01/28 10:56:00 EST |
405 |
|
Tag: blume-20020128-retraction |
406 |
|
Description: |
407 |
|
|
408 |
|
0. Retracted earlier 110.38. (The Release_110_38 tag has been replaced |
409 |
|
with blume-Release_110_38-retracted.) |
410 |
|
|
411 |
|
1. Fixed a problem with incorrect rounding modes in real64.sml. |
412 |
|
(Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.) |
413 |
|
|
414 |
|
2. A bug in ml-nlffigen related to the handling of unnamed structs, unions, |
415 |
|
and enums fixed. The naming of corresponding ML identifiers should |
416 |
|
now be consistent again. |
417 |
|
|
418 |
|
---------------------------------------------------------------------- |
419 |
|
Name: Allen Leung |
420 |
|
Date: 2002/01/27 |
421 |
|
Tag: leunga-20020127-nowhere |
422 |
|
Description: |
423 |
|
|
424 |
|
Added a target called nowhere in the configuration scripts. |
425 |
|
Enabling this will build the MLRISC 'nowhere' tool (for translating |
426 |
|
programs with where-clauses into legal SML code) during installation. |
427 |
|
|
428 |
|
---------------------------------------------------------------------- |
429 |
|
Name: Matthias Blume |
430 |
|
Date: 2002/01/25 21:27:00 EST |
431 |
|
Tag: blume-Release_110_38-retracted |
432 |
|
Description: |
433 |
|
|
434 |
|
Call it a (working) release! Version is 110.38. Bootfiles are ready. |
435 |
|
|
436 |
|
README will be added later. |
437 |
|
|
438 |
|
!!! NOTE: Re-tagged as blume-Release_110_38-retracted. Original tag |
439 |
|
(Release_110_38) removed. Reason: Last-minute bug fixes. |
440 |
|
|
441 |
|
---------------------------------------------------------------------- |
442 |
|
Name: Matthias Blume |
443 |
|
Date: 2002/01/25 |
444 |
|
Tag: blume-20020125-ffi |
445 |
|
Description: |
446 |
|
|
447 |
|
A large number of tweaks and improvements to ml-nlffi-lib and |
448 |
|
ml-nlffigen: |
449 |
|
|
450 |
|
- ML represenation types have been streamlined |
451 |
|
- getter and setter functions work with concrete values, not abstract |
452 |
|
ones where possible |
453 |
|
- ml-nlffigen command line more flexible (see README file there) |
454 |
|
- some bugs have been fixed (hopefully) |
455 |
|
|
456 |
|
---------------------------------------------------------------------- |
457 |
|
Name: Lal George |
458 |
|
Date: 2002/01/24 |
459 |
|
Tag: george-20020124-risc-ra-interface |
460 |
|
Description: |
461 |
|
|
462 |
|
There is a dramatic simplification in the interface to the |
463 |
|
register allocator for RISC architectures as a result of making |
464 |
|
parallel copy instructions explicit. |
465 |
|
|
466 |
|
---------------------------------------------------------------------- |
467 |
|
Name: Matthias Blume |
468 |
|
Date: 2002/01/22 |
469 |
|
Tag: blume-20020122-x86-ccalls |
470 |
|
Description: |
471 |
|
|
472 |
|
Bug fix for c-calls on x86 (having to do with how char- and |
473 |
|
short-arguments are being handled). |
474 |
|
|
475 |
|
---------------------------------------------------------------------- |
476 |
|
Name: Matthias Blume |
477 |
|
Date: 2002/01/21 |
478 |
|
Tag: blume-20020121-ff |
479 |
|
Description: |
480 |
|
|
481 |
|
Another day of fiddling with the FFI... |
482 |
|
|
483 |
|
1. Bug fix/workaround: CKIT does not complain about negative array |
484 |
|
dimensions, so ml-nlffigen has to guard itself against this possibility. |
485 |
|
(Otherwise a negative dimension would send it into an infinite loop.) |
486 |
|
|
487 |
|
2. Some of the abstract types (light objects, light pointers, most "base" |
488 |
|
types) in structure C are now eqtypes. |
489 |
|
|
490 |
|
3. Added constructors and test functions for NULL function pointers. |
491 |
|
|
492 |
|
---------------------------------------------------------------------- |
493 |
|
Name: Matthias Blume |
494 |
|
Date: 2002/01/18 |
495 |
|
Tag: blume-20020118-ready-for-new-release |
496 |
|
Description: |
497 |
|
|
498 |
|
Made config/srcarchiveurl point to a new place. (Will provide boot |
499 |
|
files shortly.) |
500 |
|
|
501 |
|
Maybe we christen this to be 110.38? |
502 |
|
|
503 |
|
---------------------------------------------------------------------- |
504 |
|
Name: Matthias Blume |
505 |
|
Date: 2002/01/18 |
506 |
|
Tag: blume-20020118-more-ffifiddle |
507 |
|
Description: |
508 |
|
|
509 |
|
Today's FFI fiddling: |
510 |
|
|
511 |
|
- Provided a structure CGetSet with "convenient" versions of C.Get.* and |
512 |
|
C.Set.* that use concrete (MLRep.*) arguments and results instead |
513 |
|
of abstract ones. |
514 |
|
|
515 |
|
- Provided word-style bit operations etc. for "int" representation |
516 |
|
types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short, |
517 |
|
and Long. |
518 |
|
|
519 |
|
---------------------------------------------------------------------- |
520 |
|
Name: Matthias Blume |
521 |
|
Date: 2002/01/18 |
522 |
|
Tag: blume-20020118-use-x86-fp |
523 |
|
Description: |
524 |
|
|
525 |
|
Now that x86-fast-fp seems to be working, I turned it back on again |
526 |
|
by default. (Seems to work fine now, even with the FFI.) |
527 |
|
|
528 |
|
Other than that, I added some documentation about the FFI to |
529 |
|
src/ml-nlffigen/README and updated the FFI test examples in |
530 |
|
src/ml-nlffi-lib/Tests/*. |
531 |
|
|
532 |
|
---------------------------------------------------------------------- |
533 |
|
Name: Allen Leung |
534 |
|
Date: 2002/01/17 |
535 |
|
Tag: leunga-20020117-x86-fast-fp-call |
536 |
|
Description: |
537 |
|
|
538 |
|
1. Fixed a problem with handling return fp values when x86's fast fp |
539 |
|
mode is turned on. |
540 |
|
|
541 |
|
2. Minor pretty printing fix for cellset. Print %st(0) as %st(0) instead |
542 |
|
of %f32. |
543 |
|
|
544 |
|
3. Added a constructor INT32lit to the ast of MLRISC tools. |
545 |
|
|
546 |
|
---------------------------------------------------------------------- |
547 |
|
Name: Matthias Blume |
548 |
|
Date: 2002/01/16 |
549 |
|
Tag: blume-20020116-ffifiddle |
550 |
|
Description: |
551 |
|
|
552 |
|
More fiddling with the FFI interface: |
553 |
|
|
554 |
|
- Make constness 'c instead of rw wherever possible. This eliminates |
555 |
|
the need for certain explicit coercions. (However, due to ML's |
556 |
|
value polymorphism, there will still be many cases where explicit |
557 |
|
coercions are necessary. Phantom types are not the whole answer |
558 |
|
to modeling a subtyping relationship in ML.) |
559 |
|
|
560 |
|
- ro/rw coersions for pointers added. (Avoids the detour through */&.) |
561 |
|
|
562 |
|
- "printf" test example added to src/ml-nlffi-lib/Tests. (Demonstrates |
563 |
|
clumsy workaround for varargs problem.) |
564 |
|
|
565 |
|
---------------------------------------------------------------------- |
566 |
|
Name: Lal George |
567 |
|
Date: 2002/01/15 |
568 |
|
Tag: <none> |
569 |
|
Description: |
570 |
|
|
571 |
|
1. Since COPY instructions are no longer native to the architecture, |
572 |
|
a generic functor can be used to implement the expandCopies function. |
573 |
|
|
574 |
|
2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a |
575 |
|
TEXT segment. |
576 |
|
|
577 |
|
---------------------------------------------------------------------- |
578 |
|
Name: Matthias Blume |
579 |
|
Date: 2002/01/15 |
580 |
|
Tag: blume-20020115-ffiupdates |
581 |
|
Description: |
582 |
|
|
583 |
|
1. Fix for bug resulting in single-precision float values being returned |
584 |
|
incorrectly from FFI calls. |
585 |
|
|
586 |
|
2. Small modifications to C FFI API: |
587 |
|
|
588 |
|
- memory-allocation routines return straight objects (no options) |
589 |
|
and raise an exception in out-of-memory situations |
590 |
|
- unsafe extensions to cast between function pointers and pointers |
591 |
|
from/to ints |
592 |
|
- added structure C_Debug as an alternative to structure C where |
593 |
|
pointer-dereferencing (|*| and |*!) always check for null-pointers |
594 |
|
- added open_lib' to DynLinkage; open_lib' works like open_lib |
595 |
|
but also takes a (possibly empty) list of existing library handles |
596 |
|
that the current library depends on |
597 |
|
|
598 |
|
---------------------------------------------------------------------- |
599 |
|
Name: Matthias Blume |
600 |
|
Date: 2002/01/10 |
601 |
|
Tag: blume-20020110-newffigen |
602 |
|
Description: |
603 |
|
|
604 |
|
1. Updates to portable graph code. |
605 |
|
|
606 |
|
2. Major update to ml-nlffigen and ml-nlffi-lib. Things are much |
607 |
|
more scalable now so that even huge interfaces such as the one |
608 |
|
for GTK compile in finite time and space. :-) |
609 |
|
See src/ml-nlffigen/README for details on what's new. |
610 |
|
|
611 |
|
---------------------------------------------------------------------- |
612 |
|
Name: Lal George |
613 |
|
Date: 2001/01/09 14:31:35 EST 2002 |
614 |
|
Tag: george-20011206-rm-native-copy |
615 |
|
Description: |
616 |
|
|
617 |
|
Removed the native COPY and FCOPY instructions |
618 |
|
from all the architectures and replaced it with the |
619 |
|
explicit COPY instruction from the previous commit. |
620 |
|
|
621 |
|
It is now possible to simplify many of the optimizations |
622 |
|
modules that manipulate copies. This has not been |
623 |
|
done in this change. |
624 |
|
|
625 |
|
---------------------------------------------------------------------- |
626 |
|
Name: Lal George |
627 |
|
Date: 2001/12/06 16:50:13 EST 2001 |
628 |
|
Tag: george-20011206-mlrisc-instruction |
629 |
|
Description: |
630 |
|
|
631 |
|
Changed the representation of instructions from being fully abstract |
632 |
|
to being partially concrete. That is to say: |
633 |
|
|
634 |
|
from |
635 |
|
type instruction |
636 |
|
|
637 |
|
to |
638 |
|
type instr (* machine instruction *) |
639 |
|
|
640 |
|
datatype instruction = |
641 |
|
LIVE of {regs: C.cellset, spilled: C.cellset} |
642 |
|
| KILL of {regs: C.cellset, spilled: C.cellset} |
643 |
|
| COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list} |
644 |
|
| ANNOTATION of {i: instruction, a: Annotations.annotation} |
645 |
|
| INSTR of instr |
646 |
|
|
647 |
|
This makes the handling of certain special instructions that appear on |
648 |
|
all architectures easier and uniform. |
649 |
|
|
650 |
|
LIVE and KILL say that a list of registers are live or killed at the |
651 |
|
program point where they appear. No spill code is generated when an |
652 |
|
element of the 'regs' field is spilled, but the register is moved to |
653 |
|
the 'spilled' (which is present, more for debugging than anything else). |
654 |
|
|
655 |
|
LIVE replaces the (now deprecated) DEFFREG instruction on the alpha. |
656 |
|
We used to generate: |
657 |
|
|
658 |
|
DEFFREG f1 |
659 |
|
f1 := f2 + f3 |
660 |
|
trapb |
661 |
|
|
662 |
|
but now generate: |
663 |
|
|
664 |
|
f1 := f2 + f3 |
665 |
|
trapb |
666 |
|
LIVE {regs=[f1,f2,f3], spilled=[]} |
667 |
|
|
668 |
|
Furthermore, the DEFFREG (hack) required that all floating point instruction |
669 |
|
use all registers mentioned in the instruction. Therefore f1 := f2 + f3, |
670 |
|
defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting |
671 |
|
in a cleaner alpha implementation. (Hopefully, intel will not get rid of |
672 |
|
this architecture). |
673 |
|
|
674 |
|
COPYXXX is intended to replace the parallel COPY and FCOPY available on |
675 |
|
all the architectures. This will result in further simplification of the |
676 |
|
register allocator that must be aware of them for coalescing purposes, and |
677 |
|
will also simplify certain aspects of the machine description that provides |
678 |
|
callbacks related to parallel copies. |
679 |
|
|
680 |
|
