Home My Page Projects Code Snippets Project Openings SML/NJ
Summary Activity Forums Tracker Lists Tasks Docs Surveys News SCM Files

SCM Repository

[smlnj] Annotation of /sml/trunk/HISTORY
ViewVC logotype

Annotation of /sml/trunk/HISTORY

Parent Directory Parent Directory | Revision Log Revision Log


Revision 1294 - (view) (download)

1 : dbm 570 This is the HISTORY file for the Yale SML/NJ CVS repository.
2 :    
3 :     An entry should be made for _every_ commit to the repository.
4 :     The entries in this file will be used when creating the README
5 :     for new versions, so keep that in mind when writing the
6 :     description.
7 :    
8 :     The form of an entry should be:
9 :    
10 :     Name:
11 : george 1003 Date: yyyy/mm/dd
12 : dbm 570 Tag: <post-commit CVS tag>
13 :     Description:
14 : leunga 1142
15 : george 1136 ----------------------------------------------------------------------
16 : blume 1229 Name: Matthias Blume
17 : blume 1294 Date: 2002/12/12 16:25:00 EST
18 :     Tag: blume-20021212-risc-ra
19 :     Description:
20 :    
21 :     Fixed a serious bug in the rewrite code for FP spilling/reloading that
22 :     sent the RA into an infinite loop when floating point registers get
23 :     spilled. (Because of this bug, e.g., nucleic stopped compiling between
24 :     110.37 and 110.38.)
25 :     There was another set of potential problems related to the handling of
26 :     MLRISC annotations (but those did not yet cause real problems, apparently).
27 :    
28 :     ----------------------------------------------------------------------
29 :     Name: Matthias Blume
30 : blume 1291 Date: 2002/12/06 22:40:00 EST
31 :     Tag: blume-20021206-cm-fileid
32 :     Description:
33 :    
34 :     Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
35 :     This fixes the problem of CM getting confused by files that suddenly
36 :     change their identity (e.g., by getting unlinked and recreated by some
37 :     text editor such as vi). There might be a better/cheaper/cleaner way
38 :     of doing this, but for now this will have to do.
39 :    
40 :     ----------------------------------------------------------------------
41 :     Name: Matthias Blume
42 : blume 1288 Date: 2002/10/28 09:50:00 EST
43 :     Tag: blume-20021028-typecheck
44 :     Description:
45 :    
46 :     Exported structure Typecheck from $smlnj/viscomp/core.cm.
47 :    
48 :     ----------------------------------------------------------------------
49 :     Name: Matthias Blume
50 : blume 1282 Date: 2002/10/17 09:10:00 EDT
51 : blume 1279 Tag: Release_110_42
52 :     Description:
53 :    
54 : blume 1282 In good old tradition, there has been a slight hiccup so that we have
55 :     to patch 110.42 after the fact. The old release tag has been replaced
56 :     (see below).
57 :    
58 :     The change solves a problem with two competing approaches the
59 :     configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
60 :     each other's way.
61 :    
62 :     This change only affects the runtime system code and the installer script.
63 :     (No new bootfiles.)
64 :    
65 :     ----------------------------------------------------------------------
66 :     Name: Matthias Blume
67 :     Date: 2002/10/16 12:00:00 EDT
68 :     Tag: Release_110_42_removed
69 :     Description:
70 :    
71 : blume 1279 New working release. New bootfiles.
72 :    
73 :     ----------------------------------------------------------------------
74 :     Name: Matthias Blume
75 : blume 1273 Date: 2002/10/10 13:10:00 EDT
76 :     Tag: blume-20021010-ppc-divs
77 :     Description:
78 :    
79 :     The mltree operator DIVS must be implemented with an overflow check on
80 :     the PPC because the hardware indicates divide-by-zero using "overflow" as
81 :     well.
82 :    
83 :     ----------------------------------------------------------------------
84 :     Name: Matthias Blume
85 : blume 1261 Date: 2002/07/23 11:45:00
86 :     Tag: blume-20020723-smlnj-home
87 :     Description:
88 :    
89 :     Sml now senses the SMLNJ_HOME environment variable. If this is set,
90 :     then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
91 :     CM_PATHCONFIG is also set), the path configuration file is assumed
92 :     to be in $SMLNJ_HOME/lib/pathconfig. This way one can easily move
93 :     the entire tree to some other place and everything will "just work".
94 :    
95 :     (Companion commands such as ml-build and ml-makedepend also sense this
96 :     variable.)
97 :    
98 :     ----------------------------------------------------------------------
99 :     Name: Matthias Blume
100 : blume 1259 Date: 2002/07/12 21:19:00 EDT
101 :     Tag: blume-20020712-liveness
102 :     Description:
103 :    
104 :     Exported two useful "step" functions from liveness module (MLRISC).
105 :    
106 :     ----------------------------------------------------------------------
107 :     Name: Matthias Blume
108 : blume 1253 Date: 2002/07/05 16:00 EDT
109 :     Tag: Release_110_41
110 :     Description:
111 :    
112 :     New working release. New bootfiles.
113 :    
114 :     ----------------------------------------------------------------------
115 :     Name: Matthias Blume
116 : blume 1250 Date: 2002/07/05 10:25:00 EDT
117 :     Tag: blume-20020705-btimp
118 :     Description:
119 :    
120 :     Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
121 :     other clients can set up backtracing support.
122 :    
123 :     ----------------------------------------------------------------------
124 :     Name: Matthias Blume
125 : blume 1249 Date: 2002/06/25 17:23:00 EDT
126 :     Tag: blume-20020625-fpmax
127 :     Description:
128 :    
129 :     Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
130 :     case. (The sense of the isNaN test was reversed -- which made min and
131 :     max always return their first argument.)
132 :    
133 :     ----------------------------------------------------------------------
134 :     Name: Matthias Blume
135 : blume 1240 Date: 2002/06/11
136 :     Tag: blume-20020611-unixpath
137 :     Description:
138 :    
139 :     Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
140 :    
141 :     ----------------------------------------------------------------------
142 :     Name: Matthias Blume
143 : blume 1238 Date: 2002/06/10 16:35:00 EDT
144 :     Tag: blume-20020610-ieeereal
145 :     Description:
146 :    
147 :     I back-ported my implementation of IEEEReal.fromString from the
148 :     idlbasis-devel branch so that we can test it.
149 :    
150 :     Another small change is that ppDec tries to give more information
151 :     than just "<sig>" in the case of functors. However, this code is
152 :     broken in some mysterious way if the functor's body's signature
153 :     has not been declared by ascription but gets inferred from the
154 :     implementation. This needs fixing...
155 :    
156 :     ----------------------------------------------------------------------
157 :     Name: Matthias Blume
158 : blume 1229 Date: 2002/05/31
159 :     Tag: blume-20020531-btrace-mode
160 :     Description:
161 :    
162 :     Resurrected SMLofNJ.Internals.BTrace.mode. (It accidentally fell by
163 :     the wayside when I switched over to using Controls everywhere.)
164 :    
165 :     ----------------------------------------------------------------------
166 : george 1222 Name: Lal George
167 :     Date: 2002/05/23 12:21:40 EDT
168 :     Tag: george-20020523-visual-labels
169 :     Description:
170 :    
171 :     Labels are now displayed in the graphical output to make
172 :     the fall-through and target blocks obvious.
173 :    
174 :     ----------------------------------------------------------------------
175 : blume 1204 Name: Matthias Blume
176 : blume 1220 Date: 2002/05/22 11:03:00 EDT
177 :     Tag: blume-20020522-shrink
178 :     Description:
179 :    
180 :     John tweaked yesterday's fix for 1131 to handle an out-of-memory
181 :     situation that comes up when allocating huge arrays.
182 :    
183 :     ----------------------------------------------------------------------
184 :     Name: Matthias Blume
185 : blume 1215 Date: 2002/05/21 16:00:00 EDT
186 :     Tag: Release_110_40
187 :     Description:
188 :    
189 :     New working release (110.40). New bootfiles.
190 :    
191 :     [Also: John Reppy fixed GC bug 1131.]
192 :    
193 :     ----------------------------------------------------------------------
194 :     Name: Matthias Blume
195 : blume 1212 Date: 2002/05/21 12:35:00 EDT
196 :     Tag: blume-20020521-cmdoc
197 :     Description:
198 :    
199 :     CM documentation update.
200 :    
201 :     ----------------------------------------------------------------------
202 :     Name: Matthias Blume
203 : blume 1211 Date: 2002/05/21 10:55:00 EDT
204 :     Tag: blume-20020521-misc
205 :     Description:
206 :    
207 :     - John tweaked runtime to be silent on heap export (except when
208 :     GC messages are on).
209 :     - I added a few more things (cross-compiling versions of CMB) to
210 :     config/preloads (as suggestions).
211 :    
212 :     ----------------------------------------------------------------------
213 :     Name: Matthias Blume
214 : blume 1208 Date: 2002/05/20 22:25:00 EDT
215 :     Tag: blume-20020520-controls
216 :     Description:
217 :    
218 :     - Added ControlUtil structure to control-lib.cm.
219 :     - Use it throughout.
220 :     - Used Controls facility to define MLRISC controls (as opposed to
221 :     registering MLRISC control ref cells with Controls after the fact)
222 :     - Fixed messed-up controls priorities.
223 :    
224 :     * Removed again all the stuff from config/preloads that one wouldn't
225 :     be able to preload at the time the initial heap image is built.
226 :     (Many libraries, e.g., CML, do not exist yet at this time. The
227 :     only libraries that can be preloaded via config/preloads are those
228 :     that come bundled with the bootfiles.)
229 :    
230 :     ----------------------------------------------------------------------
231 :     Name: Matthias Blume
232 : blume 1204 Date: 2002/05/20 10:59:00 EDT
233 :     Tag: blume-20020520-preloads
234 :     Description:
235 :    
236 :     Added a lot of commented-out suggestions for things to be included
237 :     in config/preloads.
238 :    
239 :     ----------------------------------------------------------------------
240 : leunga 1203 Name: Allen Leung
241 :     Date: 2002/05/18 14:20:00 EDT
242 :     Tag: leunga-20020518-mdl
243 :     Description:
244 :    
245 :     o Made the mdl tool stuff compile and run again.
246 :     o I've disabled all the stuff that depends on RTL specifications; they
247 :     are all badly broken anyway.
248 :    
249 :     ----------------------------------------------------------------------
250 : blume 1201 Name: Matthias Blume
251 :     Date: 2002/05/17 16:49:00 EDT
252 :     Tag: blume-20020517-controls
253 :     Description:
254 :    
255 :     0. John Reppy made several modifications to the SML/NJ library.
256 :     In particular, there is a shiny new controls-lib.cm.
257 :    
258 :     1. Pushed new controls interface through compiler so that everything
259 :     compiles again.
260 :    
261 :     2. Added FormatComb and FORMAT_COMB to the CML version of the
262 :     SML/NJ library (so that CML compiles again).
263 :    
264 :     3. Modified init scripts because XXX_DEFAULT environment variables
265 :     are no longer with us. (Boot-time initialization is now done
266 :     using the same environment variables that are also used for
267 :     startup-time initialization of controls.)
268 :    
269 :     ----------------------------------------------------------------------
270 : george 1192 Name: Lal George
271 :     Date: 2002/05/15 09:20:10 EDT
272 :     Tag: george-20020515-pseudo-op-decls
273 :     Description:
274 :    
275 :     All pseudo-ops emitted before the first segment declaration
276 :     such as TEXT, DATA, and BSS directives are assumed to be global
277 :     declarations and are emitted first in the assembly file. This is
278 :     useful in a number of situations where one has pseudo-ops that are not
279 :     specific to any segment, and also works around the constraint that one
280 :     cannot have client pseudo-ops in the TEXT segment.
281 :    
282 :     Because no segment is associated with these declarations it is
283 :     an error to allocate any space or objects before the first segment
284 :     directive and an exception will be raised. However, we cannot make
285 :     this check for client pseudo-ops.
286 :    
287 :     These top level declarations are a field in the CFG graph_info.
288 :     In theory you can continue to add to this field after the CFG has been
289 :     built -- provided you know what you are doing;-)
290 :    
291 :     ----------------------------------------------------------------------
292 : blume 1175 Name: Matthias Blume
293 : blume 1190 Date: 2002/05/13 16:40:00 EDT
294 :     Tag: blume-20020513-pp-etc
295 :     Description:
296 :    
297 :     A few minor bugfixes:
298 :    
299 :     - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
300 :     (Bogus printouts for redefined bindings still occur. Compiler
301 :     bug should no longer occur now. We need to redo the prettyprinter
302 :     from scratch.)
303 :    
304 :     - CM pathname printer now also adds escape sequences for ( and )
305 :    
306 :     - commend and docu fixes for ml-nlffi
307 :    
308 :     ----------------------------------------------------------------------
309 :     Name: Matthias Blume
310 : blume 1189 Date: 2002/05/10 16:40:00 EDT
311 :     Tag: blume-20020510-erg-textio
312 :     Description:
313 :    
314 :     Applied the following bugfix provided by Emden Gansner:
315 :    
316 :     Output is corrupted when outputSubstr is used rather than output.
317 :    
318 :     The problem occurs when a substring
319 :    
320 :     ss = (s, dataStart, dataLen)
321 :    
322 :     where dataStart > 0, fills a stream buffer with avail bytes left.
323 :     avail bytes of s, starting at index dataStart, are copied into the
324 :     buffer, the buffer is flushed, and then the remaining dataLen-avail
325 :     bytes of ss are copied into the beginning of the buffer. Instead of
326 :     starting this copy at index dataStart+avail in s, the current code
327 :     starts the copy at index avail.
328 :    
329 :     Fix:
330 :     In text-io-fn.sml, change line 695 from
331 :     val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
332 :     to
333 :     val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
334 :    
335 :     ----------------------------------------------------------------------
336 :     Name: Matthias Blume
337 : blume 1186 Date: 2002/04/12 13:55:00 EDT
338 :     Tag: blume-20020412-assyntax
339 :     Description:
340 :    
341 :     1. Grabbed newer assyntax.h from the XFree86 project.
342 :     2. Fiddled with how to compile X86.prim.asm without warnings.
343 :     3. (Very) Minor cleanup in CM.
344 :    
345 :     ----------------------------------------------------------------------
346 :     Name: Matthias Blume
347 : blume 1185 Date: 2002/04/01 (no joke!) 17:07:00 EST
348 :     Tag: blume-20020401-x86div
349 :     Description:
350 :    
351 :     Added full support for div/mod/rem/quot on the x86, using the machine
352 :     instruction's two results (without clumsily recomputing the remainder)
353 :     directly where appropriate.
354 :    
355 :     Some more extensive power-of-two support was added to the x86 instruction
356 :     selector (avoiding expensive divs, mods, and muls where they can be
357 :     replaced with cheaper shifts and masks). However, this sort of thing
358 :     ought to be done earlier, e.g., within the CPS optimizer so that
359 :     all architectures benefit from it.
360 :    
361 :     The compiler compiles to a fixed point, but changes might be somewhat
362 :     fragile nevertheless. Please, report any strange things that you might
363 :     see wrt. div/mod/quot/rem...
364 :    
365 :     ----------------------------------------------------------------------
366 :     Name: Matthias Blume
367 : blume 1184 Date: 2002/03/29 17:22:00
368 :     Tag: blume-20020329-div
369 :     Description:
370 :    
371 :     Fixed my broken div/mod logic. Unfortunately, this means that the
372 :     inline code for div/mod now has one more comparison than before.
373 :     Fast paths (quotient > 0 or remainder = 0) are not affected, though.
374 :     The problem was with quotient = 0, because that alone does not tell
375 :     us which way the rounding went. One then has to look at whether
376 :     remainder and divisor have the same sign... :(
377 :    
378 :     Anyway, I replaced the bootfiles with fresh ones...
379 :    
380 :     ----------------------------------------------------------------------
381 :     Name: Matthias Blume
382 : blume 1183 Date: 2002/03/29 14:10:00 EST
383 :     Tag: blume-20020329-inlprims
384 :     Description:
385 :    
386 :     NEW BOOTFILES!!! Version number bumped to 110.39.3.
387 :    
388 :     Primops have changed. This means that the bin/boot-file formats have
389 :     changed as well.
390 :    
391 :     To make sure that there is no confusion, I made a new version.
392 :    
393 :    
394 :     CHANGES:
395 :    
396 :     * removed REMT from mltree (remainder should never overflow).
397 :    
398 :     * added primops to deal with divisions of all flavors to the frontend
399 :    
400 :     * handled these primops all the way through so they map to their respective
401 :     MLRISC support
402 :    
403 :     * used these primops in the implementation of Int, Int32, Word, Word32
404 :    
405 :     * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
406 :    
407 :     * parameterized INLMIN, INLMAX, and INLABS by a numkind
408 :    
409 :     * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
410 :     floating point
411 :    
412 :     * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
413 :     and Real (but Real.abs maps to a separate floating-point-only primop)
414 :    
415 :    
416 :     TODO items:
417 :    
418 :     * Hacked Alpha32 instruction selection, disabling the selection of REMx
419 :     instructions because the machine instruction encoder cannot handle
420 :     them. (Hppa, PPC, and Sparc instruction selection did not handle
421 :     REM in the first place, and REM is supported by the x86 machine coder.)
422 :    
423 :     * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
424 :     selection phase. (The two can be streamlined because the hardware
425 :     delivers both quotient and remainder at the same time anyway.)
426 :    
427 :     * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
428 :     (Currently the behavior is inconsistent both across architectures and
429 :     wrt. the draft Basis spec.)
430 :    
431 :     * Word8 should eventually be handled natively, too.
432 :    
433 :     * There seems to be one serious bug in mltree-gen.sml. It appears, though,
434 :     as if there currently is no execution path that could trigger it in
435 :     SML/NJ. (The assumptions underlying functions arith and promotable do not
436 :     hold for things like multiplication and division.)
437 :    
438 :     ----------------------------------------------------------------------
439 :     Name: Matthias Blume
440 : blume 1181 Date: 2002/03/27 16:27:00 EST
441 :     Tag: blume-20020327-mlrisc-divisions
442 :     Description:
443 :    
444 :     Added support for all four division operations (ML's div, mod, quot,
445 :     and rem) to MLRISC. In the course of doing so, I also rationalized
446 :     the naming (no more annoying switch-around of DIV and QUOT), by
447 :     parameterizing the operation by div_rounding_mode (which can be either
448 :     DIV_TO_ZERO or DIV_TO_NEGINF).
449 :    
450 :     The generic MLTreeGen functor takes care of compiling all four
451 :     operations down to only round-to-zero div.
452 :    
453 :     Missing pieces:
454 :    
455 :     * Doing something smarter than relying on MLTreeGen on architectures
456 :     like, e.g., the x86 where hardware division delivers both quotient and
457 :     remainder at the same time. With this, the implementation of the
458 :     round-to-neginf operations could be further streamlined.
459 :    
460 :     * Remove inlining support for div/mod/rem from the frontend and replace it
461 :     with primops that get carried through to the backend. Do this for all
462 :     int and word types.
463 :    
464 :     ----------------------------------------------------------------------
465 :     Name: Matthias Blume
466 : blume 1180 Date: 2002/03/25 17:25:00 EST
467 :     Tag: blume-20020325-divmod
468 :     Description:
469 :    
470 :     I improved (hopefully without breaking them) the implementation of Int.div,
471 :     Int.mod, and Int.rem. For this, the code in translate.sml now takes
472 :     advantage of the following observations:
473 :    
474 :     Let q = x quot y r = x rem y
475 :     d = x div y m = x mod y
476 :    
477 :     where "quot" is the round-to-zero version of integer division that
478 :     hardware usually provides. Then we have:
479 :    
480 :     r = x - q * y where neither the * nor the - will overflow
481 :     d = if q >= 0 orelse x = q * y then q else q - 1
482 :     where neither the * nor the - will overflow
483 :     m = if q >= 0 orelse r = 0 then r else r + y
484 :     where the + will not overflow
485 :    
486 :     This results in substantial simplification of the generated code.
487 :     The following table shows the number of CFG nodes and edges generated
488 :     for
489 :     fun f (x, y) = x OPER y
490 :     (* with OPER \in div, mod, quot, rem *)
491 :    
492 :    
493 :     OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
494 :     --------------------------------------------------------
495 :     div | 24 | 39 | 12 | 16
496 :     mod | 41 | 71 | 12 | 16
497 :     quot | 8 | 10 | 8 | 10
498 :     rem | 10 | 14 | 8 | 10
499 :    
500 :    
501 :     ----------------------------------------------------------------------
502 :     Name: Matthias Blume
503 : blume 1179 Date: 2002/03/25 22:06:00 EST
504 :     Tag: blume-20020325-cprotobug
505 :     Description:
506 :    
507 :     Fixed a bug in cproto (c prototype decoder).
508 :    
509 :     ----------------------------------------------------------------------
510 :     Name: Matthias Blume
511 : blume 1178 Date: 2002/03/25 16:00:00 EST
512 :     Tag: blume-20020325-raw-primops
513 :     Description:
514 :    
515 :     I did some cleanup to Allen's new primop code and
516 :     replaced yesterday's bootfiles with new ones.
517 :     (But they are stored in the same place.)
518 :    
519 :     ----------------------------------------------------------------------
520 :     Name: Matthias Blume
521 : blume 1175 Date: 2002/03/24 22:40:00 EST
522 :     Tag: blume-20020324-bootfiles
523 :     Description:
524 :    
525 :     Made the bootfiles that Allen asked for.
526 :    
527 :     ----------------------------------------------------------------------
528 : leunga 1174 Name: Allen Leung
529 :     Date: 2002/03/23 15:50:00 EST
530 :     Tag: leunga-20020323-flint-cps-rcc-primops
531 :     Description:
532 :    
533 :     1. Changes to FLINT primops:
534 :    
535 :     (* make a call to a C-function;
536 :     * The primop carries C function prototype information and specifies
537 :     * which of its (ML-) arguments are floating point. C prototype
538 :     * information is for use by the backend, ML information is for
539 :     * use by the CPS converter. *)
540 :     | RAW_CCALL of { c_proto: CTypes.c_proto,
541 :     ml_args: ccall_type list,
542 :     ml_res_opt: ccall_type option,
543 :     reentrant : bool
544 :     } option
545 :     (* Allocate uninitialized storage on the heap.
546 :     * The record is meant to hold short-lived C objects, i.e., they
547 :     * are not ML pointers. With the tag, the representation is
548 :     * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
549 :     *)
550 :     | RAW_RECORD of {tag:bool,sz:int}
551 :     and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
552 :    
553 :     2. These CPS primops are now overloaded:
554 :    
555 :     rawload of {kind:numkind}
556 :     rawstore of {kind:numkind}
557 :    
558 :     The one argument form is:
559 :    
560 :     rawload {kind} address
561 :    
562 :     The two argument form is:
563 :    
564 :     rawload {kind} [ml object, byte-offset]
565 :    
566 :     3. RAW_CCALL/RCC now takes two extra arguments:
567 :    
568 :     a. The first is whether the C call is reentrant, i.e., whether
569 :     ML state should be saved and restored.
570 :     b. The second argument is a string argument specifying the name of
571 :     library and the C function.
572 :    
573 :     These things are currently not handled in the code generator, yet.
574 :    
575 :     4. In CProto,
576 :    
577 :     An encoding type of "bool" means "ml object" and is mapped into
578 :     C prototype of PTR. Note that "bool" is different than "string",
579 :     even though "string" is also mapped into PTR, because "bool"
580 :     is assigned an CPS type of BOGt, while "string" is assigned INT32t.
581 :    
582 :     5. Pickler/unpicker
583 :    
584 :     Changed to handle RAW_RECORD and newest RAW_CCALL
585 :    
586 :     6. MLRiscGen,
587 :    
588 :     1. Changed to handle the new rawload/rawstore/rawrecord operators.
589 :     2. Code for handling C Calls has been moved to a new module CPSCCalls,
590 :     in the file CodeGen/cpscompile/cps-c-calls.sml
591 :    
592 :     7. Added the conditional move operator
593 :    
594 :     condmove of branch
595 :    
596 :     to cps. Generation of this is still buggy so it is currently
597 :     disabled.
598 :    
599 :     ----------------------------------------------------------------------
600 : george 1168 Name: Lal George
601 :     Date: 2002/03/22 14:18:25 EST
602 : george 1171 Tag: george-20020322-cps-branch-prob
603 : george 1168 Description:
604 :    
605 : george 1171 Implemented the Ball-Larus branch prediction-heuristics, and
606 : george 1168 incorporated graphical viewers for control flow graphs.
607 :    
608 : george 1171 Ball-Larus Heuristics:
609 : george 1168 ---------------------
610 :     See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
611 :    
612 :     By design it uses the Dempster-Shafer theory for combining
613 :     probabilities. For example, in the function:
614 :    
615 :     fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
616 :    
617 :     the ball-larus heuristics predicts that the n=0 is unlikely
618 :     (OH-heuristic), and the 'then' branch is unlikely because of the
619 :     RH-heuristic -- giving the 'then' branch an even lower combined
620 : george 1171 probability using the Dempster-Shafer theory.
621 : george 1168
622 :     Finally, John Reppy's loop analysis in MLRISC, further lowers the
623 :     probability of the 'then' branch because of the loop in the else
624 :     branch.
625 :    
626 :    
627 :     Graphical Viewing:
628 :     ------------------
629 :     I merely plugged in Allen's graphical viewers into the compiler. The
630 :     additional code is not much. At the top level, saying:
631 :    
632 :     Control.MLRISC.getFlag "cfg-graphical-view" := true;
633 :    
634 :     will display the graphical view of the control flow graph just before
635 :     back-patching. daVinci must be in your path for this to work. If
636 :     daVinci is not available, then the default viewer can be changed
637 :     using:
638 :    
639 :     Control.MLRISC.getString "viewer"
640 :    
641 :     which can be set to "dot" or "vcg" for the corresponding viewers. Of
642 :     course, these viewers must be in your path.
643 :    
644 :     The above will display the compilation unit at the level of clusters,
645 :     many of which are small, boring, and un-interesting. Also setting:
646 :    
647 :     Control.MLRISC.getInt "cfg-graphical-view_size"
648 :    
649 :     will display clusters that are larger than the value set by the above.
650 :    
651 :    
652 :     ----------------------------------------------------------------------
653 : blume 1159 Name: Matthias Blume
654 :     Date: 2002/03/21 22:20:00 EST
655 :     Tag: blume-20020321-kmp-bugfix
656 :     Description:
657 :    
658 :     Changed the interface to the KMP routine in PreString and fixed
659 :     a minor bug in one place where it was used.
660 :    
661 :     ----------------------------------------------------------------------
662 : leunga 1156 Name: Allen Leung
663 : leunga 1158 Date: 2002/03/21 20:30:00 EST
664 :     Tag: leunga-20020321-cfg
665 :     Description:
666 :    
667 :     Fixed a potential problem in cfg edge splitting.
668 :    
669 :     ----------------------------------------------------------------------
670 :     Name: Allen Leung
671 : leunga 1156 Date: 2002/03/21 17:15:00 EST
672 :     Tag: leunga-20020321-x86-fp-cfg
673 :     Description:
674 :    
675 :     1. Recoded the buggy parts of x86-fp.
676 :    
677 :     a. All the block reordering code has been removed.
678 :     We now depend on the block placement phases to do this work.
679 :    
680 :     b. Critical edge splitting code has been simplified and moved into the
681 :     CFG modules, as where they belong.
682 :    
683 :     Both of these were quite buggy and complex. The code is now much, much
684 :     simpler.
685 :    
686 :     2. X86 backend.
687 :    
688 :     a. Added instructions for 64-bit support. Instruction selection for
689 :     64-bit has not been committed, however, since that
690 :     requires changes to MLTREE which haven't been approved by
691 :     Lal and John.
692 :    
693 :     b. Added support for FUCOMI and FUCOMIP when generating code for
694 :     PentiumPro and above. We only generate these instructions in
695 :     the fast-fp mode.
696 :    
697 :     c. Added cases for JP and JNP in X86FreqProps.
698 :    
699 :     3. CFG
700 :    
701 :     CFG now has a bunch of methods for edge splitting and merging.
702 :    
703 :     4. Machine description.
704 :    
705 :     John's simplification of MLTREE_BASIS.fcond broke a few machine
706 :     description things:
707 :    
708 :     rtl-build.{sig,sml} and hppa.mdl fixed.
709 :    
710 :     NOTE: the machine description stuff in the repository is still broken.
711 :     Again, I can't put my fixes in because that involves
712 :     changes to MLTREE.
713 :    
714 :     ----------------------------------------------------------------------
715 : blume 1145 Name: Matthias Blume
716 : blume 1155 Date: 2002/03/20 15:55:00 EST
717 :     Tag: blume-20020320-kmp
718 :     Description:
719 :    
720 :     Implemented Knuth-Morris-Pratt string matching in PreString and used
721 :     it for String.isSubstring, Substring.isSubstring, and
722 :     Substring.position.
723 :    
724 :     (Might need some stress-testing. Simple examples worked fine.)