ANNOTATION should be obvious, and now INSTR represents the honest to God |
681 |
|
machine instruction set! |
682 |
|
|
683 |
|
The <arch>/instructions/<arch>Instr.sml files define certain utility |
684 |
|
functions for making porting easier -- essentially converting upper case |
685 |
|
to lower case. All machine instructions (of type instr) are in upper case, |
686 |
|
and the lower case form generates an MLRISC instruction. For example on |
687 |
|
the alpha we have: |
688 |
|
|
689 |
|
datatype instr = |
690 |
|
LDA of {r:cell, b:cell, d:operand} |
691 |
|
| ... |
692 |
|
|
693 |
|
val lda : {r:cell, b:cell, d:operand} -> instruction |
694 |
|
... |
695 |
|
|
696 |
|
where lda is just (INSTR o LDA), etc. |
697 |
|
|
698 |
|
---------------------------------------------------------------------- |
699 |
|
Name: Matthias Blume |
700 |
|
Date: 2001/11/22 21:40:00 EST |
701 |
|
Tag: Release_110_37 |
702 |
|
Description: |
703 |
|
|
704 |
|
Release 110.37. This time for real. |
705 |
|
|
706 |
|
---------------------------------------------------------------------- |
707 |
|
Name: Matthias Blume |
708 |
|
Date: 2001/11/21 16:35:00 EST |
709 |
|
Tag: blume-20011121-foot-in-mouth |
710 |
|
Description: |
711 |
|
|
712 |
|
Removed the "Release_110_37" tag because of a serious bug. |
713 |
|
This will be re-tagged once the bug is fixed. |
714 |
|
|
715 |
|
---------------------------------------------------------------------- |
716 |
|
Name: Matthias Blume |
717 |
|
Date: 2001/11/21 16:14:00 EST |
718 |
|
Tag: blume-20011121-forgottenfile |
719 |
|
Description: |
720 |
|
|
721 |
|
Forgot to add a file. (Just a .tex-file -- part of |
722 |
|
the CM manual source.) |
723 |
|
|
724 |
|
---------------------------------------------------------------------- |
725 |
|
Name: Matthias Blume |
726 |
|
Date: 2001/11/21 16:10:00 EST |
727 |
|
Tag: blume-20011121-invalid_110_37 |
728 |
|
Description: |
729 |
|
|
730 |
|
Note: I removed the original tag "Release_110_37" from this commit |
731 |
|
because we found a serious bug in all non-x86 backends. |
732 |
|
- Matthias |
733 |
|
|
734 |
|
1. Modifications to the SML/NJ code generator and to the runtime system |
735 |
|
so that code object name strings are directly inserted into code |
736 |
|
objects at code generation time. The only business the runtime system |
737 |
|
has with this is now to read the name strings on occasions. |
738 |
|
(The encoding of the name string has also changed somewhat.) |
739 |
|
|
740 |
|
2. CM now implements a simple "set calculus" for specifying export lists. |
741 |
|
In particular, it is now possible to refer to the export lists of |
742 |
|
other libraries/groups/sources and form unions as well as differences. |
743 |
|
See the latest CM manual for details. |
744 |
|
|
745 |
|
3. An separate notion of "proxy" libraries has again be eliminated from |
746 |
|
CM's model. (Proxy libraries are now simply a special case of using |
747 |
|
the export list calculus.) |
748 |
|
|
749 |
|
4. Some of the existing libraries now take advantage of the new set |
750 |
|
calculus. |
751 |
|
(Notice that not all libraries have been converted because some |
752 |
|
of the existing .cm-files are supposed to be backward compatible |
753 |
|
with 110.0.x.) |
754 |
|
|
755 |
|
5. Some cleanup in stand-alone programs. (Don't use "exnMessage" -- use |
756 |
|
"General.exnMessage"! The former relies on a certain hook to be |
757 |
|
initialized, and that often does not happen in the stand-alone case.) |
758 |
|
|
759 |
|
---------------------------------------------------------------------- |
760 |
|
Name: Lal George |
761 |
|
Date: 2001/11/21 13:56:18 EST |
762 |
|
Tag: george-2001121-pseudo-ops |
763 |
|
Description: |
764 |
|
|
765 |
|
Implemented a complete redesign of MLRISC pseudo-ops. Now there |
766 |
|
ought to never be any question of incompatabilities with |
767 |
|
pseudo-op syntax expected by host assemblers. |
768 |
|
|
769 |
|
For now, only modules supporting GAS syntax are implemented |
770 |
|
but more should follow, such as MASM, and vendor assembler |
771 |
|
syntax, e.g. IBM as, Sun as, etc. |
772 |
|
|
773 |
|
---------------------------------------------------------------------- |
774 |
|
Name: Matthias Blume |
775 |
|
Date: 2001/11/14 11:52:00 EST |
776 |
|
Tag: blume-20011114-srcname |
777 |
|
Description: |
778 |
|
|
779 |
|
1. Routed the name of the current source file to mlriscgen where it |
780 |
|
should be directly emitted into the code object. (This last part |
781 |
|
is yet to be done.) |
782 |
|
|
783 |
|
2. Some cleanup of the pgraph code to make it match the proposal that |
784 |
|
I put out the other day. (The proposal notwithstanding, things are |
785 |
|
still in flux here.) |
786 |
|
|
787 |
|
---------------------------------------------------------------------- |
788 |
|
Name: Lal George |
789 |
|
Date: 2001/11/14 09:44:04 EST |
790 |
|
Tag: |
791 |
|
Description: |
792 |
|
|
793 |
|
Fix for a backpatching bug reported by Allen. |
794 |
|
|
795 |
|
Because the boundary between short and long span-dependent |
796 |
|
instructions is +/- 128, there are an astounding number of |
797 |
|
span-dependent instructions whose size is over estimated. |
798 |
|
|
799 |
|
Allen came up with the idea of letting the size of span |
800 |
|
dependent instructions be non-monotonic, for a maxIter |
801 |
|
number of times, after which the size must be monotonically |
802 |
|
increasing. |
803 |
|
|
804 |
|
This table shows the number of span-dependent instructions |
805 |
|
whose size was over-estimated as a function of maxIter, for the |
806 |
|
file Parse/parse/ml.grm.sml: |
807 |
|
|
808 |
|
maxIter # of instructions: |
809 |
|
10 687 |
810 |
|
20 438 |
811 |
|
30 198 |
812 |
|
40 0 |
813 |
|
|
814 |
|
In compiling the compiler, there is no significant difference in |
815 |
|
compilation speed between maxIter=10 and maxIter=40. Actually, |
816 |
|
my measurements showed that maxIter=40 was a tad faster than |
817 |
|
maxIter=10! Also 96% of the files in the compiler reach a fix |
818 |
|
point within 13 iterations, so fixing maxIter at 40, while high, |
819 |
|
is okay. |
820 |
|
|
821 |
|
---------------------------------------------------------------------- |
822 |
|
Name: Matthias Blume |
823 |
|
Date: 2001/10/31 15:25:00 EST |
824 |
|
Tag: blume-20011031-pgraph |
825 |
|
Description: |
826 |
|
|
827 |
|
CKIT: |
828 |
|
* Changed the "Function" constructor of type Ast.ctype to carry optional |
829 |
|
argument identifiers. |
830 |
|
* Changed the return type of TypeUtil.getFunction accordingly. |
831 |
|
* Type equality ignores the argument names. |
832 |
|
* TypeUtil.composite tries to preserve argument names but gives up quickly |
833 |
|
if there is a mismatch. |
834 |
|
|
835 |
|
installation script: |
836 |
|
* attempts to use "curl" if available (unless "wget" is available as well) |
837 |
|
|
838 |
|
CM: |
839 |
|
* has an experimental implementation of "portable graphs" which I will |
840 |
|
soon propose as an implementation-independent library format |
841 |
|
* there are also new libraries $/pgraph.cm and $/pgraph-util.cm |
842 |
|
|
843 |
|
NLFFI-LIB: |
844 |
|
* some cleanup (all cosmetic) |
845 |
|
|
846 |
|
NLFFIGEN: |
847 |
|
* temporarily disabled the mechanism that suppresses ML output for |
848 |
|
C definitions whose identifiers start with an underscore character |
849 |
|
* generate val bindings for enum constants |
850 |
|
* user can request that only one style (light or heavy) is being used; |
851 |
|
default is to use both (command-line arguments: -heavy and -light) |
852 |
|
* fixed bug in handling of function types involving incomplete pointers |
853 |
|
* generate ML entry points that take record arguments (i.e., using |
854 |
|
named arguments) for C functions that have a prototype with named |
855 |
|
arguments |
856 |
|
(see changes to CKIT) |
857 |
|
|
858 |
|
---------------------------------------------------------------------- |
859 |
|
Name: Allen Leung |
860 |
|
Date: 2001/10/27 20:34:00 EDT |
861 |
|
Tag: leunga-20011027-x86-fast-fp-call |
862 |
|
Description: |
863 |
|
|
864 |
|
Fixed the bug described in blume-20010920-slowfp. |
865 |
|
|
866 |
|
The fix involves |
867 |
|
1. generating FCOPYs in FSTP in ia32-svid |
868 |
|
2. marking a CALL with the appropriate annotation |
869 |
|
|
870 |
|
---------------------------------------------------------------------- |
871 |
|
Name: Matthias Blume |
872 |
|
Date: 2001/10/16 11:32:00 EDT |
873 |
|
Tag: blume-20011016-netbsd |
874 |
|
Description: |
875 |
|
|
876 |
|
Underscore patch from Chris Richards (fixing problem with compiling |
877 |
|
runtime system under recent NetBSD). |
878 |
|
|
879 |
|
---------------------------------------------------------------------- |
880 |
|
Name: Allen Leung |
881 |
|
Date: 2001/10/12 17:18:32 EDT 2001 |
882 |
|
Tag: leung-20011012-x86-printflowgraph |
883 |
|
Description: |
884 |
|
|
885 |
|
X86RA now uses a valid (instead of dummy) PrintFlowgraph module. |
886 |
|
|
887 |
|
---------------------------------------------------------------------- |
888 |
|
Name: Lal George |
889 |
|
Date: 2001/10/11 23:51:34 EDT |
890 |
|
Tag: george-20011011-too-many-instrs |
891 |
|
Description: |
892 |
|
|
893 |
|
The representation of a program point never expected to see more |
894 |
|
than 65536 instructions in a basic block! |
895 |
|
|
896 |
|
---------------------------------------------------------------------- |
897 |
|
Name: Lal George |
898 |
|
Date: 2001/10/09 09:41:37 EDT |
899 |
|
Tag: george-20011008-mlrisc-labels |
900 |
|
Description: |
901 |
|
|
902 |
|
Changed the machine description files to support printing of |
903 |
|
local and global labels in assembly code, based on host assembler |
904 |
|
conventions. |
905 |
|
|
906 |
|
---------------------------------------------------------------------- |
907 |
|
Name: Matthias Blume |
908 |
|
Date: 2001/09/25 15:25:00 EDT |
909 |
|
Tag: blume-20010925-exninfo |
910 |
|
Description: |
911 |
|
|
912 |
|
I provided a non-hook implementation of exnName (at the toplevel) and |
913 |
|
made the "dummy" implementation of exnMessage (at the toplevel) more |
914 |
|
useful: if nothing gets "hooked in", then at least you are going to |
915 |
|
see the exception name and a message indicating why you don't see more. |
916 |
|
|
917 |
|
[For the time being, programs that need exnMessage and want to use |
918 |
|
ml-build should either use General.exnMessage (strongly recommended) or |
919 |
|
refer to structure General at some other point so that CM sees a |
920 |
|
static dependency.] |
921 |
|
|
922 |
|
[Similar remarks go for "print" and "use": If you want to use their |
923 |
|
functionality in stand-alone programs generated by ml-build, then use |
924 |
|
TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).] |
925 |
|
|
926 |
|
---------------------------------------------------------------------- |
927 |
|
Name: Matthias Blume |
928 |
|
Date: 2001/09/20 17:28:00 EDT |
929 |
|
Tag: blume-20010920-slowfp |
930 |
|
Description: |
931 |
|
|
932 |
|
Allen says that x86-fast-fp is not safe yet, so I turned it off again... |
933 |
|
|
934 |
|
---------------------------------------------------------------------- |
935 |
|
Name: Matthias Blume |
936 |
|
Date: 2001/09/20 17:20:00 EDT |
937 |
|
Tag: blume-20010920-canonicalpaths |
938 |
|
Description: |
939 |
|
|
940 |
|
0. Updated the BOOT file (something that I forgot to do earlier). |
941 |
|
|
942 |
|
1. Small internal change to CM so that it avoids "/../" in filenames |
943 |
|
as much as possible (but only where it is safe). |
944 |
|
|
945 |
|
2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so |