725 :    
726 :     ----------------------------------------------------------------------
727 :     Name: Matthias Blume
728 : blume 1152 Date: 2002/03/19 16:37:00 EST
729 :     Tag: blume-20020319-witnesses
730 :     Description:
731 :    
732 :     Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
733 :    
734 :     This implements a generic mechanism for changing constness qualifiers
735 :     anywhere within big C types without resorting to outright "casts".
736 :     (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
737 :     modify the constness at the outermost level.)
738 :     The implementation of "convert" is based on the idea of "witness"
739 :     values -- values that are not used by the operation but whose types
740 :     "testify" to their applicability. On the implementation side, "convert"
741 :     is simply a projection (returning its second curried argument). With
742 :     cross-module inlining, it should not result in any machine code being
743 :     generated.
744 :    
745 :     ----------------------------------------------------------------------
746 :     Name: Matthias Blume
747 : blume 1148 Date: 2002/03/15 16:40:00 EST
748 :     Tag: blume-20020315-basis
749 :     Description:
750 :    
751 :     Provided (preliminary?) implementations for
752 :    
753 :     {String,Substring}.{concatWith,isSuffix,isSubstring}
754 :    
755 :     and
756 :    
757 :     Substring.full
758 :    
759 :     Those are in the Basis spec but they were missing in SML/NJ.
760 :    
761 :     ----------------------------------------------------------------------
762 :     Name: Matthias Blume
763 : blume 1145 Date: 2002/03/14 21:30:00 EST
764 :     Tag: blume-20020314-controls
765 :     Description:
766 :    
767 :     Controls:
768 :     ---------
769 :    
770 :     1. Factored out the recently-added Controls : CONTROLS stuff and put
771 :     it into its own library $/controls-lib.cm. The source tree for
772 :     this is under src/smlnj-lib/Controls.
773 :    
774 :     2. Changed the names of types and functions in this interface, so they
775 :     make a bit more "sense":
776 :    
777 :     module -> registry
778 :     'a registry -> 'a group
779 :    
780 :     3. The interface now deals in ref cells only. The getter/setter interface
781 :     is (mostly) gone.
782 :    
783 :     4. Added a function that lets one register an already-existing ref cell.
784 :    
785 :     5. Made the corresponding modifications to the rest of the code so that
786 :     everything compiles again.
787 :    
788 :     6. Changed the implementation of Controls.MLRISC back to something closer
789 :     to the original. In particular, this module (and therefore MLRISC)
790 :     does not depend on Controls. There now is some link-time code in
791 :     int-sys.sml that registers the MLRISC controls with the Controls
792 :     module.
793 :    
794 :     CM:
795 :     ---
796 :    
797 :     * One can now specify the lambda-split aggressiveness in init.cmi.
798 :    
799 :     ----------------------------------------------------------------------
800 : leunga 1142 Name: Allen Leung
801 :     Date: 2002/03/13 17:30:00 EST
802 :     Tag: leunga-20020313-x86-fp-unary
803 :     Description:
804 :    
805 :     Bug fix for:
806 :    
807 :     > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
808 :     > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
809 :     > - fun f(x,(y,z)) = Real.~ y;
810 :     > [autoloading]
811 :     > [autoloading done]
812 :     > fchsl (%eax), 184(%esp)
813 :     > Error: MLRisc bug: X86MCEmitter.emitInstr
814 :     >
815 :     > uncaught exception Error
816 :     > raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
817 :    
818 :     The problem was that the code generator did not generate any fp registers
819 :     in this case, and the ra didn't know that it needed to run the X86FP phase to
820 :     translate the pseudo fp instruction. This only happened with unary fp
821 :     operators in certain situations.
822 :    
823 :     ----------------------------------------------------------------------
824 : blume 1137 Name: Matthias Blume
825 : blume 1140 Date: 2002/03/13 14:00:00 EST
826 :     Tag: blume-20020313-overload-etc
827 :     Description:
828 :    
829 :     1. Added _overload as a synonym for overload for backward compatibility.
830 :     (Control.overloadKW must be true for either version to be accepted.)
831 :    
832 :     2. Fixed bug in install script that caused more things to be installed
833 :     than what was requested in config/targets.
834 :    
835 :     3. Made CM aware of the (_)overload construct so that autoloading
836 :     works.
837 :    
838 :     ----------------------------------------------------------------------
839 :     Name: Matthias Blume
840 : blume 1138 Date: 2002/03/12 22:03:00 EST
841 :     Tag: blume-20020312-url
842 :     Description:
843 :    
844 :     Forgot to update BOOT and srcarchiveurl.
845 :    
846 :     ----------------------------------------------------------------------
847 :     Name: Matthias Blume
848 : blume 1137 Date: 2002/03/12 17:30:00 EST
849 :     Tag: blume-20020312-version110392
850 :     Description:
851 :    
852 :     Yet another version number bump (because of small changes to the
853 :     binfile format). Version number is now 110.39.2. NEW BOOTFILES!
854 :    
855 :     Changes:
856 :    
857 :     The new pid generation scheme described a few weeks ago was overly
858 :     complicated. I implemented a new mechanism that is simpler and
859 :     provides a bit more "stability": Once CM has seen a compilation
860 :     unit, it keeps its identity constant (as long as you do not delete
861 :     those crucial CM/GUID/* files). This means that when you change
862 : blume 1145 an interface, compile, then go back to the old interface, and
863 : blume 1137 compile again, you arrive at the original pid.
864 :    
865 :     There now also is a mechanism that instructs CM to use the plain
866 :     environment hash as a module's pid (effectively making its GUID
867 :     the empty string). For this, "noguid" must be specified as an
868 :     option to the .sml file in question within its .cm file.
869 :     This is most useful for code that is being generated by tools such
870 :     as ml-nlffigen (because during development programmers tend to
871 :     erase the tool's entire output directory tree including CM's cached
872 :     GUIDs). "noguid" is somewhat dangerous (since it can be used to locally
873 :     revert to the old, broken behavior of SML/NJ, but in specific cases
874 :     where there is no danger of interface confusion, its use is ok
875 :     (I think).
876 :    
877 :     ml-nlffigen by default generates "noguid" annotations. They can be
878 :     turned off by specifying -guid in its command line.
879 :    
880 :     ----------------------------------------------------------------------
881 : george 1136 Name: Lal George
882 :     Date: 2002/03/12 12 14:42:36 EST
883 :     Tag: george-20020312-frequency-computation
884 :     Description:
885 :    
886 :     Integrated jump chaining and static block frequency into the
887 :     compiler. More details and numbers later.
888 : leunga 1025
889 : george 1003 ----------------------------------------------------------------------
890 : george 1133 Name: Lal George
891 :     Date: 2002/03/11 11 22:38:53 EST
892 :     Tag: george-20020311-jump-chain-elim
893 :     Description:
894 :    
895 :     Tested the jump chain elimination on all architectures (except the
896 :     hppa). This is on by default right now and is profitable for the
897 :     alpha and x86, however, it may not be profitable for the sparc and ppc
898 :     when compiling the compiler.
899 :    
900 :     The gc test will typically jump to a label at the end of the cluster,
901 :     where there is another jump to an external cluster containing the actual
902 :     code to invoke gc. This is to allow factoring of common gc invocation
903 :     sequences. That is to say, we generate:
904 :    
905 :     f:
906 :     testgc
907 :     ja L1 % jump if above to L1
908 :    
909 :     L1:
910 :     jmp L2
911 :    
912 :    
913 :     After jump chain elimination the 'ja L1' instructions is converted to
914 :     'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
915 :     up being implemented in their long form (if L2 is far away) using:
916 :    
917 :     jbe L3 % jump if below or equal to L3
918 :     jmp L2
919 :     L3:
920 :     ...
921 :    
922 :    
923 :     For large compilation units L2 may be far away.
924 :    
925 :    
926 :     ----------------------------------------------------------------------
927 : blume 1132 Name: Matthias Blume
928 :     Date: 2002/03/11 13:30:00 EST
929 :     Tag: blume-20020311-mltreeeval
930 :     Description:
931 :    
932 :     A functor parameter was missing.
933 :    
934 :     ----------------------------------------------------------------------
935 : leunga 1130 Name: Allen Leung
936 : leunga 1131 Date: 2002/03/11 10:30:00 EST
937 : leunga 1142 Tag: leunga-20020311-runtime-string0
938 : leunga 1131 Description:
939 :    
940 :     The representation of the empty string now points to a
941 :     legal null terminated C string instead of unit. It is now possible
942 :     to convert an ML string into C string with InlineT.CharVector.getData.
943 :     This compiles into one single machine instruction.
944 :    
945 :     ----------------------------------------------------------------------
946 :     Name: Allen Leung
947 : leunga 1130 Date: 2002/03/10 23:55:00 EST
948 :     Tag: leunga-20020310-x86-call
949 :     Description:
950 :    
951 :     Added machine generation for CALL instruction (relative displacement mode)
952 :    
953 :     ----------------------------------------------------------------------
954 : blume 1128 Name: Matthias Blume
955 :     Date: 2002/03/08 16:05:00
956 :     Tag: blume-20020308-entrypoints
957 :     Description:
958 :    
959 :     Version number bumped to 110.39.1. NEW BOOTFILES!
960 :    
961 :     Entrypoints: non-zero offset into a code object where execution should begin.
962 :    
963 :     - Added the notion of an entrypoint to CodeObj.
964 :     - Added reading/writing of entrypoint info to Binfile.
965 :     - Made runtime system bootloader aware of entrypoints.
966 :     - Use the address of the label of the first function given to mlriscGen
967 :     as the entrypoint. This address is currently always 0, but it will
968 :     not be 0 once we turn on block placement.
969 :     - Removed the linkage cluster code (which was The Other Way(tm) of dealing
970 :     with entry points) from mlriscGen.
971 :    
972 :     ----------------------------------------------------------------------
973 : leunga 1127 Name: Allen Leung
974 :     Date: 2002/03/07 20:45:00 EST
975 :     Tag: leunga-20020307-x86-cmov
976 :     Description:
977 :    
978 :     Bug fixes for CMOVcc on x86.
979 :    
980 :     1. Added machine code generation for CMOVcc
981 :     2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
982 :     3. CMOVcc cannot have an immediate operand as argument.
983 :    
984 :     ----------------------------------------------------------------------
985 : blume 1126 Name: Matthias Blume
986 :     Date: 2002/03/07 16:15:00 EST
987 :     Tag: blume-20020307-controls
988 :     Description:
989 :    
990 :     This is a very large but mostly boring patch which makes (almost)
991 :     every tuneable compiler knob (i.e., pretty much everything under
992 :     Control.* plus a few other things) configurable via both the command
993 :     line and environment variables in the style CM did its configuration
994 :     until now.
995 :    
996 :     Try starting sml with '-h' (or, if you are brave, '-H')
997 :    
998 :     To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
999 :     implements the underlying generic mechanism.
1000 :    
1001 :     The interface to some of the existing such facilities has changed somewhat.
1002 :     For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1003 :     (The getFoo interface is still there for backward-compatibility, but its
1004 :     use is deprecated.)
1005 :    
1006 :     The ml-build script passes -Cxxx=yyy command-line arguments through so
1007 :     that one can now twiddle the compiler settings when using this "batch"
1008 :     compiler.
1009 :    
1010 :     TODO items:
1011 :    
1012 :     We should go through and throw out all controls that are no longer
1013 :     connected to anything. Moreover, we should go through and provide
1014 :     meaningful (and correct!) documentation strings for those controls
1015 :     that still are connected.
1016 :    
1017 :     Currently, multiple calls to Controls.new are accepted (only the first
1018 :     has any effect). Eventually we should make sure that every control
1019 :     is being made (via Controls.new) exactly once. Future access can then
1020 :     be done using Controls.acc.
1021 :    
1022 :     Finally, it would probably be a good idea to use the getter-setter
1023 :     interface to controls rather than ref cells. For the time being, both
1024 :     styles are provided by the Controls module, but getter-setter pairs are
1025 :     better if thread-safety is of any concern because they can be wrapped.
1026 :    
1027 :     *****************************************
1028 :    
1029 :     One bug fix: The function blockPlacement in three of the MLRISC
1030 :     backpatch files used to be hard-wired to one of two possibilities at
1031 :     link time (according to the value of the placementFlag). But (I
1032 :     think) it should rather sense the flag every time.
1033 :    
1034 :     *****************************************
1035 :    
1036 :     Other assorted changes (by other people who did not supply a HISTORY entry):
1037 :    
1038 :     1. the cross-module inliner now works much better (Monnier)
1039 :     2. representation of weights, frequencies, and probabilities in MLRISC
1040 :     changed in preparation of using those for weighted block placement
1041 :     (Reppy, George)
1042 :    
1043 :     ----------------------------------------------------------------------
1044 : george 1116 Name: Lal George
1045 : george 1124 Date: 2002/03/07 14:44:24 EST 2002
1046 :     Tag: george-20020307-weighted-block-placement
1047 :    
1048 :     Tested the weighted block placement optimization on all architectures
1049 :     (except the hppa) using AMPL to generate the block and edge frequencies.
1050 :     Changes were required in the machine properties to correctly
1051 :     categorize trap instructions. There is an MLRISC flag
1052 :     "weighted-block-placement" that can be used to enable weighted block
1053 :     placement, but this will be ineffective without block/edge
1054 :     frequencies (coming soon).
1055 :    
1056 :    
1057 :     ----------------------------------------------------------------------
1058 :     Name: Lal George
1059 : george 1116 Date: 2002/03/05 17:24:48 EST
1060 :     Tag: george-20020305-linkage-cluster
1061 :    
1062 : george 1124 In order to support the block placement optimization, a new cluster
1063 :     is generated as the very first cluster (called the linkage cluster).
1064 :     It contains a single jump to the 'real' entry point for the compilation
1065 :     unit. Block placement has no effect on the linkage cluster itself, but
1066 :     all the other clusters have full freedom in the manner in which they
1067 :     reorder blocks or functions.
1068 : george 1116
1069 :     On the x86 the typical linkage code that is generated is:
1070 :     ----------------------
1071 :     .align 2
1072 :     L0:
1073 :     addl $L1-L0, 72(%esp)
1074 : george 1124 jmp L1
1075 : george 1116
1076 :    
1077 :     .align 2
1078 :     L1:
1079 :     ----------------------
1080 :    
1081 :     72(%esp) is the memory location for the stdlink register. This
1082 :     must contain the address of the CPS function being called. In the
1083 : george 1124 above example, it contains the address of L0; before
1084 : george 1116 calling L1 (the real entry point for the compilation unit), it
1085 :     must contain the address for L1, and hence
1086 :    
1087 :     addl $L1-L0, 72(%esp)
1088 :    
1089 : george 1124 I have tested this on all architectures except the hppa.The increase
1090 :     in code size is of course negligible
1091 : george 1116
1092 :     ----------------------------------------------------------------------
1093 : leunga 1115 Name: Allen Leung
1094 :     Date: 2002/03/03 13:20:00 EST
1095 :     Tag: leunga-20020303-mlrisc-tools
1096 :    
1097 :     Added #[ ... ] expressions to mlrisc tools
1098 :    
1099 :     ----------------------------------------------------------------------
1100 : blume 1096 Name: Matthias Blume
1101 : blume 1098 Date: 2002/02/27 12:29:00 EST
1102 :     Tag: blume-20020227-cdebug
1103 :     Description:
1104 :    
1105 :     - made types in structure C and C_Debug to be equal
1106 :     - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1107 :     - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1108 :    
1109 :     ----------------------------------------------------------------------
1110 :     Name: Matthias Blume
1111 : blume 1096 Date: 2002/02/26 12:00:00 EST
1112 :     Tag: blume-20020226-ffi
1113 :     Description:
1114 :    
1115 :     1. Fixed a minor bug in CM's "noweb" tool:
1116 :     If numbering is turned off, then truly don't number (i.e., do not
1117 :     supply the -L option to noweb). The previous behavior was to supply
1118 :     -L'' -- which caused noweb to use the "default" line numbering scheme.
1119 :     Thanks to Chris Richards for pointing this out (and supplying the fix).
1120 :    
1121 :     2. Once again, I reworked some aspects of the FFI:
1122 :    
1123 :     A. The incomplete/complete type business:
1124 :    
1125 :     - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1126 :     gone!
1127 :     - ML types representing an incomplete type are now *equal* to
1128 :     ML types representing their corresponding complete types (just like
1129 :     in C). This is still safe because ml-nlffigen will not generate
1130 :     RTTI for incomplete types, nor will it generate functions that
1131 :     require access to such RTTI. But when ML code generated from both
1132 :     incomplete and complete versions of the C type meet, the ML types
1133 :     are trivially interoperable.
1134 :    
1135 :     NOTE: These changes restore the full generality of the translation
1136 :     (which was previously lost when I eliminated functorization)!
1137 :    
1138 :     B. Enum types:
1139 :    
1140 :     - Structure C now has a type constructor "enum" that is similar to
1141 :     how the "su" constructor works. However, "enum" is not a phantom
1142 :     type because each "T enum" has values (and is isomorphic to
1143 :     MLRep.Signed.int).
1144 :     - There are generic access operations for enum objects (using
1145 :     MLRep.Signed.int).
1146 :     - ml-nlffigen will generate a structure E_foo for each "enum foo".
1147 :     * The structure contains the definition of type "mlrep" (the ML-side
1148 :     representation type of the enum). Normally, mlrep is the same
1149 :     as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1150 :     then mlrep will be defined as a datatype -- thus facilitating
1151 :     pattern matching on mlrep values.
1152 :     ("-ec" will be suppressed if there are duplicate values in an
1153 :     enumeration.)
1154 :     * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1155 :     will be generated for each C enum constant xxx.
1156 :     * Conversion functions m2i and i2m convert between mlrep and
1157 :     MLRep.Signed.int. (Without "-ec", these functions are identities.)
1158 :     * Coversion functions c and ml convert between mlrep and "tag enum".
1159 :     * Access functions (get/set) fetch and store mlrep values.
1160 :     - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1161 :     enumerations are merged into one single enumeration represented by
1162 :     structure E_'.
1163 :    
1164 :     ----------------------------------------------------------------------
1165 : leunga 1080 Name: Allen Leung
1166 : leunga 1094 Date: 2002/02/25 04:45:00 EST
1167 :     Tag: leunga-20020225-cps-spill
1168 :    
1169 :     This is a new implementation of the CPS spill phase.
1170 :     The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1171 :     In case of problems, replace it with the old file spill.sml
1172 :    
1173 :     The current compiler runs into some serious performance problems when
1174 :     constructing a large record. This can happen when we try to compile a
1175 :     structure with many items. Even a very simple structure like the following
1176 :     makes the compiler slow down.
1177 :    
1178 :     structure Foo = struct
1179 :     val x_1 = 0w1 : Word32.int
1180 :     val x_2 = 0w2 : Word32.int
1181 :     val x_3 = 0w3 : Word32.int
1182 :     ...
1183 :     val x_N = 0wN : Word32.int
1184 :     end
1185 :    
1186 :     The following table shows the compile time, from N=1000 to N=4000,
1187 :     with the old compiler:
1188 :    
1189 :     N
1190 :     1000 CPS 100 spill 0.04u 0.00s 0.00g
1191 :     MLRISC ra 0.06u 0.00s 0.05g
1192 :     (spills = 0 reloads = 0)
1193 :     TOTAL 0.63u 0.07s 0.21g
1194 :    
1195 :     1100 CPS 100 spill 8.25u 0.32s 0.64g
1196 :     MLRISC ra 5.68u 0.59s 3.93g
1197 :     (spills = 0 reloads = 0)
1198 :     TOTAL 14.71u 0.99s 4.81g
1199 :    
1200 :     1500 CPS 100 spill 58.55u 2.34s 1.74g
1201 :     MLRISC ra 5.54u 0.65s 3.91g
1202 :     (spills = 543 reloads = 1082)
1203 :     TOTAL 65.40u 3.13s 6.00g
1204 :    
1205 :     2000 CPS 100 spill 126.69u 4.84s 3.08g
1206 :     MLRISC ra 0.80u 0.10s 0.55g
1207 :     (spills = 42 reloads = 84)
1208 :     TOTAL 129.42u 5.10s 4.13g
1209 :    
1210 :     3000 CPS 100 spill 675.59u 19.03s 11.64g
1211 :     MLRISC ra 2.69u 0.27s 1.38g
1212 :     (spills = 62 reloads = 124)
1213 :     TOTAL 682.48u 19.61s 13.99g
1214 :    
1215 :     4000 CPS 100 spill 2362.82u 56.28s 43.60g
1216 :     MLRISC ra 4.96u 0.27s 2.72g
1217 :     (spills = 85 reloads = 170)
1218 :     TOTAL 2375.26u 57.21s 48.00g
1219 :    
1220 :     As you can see the old cps spill module suffers from some serious
1221 : leunga 1115 performance problem. But since I cannot decipher the old code fully,
1222 :     instead of patching the problems up, I'm reimplementing it
1223 : leunga 1094 with a different algorithm. The new code is more modular,
1224 :     smaller when compiled, and substantially faster
1225 :     (O(n log n) time and O(n) space). Timing of the new spill module:
1226 :    
1227 :     4000 CPS 100 spill 0.02u 0.00s 0.00g
1228 :     MLRISC ra 0.25u 0.02s 0.15g
1229 :     (spills=1 reloads=3)
1230 :     TOTAL 7.74u 0.34s 1.62g
1231 :    
1232 :     Implementation details:
1233 :    
1234 :     As far as I can tell, the purpose of the CPS spill module is to make sure the
1235 :     number of live variables at any program point (the bandwidth)
1236 :     does not exceed a certain limit, which is determined by the
1237 :     size of the spill area.
1238 :    
1239 :     When the bandwidth is too large, we decrease the register pressure by
1240 :     packing live variables into spill records. How we achieve this is
1241 :     completely different than what we did in the old code.
1242 :    
1243 :     First, there is something about the MLRiscGen code generator
1244 :     that we should be aware of:
1245 :    
1246 :     o MLRiscGen performs code motion!
1247 :    
1248 :     In particular, it will move floating point computations and
1249 :     address computations involving only the heap pointer to
1250 :     their use sites (if there is only a single use).
1251 :     What this means is that if we have a CPS record construction
1252 :     statement
1253 :    
1254 :     RECORD(k,vl,w,e)
1255 :    
1256 :     we should never count the new record address w as live if w
1257 :     has only one use (which is often the case).
1258 :    
1259 :     We should do something similar to floating point, but the transformation
1260 :     there is much more complex, so I won't deal with that.
1261 :    
1262 :     Secondly, there are now two new cps primops at our disposal:
1263 :    
1264 :     1. rawrecord of record_kind option
1265 :     This pure operator allocates some uninitialized storage from the heap.
1266 :     There are two forms:
1267 :    
1268 :     rawrecord NONE [INT n] allocates a tagless record of length n
1269 :     rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1270 :     and initializes the tag.
1271 :    
1272 :     2. rawupdate of cty
1273 :     rawupdate cty (v,i,x)
1274 :     Assigns to x to the ith component of record v.
1275 :     The storelist is not updated.
1276 :    
1277 :     We use these new primops for both spilling and increment record construction.
1278 :    
1279 :     1. Spilling.
1280 :    
1281 :     This is implemented with a linear scan algorithm (but generalized
1282 :     to trees). The algorithm will create a single spill record at the
1283 :     beginning of the cps function and use rawupdate to spill to it,
1284 :     and SELECT or SELp to reload from it. So both spills and reloads
1285 :     are fine-grain operations. In contrast, in the old algorithm
1286 :     "spills" have to be bundled together in records.
1287 :    
1288 :     Ideally, we should sink the spill record construction to where
1289 :     it is needed. We can even split the spill record into multiple ones
1290 :     at the places where they are needed. But CPS is not a good
1291 :     representation for global code motion, so I'll keep it simple and
1292 :     am not attempting this.
1293 :    
1294 :     2. Incremental record construction (aka record splitting).
1295 :    
1296 :     Long records with many component values which are simulatenously live
1297 :     (recall that single use record addresses are not considered to
1298 :     be live) are constructed with rawrecord and rawupdate.
1299 :     We allocate space on the heap with rawrecord first, then gradually
1300 :     fill it in with rawupdate. This is the technique suggested to me
1301 :     by Matthias.
1302 :    
1303 :     Some restrictions on when this is applicable:
1304 :     1. It is not a VECTOR record. The code generator currently does not handle
1305 :     this case. VECTOR record uses double indirection like arrays.
1306 :     2. All the record component values are defined in the same "basic block"
1307 :     as the record constructor. This is to prevent speculative
1308 :     record construction.
1309 :    
1310 :     ----------------------------------------------------------------------
1311 :     Name: Allen Leung
1312 : leunga 1086 Date: 2002/02/22 01:02:00 EST
1313 :     Tag: leunga-20020222-mlrisc-tools
1314 :    
1315 :     Minor bug fixes in the parser and rewriter
1316 :    
1317 :     ----------------------------------------------------------------------
1318 :     Name: Allen Leung
1319 : leunga 1085 Date: 2002/02/21 20:20:00 EST
1320 :     Tag: leunga-20020221-peephole
1321 :    
1322 :     Regenerated the peephole files. Some contained typos in the specification
1323 :     and some didn't compile because of pretty printing bugs in the old version
1324 :     of 'nowhere'.
1325 :    
1326 :     ----------------------------------------------------------------------
1327 :     Name: Allen Leung
1328 : leunga 1079 Date: 2002/02/19 20:20:00 EST
1329 :     Tag: leunga-20020219-mlrisc-tools
1330 :     Description:
1331 :    
1332 :     Minor bug fixes to the mlrisc-tools library:
1333 :    
1334 :     1. Fixed up parsing colon suffixed keywords
1335 :     2. Added the ability to shut the error messages up
1336 :     3. Reimplemented the pretty printer and fixed up/improved
1337 :     the pretty printing of handle and -> types.
1338 :     4. Fixed up generation of literal symbols in the nowhere tool.
1339 :     5. Added some SML keywords to to sml.sty
1340 :    
1341 :     ----------------------------------------------------------------------
1342 :     Name: Matthias Blume
1343 : blume 1078 Date: 2002/02/19 16:20:00 EST
1344 :     Tag: blume-20020219-cmffi
1345 :     Description:
1346 :    
1347 :     A wild mix of changes, some minor, some major:
1348 :    
1349 :     * All C FFI-related libraries are now anchored under $c:
1350 :     $/c.cm --> $c/c.cm
1351 :     $/c-int.cm --> $c/internals/c-int.cm
1352 :     $/memory.cm --> $c/memory/memory.cm
1353 :    
1354 :     * "make" tool (in CM) now treats its argument pathname slightly
1355 :     differently:
1356 :     1. If the native expansion is an absolute name, then before invoking
1357 :     the "make" command on it, CM will apply OS.Path.mkRelative
1358 :     (with relativeTo = OS.FileSys.getDir()) to it.
1359 :     2. The argument will be passed through to subsequent phases of CM
1360 :     processing without "going native". In particular, if the argument
1361 :     was an anchored path, then "make" will not lose track of that anchor.
1362 :    
1363 :     * Compiler backends now "know" their respective C calling conventions
1364 :     instead of having to be told about it by ml-nlffigen. This relieves
1365 :     ml-nlffigen from one of its burdens.
1366 :    
1367 :     * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1368 :    
1369 :     * Export C_DEBUG and C_Debug from $c/c.cm.
1370 :    
1371 :     * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1372 :     subtyping relationship between incomplete pointers and their complete
1373 :     counterparts. For this, ('t, 'c) ptr has been changed to 'o ptr --
1374 :     with the convention of instantiating 'o with ('t, 'c) obj whenever
1375 :     the pointer target type is complete. In the incomplete case, 'o
1376 :     will be instantiated with some "'c iobj" -- a type obtained by
1377 :     using one of the functors PointerToIncompleteType or PointerToCompleteType.
1378 :    
1379 :     Operations that work on both incomplete and complete pointer types are
1380 :     typed as taking an 'o ptr while operations that require the target to
1381 :     be known are typed as taking some ('t, 'c) obj ptr.
1382 :    
1383 :     voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1384 :     where void is an eqtype without any values. This makes it possible
1385 :     to work on voidptr values using functions meant to operate on light
1386 :     incomplete pointers.
1387 :    
1388 :     * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1389 :     been vastly simplified.
1390 :    
1391 :     ----------------------------------------------------------------------
1392 :     Name: Matthias Blume
1393 : blume 1076 Date: 2002/02/19 10:48:00 EST
1394 :     Tag: blume-20020219-pqfix
1395 :     Description:
1396 :    
1397 :     Applied Chris Okasaki's bug fix for priority queues.
1398 :    
1399 :     ----------------------------------------------------------------------
1400 :     Name: Matthias Blume
1401 : blume 1073 Date: 2002/02/15 17:05:00
1402 : blume 1069 Tag: Release_110_39
1403 :     Description:
1404 :    
1405 : blume 1073 Last-minute retagging is becoming a tradition... :-(
1406 :    
1407 :     This is the working release 110.39.