946 |
|
that arguments that contain delimiters are passed through correctly. |
947 |
|
This change also means that all "special" arguments of the form |
948 |
|
@SMLxxx... must come first. |
949 |
|
|
950 |
|
3. Changed install script to put relative anchor names for tool commands |
951 |
|
into pathconfig. |
952 |
|
|
953 |
|
---------------------------------------------------------------------- |
954 |
|
Name: Matthias Blume |
955 |
|
Date: 2001/09/18 15:35:00 EDT |
956 |
|
Tag: blume-20010918-readme11036 |
957 |
|
Description: |
958 |
|
|
959 |
|
Added README files. |
960 |
|
|
961 |
|
---------------------------------------------------------------------- |
962 |
|
Name: Matthias Blume |
963 |
|
Date: 2001/09/18 11:45:00 EDT |
964 |
|
Tag: Release_110_36 (retag) |
965 |
|
Description: |
966 |
|
|
967 |
|
Fixed mistake in config/preloads. Retagged as 110.36. |
968 |
|
|
969 |
|
---------------------------------------------------------------------- |
970 |
|
Name: Matthias Blume |
971 |
|
Date: 2001/09/18 09:40:00 EDT |
972 |
|
Tag: Release_110_36_orig (tag changed) |
973 |
|
Description: |
974 |
|
|
975 |
|
New version (110.36). New bootfiles. |
976 |
|
|
977 |
|
---------------------------------------------------------------------- |
978 |
|
Name: Matthias Blume |
979 |
|
Date: 2001/09/14 16:15:00 EDT |
980 |
|
Tag: blume-20010914-x86fastfp |
981 |
|
Description: |
982 |
|
|
983 |
|
John committed some changes that Allen made, in particular a (hopefully) |
984 |
|
correctly working version of the x86-fp module. |
985 |
|
|
986 |
|
I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp" |
987 |
|
flag to "true". Everything seems to compile to a fixpoint ok, and |
988 |
|
"mandelbrot" speeds up by about 15%. |
989 |
|
|
990 |
|
---------------------------------------------------------------------- |
991 |
|
Name: Matthias Blume |
992 |
|
Date: 2001/09/13 11:20:00 EDT |
993 |
|
Tag: blume-20010913-minimal |
994 |
|
Description: |
995 |
|
|
996 |
|
1. Stefan Monnier's patch to fix a miscompilation problem that |
997 |
|
was brought to light by John Reppy's work on Moby. |
998 |
|
|
999 |
|
2. Implemented a minimal "structure Compiler" that contains just |
1000 |
|
"version" and "architecture". The minimal version will be |
1001 |
|
available when the full version is not. This is for backward- |
1002 |
|
compatibility with code that wants to test Compiler.version. |
1003 |
|
|
1004 |
|
---------------------------------------------------------------------- |
1005 |
|
Name: Matthias Blume |
1006 |
|
Date: 2001/08/28 14:03:00 EDT |
1007 |
|
Tag: blume-20010828-ml-lex |
1008 |
|
Description: |
1009 |
|
|
1010 |
|
Fix for bug 1581, received from Neophytos Michael. |
1011 |
|
|
1012 |
|
---------------------------------------------------------------------- |
1013 |
|
Name: Matthias Blume |
1014 |
|
Date: 2001/08/27 11:20:00 EDT |
1015 |
|
Tag: blume-20010827-readme11035 |
1016 |
|
Description: |
1017 |
|
|
1018 |
|
Fleshed out the README file for 110.35. |
1019 |
|
|
1020 |
|
---------------------------------------------------------------------- |
1021 |
|
Name: Matthias Blume |
1022 |
|
Date: 2001/08/24 17:10:00 EDT |
1023 |
|
Tag: Release_110_35 |
1024 |
|
Description: |
1025 |
|
|
1026 |
|
New version number (110.35). New bootfiles. |
1027 |
|
|
1028 |
|
---------------------------------------------------------------------- |
1029 |
|
Name: Lal George |
1030 |
|
Date: 2001/08/24 13:47:18 EDT 2001 |
1031 |
|
Tag: george-20010824-MLRISC-graphs |
1032 |
|
Description: |
1033 |
|
|
1034 |
|
removed clusters from MLRISC completely and replaced with graphs. |
1035 |
|
|
1036 |
|
---------------------------------------------------------------------- |
1037 |
|
Name: Matthias Blume |
1038 |
|
Date: 2001/08/23 17:50:00 EDT |
1039 |
|
Tag: blume-20010823-toplevel |
1040 |
|
Description: |
1041 |
|
|
1042 |
|
- some reorganization of the code that implements various kinds of |
1043 |
|
environments in the compiler (static, dynamic, symbolic, combined) |
1044 |
|
- re-implemented the EnvRef module so that evalStream works properly |
1045 |
|
(if the stream contains references to "use", "CM.make", etc.) |
1046 |
|
- cleaned up evalloop.sml and interact.sml (but they need more cleaning) |
1047 |
|
|
1048 |
|
---------------------------------------------------------------------- |
1049 |
|
Name: Matthias Blume |
1050 |
|
Date: 2001/08/20 15:50 EDT |
1051 |
|
Tag: blume20010820-slipup |
1052 |
|
Description: |
1053 |
|
|
1054 |
|
I forgot to commit a few files. Here they are... |
1055 |
|
|
1056 |
|
---------------------------------------------------------------------- |
1057 |
|
Name: Matthias Blume |
1058 |
|
Date: 2001/08/20 15:35:00 EDT |
1059 |
|
Tag: blume-20010820-debugprof |
1060 |
|
Description: |
1061 |
|
|
1062 |
|
!!!! NEW BOOTFILES !!!! |
1063 |
|
|
1064 |
|
This is another round of reorganizing the compiler sources. This |
1065 |
|
time the main goal was to factor out all the "instrumentation" |
1066 |
|
passes (for profiling and backtracing) into their own library. |
1067 |
|
The difficulty was to do it in such a way that it does not depend |
1068 |
|
on elaborate.cm but only on elabdata.cm. |
1069 |
|
|
1070 |
|
Therefore there have been further changes to both elaborate.cm and |
1071 |
|
elabdata.cm -- more "generic" things have been moved from the former |
1072 |
|
to the latter. As a result, I was forced to split the assignment |
1073 |
|
of numbers indicating "primtyc"s into two portions: SML-generic and |
1074 |
|
SML/NJ-specific. Since it would have been awkward to maintain, |
1075 |
|
I bit the bullet and actually _changed_ the mapping between these |
1076 |
|
numbers and primtycs. The bottom line of this is that you need |
1077 |
|
a new set of bin- and bootfiles. |
1078 |
|
|
1079 |
|
I have built new bootfiles for all architectures, so doing a fresh |
1080 |
|
checkout and config/install.sh should be all you need. |
1081 |
|
|
1082 |
|
The newly created library's name is |
1083 |
|
|
1084 |
|
$smlnj/viscomp/debugprof.cm |
1085 |
|
|
1086 |
|
and its sources live under |
1087 |
|
|
1088 |
|
src/compiler/DebugProf |
1089 |
|
|
1090 |
|
---------------------------------------------------------------------- |
1091 |
|
Name: Matthias Blume |
1092 |
|
Date: 2001/08/15 17:15:00 EDT |
1093 |
|
Tag: blume-20010815-compreorg |
1094 |
|
Description: |
1095 |
|
|
1096 |
|
This is a first cut at reorganizing the CM libraries that make up the |
1097 |
|
core of the compiler. The idea is to separate out pieces that could |
1098 |
|
be used independently by tools, e.g., the parser, the typechecker, etc. |
1099 |
|
|
1100 |
|
The current status is a step in this direction, but it is not quite |
1101 |
|
satisfactory yet. Expect more changes in the future. |
1102 |
|
|
1103 |
|
Here is the current (new) organization... |
1104 |
|
|
1105 |
|
What used to be $smlnj/viscomp/core.cm is now divided into |
1106 |
|
six CM libraries: |
1107 |
|
|
1108 |
|
$smlnj/viscomp/basics.cm |
1109 |
|
/parser.cm |
1110 |
|
/elabdata.cm |
1111 |
|
/elaborate.cm |
1112 |
|
/execute.cm |
1113 |
|
/core.cm |
1114 |
|
|
1115 |
|
The CM files for these libraries live under src/system/smlnj/viscomp. |
1116 |
|
All these libraries are proxy libraries that contain precisely |
1117 |
|
one CM library component. Here are the locations of the components |
1118 |
|
(all within the src/compiler tree): |
1119 |
|
|
1120 |
|
Basics/basics.cm |
1121 |
|
Parse/parser.cm |
1122 |
|
ElabData/elabdata.cm |
1123 |
|
Elaborator/elaborate.cm |
1124 |
|
Execution/execute.cm |
1125 |
|
core.cm |
1126 |
|
|
1127 |
|
[This organization is the same that has been used already |
1128 |
|
for a while for the architecture-specific parts of the visible |
1129 |
|
compiler and for the old version of core.cm.] |
1130 |
|
|
1131 |
|
As you will notice, many source files have been moved from their |
1132 |
|
respective original locations to a new home in one of the above |
1133 |
|
subtrees. |
1134 |
|
|
1135 |
|
The division of labor between the new libraries is the following: |
1136 |
|
|
1137 |
|
basics.cm: |
1138 |
|
- Simple, basic definitions that pertain to many (or all) of |
1139 |
|
the other libraries. |
1140 |
|
parser.cm: |
1141 |
|
- The SML parser, producing output of type Ast.dec. |
1142 |
|
- The type family for Ast is also defined and exported here. |
1143 |
|
elabdata.cm: |
1144 |
|
- The datatypes that describe input and output of the elaborator. |
1145 |
|
This includes types, absyn, and static environments. |
1146 |
|
elaborator.cm: |
1147 |
|
- The SML/NJ type checker and elaborator. |
1148 |
|
This maps an Ast.dec (with a given static environment) to |
1149 |
|
an Absyn.dec (with a new static environment). |
1150 |
|
- This libraries implements certain modules that used to be |
1151 |
|
structures as functors (to remove dependencies on FLINT). |
1152 |
|
execute.cm: |
1153 |
|
- Everything having to do with executing binary code objects. |
1154 |
|
- Dynamic environments. |
1155 |
|
core.cm: |
1156 |
|
- SML/NJ-specific instantiations of the elaborator and MLRISC. |
1157 |
|
- Top-level modules. |
1158 |
|
- FLINT (this should eventually become its own library) |
1159 |
|
|
1160 |
|
Notes: |
1161 |
|
|
1162 |
|
I am not 100% happy with the way I separated the elaborator (and its |
1163 |
|
data structures) from FLINT. Two instances of the same problem: |
1164 |
|
|
1165 |
|
1. Data structures contain certain fields that carry FLINT-specific |
1166 |
|
information. I hacked around this using exn and the property list |
1167 |
|
module from smlnj-lib. But the fact that there are middle-end |
1168 |
|
specific fields around at all is a bit annoying. |
1169 |
|
|
1170 |
|
2. The elaborator calculates certain FLINT-related information. I tried |
1171 |
|
to make this as abstract as I could using functorization, but, again, |
1172 |
|
the fact that the elaborator has to perform calculations on behalf |
1173 |
|
of the middle-end at all is not nice. |
1174 |
|
|
1175 |
|
3. Having to used exn and property lists is unfortunate because it |
1176 |
|
weakens type checking. The other alternative (parameterizing |
1177 |
|
nearly *everything*) is not appealing, though. |
1178 |
|
|
1179 |
|
I removed the "rebinding =" warning hack because due to the new organization |
1180 |
|
it was awkward to maintain it. As a result, the compiler now issues some of |
1181 |
|
these warnings when compiling init.cmi during bootstrap compilation. On |
1182 |
|
the plus side, you also get a warning when you do, for example: |
1183 |
|
val op = = Int32.+ |
1184 |
|
which was not the case up to now. |
1185 |
|
|
1186 |
|
I placed "assign" and "deref" into the _Core structure so that the |
1187 |
|
code that deals with the "lazy" keyword can find them there. This |
1188 |
|
removes the need for having access to the primitive environment |
1189 |
|
during elaboration. |
1190 |
|
|
1191 |
|
---------------------------------------------------------------------- |
1192 |
|
Name: Matthias Blume |
1193 |
|
Date: 2001/08/13 |
1194 |
|
Tag: blume-20010813-closures |
1195 |
|
Description: |
1196 |
|
|
1197 |
|
This fix was sent to us by Zhong Shao. It is supposed to improve the |
1198 |
|
performance of certain loops by avoiding needless closure allocation. |
1199 |
|
|
1200 |
|
---------------------------------------------------------------------- |
1201 |
|
Name: Lal George |
1202 |
|
Date: 2001/07/31 10:03:23 EDT 2001 |
1203 |
|
Tag: george-20010731-x86-fmalloc |
1204 |
|
Description: Fixed bug in x86 calls |
1205 |
|
|
1206 |
|
There was a bug where call instructions would mysteriously |
1207 |
|