1408 :    
1409 :     ----------------------------------------------------------------------
1410 :     Name: Matthias Blume
1411 :     Date: 2002/02/15 16:00:00 EST
1412 :     Tag: Release_110_39-orig
1413 :     Description:
1414 :    
1415 : blume 1069 Working release 110.39. New bootfiles.
1416 :    
1417 : blume 1073 (Update: There was a small bug in the installer so it wouldn't work
1418 :     with all shells. So I retagged. -Matthias)
1419 :    
1420 : blume 1069 ----------------------------------------------------------------------
1421 :     Name: Matthias Blume
1422 : blume 1068 Date: 2002/02/15 14:17:00 EST
1423 :     Tag: blume-20020215-showbindings
1424 :     Description:
1425 :    
1426 :     Added EnvRef.listBoundSymbols and CM.State.showBindings. Especially
1427 :     the latter can be useful for exploring what bindings are available at
1428 :     the interactive prompt. (The first function returns only the list
1429 :     of symbols that are really bound, the second prints those but also the
1430 :     ones that CM's autoloading mechanism knows about.)
1431 :    
1432 :     ----------------------------------------------------------------------
1433 :     Name: Matthias Blume
1434 : blume 1067 Date: 2002/02/15 12:08:00 EST
1435 :     Tag: blume-20020215-iptrs
1436 :     Description:
1437 :    
1438 :     Two improvements to ml-nlffigen:
1439 :    
1440 :     1. Write files only if they do not exist or if their current contents
1441 :     do not coincide with what's being written. (That is, avoid messing
1442 :     with the time stamps unless absolutely necessary.)
1443 :    
1444 :     2. Implement a "repository" mechanism for generated files related
1445 :     to "incomplete pointer types". See the README file for details.
1446 :    
1447 :     ----------------------------------------------------------------------
1448 :     Name: Matthias Blume
1449 : blume 1066 Date: 2002/02/14 11:50:00 EST
1450 :     Tag: blume-20020214-quote
1451 :     Description:
1452 :    
1453 :     Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm). This is required
1454 :     because of the new and improved tag generation scheme. (Thanks to Allen
1455 :     Leung for pointing it out.)
1456 :    
1457 :     ----------------------------------------------------------------------
1458 : george 1065 Name: Lal George
1459 :     Date: 2002/02/14 09:55:27 EST 2002
1460 :     Tag: george-20020214-isabelle-bug
1461 :     Description:
1462 :    
1463 :     Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1464 :     of Isabelle on the x86.
1465 :    
1466 :     From Allen:
1467 :     -----------
1468 :     I've found the problem:
1469 :    
1470 :     in ra-core.sml, I use the counter "blocked" to keep track of the
1471 :     true number of elements in the freeze queue. When the counter goes
1472 :     to zero, I skip examining the queue. But I've messed up the
1473 :     bookkeeping in combine():
1474 :    
1475 :     else ();
1476 :     case !ucol of
1477 :     PSEUDO => (if !cntv > 0 then
1478 :     (if !cntu > 0 then blocked := !blocked - 1 else ();
1479 :     ^^^^^^^^^^^^^^^^^^^^^^^
1480 :     moveu := mergeMoveList(!movev, !moveu)
1481 :     )
1482 :     else ();
1483 :    
1484 :     combine() is called to coalesce two nodes u and v.
1485 :     I think I was thinking that if the move counts of u and v are both
1486 :     greater than zero then after they are coalesced then one node is
1487 :     removed from the freeze queue. Apparently I was thinking that
1488 :     both u and v are of low degree, but that's clearly not necessarily true.
1489 :    
1490 :    
1491 :     02/12/2002:
1492 :     Here's the patch. HOL now compiles.
1493 :    
1494 :     I don't know how this impact on performance (compile
1495 :     time or runtime). This bug caused the RA (especially on the x86)
1496 :     to go thru the potential spill phase when there are still nodes on the
1497 :     freeze queue.
1498 :    
1499 :    
1500 :    
1501 :    
1502 :     ----------------------------------------------------------------------
1503 : blume 1058 Name: Matthias Blume
1504 : blume 1064 Date: 2002/02/13 22:40:00 EST
1505 :     Tag: blume-20020213-fptr-rtti
1506 :     Description:
1507 :    
1508 :     Fixed a bug in ml-nlffigen that was introduced with one of the previous
1509 :     updates.
1510 :    
1511 :     ----------------------------------------------------------------------
1512 :     Name: Matthias Blume
1513 : blume 1063 Date: 2002/02/13 16:41:00 EST
1514 :     Tag: blume-20020213-cmlpq
1515 :     Description:
1516 :    
1517 :     Added new priority queue export symbols (which have just been added to
1518 :     smlnj-lib.cm) to CML's version of smlnj-lib.cm. (Otherwise CML would
1519 :     not compile and the installer would choke.)
1520 :    
1521 :     ----------------------------------------------------------------------
1522 :     Name: Matthias Blume
1523 : blume 1062 Date: 2002/02/13 16:15:00 EST
1524 :     Tag: blume-20020213-various
1525 :     Description:
1526 :    
1527 :     1. More tweaks to ml-nlffigen:
1528 :    
1529 :     - better internal datastructures (resulting in slight speedup)
1530 :     - "-match" option requires exact match
1531 :     - "localized" gensym counters (untagged structs/unions nested within
1532 :     other structs/unions or within typedefs get a fresh counter; their
1533 :     tag will be prefixed by a concatenation of their parents' tags)
1534 :     - bug fixes (related to calculation of transitive closure of types
1535 :     to be included in the output)
1536 :    
1537 :     2. Minor Basis updates:
1538 :    
1539 :     - added implementations for List.collate and Option.app
1540 :    
1541 :     ----------------------------------------------------------------------
1542 :     Name: Matthias Blume
1543 : blume 1060 Date: 2002/02/11 15:55:00 EST
1544 :     Tag: blume-20020211-gensym
1545 :     Description:
1546 :    
1547 :     Added a "-gensym" option to command line of ml-nlffigen. This can be
1548 :     used to specify a "stem" -- a string that is inserted in all "gensym'd"
1549 :     names (ML structure names that correspond to unnamed C structs, unions,
1550 :     and enums), so that separate runs of ml-nlffigen do not clash.
1551 :    
1552 :     ----------------------------------------------------------------------
1553 :     Name: Matthias Blume
1554 : blume 1059 Date: 2002/02/11 12:05:00 EST
1555 :     Tag: blume-20020211-gensml
1556 :     Description:
1557 :    
1558 :     A quick fix for a problem with GenSML (in the pgraph-util library):
1559 :     Make generation of toplevel "local" optional. (Strictly speaking,
1560 :     signature definitions within "local" are not legal SML.)
1561 :    
1562 :     Other than that: updates to INSTALL and cm/TODO.
1563 :    
1564 :     ----------------------------------------------------------------------
1565 :     Name: Matthias Blume
1566 : blume 1058 Date: 2002/02/08 15:00:00 EST
1567 :     Tag: blume-20020208-uniquepid
1568 :     Description:
1569 :    
1570 :     0. Version number has been bumped to 110.38.1. NEW BOOTFILES!!!
1571 :    
1572 :     1. The installer (config/install.sh) has gotten smarter:
1573 :    
1574 :     - Configuration options are a bit easier to specify now
1575 :     (in config/targets).
1576 :     - Bug in recognizing .tar.bz2 files fixed.
1577 :     - Installer automatically resolves dependencies between
1578 :     configuration options (e.g., if you ask for eXene, you will
1579 :     also get cml -- regardless whether you asked for it or not).
1580 :     - Installer can run in "quieter mode" by setting the environment
1581 :     variable INSTALL_QUIETLY to "true". "Quieter" does not mean
1582 :     "completely silent", though.
1583 :     - Build HashCons library as part of smlnj-lib.
1584 :    
1585 :     2. A new scheme for assigning persistent identifiers to compilation
1586 :     units (and, by extension, to types etc.) has been put into place.
1587 :     This fixes a long-standing bug where types and even dynamic values
1588 :     can get internally confused, thereby compromising type safety
1589 :     (abstraction) and dynamic correctness. See
1590 :    
1591 :     http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1592 :    
1593 :     for an example of how things could go wrong until now.
1594 :    
1595 :     The downside of the new scheme is that pids are not quite as
1596 :     persistent as they used to be: CM will generate a fresh pid
1597 :     for every compilation unit that it thinks it sees for the first
1598 :     time. That means that if you compile starting from a clean, fresh
1599 :     source tree at two different times, you end up with different
1600 :     binaries.
1601 :    
1602 :     Cutoff recompilation, however, has not been compromised because
1603 :     CM keeps pid information in special caches between runs.
1604 :    
1605 :     ----------------------------------------------------------------------
1606 : george 1053 Name: Lal George
1607 : george 1055 Date: 2002/02/07 15:34:13 EST 2002
1608 :     Tag: <none>
1609 :     Description:
1610 :    
1611 :     Compilers that generate assembly code may produce global labels
1612 :     whose value is resolved at link time. The various peephole optimization
1613 :     modules did not take this in account.
1614 :    
1615 :     TODO. The Labels.addrOf function should really return an option
1616 :     type so that clients are forced to deal with this issue, rather
1617 :     than an exception being raised.
1618 :    
1619 :     ----------------------------------------------------------------------
1620 :     Name: Lal George
1621 : george 1053 Date: 2002/02/06 13:55:02 EST
1622 :     Tag: george-20020206-ra-breakup
1623 :     Description:
1624 :    
1625 :     1. A bug fix from Allen.
1626 :    
1627 :     A typo causes extra fstp %st(0)'s to be generated at compensation
1628 :     edges, which might cause stack underflow traps at runtime. This
1629 :     occurs in fft where there are extraneous fstps right before the 'into'
1630 :     trap instruction (in this case they are harmless since none of the
1631 :     integers overflow.)
1632 :    
1633 : george 1054 2. Pulled out various utility modules that were embedded in the modules
1634 : george 1053 of the register allocator. I need these modules for other purposes, but
1635 : george 1054 they are not complete enough to put into a library (just yet).
1636 : george 1053 ----------------------------------------------------------------------
1637 : blume 1049 Name: Matthias Blume
1638 :     Date: 2002/01/31 16:05:00 EST
1639 :     Tag: blume-20020131-sparc-ccalls
1640 :     Description:
1641 :    
1642 :     1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1643 :     of extra stack space by mistake. Fixed.
1644 :    
1645 :     2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1646 :    
1647 :     ----------------------------------------------------------------------
1648 : leunga 1048 Name: Allen Leung
1649 :     Date: 2002/01/30
1650 :     Tag: leunga-20020130-nowhere-bug-fix
1651 :     Description:
1652 :    
1653 :     MLRISC bug fixes:
1654 :     1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1655 :     2. MachineInt.fromString was negating its value.
1656 :    
1657 :     ----------------------------------------------------------------------
1658 : blume 1039 Name: Matthias Blume
1659 : blume 1047 Date: 2002/01/29
1660 :     Tag: blume-20020129-INSTALL
1661 :     Description:
1662 :    
1663 :     - Added somewhat detailed installation instructions (file INSTALL).
1664 :     - Fixed curl-detection bug in config/install.sh.
1665 :     - It is now possible to select the URL getter using the URLGETTER
1666 :     environment variable:
1667 :    
1668 :     not set / "unknown" --> automatic detection (script tries wget,
1669 :     curl, and lynx)
1670 :     "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1671 :     how to properly invoke them)
1672 :     other --> use $URLGETTER directly, it must take
1673 :     precisely two command-line arguments
1674 :     (source URL and destination file name)
1675 :    
1676 :     ----------------------------------------------------------------------
1677 :     Name: Matthias Blume
1678 : blume 1044 Date: 2002/01/28
1679 : blume 1045 Tag: blume-20020128-sparc-ccalls
1680 :     Description:
1681 :    
1682 :     - Fixed problem with calculation of "used" registers in sparc-c-calls.
1683 :     - Make use of the allocParam argument in sparc-c-calls.
1684 :    
1685 :     ----------------------------------------------------------------------
1686 :     Name: Matthias Blume
1687 :     Date: 2002/01/28
1688 : blume 1044 Tag: blume-20020128-allocParam
1689 :     Description:
1690 :    
1691 :     John Reppy: Changes c-calls API to accept client-callback for
1692 :     allocating extra stack space.
1693 :     me: Corresponding changes to mlriscGen (using a dummy argument that
1694 :     does not change the current behavior).
1695 :    
1696 :     ----------------------------------------------------------------------
1697 :     Name: Matthias Blume
1698 : blume 1040 Date: 2002/01/28 12:00:00
1699 :     Tag: Release_110_38
1700 :     Description:
1701 :    
1702 :     This time for real!!!
1703 :    
1704 :     ----------------------------------------------------------------------
1705 :     Name: Matthias Blume
1706 : blume 1039 Date: 2002/01/28 10:56:00 EST
1707 :     Tag: blume-20020128-retraction
1708 :     Description:
1709 :    
1710 :     0. Retracted earlier 110.38. (The Release_110_38 tag has been replaced
1711 :     with blume-Release_110_38-retracted.)
1712 :    
1713 :     1. Fixed a problem with incorrect rounding modes in real64.sml.
1714 :     (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1715 :    
1716 :     2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1717 :     and enums fixed. The naming of corresponding ML identifiers should
1718 :     now be consistent again.
1719 :    
1720 :     ----------------------------------------------------------------------
1721 : leunga 1038 Name: Allen Leung
1722 :     Date: 2002/01/27
1723 :     Tag: leunga-20020127-nowhere
1724 :     Description:
1725 :    
1726 :     Added a target called nowhere in the configuration scripts.
1727 :     Enabling this will build the MLRISC 'nowhere' tool (for translating
1728 :     programs with where-clauses into legal SML code) during installation.
1729 :    
1730 :     ----------------------------------------------------------------------
1731 : blume 1036 Name: Matthias Blume
1732 : leunga 1038 Date: 2002/01/25 21:27:00 EST
1733 : blume 1039 Tag: blume-Release_110_38-retracted
1734 : blume 1037 Description:
1735 :    
1736 :     Call it a (working) release! Version is 110.38. Bootfiles are ready.
1737 :    
1738 :     README will be added later.
1739 :    
1740 : blume 1039 !!! NOTE: Re-tagged as blume-Release_110_38-retracted. Original tag
1741 :     (Release_110_38) removed. Reason: Last-minute bug fixes.
1742 :    
1743 : blume 1037 ----------------------------------------------------------------------
1744 :     Name: Matthias Blume
1745 : blume 1036 Date: 2002/01/25
1746 :     Tag: blume-20020125-ffi
1747 :     Description:
1748 :    
1749 :     A large number of tweaks and improvements to ml-nlffi-lib and
1750 :     ml-nlffigen:
1751 :    
1752 :     - ML represenation types have been streamlined
1753 :     - getter and setter functions work with concrete values, not abstract
1754 :     ones where possible
1755 :     - ml-nlffigen command line more flexible (see README file there)
1756 :     - some bugs have been fixed (hopefully)
1757 :    
1758 :     ----------------------------------------------------------------------
1759 : george 1034 Name: Lal George
1760 :     Date: 2002/01/24
1761 :     Tag: george-20020124-risc-ra-interface
1762 :     Description:
1763 :    
1764 :     There is a dramatic simplification in the interface to the
1765 :     register allocator for RISC architectures as a result of making
1766 :     parallel copy instructions explicit.
1767 :    
1768 :     ----------------------------------------------------------------------
1769 : blume 1028 Name: Matthias Blume
1770 : blume 1032 Date: 2002/01/22
1771 :     Tag: blume-20020122-x86-ccalls
1772 :     Description:
1773 :    
1774 :     Bug fix for c-calls on x86 (having to do with how char- and
1775 :     short-arguments are being handled).
1776 :    
1777 :     ----------------------------------------------------------------------
1778 :     Name: Matthias Blume
1779 : blume 1031 Date: 2002/01/21
1780 :     Tag: blume-20020121-ff
1781 :     Description:
1782 :    
1783 :     Another day of fiddling with the FFI...
1784 :    
1785 :     1. Bug fix/workaround: CKIT does not complain about negative array
1786 :     dimensions, so ml-nlffigen has to guard itself against this possibility.
1787 :     (Otherwise a negative dimension would send it into an infinite loop.)
1788 :    
1789 :     2. Some of the abstract types (light objects, light pointers, most "base"
1790 :     types) in structure C are now eqtypes.
1791 :    
1792 :     3. Added constructors and test functions for NULL function pointers.
1793 :    
1794 :     ----------------------------------------------------------------------
1795 :     Name: Matthias Blume
1796 : blume 1028 Date: 2002/01/18
1797 : blume 1030 Tag: blume-20020118-ready-for-new-release
1798 :     Description:
1799 :    
1800 :     Made config/srcarchiveurl point to a new place. (Will provide boot
1801 :     files shortly.)
1802 :    
1803 :     Maybe we christen this to be 110.38?
1804 :    
1805 :     ----------------------------------------------------------------------
1806 :     Name: Matthias Blume
1807 :     Date: 2002/01/18
1808 : blume 1029 Tag: blume-20020118-more-ffifiddle
1809 :     Description:
1810 :    
1811 :     Today's FFI fiddling:
1812 :    
1813 :     - Provided a structure CGetSet with "convenient" versions of C.Get.* and
1814 :     C.Set.* that use concrete (MLRep.*) arguments and results instead
1815 :     of abstract ones.
1816 :    
1817 :     - Provided word-style bit operations etc. for "int" representation
1818 :     types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1819 :     and Long.
1820 :    
1821 :     ----------------------------------------------------------------------
1822 :     Name: Matthias Blume
1823 :     Date: 2002/01/18
1824 : blume 1028 Tag: blume-20020118-use-x86-fp
1825 :     Description:
1826 :    
1827 :     Now that x86-fast-fp seems to be working, I turned it back on again
1828 :     by default. (Seems to work fine now, even with the FFI.)
1829 :    
1830 :     Other than that, I added some documentation about the FFI to
1831 :     src/ml-nlffigen/README and updated the FFI test examples in
1832 :     src/ml-nlffi-lib/Tests/*.
1833 :    
1834 :     ----------------------------------------------------------------------
1835 : leunga 1025 Name: Allen Leung
1836 :     Date: 2002/01/17
1837 :     Tag: leunga-20020117-x86-fast-fp-call
1838 :     Description:
1839 :    
1840 :     1. Fixed a problem with handling return fp values when x86's fast fp
1841 :     mode is turned on.
1842 :    
1843 :     2. Minor pretty printing fix for cellset. Print %st(0) as %st(0) instead
1844 :     of %f32.
1845 :    
1846 :     3. Added a constructor INT32lit to the ast of MLRISC tools.
1847 :    
1848 :     ----------------------------------------------------------------------
1849 : blume 1021 Name: Matthias Blume
1850 :     Date: 2002/01/16
1851 :     Tag: blume-20020116-ffifiddle
1852 :     Description:
1853 :    
1854 :     More fiddling with the FFI interface:
1855 :    
1856 :     - Make constness 'c instead of rw wherever possible. This eliminates
1857 :     the need for certain explicit coercions. (However, due to ML's
1858 :     value polymorphism, there will still be many cases where explicit
1859 :     coercions are necessary. Phantom types are not the whole answer
1860 :     to modeling a subtyping relationship in ML.)
1861 :    
1862 :     - ro/rw coersions for pointers added. (Avoids the detour through */&.)
1863 :    
1864 :     - "printf" test example added to src/ml-nlffi-lib/Tests. (Demonstrates
1865 :     clumsy workaround for varargs problem.)
1866 :    
1867 :     ----------------------------------------------------------------------
1868 : george 1016 Name: Lal George
1869 :     Date: 2002/01/15
1870 :     Tag: <none>
1871 :     Description:
1872 :    
1873 :     1. Since COPY instructions are no longer native to the architecture,
1874 :     a generic functor can be used to implement the expandCopies function.
1875 :    
1876 :     2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1877 :     TEXT segment.
1878 :    
1879 :     ----------------------------------------------------------------------
1880 : blume 1011 Name: Matthias Blume
1881 : blume 1015 Date: 2002/01/15
1882 :     Tag: blume-20020115-ffiupdates
1883 :     Description:
1884 :    
1885 :     1. Fix for bug resulting in single-precision float values being returned
1886 :     incorrectly from FFI calls.
1887 :    
1888 :     2. Small modifications to C FFI API:
1889 :    
1890 :     - memory-allocation routines return straight objects (no options)
1891 :     and raise an exception in out-of-memory situations
1892 :     - unsafe extensions to cast between function pointers and pointers
1893 :     from/to ints
1894 :     - added structure C_Debug as an alternative to structure C where
1895 :     pointer-dereferencing (|*| and |*!) always check for null-pointers
1896 :     - added open_lib' to DynLinkage; open_lib' works like open_lib
1897 :     but also takes a (possibly empty) list of existing library handles
1898 :     that the current library depends on
1899 :    
1900 :     ----------------------------------------------------------------------
1901 :     Name: Matthias Blume
1902 : blume 1011 Date: 2002/01/10
1903 :     Tag: blume-20020110-newffigen
1904 :     Description:
1905 :    
1906 :     1. Updates to portable graph code.
1907 :    
1908 :     2. Major update to ml-nlffigen and ml-nlffi-lib. Things are much
1909 :     more scalable now so that even huge interfaces such as the one
1910 :     for GTK compile in finite time and space. :-)
1911 :     See src/ml-nlffigen/README for details on what's new.
1912 :    
1913 :     ----------------------------------------------------------------------
1914 : george 1003 Name: Lal George
1915 : george 1009 Date: 2001/01/09 14:31:35 EST 2002
1916 :     Tag: george-20011206-rm-native-copy
1917 :     Description:
1918 :    
1919 :     Removed the native COPY and FCOPY instructions
1920 :     from all the architectures and replaced it with the
1921 :     explicit COPY instruction from the previous commit.
1922 :    
1923 :     It is now possible to simplify many of the optimizations
1924 :     modules that manipulate copies. This has not been
1925 :     done in this change.
1926 :    
1927 :     ----------------------------------------------------------------------
1928 :     Name: Lal George
1929 : george 1003 Date: 2001/12/06 16:50:13 EST 2001
1930 :     Tag: george-20011206-mlrisc-instruction
1931 :     Description:
1932 : george 824
1933 : george 1003 Changed the representation of instructions from being fully abstract
1934 :     to being partially concrete. That is to say:
1935 :    
1936 :     from
1937 :     type instruction
1938 :    
1939 :     to
1940 :     type instr (* machine instruction *)
1941 :    
1942 :     datatype instruction =
1943 :     LIVE of {regs: C.cellset, spilled: C.cellset}
1944 :     | KILL of {regs: C.cellset, spilled: C.cellset}
1945 :     | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
1946 :     | ANNOTATION of {i: instruction, a: Annotations.annotation}
1947 :     | INSTR of instr
1948 :    
1949 :     This makes the handling of certain special instructions that appear on
1950 :     all architectures easier and uniform.
1951 :    
1952 :     LIVE and KILL say that a list of registers are live or killed at the
1953 :     program point where they appear. No spill code is generated when an
1954 :     element of the 'regs' field is spilled, but the register is moved to
1955 :     the 'spilled' (which is present, more for debugging than anything else).
1956 :    
1957 :     LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
1958 :     We used to generate:
1959 :    
1960 :     DEFFREG f1
1961 :     f1 := f2 + f3
1962 :     trapb
1963 :    
1964 :     but now generate:
1965 :    
1966 :     f1 := f2 + f3
1967 :     trapb
1968 :     LIVE {regs=[f1,f2,f3], spilled=[]}
1969 :    
1970 :     Furthermore, the DEFFREG (hack) required that all floating point instruction
1971 :     use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
1972 :     defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
1973 :     in a cleaner alpha implementation. (Hopefully, intel will not get rid of
1974 :     this architecture).
1975 :    
1976 :     COPYXXX is intended to replace the parallel COPY and FCOPY available on
1977 :     all the architectures. This will result in further simplification of the
1978 :     register allocator that must be aware of them for coalescing purposes, and
1979 :     will also simplify certain aspects of the machine description that provides
1980 :     callbacks related to parallel copies.
1981 :    
1982 :     ANNOTATION should be obvious, and now INSTR represents the honest to God
1983 :     machine instruction set!
1984 :    
1985 :     The <arch>/instructions/<arch>Instr.sml files define certain utility
1986 :     functions for making porting easier -- essentially converting upper case
1987 :     to lower case. All machine instructions (of type instr) are in upper case,
1988 :     and the lower case form generates an MLRISC instruction. For example on
1989 :     the alpha we have:
1990 :    
1991 :     datatype instr =
1992 :     LDA of {r:cell, b:cell, d:operand}
1993 :     | ...
1994 :    
1995 :     val lda : {r:cell, b:cell, d:operand} -> instruction
1996 :     ...
1997 :    
1998 :     where lda is just (INSTR o LDA), etc.
1999 :    
2000 : leunga 815 ----------------------------------------------------------------------
2001 : blume 986 Name: Matthias Blume
2002 : blume 993 Date: 2001/11/22 21:40:00 EST
2003 :     Tag: Release_110_37
2004 :     Description:
2005 :    
2006 :     Release 110.37. This time for real.
2007 :    
2008 :     ----------------------------------------------------------------------
2009 :     Name: Matthias Blume
2010 : blume 988 Date: 2001/11/21 16:35:00 EST
2011 :     Tag: blume-20011121-foot-in-mouth
2012 :     Description:
2013 :    
2014 :     Removed the "Release_110_37" tag because of a serious bug.
2015 :     This will be re-tagged once the bug is fixed.
2016 :    
2017 :     ----------------------------------------------------------------------
2018 :     Name: Matthias Blume
2019 : blume 987 Date: 2001/11/21 16:14:00 EST
2020 :     Tag: blume-20011121-forgottenfile
2021 :     Description:
2022 :    
2023 :     Forgot to add a file. (Just a .tex-file -- part of
2024 :     the CM manual source.)
2025 :    
2026 :     ----------------------------------------------------------------------
2027 :     Name: Matthias Blume
2028 : blume 986 Date: 2001/11/21 16:10:00 EST
2029 : blume 988 Tag: blume-20011121-invalid_110_37
2030 : blume 986 Description:
2031 :    
2032 : blume 988 Note: I removed the original tag "Release_110_37" from this commit
2033 :     because we found a serious bug in all non-x86 backends.
2034 :     - Matthias
2035 :    
2036 : blume 986 1. Modifications to the SML/NJ code generator and to the runtime system
2037 :     so that code object name strings are directly inserted into code
2038 :     objects at code generation time. The only business the runtime system
2039 :     has with this is now to read the name strings on occasions.
2040 :     (The encoding of the name string has also changed somewhat.)
2041 :    
2042 :     2. CM now implements a simple "set calculus" for specifying export lists.
2043 :     In particular, it is now possible to refer to the export lists of
2044 :     other libraries/groups/sources and form unions as well as differences.
2045 :     See the latest CM manual for details.
2046 :    
2047 :     3. An separate notion of "proxy" libraries has again be eliminated from
2048 :     CM's model. (Proxy libraries are now simply a special case of using
2049 :     the export list calculus.)
2050 :    
2051 :     4. Some of the existing libraries now take advantage of the new set
2052 :     calculus.
2053 :     (Notice that not all libraries have been converted because some
2054 :     of the existing .cm-files are supposed to be backward compatible
2055 :     with 110.0.x.)
2056 :    
2057 :     5. Some cleanup in stand-alone programs. (Don't use "exnMessage" -- use
2058 :     "General.exnMessage"! The former relies on a certain hook to be
2059 :     initialized, and that often does not happen in the stand-alone case.)
2060 :    
2061 :     ----------------------------------------------------------------------
2062 : george 984 Name: Lal George
2063 :     Date: 2001/11/21 13:56:18 EST
2064 :     Tag: george-2001121-pseudo-ops
2065 :     Description:
2066 :    
2067 :     Implemented a complete redesign of MLRISC pseudo-ops. Now there
2068 :     ought to never be any question of incompatabilities with
2069 :     pseudo-op syntax expected by host assemblers.
2070 :    
2071 :     For now, only modules supporting GAS syntax are implemented
2072 :     but more should follow, such as MASM, and vendor assembler
2073 :     syntax, e.g. IBM as, Sun as, etc.
2074 :    
2075 :     ----------------------------------------------------------------------
2076 : blume 977 Name: Matthias Blume
2077 :     Date: 2001/11/14 11:52:00 EST
2078 :     Tag: blume-20011114-srcname
2079 :     Description:
2080 :    
2081 :     1. Routed the name of the current source file to mlriscgen where it
2082 :     should be directly emitted into the code object. (This last part
2083 :     is yet to be done.)
2084 :    
2085 :     2. Some cleanup of the pgraph code to make it match the proposal that
2086 :     I put out the other day. (The proposal notwithstanding, things are
2087 :     still in flux here.)
2088 :    
2089 :     ----------------------------------------------------------------------
2090 : george 976 Name: Lal George
2091 :     Date: 2001/11/14 09:44:04 EST
2092 :     Tag:
2093 :     Description:
2094 :    
2095 :     Fix for a backpatching bug reported by Allen.