vanish. The call instruction had to be one that returned |
1208 |
|
a floating point value. |
1209 |
|
|
1210 |
|
---------------------------------------------------------------------- |
1211 |
|
Name: Lal George |
1212 |
|
Date: 2001/07/19 16:36:29 EDT 2001 |
1213 |
|
Tag: george-20010719-simple-cells |
1214 |
|
Description: |
1215 |
|
|
1216 |
|
I have dramatically simplified the interface for CELLS in MLRISC. |
1217 |
|
|
1218 |
|
In summary, the cells interface is broken up into three parts: |
1219 |
|
|
1220 |
|
1. CellsBasis : CELLS_BASIS |
1221 |
|
|
1222 |
|
CellsBasis is a top level structure and common for all |
1223 |
|
architectures. it contains the definitions of basic datatypes |
1224 |
|
and utility functions over these types. |
1225 |
|
|
1226 |
|
2. functor Cells() : CELLS |
1227 |
|
|
1228 |
|
Cells generates an interface for CELLS that incorporates the |
1229 |
|
specific resources on the target architecture, such as the |
1230 |
|
presence of special register classes, their number and size, |
1231 |
|
and various useful substructures. |
1232 |
|
|
1233 |
|
3. <ARCH>CELLS |
1234 |
|
|
1235 |
|
e.g. SparcCells: SPARCCELLS |
1236 |
|
|
1237 |
|
<ARCH>CELLS usually contains additional bindings for special |
1238 |
|
registers on the architecture, such as: |
1239 |
|
|
1240 |
|
val r0 : cell (* register zero *) |
1241 |
|
val y : cell (* Y register *) |
1242 |
|
val psr : cell (* processor status register *) |
1243 |
|
... |
1244 |
|
|
1245 |
|
The structure returned by applying the Cells functor is opened |
1246 |
|
in this interface. |
1247 |
|
|
1248 |
|
The main implication of all this is that the datatypes for cells is |
1249 |
|
split between CellsBasis and CELLS -- a fairly simple change for user |
1250 |
|
code. |
1251 |
|
|
1252 |
|
In the old scheme the CELLS interface had a definitional binding of |
1253 |
|
the form: |
1254 |
|
|
1255 |
|
signature CELLS = sig |
1256 |
|
|
1257 |
|
structure CellsBasis = CellsBasis |
1258 |
|
|
1259 |
|
... |
1260 |
|
|
1261 |
|
end |
1262 |
|
|
1263 |
|
With all the sharing constraints that goes on in MLRISC, this old |
1264 |
|
design quickly leads to errors such as: |
1265 |
|
|
1266 |
|
"structure definition spec inside of sharing ... " |
1267 |
|
|
1268 |
|
|
1269 |
|
and appears to require an unacceptable amount of sharing and where |
1270 |
|
constraint hackery. |
1271 |
|
|
1272 |
|
I think this error message (the interaction of definitional specs and |
1273 |
|
sharing) requires more explanation on our web page. |
1274 |
|
|
1275 |
|
---------------------------------------------------------------------- |
1276 |
|
Name: Matthias Blume |
1277 |
|
Date: 2001/07/19 15:00:00 EDT |
1278 |
|
Tag: blume-20010719-libreorg |
1279 |
|
Description: |
1280 |
|
|
1281 |
|
This update puts together a fairly extensive but straightforward change |
1282 |
|
to the way the libraries that implement the interactive system are |
1283 |
|
organized: |
1284 |
|
|
1285 |
|
The biggest change is the elimination of structure Compiler. As a |
1286 |
|
replacement for this structure, there is now a CM library |
1287 |
|
(known as $smlnj/compiler.cm or $smlnj/compiler/current.cm) |
1288 |
|
that exports all the substructures of the original structure Compiler |
1289 |
|
directly. So instead of saying Compiler.Foo.bar one now simply |
1290 |
|
says Foo.bar. (The CM libraries actually export a collection of |
1291 |
|
structures that is richer than the collection of substructures of |
1292 |
|
structure Compiler.) |
1293 |
|
|
1294 |
|
To make the transition smooth, there is a separate library called |
1295 |
|
$smlnj/compiler/compiler.cm which puts together and exports the |
1296 |
|
original structure Compiler (or at least something very close to it). |
1297 |
|
|
1298 |
|
There are five members of the original structure Compiler |
1299 |
|
that are not exported directly but which instead became members |
1300 |
|
of a new structure Backend (described by signature BACKEND). These are: |
1301 |
|
structure Profile (: PROFILE), structure Compile (: COMPILE), structure |
1302 |
|
Interact (: INTERACT), structure Machine (: MACHINE), and val |
1303 |
|
architecture (: string). |
1304 |
|
|
1305 |
|
Structure Compiler.Version has become structure CompilerVersion. |
1306 |
|
|
1307 |
|
Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided |
1308 |
|
by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc, |
1309 |
|
or x86, respectively. |
1310 |
|
Each of these exports the same frontend structures that |
1311 |
|
$smlnj/compiler.cm exports. But they do not have a structure Backend |
1312 |
|
and instead export some structure <Arch>Backend where <Arch> is Alpha32, |
1313 |
|
Hppa, PPC, Sparc, or X86, respectively. |
1314 |
|
|
1315 |
|
Library $smlnj/compiler/all.cm exports the union of the exports of |
1316 |
|
$smlnj/compiler/<arch>.cm |
1317 |
|
|
1318 |
|
There are no structures <Arch>Compiler anymore, use |
1319 |
|
$smlnj/compiler/<arch>.cm instead. |
1320 |
|
|
1321 |
|
Library host-compiler-0.cm is gone. Instead, the internal library |
1322 |
|
that instantiates CM is now called cm0.cm. Selection of the host |
1323 |
|
compiler (backend) is no longer done here but. (Responsibility for it |
1324 |
|
now lies with $smlnj/compiler/current.cm. This seems to be more |
1325 |
|
logical.) |
1326 |
|
|
1327 |
|
Many individual files have been moved or renamed. Some files have |
1328 |
|
been split into multiple files, and some "dead" files have been deleted. |
1329 |
|
|
1330 |
|
Aside from these changes to library organization, there are also changes |
1331 |
|
to the way the code itself is organized: |
1332 |
|
|
1333 |
|
Structure Binfile has been re-implemented in such a way that it no |
1334 |
|
longer needs any knowledge of the compiler. It exclusively deals |
1335 |
|
with the details of binfile layout. It no longer invokes the |
1336 |
|
compiler (for the purpose of creating new prospective binfile |
1337 |
|
content), and it no longer has any knowledge of how to interpret |
1338 |
|
pickles. |
1339 |
|
|
1340 |
|
Structure Compile (: COMPILE) has been stripped down to the bare |
1341 |
|
essentials of compilation. It no longer deals with linking/execution. |
1342 |
|
The interface has been cleaned up considerably. |
1343 |
|
|
1344 |
|
Utility routines for dealing with linking and execution have been |
1345 |
|
moved into their own substructures. |
1346 |
|
|
1347 |
|
(The ultimate goal of these changes is to provide a light-weight |
1348 |
|
binfile loader/linker (at least for, e.g., stable libraries) that |
1349 |
|
does not require CM or the compiler to be present.) |
1350 |
|
|
1351 |
|
CM documentation has been updated to reflect the changes to library |
1352 |
|
organization. |
1353 |
|
|
1354 |
|
---------------------------------------------------------------------- |
1355 |
|
Name: Matthias Blume |
1356 |
|
Date: 2001/07/10 17:30:00 EDT |
1357 |
|
Tag: Release_110_34 |
1358 |
|
Description: |
1359 |
|
|
1360 |
|
Minor tweak to 110.34 (re-tagged): |
1361 |
|
|
1362 |
|
- README.html file added to CVS repository |
1363 |
|
- runtime compiles properly under FreeBSD 3.X and 4.X |
1364 |
|
|
1365 |
|
---------------------------------------------------------------------- |
1366 |
|
Name: Matthias Blume |
1367 |
|
Date: 2001/07/10 17:30:00 EDT |
1368 |
|
Tag: Release_110_34 |
1369 |
|
Description: |
1370 |
|
|
1371 |
|
New version number (110.34). New bootfiles. |
1372 |
|
|
1373 |
|
---------------------------------------------------------------------- |
1374 |
|
Name: Matthias Blume |
1375 |
|
Date: 2001/07/09 16:00:00 EDT |
1376 |
|
Tag: blume-20010709-more-varargs |
1377 |
|
Description: |
1378 |
|
|
1379 |
|
I changed the handling of varargs in ml-nlffigen again: |
1380 |
|
The ellipsis ... will now simply be ignored (with an accompanying warning). |
1381 |
|
|
1382 |
|
The immediate effect is that you can actually call a varargs function |
1383 |
|
from ML -- but you can't actually supply any arguments beyond the ones |
1384 |
|
specified explicitly. (For example, you can call printf with its format |
1385 |
|
string, but you cannot pass additional arguments.) |
1386 |
|
|
1387 |
|
This behavior is only marginally more useful than the one before, but |
1388 |
|
it has the advantage that a function or, more importantly, a function |
1389 |
|
type never gets dropped on the floor, thus avoiding follow-up problems with |
1390 |
|
other types that refer to the offending one. |
1391 |
|
|
1392 |
|
---------------------------------------------------------------------- |
1393 |
|
Name: Matthias Blume |
1394 |
|
Date: 2001/07/09 11:25:00 EDT |
1395 |
|
Tag: blume-20010709-varargs |
1396 |
|
Description: |
1397 |
|
|
1398 |
|
1. ckit-lib.cm now exports structure Error |
1399 |
|
2. ml-nlffigen reports occurences of "..." (i.e., varargs function types) |
1400 |
|
with a warning accompanied by a source location. Moreover, it |
1401 |
|
merely skips the offending function or type and proceeds with the |
1402 |
|
rest of its work.u As a result, one can safely feed C code containing |
1403 |
|
"..." to ml-nlffigen. |
1404 |
|
3. There are some internal improvements to CM, providing slightly |
1405 |
|
more general string substitutions in the tools subsystem. |
1406 |
|
|
1407 |
|
---------------------------------------------------------------------- |
1408 |
|
Name: Matthias Blume |
1409 |
|
Date: 2001/06/27 15:10:00 EDT |
1410 |
|
Tag: blume-20010627-concur |
1411 |
|
Description: |
1412 |
|
|
1413 |
|
Fixed a small bug in CM's handling of parallel compilation. |
1414 |
|
(You could observe the bug by Control-C-interrupting an ordinary |
1415 |
|
CMB.make or CM.stabilize and then attaching some compile servers. |
1416 |
|
The result was that all of a sudden the previously interrupted |
1417 |
|
compilation would continue on its own. This was because of |
1418 |
|
an over-optimization: CM did not bother to clean out certain queues |
1419 |
|
when no servers were attached "anyway", resulting in the contents |
1420 |
|
of these queues to grab control when new servers did get attached.) |
1421 |
|
|
1422 |
|
There is also another minor update to the CM manual. |
1423 |
|
|
1424 |
|
---------------------------------------------------------------------- |
1425 |
|
Name: Matthias Blume |
1426 |
|
Date: 2001/06/26 16:15:00 EDT |
1427 |
|
Tag: blume-20010626-cmdoc |
1428 |
|
Description: |
1429 |
|
|
1430 |
|
Minor typo fixed in CM manual (syntax diagram for libraries). |
1431 |
|
|
1432 |
|
---------------------------------------------------------------------- |
1433 |
|
Name: Matthias Blume |
1434 |
|
Date: 2001/06/25 22:55:00 EDT |
1435 |
|
Tag: blume-20010625-x86pc |
1436 |
|
Description: |
1437 |
|
|
1438 |
|
Fixed a nasty bug in the X86 assembly code that caused signal |
1439 |
|
handlers to fail (crash) randomly. |
1440 |
|
|
1441 |
|
---------------------------------------------------------------------- |
1442 |
|
Name: Matthias Blume |
1443 |
|
Date: 2001/06/25 12:05:00 EDT |
1444 |
|
Tag: blume-20010625-nlffigen |
1445 |
|
Description: |
1446 |
|
|
1447 |
|
This update fixes a number of minor bugs in ml-nlffigen as reported by |
1448 |
|
Nick Carter <nbc@andrew.cmu.edu>. |
1449 |
|
|
1450 |
|
1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted. |
1451 |
|
2. Default names for generated files are now derived from the name of |
1452 |
|
the C file *without its directory*. In particular, this causes generated |
1453 |
|
files to be placed locally even if the C file is in some system directory. |
1454 |
|
3. Default names for generated signatures and structures are also derived |
1455 |
|
from the C file name without its directory. This avoids silly things |
1456 |
|
like "structure GL/GL". |
1457 |
|
(Other silly names are still possible because ml-nlffigen does not do |
1458 |
|
a thorough check of whether generated names are legal ML identifiers. |
1459 |
|
When in doubt, use command line arguments to force particular names.) |
1460 |
|
|
1461 |
|
---------------------------------------------------------------------- |
1462 |
|
Name: Matthias Blume |
1463 |
|
Date: 2001/06/21 12:25:00 EDT |
1464 |
|
Tag: blume-20010621-eXene |
1465 |
|
Description: |
1466 |
|
|
1467 |
|
eXene now compiles and (sort of) works again. |
1468 |
|
|
1469 |
|
The library name (for version > 110.33) is $/eXene.cm. |
1470 |
|
|
1471 |
|
I also added an new example in src/eXene/examples/nbody. See the |
1472 |
|
README file there for details. |
1473 |
|
|
1474 |
|
---------------------------------------------------------------------- |
1475 |
|
Name: Matthias Blume |
1476 |
|
Date: 2001/06/20 16:40:00 EDT |
1477 |
|
Tag: blume-20010620-cml |
1478 |
|
Description: |
1479 |
|
|
1480 |
|
CML now compiles and works again. |
1481 |
|
|
1482 |
|
Libraries (for version > 110.33): |
1483 |
|
|
1484 |
|
$cml/cml.cm Main CML library. |
1485 |
|
$cml/basis.cm CML's version of $/basis.cm. |
1486 |
|
$cml/cml-internal.cm Internal helper library. |
1487 |
|
$cml/core-cml.cm Internal helper library. |
1488 |
|
$cml-lib/trace-cml.cm Tracing facility. |
1489 |
|
$cml-lib/smlnj-lib.cm CML's version of $/smlnj-lib.cm |
1490 |
|
|
1491 |
|
The installer (config/install.sh) has been taught how to properly |
1492 |
|
install this stuff. |
1493 |
|
|
1494 |
|
---------------------------------------------------------------------- |
1495 |
|
Name: Matthias Blume |
1496 |
|
Date: 2001/06/19 17:55:00 EDT |
1497 |
|
Tag: blume-20010619-instantiate |
1498 |
|
Description: |
1499 |
|
|
1500 |
|
This un-breaks the fix for bug 1432. |
1501 |
|
(The bug was originally fixed in 110.9 but I broke it again some |
1502 |
|
time after that.) |
1503 |
|
|
1504 |
|
---------------------------------------------------------------------- |
1505 |
|
Name: Matthias Blume |
1506 |
|
Date: 2001/06/19 17:25:00 EDT |
1507 |
|
Tag: blume-20010619-signals |
1508 |
|
Description: |
1509 |
|
|
1510 |
|
This should (hopefully) fix the long-standing signal handling bug. |
1511 |
|
(The runtime system was constructing a continuation record with an |
1512 |
|
incorrect descriptor which would cause the GC to drop data on the floor...) |
1513 |
|
|
1514 |
|
---------------------------------------------------------------------- |
1515 |
|
Name: Matthias Blume |
1516 |
|
Date: 2001/06/15 15:05:00 EDT |
1517 |
|
Tag: blume-20010615-moresparc |
1518 |
|
Description: |
1519 |
|
|
1520 |
|
Here is a short late-hour update related to Sparc c-calls: |
1521 |
|
|
1522 |
|
-- made handling of double-word arguments a bit smarter |
1523 |
|
|
1524 |
|
-- instruction selection phase tries to collapse certain clumsily |
1525 |
|
constructed ML-Trees; typical example: |
1526 |
|
|
1527 |
|
ADD(ty,ADD(_,e,LI d1),LI d2) -> ADD(ty,e,LI(d1+d2)) |
1528 |
|
|
1529 |
|
This currently has no further impact on SML/NJ since mlriscGen does |
1530 |
|
not seem to generate such patterns in the first place, and c-calls |
1531 |
|
(which did generate them in the beginning) has meanwhile been fixed |
1532 |
|
so as to avoid them as well. |
1533 |
|
|
1534 |
|
---------------------------------------------------------------------- |
1535 |
|
Name: Matthias Blume |
1536 |
|
Date: 2001/06/15 15:05:00 EDT |
1537 |
|
Tag: blume-20010615-sparc |
1538 |
|
Description: |
1539 |
|
|
1540 |
|
The purpose of this update is to provide an implementation of NLFFI |
1541 |
|
on Sparc machines. |
1542 |
|
|
1543 |
|
Here are the changes in detail: |
1544 |
|
|
1545 |
|
* src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing |
1546 |
|
the Sparc implementation of the c-calls API. |
1547 |
|
* The Sparc backend of SML/NJ has been modified to uniformely use %fp |
1548 |
|
for accessing the ML frame. Thus, we have a real frame pointer and |
1549 |
|
can freely modify %sp without need for an omit-frame-ptr phase. |
1550 |
|
The vfp logic in src/compiler/CodeGen/* has been changed to accomodate |
1551 |
|
this case. |
1552 |
|
* ml-nlffigen has been taught to produce code for different architectures |
1553 |
|
and calling conventions. |
1554 |
|
* In a way similar to what was done in the x86 case, the Sparc |
1555 |
|
backend uses its own specific extension to mltree. (For example, |
1556 |
|
it needs to be able to generate UNIMP instructions which are part |
1557 |
|
of the calling convention.) |
1558 |
|
* ml-nlffi-lib was reorganized to make it more modular (in particular, |
1559 |
|
to make it easier to plug in new machine- and os-dependent parts). |
1560 |
|
|
1561 |
|
There are some other fairly unrelated bug fixes and cleanups as well: |
1562 |
|
|
1563 |
|
* I further hacked the .cm files for MLRISC tools (like MDLGen) so |
1564 |
|
that they properly share their libraries with existing SML/NJ libraries. |
1565 |
|
* I fixed a minor cosmetic bug in CM, supressing certain spurious |
1566 |
|
follow-up error messages. |
1567 |
|
* Updates to CM/CMB documentation. |
1568 |
|
|
1569 |
|
TODO items: |
1570 |
|
|
1571 |
|
* MLRISC should use a different register as its asmTemp on the Sparc. |
1572 |
|
(The current %o2 is a really bad choice because it is part of the |
1573 |
|
calling conventions, so things might interfere in unexpected ways.) |
1574 |
|
|
1575 |
|
---------------------------------------------------------------------- |
1576 |
|
Name: Matthias Blume |
1577 |
|
Date: 2001/06/07 |
1578 |
|
Tag: blume-20010607-calls |
1579 |
|
Description: |
1580 |
|
|
1581 |
|
A number of internal changes related to C calls and calling conventions: |
1582 |
|
|
1583 |
|
1. ML-Tree CALL statements now carry a "pops" field. It indicates the |
1584 |
|
number of bytes popped implicitly (by the callee). In most cases |
1585 |
|
this field is 0 but on x86/win32 it is some non-zero value. This |
1586 |
|
is information provided for the benefit of the "omit-frameptr" pass. |
1587 |
|
2. The CALL instruction on the x86 carries a similar "pops" field. |
1588 |
|
The instruction selection phase copies its value from the ML-Tree |
1589 |
|
CALL statement. |
1590 |
|
3. On all other architectures, the instruction selection phase checks |
1591 |
|
whether "pops=0" and complains if not. |
1592 |
|
4. The c-calls implementation for x86 now accepts two calling conventions: |
1593 |
|
"ccall" and "stdcall". When "ccall" is selected, the caller cleans |
1594 |
|
up after the call and pops is set to 0. For "stdcall", the caller |
1595 |
|
does nothing, leaving the cleanup to the callee; pops is set to |
1596 |
|
the number of bytes that were pushed onto the stack. |
1597 |
|
5. The cproto decoder (compiler/Semant/types/cproto.sml) now can |
1598 |
|
distinguish between "ccall" and "stdcall". |
1599 |
|
6. The UNIMP instruction has been added to the supported Sparc instruction |
1600 |
|
set. (This is needed for implementing the official C calling convention |
1601 |
|
on this architecture.) |
1602 |
|
7. I fixed some of the .cm files under src/MLRISC/Tools to make them |
1603 |
|
work with the latest CM. |
1604 |
|
|
1605 |
|
---------------------------------------------------------------------- |
1606 |
|
Name: Matthias Blume |
1607 |
|
Date: 2001/06/05 15:10:00 EDT |
1608 |
|
Tag: blume-20010605-cm-index |
1609 |
|
Description: |
1610 |
|
|
1611 |
|
0. The "lambdasplit" parameter for class "sml" in CM has been documented. |
1612 |
|
|
1613 |
|
1. CM can now generate "index files". These are human-readable files |
1614 |
|
that list on a per-.cm-file basis each toplevel symbol defined or |
1615 |
|
imported. The location of the index file for |
1616 |
|
<p>/<d>.cm is <p>/CM/INDEX/<d>.cm. |
1617 |
|
To enable index-file generation, set CM.Control.generate_index to true |
1618 |
|
or export an environment-symbol: export CM_GENERATE_INDEX=true. |
1619 |
|
|
1620 |
|
The CM manual has been updated accordingly. |
1621 |
|
|
1622 |
|
2. I made some slight modifications to the c-calls API in MLRISC. |
1623 |
|
|
1624 |
|
a) There is now a callback to support saving/restoring of |
1625 |
|
dedicated but caller-save registers around the actual call |
1626 |
|
instruction. |
1627 |
|
b) One can optionally specify a comment-annotation for the |
1628 |
|
call instruction. |
1629 |
|
|
1630 |
|
3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop. |
1631 |
|
(For example, the comment annotation shows the C prototype of |
1632 |
|
the function being called.) |
1633 |
|
|
1634 |
|
---------------------------------------------------------------------- |
1635 |
|
Name: Matthias Blume |
1636 |
|
Date: 2001/06/01 13:30:00 EDT |
1637 |
|
Tag: blume-20010601-nlffi-cleanup |
1638 |
|
Description: |
1639 |
|
|
1640 |
|
This is mostly a cleanup of MLFFI stuff: |
1641 |
|
|
1642 |
|
- some signature files have been put into a more exposed place |
1643 |
|
- the ugly 'f type parameter is gone (simplifies types tremendously!) |
1644 |
|
- ml-nlffigen changed accordingly |
1645 |
|
- tutorial updated |
1646 |
|
|
1647 |
|
Other changes: |
1648 |
|
|
1649 |
|
- author's affiliation in CM manual(s) updated |
1650 |
|
- some more recognized keywords added to Allen's sml.sty |
1651 |
|
|
1652 |
|
---------------------------------------------------------------------- |
1653 |
|
Name: Matthias Blume |
1654 |
|
Date: 2001/05/25 15:30:00 EDT |
1655 |
|
Tag: blume-20010525-iptr |
1656 |
|
Description: |
1657 |
|
|
1658 |
|
- put the official 110.33-README (as it appears on the ftp server) under |
1659 |
|
CVS |
1660 |
|
- fixed a small bug related to incomplete pointer types in |
1661 |
|
ml-nlffigen |
1662 |
|
- small cosmetic change to the ml-nlffi-lib's "arr" type constructor |
1663 |
|
(it does not need the 'f type parameter) |
1664 |
|
|
1665 |
|
---------------------------------------------------------------------- |
1666 |
|
Name: Matthias Blume |
1667 |
|
Date: 2001/05/23 14:30:00 EDT |
1668 |
|
Tag: Release_110_33 |
1669 |
|
Description: |
1670 |
|
|
1671 |
|
New version number (110.33). New bootfiles. |
1672 |
|
|
1673 |
|
---------------------------------------------------------------------- |
1674 |
|
Name: Matthias Blume |
1675 |
|
Date: 2001/05/22 18:06:00 EDT |
1676 |
|
Tag: blume-20010522-targets |
1677 |
|
Description: |
1678 |
|
|
1679 |
|
Made install.sh use file config/targets.customized if it exists, falling |
1680 |
|
back to config/targets if it doesn't. This way one can have a customized |
1681 |
|
version of the targets file without touching the "real thing", thus |
1682 |
|
eliminating the constant fear of accidentally checking something bogus |
1683 |
|
back into the CVS repository... (File config/targets.customized must |
1684 |
|
not be added to the repository!) |
1685 |
|
|
1686 |
|
---------------------------------------------------------------------- |
1687 |
|
Name: Matthias Blume |
1688 |
|
Date: 2001/05/22 16:30:00 EDT |
1689 |
|
Tag: blume-20010522-minitut |
1690 |
|
Description: |
1691 |
|
|
1692 |
|
1. Bug fix in ml-nlffigen; now (hopefully) correctly handling |
1693 |
|
struct returns. |
1694 |
|
2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt. This is some very |
1695 |
|
incomplete, preliminary documentation for NLFFI. |
1696 |
|
|
1697 |
|
---------------------------------------------------------------------- |
1698 |
|
Name: Matthias Blume |
1699 |
|
Date: 2001/05/14 11:30:00 EDT |
1700 |
|
Tag: blume-20010514-script |
1701 |
|
Description: |
1702 |
|
|
1703 |
|
Some bugs in install script fixed. |