2096 :    
2097 :     Because the boundary between short and long span-dependent
2098 :     instructions is +/- 128, there are an astounding number of
2099 :     span-dependent instructions whose size is over estimated.
2100 :    
2101 :     Allen came up with the idea of letting the size of span
2102 :     dependent instructions be non-monotonic, for a maxIter
2103 :     number of times, after which the size must be monotonically
2104 :     increasing.
2105 :    
2106 :     This table shows the number of span-dependent instructions
2107 :     whose size was over-estimated as a function of maxIter, for the
2108 :     file Parse/parse/ml.grm.sml:
2109 :    
2110 :     maxIter # of instructions:
2111 :     10 687
2112 :     20 438
2113 :     30 198
2114 :     40 0
2115 :    
2116 :     In compiling the compiler, there is no significant difference in
2117 :     compilation speed between maxIter=10 and maxIter=40. Actually,
2118 :     my measurements showed that maxIter=40 was a tad faster than
2119 :     maxIter=10! Also 96% of the files in the compiler reach a fix
2120 :     point within 13 iterations, so fixing maxIter at 40, while high,
2121 :     is okay.
2122 :    
2123 :     ----------------------------------------------------------------------
2124 : blume 975 Name: Matthias Blume
2125 :     Date: 2001/10/31 15:25:00 EST
2126 :     Tag: blume-20011031-pgraph
2127 :     Description:
2128 :    
2129 :     CKIT:
2130 :     * Changed the "Function" constructor of type Ast.ctype to carry optional
2131 :     argument identifiers.
2132 :     * Changed the return type of TypeUtil.getFunction accordingly.
2133 :     * Type equality ignores the argument names.
2134 :     * TypeUtil.composite tries to preserve argument names but gives up quickly
2135 :     if there is a mismatch.
2136 :    
2137 :     installation script:
2138 :     * attempts to use "curl" if available (unless "wget" is available as well)
2139 :    
2140 :     CM:
2141 :     * has an experimental implementation of "portable graphs" which I will
2142 :     soon propose as an implementation-independent library format
2143 :     * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2144 :    
2145 :     NLFFI-LIB:
2146 :     * some cleanup (all cosmetic)
2147 :    
2148 :     NLFFIGEN:
2149 :     * temporarily disabled the mechanism that suppresses ML output for
2150 :     C definitions whose identifiers start with an underscore character
2151 :     * generate val bindings for enum constants
2152 :     * user can request that only one style (light or heavy) is being used;
2153 :     default is to use both (command-line arguments: -heavy and -light)
2154 :     * fixed bug in handling of function types involving incomplete pointers
2155 :     * generate ML entry points that take record arguments (i.e., using
2156 :     named arguments) for C functions that have a prototype with named
2157 :     arguments
2158 :     (see changes to CKIT)
2159 :    
2160 :     ----------------------------------------------------------------------
2161 : leunga 972 Name: Allen Leung
2162 :     Date: 2001/10/27 20:34:00 EDT
2163 :     Tag: leunga-20011027-x86-fast-fp-call
2164 :     Description:
2165 :    
2166 :     Fixed the bug described in blume-20010920-slowfp.
2167 :    
2168 :     The fix involves
2169 :     1. generating FCOPYs in FSTP in ia32-svid
2170 :     2. marking a CALL with the appropriate annotation
2171 :    
2172 :     ----------------------------------------------------------------------
2173 : blume 961 Name: Matthias Blume
2174 :     Date: 2001/10/16 11:32:00 EDT
2175 :     Tag: blume-20011016-netbsd
2176 :     Description:
2177 :    
2178 :     Underscore patch from Chris Richards (fixing problem with compiling
2179 :     runtime system under recent NetBSD).
2180 :    
2181 :     ----------------------------------------------------------------------
2182 : george 959 Name: Allen Leung
2183 :     Date: 2001/10/12 17:18:32 EDT 2001
2184 :     Tag: leung-20011012-x86-printflowgraph
2185 :     Description:
2186 :    
2187 :     X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2188 :    
2189 :     ----------------------------------------------------------------------
2190 : george 952 Name: Lal George
2191 : george 958 Date: 2001/10/11 23:51:34 EDT
2192 :     Tag: george-20011011-too-many-instrs
2193 :     Description:
2194 :    
2195 :     The representation of a program point never expected to see more
2196 :     than 65536 instructions in a basic block!
2197 :    
2198 :     ----------------------------------------------------------------------
2199 :     Name: Lal George
2200 : george 952 Date: 2001/10/09 09:41:37 EDT
2201 :     Tag: george-20011008-mlrisc-labels
2202 :     Description:
2203 :    
2204 :     Changed the machine description files to support printing of
2205 :     local and global labels in assembly code, based on host assembler
2206 :     conventions.
2207 :    
2208 :     ----------------------------------------------------------------------
2209 : blume 916 Name: Matthias Blume
2210 : blume 939 Date: 2001/09/25 15:25:00 EDT
2211 :     Tag: blume-20010925-exninfo
2212 :     Description:
2213 :    
2214 :     I provided a non-hook implementation of exnName (at the toplevel) and
2215 :     made the "dummy" implementation of exnMessage (at the toplevel) more
2216 :     useful: if nothing gets "hooked in", then at least you are going to
2217 :     see the exception name and a message indicating why you don't see more.
2218 :    
2219 :     [For the time being, programs that need exnMessage and want to use
2220 :     ml-build should either use General.exnMessage (strongly recommended) or
2221 :     refer to structure General at some other point so that CM sees a
2222 :     static dependency.]
2223 :    
2224 :     [Similar remarks go for "print" and "use": If you want to use their
2225 :     functionality in stand-alone programs generated by ml-build, then use
2226 :     TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2227 :    
2228 :     ----------------------------------------------------------------------
2229 :     Name: Matthias Blume
2230 : blume 938 Date: 2001/09/20 17:28:00 EDT
2231 :     Tag: blume-20010920-slowfp
2232 :     Description:
2233 :    
2234 :     Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2235 :    
2236 :     ----------------------------------------------------------------------
2237 :     Name: Matthias Blume
2238 : blume 937 Date: 2001/09/20 17:20:00 EDT
2239 :     Tag: blume-20010920-canonicalpaths
2240 :     Description:
2241 :    
2242 :     0. Updated the BOOT file (something that I forgot to do earlier).
2243 :    
2244 :     1. Small internal change to CM so that it avoids "/../" in filenames
2245 :     as much as possible (but only where it is safe).
2246 :    
2247 :     2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2248 :     that arguments that contain delimiters are passed through correctly.
2249 :     This change also means that all "special" arguments of the form
2250 :     @SMLxxx... must come first.
2251 :    
2252 :     3. Changed install script to put relative anchor names for tool commands
2253 :     into pathconfig.
2254 :    
2255 :     ----------------------------------------------------------------------
2256 :     Name: Matthias Blume
2257 : blume 932 Date: 2001/09/18 15:35:00 EDT
2258 :     Tag: blume-20010918-readme11036
2259 :     Description:
2260 :    
2261 :     Added README files.
2262 :    
2263 :     ----------------------------------------------------------------------
2264 :     Name: Matthias Blume
2265 : blume 929 Date: 2001/09/18 11:45:00 EDT
2266 :     Tag: Release_110_36 (retag)
2267 :     Description:
2268 :    
2269 :     Fixed mistake in config/preloads. Retagged as 110.36.
2270 :    
2271 :     ----------------------------------------------------------------------
2272 :     Name: Matthias Blume
2273 : blume 928 Date: 2001/09/18 09:40:00 EDT
2274 : blume 929 Tag: Release_110_36_orig (tag changed)
2275 : blume 928 Description:
2276 :    
2277 :     New version (110.36). New bootfiles.
2278 :    
2279 :     ----------------------------------------------------------------------
2280 :     Name: Matthias Blume
2281 : blume 927 Date: 2001/09/14 16:15:00 EDT
2282 :     Tag: blume-20010914-x86fastfp
2283 :     Description:
2284 :    
2285 :     John committed some changes that Allen made, in particular a (hopefully)
2286 :     correctly working version of the x86-fp module.
2287 :    
2288 :     I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
2289 :     flag to "true". Everything seems to compile to a fixpoint ok, and
2290 :     "mandelbrot" speeds up by about 15%.
2291 :    
2292 :     ----------------------------------------------------------------------
2293 :     Name: Matthias Blume
2294 : blume 923 Date: 2001/09/13 11:20:00 EDT
2295 :     Tag: blume-20010913-minimal
2296 :     Description:
2297 :    
2298 :     1. Stefan Monnier's patch to fix a miscompilation problem that
2299 :     was brought to light by John Reppy's work on Moby.
2300 :    
2301 :     2. Implemented a minimal "structure Compiler" that contains just
2302 :     "version" and "architecture". The minimal version will be
2303 :     available when the full version is not. This is for backward-
2304 :     compatibility with code that wants to test Compiler.version.
2305 :    
2306 :     ----------------------------------------------------------------------
2307 :     Name: Matthias Blume
2308 : blume 920 Date: 2001/08/28 14:03:00 EDT
2309 :     Tag: blume-20010828-ml-lex
2310 :     Description:
2311 :    
2312 :     Fix for bug 1581, received from Neophytos Michael.
2313 :    
2314 :     ----------------------------------------------------------------------
2315 :     Name: Matthias Blume
2316 : blume 918 Date: 2001/08/27 11:20:00 EDT
2317 :     Tag: blume-20010827-readme11035
2318 :     Description:
2319 :    
2320 :     Fleshed out the README file for 110.35.
2321 :    
2322 :     ----------------------------------------------------------------------
2323 :     Name: Matthias Blume
2324 : blume 916 Date: 2001/08/24 17:10:00 EDT
2325 :     Tag: Release_110_35
2326 :     Description:
2327 :    
2328 :     New version number (110.35). New bootfiles.
2329 :    
2330 :     ----------------------------------------------------------------------
2331 : george 915 Name: Lal George
2332 : george 909 Date: 2001/08/24 13:47:18 EDT 2001
2333 :     Tag: george-20010824-MLRISC-graphs
2334 :     Description:
2335 :    
2336 :     removed clusters from MLRISC completely and replaced with graphs.
2337 :    
2338 :     ----------------------------------------------------------------------
2339 :     Name: Matthias Blume
2340 : blume 905 Date: 2001/08/23 17:50:00 EDT
2341 :     Tag: blume-20010823-toplevel
2342 :     Description:
2343 :    
2344 :     - some reorganization of the code that implements various kinds of
2345 :     environments in the compiler (static, dynamic, symbolic, combined)
2346 :     - re-implemented the EnvRef module so that evalStream works properly
2347 :     (if the stream contains references to "use", "CM.make", etc.)
2348 :     - cleaned up evalloop.sml and interact.sml (but they need more cleaning)
2349 :    
2350 :     ----------------------------------------------------------------------
2351 :     Name: Matthias Blume
2352 : blume 904 Date: 2001/08/20 15:50 EDT
2353 :     Tag: blume20010820-slipup
2354 :     Description:
2355 :    
2356 :     I forgot to commit a few files. Here they are...
2357 :    
2358 :     ----------------------------------------------------------------------
2359 :     Name: Matthias Blume
2360 : blume 903 Date: 2001/08/20 15:35:00 EDT
2361 :     Tag: blume-20010820-debugprof
2362 :     Description:
2363 :    
2364 :     !!!! NEW BOOTFILES !!!!
2365 :    
2366 :     This is another round of reorganizing the compiler sources. This
2367 :     time the main goal was to factor out all the "instrumentation"
2368 :     passes (for profiling and backtracing) into their own library.
2369 :     The difficulty was to do it in such a way that it does not depend
2370 :     on elaborate.cm but only on elabdata.cm.
2371 :    
2372 :     Therefore there have been further changes to both elaborate.cm and
2373 :     elabdata.cm -- more "generic" things have been moved from the former
2374 :     to the latter. As a result, I was forced to split the assignment
2375 :     of numbers indicating "primtyc"s into two portions: SML-generic and
2376 :     SML/NJ-specific. Since it would have been awkward to maintain,
2377 :     I bit the bullet and actually _changed_ the mapping between these
2378 :     numbers and primtycs. The bottom line of this is that you need
2379 :     a new set of bin- and bootfiles.
2380 :    
2381 :     I have built new bootfiles for all architectures, so doing a fresh
2382 :     checkout and config/install.sh should be all you need.
2383 :    
2384 :     The newly created library's name is
2385 :    
2386 :     $smlnj/viscomp/debugprof.cm
2387 :    
2388 :     and its sources live under
2389 :    
2390 :     src/compiler/DebugProf
2391 :    
2392 :     ----------------------------------------------------------------------
2393 :     Name: Matthias Blume
2394 : blume 902 Date: 2001/08/15 17:15:00 EDT
2395 :     Tag: blume-20010815-compreorg
2396 :     Description:
2397 :    
2398 :     This is a first cut at reorganizing the CM libraries that make up the
2399 :     core of the compiler. The idea is to separate out pieces that could
2400 :     be used independently by tools, e.g., the parser, the typechecker, etc.
2401 :    
2402 :     The current status is a step in this direction, but it is not quite
2403 :     satisfactory yet. Expect more changes in the future.
2404 :    
2405 :     Here is the current (new) organization...
2406 :    
2407 :     What used to be $smlnj/viscomp/core.cm is now divided into
2408 :     six CM libraries:
2409 :    
2410 :     $smlnj/viscomp/basics.cm
2411 :     /parser.cm
2412 :     /elabdata.cm
2413 :     /elaborate.cm
2414 :     /execute.cm
2415 :     /core.cm
2416 :    
2417 :     The CM files for these libraries live under src/system/smlnj/viscomp.
2418 :     All these libraries are proxy libraries that contain precisely
2419 :     one CM library component. Here are the locations of the components
2420 :     (all within the src/compiler tree):
2421 :    
2422 :     Basics/basics.cm
2423 :     Parse/parser.cm
2424 :     ElabData/elabdata.cm
2425 :     Elaborator/elaborate.cm
2426 :     Execution/execute.cm
2427 :     core.cm
2428 :    
2429 :     [This organization is the same that has been used already
2430 :     for a while for the architecture-specific parts of the visible
2431 :     compiler and for the old version of core.cm.]
2432 :    
2433 :     As you will notice, many source files have been moved from their
2434 :     respective original locations to a new home in one of the above
2435 :     subtrees.
2436 :    
2437 :     The division of labor between the new libraries is the following:
2438 :    
2439 :     basics.cm:
2440 :     - Simple, basic definitions that pertain to many (or all) of
2441 :     the other libraries.