1704 |
|
|
1705 |
|
In addition to that I also made a slight change to the NLFFI API: |
1706 |
|
Functors generated by ml-nlffigen now take the dynamic library as a |
1707 |
|
straight functor argument, not as a suspended one. (The original |
1708 |
|
functor code used to force the suspension right away anyway, so there |
1709 |
|
was nothing gained by this complication of the interface.) |
1710 |
|
|
1711 |
|
---------------------------------------------------------------------- |
1712 |
|
Name: Matthias Blume |
1713 |
|
Date: 2001/05/11 14:35:00 EDT |
1714 |
|
Tag: blume-20010511-ml-nlffi |
1715 |
|
Description: |
1716 |
|
|
1717 |
|
I finally took the plunge and added my new FFI code to the main |
1718 |
|
repository. For x86-linux it is now ready for prime-time. |
1719 |
|
|
1720 |
|
There are two new subdirectories of "src": |
1721 |
|
|
1722 |
|
- ml-nlffi-lib: |
1723 |
|
The utility library for programs using the FFI interface. |
1724 |
|
Here is the implementation of $/c.cm and its associated low-level |
1725 |
|
partners $/c-int.cm and $/memory.cm. |
1726 |
|
- ml-nlffigen: |
1727 |
|
A stand-alone program for generating ML glue code from C source |
1728 |
|
code. |
1729 |
|
|
1730 |
|
Building ml-nlffigen requires $/ckit-lib.cm. |
1731 |
|
|
1732 |
|
The config/install.sh script has been updates to do the Right Thing |
1733 |
|
(hopefully). |
1734 |
|
|
1735 |
|
Notice that the source tree for the C-Kit will not be put under "src" |
1736 |
|
but directly under the installation root directory. (This is the |
1737 |
|
structure that currently exists on the CVS server when you check out |
1738 |
|
module "sml".) Fortunately, config/install.sh knows about this oddity. |
1739 |
|
|
1740 |
|
Bugs: No documentation yet. |
1741 |
|
|
1742 |
|
---------------------------------------------------------------------- |
1743 |
|
Name: Matthias Blume |
1744 |
|
Date: 2001/05/09 16:35:00 EDT |
1745 |
|
Tag: blume-20010509-cpscontract |
1746 |
|
Description: |
1747 |
|
|
1748 |
|
Fixed a bug in the accounting code in cpsopt/contract.sml. (The |
1749 |
|
wrapper/unwrapper elimination did not decrement usage counts and some |
1750 |
|
dead variables got overlooked by the dead-up logic.) |
1751 |
|
|
1752 |
|
---------------------------------------------------------------------- |
1753 |
|
Name: Lal George |
1754 |
|
Date: 2001/05/08 17:26:09 EDT |
1755 |
|
Tag: george-20010508-omit-frameptr |
1756 |
|
Description: |
1757 |
|
|
1758 |
|
Changes to implement the omit-frame-pointer optimization to support |
1759 |
|
raw C calls. For now, there is only support on the Intel x86, but |
1760 |
|
other architectures will follow as more experience is gained with this. |
1761 |
|
|
1762 |
|
|
1763 |
|
---------------------------------------------------------------------- |
1764 |
|
Name: Matthias Blume |
1765 |
|
Date: 2001/05/07 14:40:00 EDT |
1766 |
|
Tag: blume-20010507-proxies |
1767 |
|
Description: |
1768 |
|
|
1769 |
|
I made into "proxy libraries" all libraries that qualify for such a |
1770 |
|
change. (A qualifying library is a library that has another library or |
1771 |
|
groups as its sole member and repeats that member's export list |
1772 |
|
verbatim. A proxy library avoids this repetition by omitting its export |
1773 |
|
list, effectively inheriting the list that its (only) member exports. |
1774 |
|
See the CM manual for more explanation.) |
1775 |
|
The main effect is that explicit export lists for these libraries |
1776 |
|
do not have to be kepts in sync, making maintenance a bit easier. |
1777 |
|
|
1778 |
|
I also added copyright notices to many .cm-files. |
1779 |
|
|
1780 |
|
Last but not least, I made a new set of bootfiles. |
1781 |
|
|
1782 |
|
---------------------------------------------------------------------- |
1783 |
|
Name: Matthias Blume |
1784 |
|
Date: 2001/05/04 17:00:00 EDT |
1785 |
|
Tag: blume-20010504-cm-lsplit |
1786 |
|
Description: |
1787 |
|
|
1788 |
|
0. John merged pending changes to $/smlnj-lib.cm |
1789 |
|
|
1790 |
|
1. Allen's previous change accidentally backed out of one of Lal's |
1791 |
|
earlier changes. I undid this mistake (re-introducing Lal's change). |
1792 |
|
|
1793 |
|
2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm) |
1794 |
|
within the compiler where applicable. There is some code simplification |
1795 |
|
because of that. |
1796 |
|
|
1797 |
|
3. The "split" phase (in FLINT) is now part of the default list of phases. |
1798 |
|
Compiler.Control.LambdaSplitting.* can be used to globally control the |
1799 |
|
lambda-splitting (cross-module-inlining) engine. In addition to that, |
1800 |
|
it can now also be controlled on a per-source basis: CM has been taught |
1801 |
|
a new tool parameter applicable to ML source files. |
1802 |
|
|
1803 |
|
- To turn lambda-splitting off completely: |
1804 |
|
local open Compiler.Control.LambdaSplitting in |
1805 |
|
val _ = set Off |
1806 |
|
end |
1807 |
|
- To make "no lambda-splitting" the global default (but allow per-source |
1808 |
|
overriding); this is the initial setting: |
1809 |
|
local open Compiler.Control.LambdaSplitting in |
1810 |
|
val _ = set (Default NONE) |
1811 |
|
end |
1812 |
|
- To make "lambda-splitting with aggressiveness a" the global default |
1813 |
|
(and allow per-source overriding): |
1814 |
|
local open Compiler.Control.LambdaSplitting in |
1815 |
|
val _ = set (Default (SOME a)) |
1816 |
|
end |
1817 |
|
|
1818 |
|
- To turn lambda-splitting off for a given ML souce file (say: a.sml) |
1819 |
|
write (in the respective .cm-file): |
1820 |
|
a.sml (lambdasplitting:off) |
1821 |
|
- To turn lambda-splitting for a.sml on with minimal aggressiveness: |
1822 |
|
a.sml (lambdasplitting:on) |
1823 |
|
- To turn lambda-splitting for a.sml on with aggressiveness <a> (where |
1824 |
|
<a> is a decimal non-negative integer): |
1825 |
|
a.sml (lambdasplitting:<a>) |
1826 |
|
- To turn lambda-splitting for a.sml on with maximal aggressiveness: |
1827 |
|
a.sml (lambdasplitting:infinity) |
1828 |
|
- To use the global default for a.sml: |
1829 |
|
a.sml (lambdasplitting:default) |
1830 |
|
or simply |
1831 |
|
a.sml |
1832 |
|
|
1833 |
|
---------------------------------------------------------------------- |
1834 |
|
Name: Allen Leung |
1835 |
|
Date: 2001/05/04 01:57:00 EDT |
1836 |
|
Tag: leunga-20010504-sync |
1837 |
|
Description: |
1838 |
|
|
1839 |
|
MLRISC features. |
1840 |
|
|
1841 |
|
1. Fix to CMPXCHG instructions. |
1842 |
|
2. Changed RA interface to allow annotations in callbacks. |
1843 |
|
3. Added a new method to the stream interface to allow annotations updates. |
1844 |
|
|
1845 |
|
---------------------------------------------------------------------- |
1846 |
|
Name: Matthias Blume |
1847 |
|
Date: 2001/05/01 11:45:00 EDT |
1848 |
|
Tag: blume-20010501-pcedittmp |
1849 |
|
Description: |
1850 |
|
|
1851 |
|
Changed install.sh to use the current working directory instead of |
1852 |
|
/usr/tmp for a temporary file (pcedittmp). The previous choice |
1853 |
|
of /usr/tmp caused trouble with MacOS X because of file premission |
1854 |
|
problems. |
1855 |
|
|
1856 |
|
---------------------------------------------------------------------- |
1857 |
|
Name: Matthias Blume |
1858 |
|
Date: 2001/04/20 11:10:00 EDT |
1859 |
|
Tag: blume-20010420-inMLflag |
1860 |
|
Description: |
1861 |
|
|
1862 |
|
- added vp_limitPtrMask to vproc-state.h |
1863 |
|
(for use by the raw-C-calls mechanism to implement proper interrupt |
1864 |
|
handling) |
1865 |
|
- made the ML compiler aware of various data-structure offsets so it |
1866 |
|
can generate code for accessing the vp_inML flag and vp_limitPtrMask |
1867 |
|
- tweaked mlriscGen.sml to have it emit interrupt-handling code for |
1868 |
|
raw C-calls |
1869 |
|
|
1870 |
|
---------------------------------------------------------------------- |
1871 |
|
Name: Lal George |
1872 |
|
Date: 2001/04/20 09:15:28 EDT |
1873 |
|
Tag: george-20010420-macosX |
1874 |
|
Description: |
1875 |
|
|
1876 |
|
- Changes to port to Mac OS X; Darwin. |
1877 |
|
|
1878 |
|
- In the process I found that sqrt was broken on the PPC, because the |
1879 |
|
fsqrt instruction is not implemented. |
1880 |
|
|
1881 |
|
---------------------------------------------------------------------- |
1882 |
|
Name: Matthias Blume |
1883 |
|
Date: 2001/04/18 12:45:00 EDT |
1884 |
|
Tag: blume-20010418-ccalls |
1885 |
|
Description: |
1886 |
|
|
1887 |
|
- fixed two off-by-4 errors in the x86-specific c-calls implementation |
1888 |
|
(this bug prevented structure arguments containing pointers from being |
1889 |
|
passed correctly) |
1890 |
|
- changed the raw-C-call code in mlriscGen.sml in such a way that |
1891 |
|
structure arguments are represented as a pointer to the beginning |
1892 |
|
of the structure (instead of having a series of synthesized arguments, |
1893 |
|
one for each structure member) |
1894 |
|
|
1895 |
|
- made makeml script's verbosity level configurable via environment |
1896 |
|
variable (MAKEML_VERBOSITY) |
1897 |
|
|
1898 |
|
- eliminated placeholder implementations for f32l, w16s, i16s, and f32s |
1899 |
|
in rawmem-x86.sml; we are now using the real thing |
1900 |
|
|
1901 |
|
---------------------------------------------------------------------- |
1902 |
|
Name: Matthias Blume |
1903 |
|
Date: 2001/03/22 16:25:00 EST |
1904 |
|
Tag: blume-20010322-bootfiles |
1905 |
|
Description: |
1906 |
|
|
1907 |
|
Created a new set of bootfiles (for your automatic installation convenience). |
1908 |
|
|
1909 |
|
---------------------------------------------------------------------- |
1910 |
|
Name: Matthias Blume |
1911 |
|
Date: 2001/03/22 15:10:00 EST |
1912 |
|
Tag: blume-20010322-rawmem-parcm |
1913 |
|
Description: |
1914 |
|
|
1915 |
|
1. All "raw memory access" primitives for the new FFI are implemented now |
1916 |
|
(at least on the x86). |
1917 |
|
2. Some further cleanup of CM's parallel make mechanism. |
1918 |
|
|
1919 |
|
---------------------------------------------------------------------- |
1920 |
|
Name: Matthias Blume |
1921 |
|
Date: 2001/03/19 17:53:00 EST |
1922 |
|
Tag: blume-20010319-parallel |
1923 |
|
Description: |
1924 |
|
|
1925 |
|
Parallel make (using compile servers) now works again. |
1926 |
|
|
1927 |
|
To this end, CM.stabilize and CMB.make have been modified to work in |
1928 |
|
two passes when compile servers are attached: |
1929 |
|
1. Compile everything, do not perform stabilization; this pass |
1930 |
|
uses compile servers |
1931 |
|
2. Stabilize everything; this pass does not use compile servers |
1932 |
|
If there are no compile servers, the two passes are combined into one |
1933 |
|
(as before). Splitting the passes increases the inherent parallelism |
1934 |
|
in the dependency graph because the entire graph including all |
1935 |
|
libraries is available at the same time. This, in turn, improves |
1936 |
|
server utilization. The downside is that the master process will |
1937 |
|
have to do some extra work after compilation is done (because for |
1938 |
|
technical reasons it must re-read all the binfiles during stabilization). |
1939 |
|
|
1940 |
|
---------------------------------------------------------------------- |
1941 |
|
Name: Matthias Blume |
1942 |
|
Date: 2001/03/16 12:22:00 EST |
1943 |
|
Tag: blume-20010316-bootfiles |
1944 |
|
Description: |
1945 |
|
|
1946 |
|
Created a new set of bootfiles (for your automatic installation convenience). |
1947 |
|
|
1948 |
|
---------------------------------------------------------------------- |
1949 |
|
Name: Matthias Blume |
1950 |
|
Date: 2001/03/16 11:00:00 EST |
1951 |
|
Tag: blume-20010316-MLTREE-fixup |
1952 |
|
Description: |
1953 |
|
|
1954 |
|
This is a minor fixup for an (untagged) earlier commit by Allen. |
1955 |
|
(A file was missing). |
1956 |
|
|
1957 |
|
---------------------------------------------------------------------- |
1958 |
|
Name: Allen Leung |
1959 |
|
Date: Mon Mar 5 18:54:57 EST 2001 |
1960 |
|
Tag: leunga-20010305-cut-support |
1961 |
|
|
1962 |
|
1. New support for alternative control-flow in MLTREE. |
1963 |
|
Currently we support |
1964 |
|
|
1965 |
|
FLOW_TO(CALL ...., [k1,...,kn]) |
1966 |
|
|
1967 |
|
This is needed for 'cuts to' in C-- and try/handle-like constructs |
1968 |
|
in Moby |
1969 |
|
|
1970 |
|
New assembler flag "asm-show-cutsto" to turn on control-flow debugging. |
1971 |
|
|
1972 |
|
2. Register Allocator |
1973 |
|
|
1974 |
|
Changes in interface [from Fermin, John] |
1975 |
|
|
1976 |
|
3. Alpha 8-bit SLL support [Fermin] |
1977 |
|
|
1978 |
|
4. All architectures |
1979 |
|
|
1980 |
|
A new module (ClusterExpandCopies) for expanding parallel copies. |
1981 |
|
|
1982 |
|
---------------------------------------------------------------------- |
1983 |
|
Name: Allen Leung |
1984 |
|
Date: 2001/02/27 23:07:00 EST |
1985 |
|
Tag: leunga-20010227-minor-stuff |
1986 |
|
|
1987 |
|
1. Alpha bug fix for CMOVNE |
1988 |
|
2. Handle mltree COND(..,FCMP ...,...) |
1989 |
|
3. Bug fix in simplifier |
1990 |
|
|
1991 |
|
---------------------------------------------------------------------- |
1992 |
|
Name: Matthias Blume |
1993 |
|
Date: 2001/01/30 17:50:00 EST |
1994 |
|
Tag: blume-20010130-sync |
1995 |
|
Description: |
1996 |
|
|
1997 |
|
This is just a minor update to sync my devel branch with the main brach. |
1998 |
|
The only visible change is the addition of some README files. |
1999 |
|
|
2000 |
|
---------------------------------------------------------------------- |
2001 |
|
Name: Matthias Blume |
2002 |
|
Date: 2001/01/12 23:30:00 JST |
2003 |
|
Tag: blume-20010112-bootfiles |
2004 |
|
Description: |
2005 |
|
|
2006 |
|
Made a new set of bootfiles that goes with the current state of the |
2007 |
|
repository. |
2008 |
|
|
2009 |
|
---------------------------------------------------------------------- |
2010 |
|
Name: Matthias Blume |
2011 |
|
Date: 2001/01/12 21:20:00 JST |
2012 |
|
Tag: blume-20010112-sync |
2013 |
|
Description: |
2014 |
|
|
2015 |
|
I am just flushing out some minor changes that had accumulated in |
2016 |
|
my private branch in order to sync with the main tree. (This is |
2017 |
|
mainly because I had CVS trouble when trying to merge _into_ my |
2018 |
|
private branch.) |
2019 |
|
|
2020 |
|
Most people should be completely unaffected by this. |
2021 |
|
|
2022 |
|
---------------------------------------------------------------------- |
2023 |
|
Name: Allen Leung |
2024 |
|
Date: Thu Jan 11 21:03:00 EST 2001 |
2025 |
|
Tag: leunga-20010111-labexp=mltree |
2026 |
|
Description: |
2027 |
|
|
2028 |
|
1. Removed the type LabelExp and replace it by MLTree. |
2029 |
|
2. Rewritten mltree-simplify with the pattern matcher tool. |
2030 |
|
3. There were some bugs in alpha code generator which would break |
2031 |
|
64-bit code generation. |
2032 |
|
4. Redo the tools to generate code with the |
2033 |
|
5. The CM files in MLRISC (and in src/system/smlnj/MLRISC) |
2034 |
|
are now generated by perl scripts. |
2035 |
|
|
2036 |
|
---------------------------------------------------------------------- |
2037 |
|
Name: Matthias Blume |
2038 |
|
Date: 2001/01/10 21:55:00 JST |
2039 |
|
Tag: blume-20010110-rcc |
2040 |
|
Description: |
2041 |
|
|
2042 |
|
The RCC stuff now seems to work (but only on the x86). |
2043 |
|
This required hacking of the c-calls interface (and -implementation) in |
2044 |
|
MLRISC. |
2045 |
|
|
2046 |
|
Normal compiler users should be unaffected. |
2047 |
|
|
2048 |
|
---------------------------------------------------------------------- |
2049 |
|
Name: Matthias Blume |
2050 |
|
Date: 2001/01/09 01:20:00 JST |
2051 |
|
Tag: blume-20010109-rcc |
2052 |
|
Description: |
2053 |
|
|
2054 |
|
This is a fairly big patch, flushing out a large number of pending |
2055 |
|
changes that I made to my development copy over the last couple of days. |
2056 |
|
|
2057 |
|
Of practical relevance at this moment is a workaround for a pickling |
2058 |
|
bug that Allen ran into the other day. The cause of the bug itself is |
2059 |
|
still unknown and it might be hard to fix it properly, but the |
2060 |
|
workaround has some merits of its own (namely somewhat reducing pickling |
2061 |
|
overhead for certain libraries). Therefore, I think this solution should |
2062 |
|
be satisfactory at this time. |
2063 |
|
|
2064 |
|
The rest of the changes (i.e., the vast majority) has to do with my |
2065 |
|
ongoing efforts of providing direct support for C function calls from |
2066 |
|
ML. At the moment there is a new primop "RAW_CCALL", typing magic |
2067 |
|
in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new |
2068 |
|
case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to |
2069 |
|
translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to |
2070 |
|
handle RCC. |
2071 |
|
|
2072 |
|
The last part (the changes to mlriscGen.sml) are still known to be |
2073 |
|
wrong on the x86 and not implemented on all other architectures. But |
2074 |
|
the infrastructure is in place. I had to change a few functor |
2075 |
|
signatures in the backend to be able to route the CCalls interface |
2076 |
|
from MLRISC there, and I had to specialize the mltree type (on the |
2077 |
|
x86) to include the necessary extensions. (The extensions themselves |
2078 |
|
were already there and redy to go in MLRISC/x86). |
2079 |
|
|
2080 |
|
Everything should be very happy as soon as someone helps me with |
2081 |
|
mlriscGen.sml... |
2082 |
|
|
2083 |
|
In any case, nothing of this should matter to anyone as long as the |
2084 |
|
new primop is not being used (which is going to be the case unless you |
2085 |
|
find it where I hid it :). The rest of the compiler is completely |
2086 |
|
unaffected. |
2087 |
|
|
2088 |
|
---------------------------------------------------------------------- |
2089 |
|
Name: Matthias Blume |
2090 |
|
Date: 2001/01/05 00:30:00 JST |
2091 |
|
Tag: blume-20010105-primops |
2092 |
|
Description: |
2093 |
|
|
2094 |
|
Added some experimental support for work that I am doing right now. |
2095 |
|
These changes mostly concern added primops, but there is also a new |
2096 |
|
experimental C library in the runtime system (but currently not enabled |
2097 |
|
anywhere except on Linux/X86). |
2098 |
|
|
2099 |
|
In the course of adding primops (and playing with them), I discovered that |
2100 |
|
Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly |
2101 |
|
broken. (Zhong was very right he labeled this stuff as "major gross hack".) |
2102 |
|
To recover, I made type information in INL_PRIM mandatory and changed |
2103 |
|
prim.sml as well as built-in.sml accordingly. The InLine structure now |
2104 |
|
has complete, correct type information (i.e., no bottom types). |
2105 |
|
|
2106 |
|
Since all these changes mean that we need new binfiles, I also bumped the |
2107 |
|
version number to 110.32.1. |
2108 |
|
|
2109 |
|
---------------------------------------------------------------------- |
2110 |
|
Name: Matthias Blume |
2111 |
|
Date: 2000/12/30 22:10:00 JST |
2112 |
|
Tag: blume-20001230-various |
2113 |
|
Description: |
2114 |
|
|
2115 |
|
Added proxy libraries for MLRISC and let MLRISC libraries refer |
2116 |
|
to each other using path anchors. (See CM manual for explanation.) |
2117 |
|
|
2118 |
|
Updated CM documentation. |
2119 |
|
|
2120 |
|
Fixed some bugs in CM. |
2121 |
|
|
2122 |
|
Implemented "proxy" libraries (= syntactic sugar for CM). |
2123 |
|
|
2124 |
|
Added "-quiet" option to makeml and changed runtime system accordingly. |
2125 |
|
|
2126 |
|
Added cleanup handler for exportML to reset timers and compiler stats. |
2127 |
|
|
2128 |
|
---------------------------------------------------------------------- |
2129 |
|
Name: Lal George |
2130 |
|
Date: 2000/12/22 22:22:58 EST 2000 |
2131 |
|
Tag: Release_110_32 |
2132 |
|
Description: |
2133 |
|
|
2134 |
|
Infinite precision used throughout MLRISC. |
2135 |
|
see MLRISC/mltree/machine-int.sig |
2136 |
|
|
2137 |
|
---------------------------------------------------------------------- |
2138 |
|
Name: Matthias Blume |
2139 |
|
Date: 2000/12/22 23:16:00 JST |
2140 |
|
Tag: blume-20001222-warn |
2141 |
|
Description: |
2142 |
|
|
2143 |
|
Corrected wording and formatting of some CM warning message which I |
2144 |
|
broke in my previous patch. |
2145 |
|
|
2146 |
|
---------------------------------------------------------------------- |
2147 |
|
Name: Matthias Blume |
2148 |
|
Date: 2000/12/22 21:20:00 JST |
2149 |
|
Tag: blume-20001222-anchorenv |
2150 |
|
Description: |
2151 |
|
|
2152 |
|
Fixed CM's handling of anchor environments in connection with CMB.make. |
2153 |
|
|
2154 |
|
---------------------------------------------------------------------- |
2155 |
|
Name: Matthias Blume |
2156 |
|
Date: 2000/12/22 13:15:00 JST |
2157 |
|
Tag: blume-20001222-cleanup |
2158 |
|
Description: |
2159 |
|
|
2160 |
|
Removed src/cm/ffi which does not (and did not) belong here. |
2161 |
|
|
2162 |
|
---------------------------------------------------------------------- |
2163 |
|
Name: Matthias Blume |
2164 |
|
Date: 2000/12/21 23:55:00 JST |
2165 |
|
Tag: blume-20001221-exn |
2166 |
|
Description: |
2167 |
|
|
2168 |
|
Probably most important: CM no longer silently swallows all exceptions |
2169 |
|
in the compiler. |
2170 |
|
Plus: some other minor CM changes. For example, CM now reports some |
2171 |
|
sizes for generated binfiles (code, data, envpickle, lambdapickle). |
2172 |
|
|
2173 |
|
---------------------------------------------------------------------- |
2174 |
|
Name: Matthias Blume |
2175 |
|
Date: 2000/12/15 00:01:05 JST |
2176 |
|
Tag: blume-20001215-dirtool |
2177 |
|
Description: |
2178 |
|
|
2179 |
|
- "dir" tool added. |
2180 |
|
- improvements and cleanup to Tools structure |
2181 |
|
- documentation updates |
2182 |
|
|
2183 |
|
---------------------------------------------------------------------- |
2184 |
|
Name: Allen Leung |
2185 |
|
Date: Thu Dec 14 03:45:24 EST 2000 |
2186 |
|
Description: |
2187 |
|
Tag: leunga-20001214-int-inf |
2188 |
|
Description: |
2189 |
|
|
2190 |
|
In IntInf, added these standard functions, which are missing from our |
2191 |
|
implementation: |
2192 |
|
|
2193 |
|
andb : int * int -> int |
2194 |
|
xorb : int * int -> int |
2195 |
|
orb : int * int -> int |
2196 |
|
notb : int -> int |
2197 |
|
<< : int * word -> int |
2198 |
|
~>> : int * word -> int |
2199 |
|
|
2200 |
|
Not tested, I hope they are correct. |
2201 |
|
|
2202 |
|
---------------------------------------------------------------------- |
2203 |
|
Name: Allen Leung |
2204 |
|
Date: Fri Dec 8 19:23:26 EST 2000 |
2205 |
|
Description: |
2206 |
|
Tag: leunga-20001208-nowhere |
2207 |
|
Description: |
2208 |
|
|
2209 |
|
Slight improvements to the 'nowhere' tool to handle OR-patterns, |
2210 |
|
to generate better error messages etc. Plus a brief manual. |
2211 |
|
|
2212 |
|
---------------------------------------------------------------------- |
2213 |
|
Name: Lal George |
2214 |
|
Date: 2000/12/08 09:54:02 EST 2000 |
2215 |
|
Tag: Release_110_31 |
2216 |
|
Description: |
2217 |
|
|
2218 |
|
- Version 110.31 |
2219 |
|
---------------------------------------------------------------------- |
2220 |
|
Name: Allen Leung |
2221 |
|
Date: Thu Dec 7 22:01:04 EST 2000 |
2222 |
|
Tag: leunga-20001207-cell-monster-hack |
2223 |
|
Description: |
2224 |
|
|
2225 |
|
Major MLRISC internal changes. Affect all clients. |
2226 |
|
Summary: |
2227 |
|
|
2228 |
|
1. Type CELLS.cell = int is now replaced by a datatype. |
2229 |
|
As a result, the old regmap is now gone. Almost all interfaces |
2230 |
|
in MLRISC change as a consequence. |
2231 |
|
|
2232 |
|
2. A new brand version of machine description tool (v3.0) that generates |
2233 |
|
modules expecting the new interface. The old version is removed. |
2234 |
|
|
2235 |
|
3. The RA interface has been further abstracted into two new functors. |
2236 |
|
RISC_RA and X86RA. These functors have much simpler interfaces. |
2237 |
|
[See also directory MLRISC/demo.] |
2238 |
|
|
2239 |
|
4. Some other new source->source code generation tools are available: |
2240 |
|
|
2241 |
|
a. MLRISC/Tools/RewriteGen -- generate rewriters from rules. |
2242 |
|
b. MLRISC/Tools/WhereGen -- expands conditional pattern matching rules. |
2243 |
|
I use this tool to generate the peephole optimizers---with the new |
2244 |
|
cell type changes, peephole rules are becoming difficult to write |
2245 |
|
without conditional pattern matching. |
2246 |
|
|
2247 |
|
5. More Intmap -> IntHashTable change. Previous changes by Matthias didn't |
2248 |
|
cover the entire MLRISC source tree so many things broke. |
2249 |
|
|
2250 |
|
6. CM files have been moved to the subdirectory MLRISC/cm. |
2251 |
|
They are moved because there are a lot of them and they clutter up the |
2252 |
|
root dir. |
2253 |
|
|
2254 |
|
7. More detailed documentation to come... |
2255 |
|
|
2256 |
|
NOTE: To rebuild from 110.30 (ftp distribution), you'll have to do |
2257 |
|
a makeml -rebuild first. This is because of other other |
2258 |
|
changes that Matthias has made (see below). |
2259 |
|
|
2260 |
|
|
2261 |
|
---------------------------------------------------------------------- |
2262 |
|
Name: Matthias Blume |
2263 |
|
Date: 2000/11/30 23:12:00 JST |
2264 |
|
Tag: blume-20001130-filereorg |
2265 |
|
Description: |
2266 |
|
|
2267 |
|
Some manual updates and some file reorganizations in CM. |
2268 |
|
|
2269 |
|
---------------------------------------------------------------------- |
2270 |
|
Name: Matthias Blume |
2271 |
|
Date: 2000/11/24 17:45:00 JST |
2272 |
|
Tag: blume-20001124-link |
2273 |
|
Description: |
2274 |
|
|
2275 |
|
Drastically improved link traversal code for the case that the dynamic |
2276 |
|
value was already loaded at bootstrap time. As a result, CM and CMB |
2277 |
|
now both load blazingly fast -- even on a very slow machine. Also, |
2278 |
|
memory consumption has been further reduced by this. |
2279 |
|
|
2280 |
|
Warning: The format of the PIDMAP file has changed. THerefore, to |
2281 |
|
bootstrap you have to do this: |
2282 |
|
|
2283 |
|
1. Run CMB.make |
2284 |
|
2. Make a symbolic link for the boot directory: |
2285 |
|
ln -s sml.boot.ARCH-OS xxx |
2286 |
|
3. "Rebuild" the boot directory: |
2287 |
|
./makeml -boot xxx -rebuild sml ; rm xxx |
2288 |
|
4. Boot normally: |
2289 |
|
./makeml |
2290 |
|
|
2291 |
|
---------------------------------------------------------------------- |
2292 |
|
Name: Matthias Blume |
2293 |
|
Date: 2000/11/21 21:20:00 JST |
2294 |
|
Tag: blume-20001121-tools |
2295 |
|
Description: |
2296 |
|
|
2297 |
|
Continued hacking on autoloading problem -- with success this time. |
2298 |
|
Also changed tool-plugin mechanism. See new CM manual. |
2299 |
|
|
2300 |
|
---------------------------------------------------------------------- |
2301 |
|
Name: Matthias Blume |
2302 |
|
Date: 2000/11/19 14:30:00 JST |
2303 |
|
Tag: blume-20001119-autoload |
2304 |
|
Description: |
2305 |
|
|
2306 |
|
Some hacking to make autoloading faster. Success for CMB, no success |
2307 |
|
so far for CM. There is a reduced structure CM' that autoloads faster. |
2308 |
|
(This is a temporary, non-documented hack to be eliminated again when |
2309 |
|
the general problem is solved.) |
2310 |
|
|
2311 |
|
---------------------------------------------------------------------- |
2312 |
|
Name: Matthias Blume |
2313 |
|
Date: 2000/11/17 14:10:00 JST |
2314 |
|
Tag: blume-20001117-pickle-lib |
2315 |
|
Description: |
2316 |
|
|
2317 |
|
1. Eliminated comp-lib.cm |
2318 |
|
2. Made pickle-lib.cm |
2319 |
|
3. Eliminated all uses of intset.sml (from comp-lib.cm) |
2320 |
|
4. Replaced all uses of intmap.{sig,sml} (from comp-lib.cm) with |
2321 |
|
equivalent constructs from smlnj-lib.cm (INtHashTable). |
2322 |
|
5. Point 4. also goes for those uses of intmap.* in MLRISC. |
2323 |
|
Duplicated intmap modules thrown out. |
2324 |
|
6. Hunted down all duplicated SCC code and replaced it with |
2325 |
|
equivalent stuff (GraphSCCFn from smlnj-lib.cm). |
2326 |
|
7. Rewrote Feedback module. |
2327 |
|
8. Moved sortedlist.sml into viscomp-lib.cm. Eventually it |
2328 |
|
should be thrown out and equivalent modules from smlnj-lib.cm |
2329 |
|
should be used (IntRedBlackSet, IntListSet, ...). |
2330 |
|
|
2331 |
|
Confirmed that compiler compiles to fixpoint. |
2332 |
|
|
2333 |
|
---------------------------------------------------------------------- |
2334 |
|
Name: Allen Leung |
2335 |
|
Date: 2000/11/10 18:00:00 |
2336 |
|
Tag: leunga-20001110-new-x86-fp |
2337 |
|
|
2338 |
|
A new x86 floating point code generator has been added. |
2339 |
|
By default this is turned off. To turn this on, do: |
2340 |
|
|
2341 |
|
CM.autoload "$smlnj/compiler.cm"; |
2342 |
|
Compiler.Control.MLRISC.getFlag "x86-fast-fp" := true; |
2343 |
|
|
2344 |
|
Changes: |
2345 |
|
|
2346 |
|
1. Changed FTAN to FPTAN so that the assembly output is correct. |
2347 |
|
2. Changed the extension callback for FTANGENT to generate: |
2348 |
|
|
2349 |
|
fptan |
2350 |
|
fstp %st(0) |
2351 |
|
instead of |
2352 |
|
fptan |
2353 |
|
fstpl ftempmem |
2354 |
|
|
2355 |
|
3. Numerous assembly fixes for x86. |
2356 |
|
|
2357 |
|
5. Cleaned up the machine code output module x86/x86MC.sml and added |
2358 |
|
support for a whole bunch of instructions and addressing modes: |
2359 |
|
|
2360 |
|
fadd/fsub/fsubr/fmul/fdiv/fdivr %st, %st(n) |
2361 |
|
faddp/fsubp/fsubrp/fmulp/fdivp/fdivrp %st, %st(n) |
2362 |
|
fadd/fsub/fsubr/fmul/fdiv/fdivr %st(n), %st |
2363 |
|
fiadd/fisub/fisubr/fimul/fidiv/fidivr mem |
2364 |
|
fxch %st(n) |
2365 |
|
fld %st(n) |
2366 |
|
fst %st(n) |
2367 |
|
fst mem |
2368 |
|
fstp %st(n) |
2369 |
|
fucom %st(n) |
2370 |
|
fucomp %st(n) |
2371 |
|
|
2372 |
|
All these are now generated when the fast fp mode is turned on. |
2373 |
|
|
2374 |
|
6. Removed the dedicated registers %st(0), ..., %st(7) from X86CpsRegs |
2375 |
|
|
2376 |
|
---------------------------------------------------------------------- |
2377 |
|
Name: Matthias Blume |
2378 |
|
Date: 2000/11/09 11:20:00 JST |
2379 |
|
Tag: blume-20001109-scc |
2380 |
|
Description: |
2381 |
|
|
2382 |
|
Eliminated some code duplication: |
2383 |
|
|
2384 |
|
1. Added "where" clause to GraphSCCFn in SML/NJ Library. |
2385 |
|
(Otherwise the functor is useless.) |
2386 |
|
2. Used GraphSCCFn where SCCUtilFun was used previously. |
2387 |
|
3. Got rid of SCCUtilFun (in comp-lib.cm). |
2388 |
|
|
2389 |
|
---------------------------------------------------------------------- |
2390 |
|
Name: Lal George |
2391 |
|
Date: 2000/11/06 09:02:21 EST 2000 |
2392 |
|
Tag: Release_110_30 |
2393 |
|
Description: |
2394 |
|
|
2395 |
|
- Version 110.30 |
2396 |
|
---------------------------------------------------------------------- |
2397 |
|
Name: Matthias Blume |
2398 |
|
Date: 2000/11/04 14:45:00 |
2399 |
|
Tag: blume-20001104-mlbuild |
2400 |
|
Description: |
2401 |
|
|
2402 |
|
- Made ml-build faster on startup. |
2403 |
|
- Documentation fixes. |
2404 |
|
|
2405 |
|
---------------------------------------------------------------------- |
2406 |
|
Name: Matthias Blume |
2407 |
|
Date: 2000/11/02 17:00:00 JST |
2408 |
|
Tag: blume-20001102-condcomp |
2409 |
|
Description: |
2410 |
|
|
2411 |
|
- Small tweaks to pickler -- new BOOTFILES! |
2412 |
|
- Version bumped to 110.29.2. |
2413 |
|
- Added conditional compilation facility to init.cmi (see comment there). |
2414 |
|
---------------------------------------------------------------------- |
2415 |
|
Name: Allen Leung |
2416 |
|
Date: 2000/10/23 19:31:00 |
2417 |
|
Tag: leunga-20001023-demo-ra |
2418 |
|
|
2419 |
|
1. Minor RA changes that improves spilling on x86 (affects Moby and C-- only) |
2420 |
|
2. Test programs for the graph library updated |
2421 |
|
3. Some new MLRISC demo programs added |
2422 |
|
|
2423 |
|
---------------------------------------------------------------------- |
2424 |
|
Name: Matthias Blume |
2425 |
|
Date: 2000/08/31 22:15:00 JST |
2426 |
|
Tag: blume-20001017-errmsg |
2427 |
|
Description: |
2428 |
|
|
2429 |
|
More error message grief: Where there used to be no messages, there |
2430 |
|
now were some that had bogus error regions. Fixed. |
2431 |
|
|
2432 |
|
---------------------------------------------------------------------- |
2433 |
|
Name: Matthias Blume |
2434 |
|
Date: 2000/08/31 17:30:00 JST |
2435 |
|
Tag: blume-20001017-v110p29p1 |
2436 |
|
Description: |
2437 |
|
|
2438 |
|
I made a version 110.29.1 with new bootfiles. |
2439 |
|
|
2440 |
|
Changes: Modified pickler/unpickler for faster and leaner unpickling. |
2441 |
|
CM documentation changes and a small bugfix in CM's error reporting. |
2442 |
|
|
2443 |
|
---------------------------------------------------------------------- |
2444 |
|
Name: Lal George |
2445 |
|
Date: 2000/09/27 14:42:35 EDT |
2446 |
|
Tag: george-20000927-nodestatus |
2447 |
|
Description: |
2448 |
|
|
2449 |
|
Changed the type of the nodestatus, so that: |
2450 |
|
|
2451 |
|
SPILLED(~1) is now SPILLED |
2452 |
|
SPILLED(m) where m>=0 is now MEMREG(m) |
2453 |
|
SPILLED(s) where s<~1 is now SPILL_LOC(~s) |
2454 |
|
|
2455 |
|
---------------------------------------------------------------------- |
2456 |
|
Name: Matthias Blume |
2457 |
|
Date: 2000/09/07 14:45:00 JST |
2458 |
|
Tag: blume-20000907-cmerrmsg |
2459 |
|
Description: |
2460 |
|
|
2461 |
|
Small tweak to CM to avoid getting ML syntax error messages twice. |
2462 |
|
|
2463 |
|
---------------------------------------------------------------------- |
2464 |
|
Name: Matthias Blume |
2465 |
|
Date: 2000/08/31 18:00:00 JST |
2466 |
|
Tag: blume-20000831-cvsbootfiles |
2467 |
|
Description: |
2468 |
|
|
2469 |
|
New URL for boot files (because the 110.29 files on the BL server do |
2470 |
|
now work correctly with my updated install scripts for yacc and lex). |
2471 |
|
|
2472 |
---------------------------------------------------------------------- |
---------------------------------------------------------------------- |
2473 |
Name: Matthias Blume |
Name: Matthias Blume |
2474 |
Date: 2000/08/08 12:33:00 JST |
Date: 2000/08/08 12:33:00 JST |
3928 |
elaborator). There were a lot of changes during my "linkpath" trials |
elaborator). There were a lot of changes during my "linkpath" trials |
3929 |
that could have been reverted to their original state but weren't. |
that could have been reverted to their original state but weren't. |
3930 |
Please, don't be too harsh on me for messing with this code a bit more |
Please, don't be too harsh on me for messing with this code a bit more |
3931 |
than what was strictly necessary... (I _did_ resist the tempation |
than what was strictly necessary... (I _did_ resist the temptation |
3932 |
of doing any "global reformatting" to avoid an untimely death at |
of doing any "global reformatting" to avoid an untimely death at |
3933 |
Dave's hands. :) |
Dave's hands. :) |
3934 |
|
|