2442 :     parser.cm:
2443 :     - The SML parser, producing output of type Ast.dec.
2444 :     - The type family for Ast is also defined and exported here.
2445 :     elabdata.cm:
2446 :     - The datatypes that describe input and output of the elaborator.
2447 :     This includes types, absyn, and static environments.
2448 :     elaborator.cm:
2449 :     - The SML/NJ type checker and elaborator.
2450 :     This maps an Ast.dec (with a given static environment) to
2451 :     an Absyn.dec (with a new static environment).
2452 :     - This libraries implements certain modules that used to be
2453 :     structures as functors (to remove dependencies on FLINT).
2454 :     execute.cm:
2455 :     - Everything having to do with executing binary code objects.
2456 :     - Dynamic environments.
2457 :     core.cm:
2458 :     - SML/NJ-specific instantiations of the elaborator and MLRISC.
2459 :     - Top-level modules.
2460 :     - FLINT (this should eventually become its own library)
2461 :    
2462 :     Notes:
2463 :    
2464 :     I am not 100% happy with the way I separated the elaborator (and its
2465 :     data structures) from FLINT. Two instances of the same problem:
2466 :    
2467 :     1. Data structures contain certain fields that carry FLINT-specific
2468 :     information. I hacked around this using exn and the property list
2469 :     module from smlnj-lib. But the fact that there are middle-end
2470 :     specific fields around at all is a bit annoying.
2471 :    
2472 :     2. The elaborator calculates certain FLINT-related information. I tried
2473 :     to make this as abstract as I could using functorization, but, again,
2474 :     the fact that the elaborator has to perform calculations on behalf
2475 :     of the middle-end at all is not nice.
2476 :    
2477 :     3. Having to used exn and property lists is unfortunate because it
2478 :     weakens type checking. The other alternative (parameterizing
2479 :     nearly *everything*) is not appealing, though.
2480 :    
2481 :     I removed the "rebinding =" warning hack because due to the new organization
2482 :     it was awkward to maintain it. As a result, the compiler now issues some of
2483 :     these warnings when compiling init.cmi during bootstrap compilation. On
2484 :     the plus side, you also get a warning when you do, for example:
2485 :     val op = = Int32.+
2486 :     which was not the case up to now.
2487 :    
2488 :     I placed "assign" and "deref" into the _Core structure so that the
2489 :     code that deals with the "lazy" keyword can find them there. This
2490 :     removes the need for having access to the primitive environment
2491 :     during elaboration.
2492 :    
2493 :     ----------------------------------------------------------------------
2494 :     Name: Matthias Blume
2495 : blume 898 Date: 2001/08/13
2496 :     Tag: blume-20010813-closures
2497 :     Description:
2498 :    
2499 :     This fix was sent to us by Zhong Shao. It is supposed to improve the
2500 :     performance of certain loops by avoiding needless closure allocation.
2501 :    
2502 :     ----------------------------------------------------------------------
2503 : george 890 Name: Lal George
2504 : george 896 Date: 2001/07/31 10:03:23 EDT 2001
2505 :     Tag: george-20010731-x86-fmalloc
2506 :     Description: Fixed bug in x86 calls
2507 :    
2508 :     There was a bug where call instructions would mysteriously
2509 :     vanish. The call instruction had to be one that returned
2510 :     a floating point value.
2511 :    
2512 :     ----------------------------------------------------------------------
2513 :     Name: Lal George
2514 : george 890 Date: 2001/07/19 16:36:29 EDT 2001
2515 :     Tag: george-20010719-simple-cells
2516 :     Description:
2517 :    
2518 :     I have dramatically simplified the interface for CELLS in MLRISC.
2519 :    
2520 :     In summary, the cells interface is broken up into three parts:
2521 :    
2522 :     1. CellsBasis : CELLS_BASIS
2523 :    
2524 :     CellsBasis is a top level structure and common for all
2525 :     architectures. it contains the definitions of basic datatypes
2526 :     and utility functions over these types.
2527 :    
2528 :     2. functor Cells() : CELLS
2529 :    
2530 :     Cells generates an interface for CELLS that incorporates the
2531 :     specific resources on the target architecture, such as the
2532 :     presence of special register classes, their number and size,
2533 :     and various useful substructures.
2534 :    
2535 :     3. <ARCH>CELLS
2536 :    
2537 :     e.g. SparcCells: SPARCCELLS
2538 :    
2539 :     <ARCH>CELLS usually contains additional bindings for special
2540 :     registers on the architecture, such as:
2541 :    
2542 :     val r0 : cell (* register zero *)
2543 :     val y : cell (* Y register *)
2544 :     val psr : cell (* processor status register *)
2545 :     ...
2546 :    
2547 :     The structure returned by applying the Cells functor is opened
2548 :     in this interface.
2549 :    
2550 :     The main implication of all this is that the datatypes for cells is
2551 :     split between CellsBasis and CELLS -- a fairly simple change for user
2552 :     code.
2553 :    
2554 :     In the old scheme the CELLS interface had a definitional binding of
2555 :     the form:
2556 :    
2557 :     signature CELLS = sig
2558 :    
2559 :     structure CellsBasis = CellsBasis
2560 :    
2561 :     ...
2562 :    
2563 :     end
2564 :    
2565 :     With all the sharing constraints that goes on in MLRISC, this old
2566 :     design quickly leads to errors such as:
2567 :    
2568 :     "structure definition spec inside of sharing ... "
2569 :    
2570 :    
2571 :     and appears to require an unacceptable amount of sharing and where
2572 :     constraint hackery.
2573 :    
2574 :     I think this error message (the interaction of definitional specs and
2575 :     sharing) requires more explanation on our web page.
2576 :    
2577 :     ----------------------------------------------------------------------
2578 : blume 826 Name: Matthias Blume
2579 : blume 879 Date: 2001/07/19 15:00:00 EDT
2580 :     Tag: blume-20010719-libreorg
2581 :     Description:
2582 :    
2583 :     This update puts together a fairly extensive but straightforward change
2584 :     to the way the libraries that implement the interactive system are
2585 :     organized:
2586 :    
2587 :     The biggest change is the elimination of structure Compiler. As a
2588 :     replacement for this structure, there is now a CM library
2589 :     (known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
2590 :     that exports all the substructures of the original structure Compiler
2591 :     directly. So instead of saying Compiler.Foo.bar one now simply
2592 :     says Foo.bar. (The CM libraries actually export a collection of
2593 :     structures that is richer than the collection of substructures of
2594 :     structure Compiler.)
2595 :    
2596 :     To make the transition smooth, there is a separate library called
2597 :     $smlnj/compiler/compiler.cm which puts together and exports the
2598 :     original structure Compiler (or at least something very close to it).
2599 :    
2600 :     There are five members of the original structure Compiler
2601 :     that are not exported directly but which instead became members
2602 :     of a new structure Backend (described by signature BACKEND). These are:
2603 :     structure Profile (: PROFILE), structure Compile (: COMPILE), structure
2604 :     Interact (: INTERACT), structure Machine (: MACHINE), and val
2605 :     architecture (: string).
2606 :    
2607 :     Structure Compiler.Version has become structure CompilerVersion.
2608 :    
2609 :     Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
2610 :     by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
2611 :     or x86, respectively.
2612 :     Each of these exports the same frontend structures that
2613 :     $smlnj/compiler.cm exports. But they do not have a structure Backend
2614 :     and instead export some structure <Arch>Backend where <Arch> is Alpha32,
2615 :     Hppa, PPC, Sparc, or X86, respectively.
2616 :    
2617 :     Library $smlnj/compiler/all.cm exports the union of the exports of
2618 :     $smlnj/compiler/<arch>.cm
2619 :    
2620 :     There are no structures <Arch>Compiler anymore, use
2621 :     $smlnj/compiler/<arch>.cm instead.
2622 :    
2623 :     Library host-compiler-0.cm is gone. Instead, the internal library
2624 :     that instantiates CM is now called cm0.cm. Selection of the host
2625 :     compiler (backend) is no longer done here but. (Responsibility for it
2626 :     now lies with $smlnj/compiler/current.cm. This seems to be more
2627 :     logical.)
2628 :    
2629 :     Many individual files have been moved or renamed. Some files have
2630 :     been split into multiple files, and some "dead" files have been deleted.
2631 :    
2632 :     Aside from these changes to library organization, there are also changes
2633 :     to the way the code itself is organized:
2634 :    
2635 :     Structure Binfile has been re-implemented in such a way that it no
2636 :     longer needs any knowledge of the compiler. It exclusively deals
2637 :     with the details of binfile layout. It no longer invokes the
2638 :     compiler (for the purpose of creating new prospective binfile
2639 :     content), and it no longer has any knowledge of how to interpret
2640 :     pickles.
2641 :    
2642 :     Structure Compile (: COMPILE) has been stripped down to the bare
2643 :     essentials of compilation. It no longer deals with linking/execution.
2644 :     The interface has been cleaned up considerably.
2645 :    
2646 :     Utility routines for dealing with linking and execution have been
2647 :     moved into their own substructures.
2648 :    
2649 :     (The ultimate goal of these changes is to provide a light-weight
2650 :     binfile loader/linker (at least for, e.g., stable libraries) that
2651 :     does not require CM or the compiler to be present.)
2652 :    
2653 :     CM documentation has been updated to reflect the changes to library
2654 :     organization.
2655 :    
2656 :     ----------------------------------------------------------------------
2657 :     Name: Matthias Blume
2658 : blume 875 Date: 2001/07/10 17:30:00 EDT
2659 :     Tag: Release_110_34
2660 :     Description:
2661 :    
2662 : blume 876 Minor tweak to 110.34 (re-tagged):
2663 :    
2664 :     - README.html file added to CVS repository
2665 :     - runtime compiles properly under FreeBSD 3.X and 4.X
2666 :    
2667 :     ----------------------------------------------------------------------
2668 :     Name: Matthias Blume
2669 :     Date: 2001/07/10 17:30:00 EDT
2670 :     Tag: Release_110_34
2671 :     Description:
2672 :    
2673 : blume 875 New version number (110.34). New bootfiles.
2674 :    
2675 :     ----------------------------------------------------------------------
2676 :     Name: Matthias Blume
2677 : blume 874 Date: 2001/07/09 16:00:00 EDT
2678 :     Tag: blume-20010709-more-varargs
2679 :     Description:
2680 :    
2681 :     I changed the handling of varargs in ml-nlffigen again:
2682 :     The ellipsis ... will now simply be ignored (with an accompanying warning).
2683 :    
2684 :     The immediate effect is that you can actually call a varargs function
2685 :     from ML -- but you can't actually supply any arguments beyond the ones
2686 :     specified explicitly. (For example, you can call printf with its format
2687 :     string, but you cannot pass additional arguments.)
2688 :    
2689 :     This behavior is only marginally more useful than the one before, but
2690 :     it has the advantage that a function or, more importantly, a function
2691 :     type never gets dropped on the floor, thus avoiding follow-up problems with
2692 :     other types that refer to the offending one.
2693 :    
2694 :     ----------------------------------------------------------------------
2695 :     Name: Matthias Blume
2696 : blume 873 Date: 2001/07/09 11:25:00 EDT
2697 :     Tag: blume-20010709-varargs
2698 :     Description:
2699 :    
2700 :     1. ckit-lib.cm now exports structure Error
2701 :     2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
2702 :     with a warning accompanied by a source location. Moreover, it
2703 :     merely skips the offending function or type and proceeds with the
2704 :     rest of its work.u As a result, one can safely feed C code containing
2705 :     "..." to ml-nlffigen.
2706 :     3. There are some internal improvements to CM, providing slightly
2707 :     more general string substitutions in the tools subsystem.
2708 :    
2709 :     ----------------------------------------------------------------------
2710 :     Name: Matthias Blume
2711 : blume 854 Date: 2001/06/27 15:10:00 EDT
2712 :     Tag: blume-20010627-concur
2713 :     Description:
2714 :    
2715 :     Fixed a small bug in CM's handling of parallel compilation.
2716 :     (You could observe the bug by Control-C-interrupting an ordinary
2717 :     CMB.make or CM.stabilize and then attaching some compile servers.
2718 :     The result was that all of a sudden the previously interrupted
2719 :     compilation would continue on its own. This was because of
2720 :     an over-optimization: CM did not bother to clean out certain queues
2721 :     when no servers were attached "anyway", resulting in the contents
2722 :     of these queues to grab control when new servers did get attached.)
2723 :    
2724 :     There is also another minor update to the CM manual.
2725 :    
2726 :     ----------------------------------------------------------------------
2727 :     Name: Matthias Blume
2728 : blume 853 Date: 2001/06/26 16:15:00 EDT
2729 :     Tag: blume-20010626-cmdoc
2730 :     Description:
2731 :    
2732 :     Minor typo fixed in CM manual (syntax diagram for libraries).
2733 :    
2734 :     ----------------------------------------------------------------------
2735 :     Name: Matthias Blume
2736 : blume 852 Date: 2001/06/25 22:55:00 EDT
2737 :     Tag: blume-20010625-x86pc
2738 :     Description:
2739 :    
2740 :     Fixed a nasty bug in the X86 assembly code that caused signal
2741 :     handlers to fail (crash) randomly.
2742 :    
2743 :     ----------------------------------------------------------------------
2744 :     Name: Matthias Blume
2745 : blume 846 Date: 2001/06/25 12:05:00 EDT
2746 :     Tag: blume-20010625-nlffigen
2747 :     Description:
2748 :    
2749 :     This update fixes a number of minor bugs in ml-nlffigen as reported by
2750 :     Nick Carter <nbc@andrew.cmu.edu>.
2751 :    
2752 :     1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
2753 :     2. Default names for generated files are now derived from the name of
2754 :     the C file *without its directory*. In particular, this causes generated
2755 :     files to be placed locally even if the C file is in some system directory.
2756 :     3. Default names for generated signatures and structures are also derived
2757 :     from the C file name without its directory. This avoids silly things
2758 :     like "structure GL/GL".
2759 :     (Other silly names are still possible because ml-nlffigen does not do
2760 : blume 847 a thorough check of whether generated names are legal ML identifiers.
2761 :     When in doubt, use command line arguments to force particular names.)
2762 : blume 846
2763 :     ----------------------------------------------------------------------
2764 :     Name: Matthias Blume
2765 : blume 845 Date: 2001/06/21 12:25:00 EDT
2766 :     Tag: blume-20010621-eXene
2767 :     Description:
2768 :    
2769 :     eXene now compiles and (sort of) works again.
2770 :    
2771 :     The library name (for version > 110.33) is $/eXene.cm.
2772 :    
2773 :     I also added an new example in src/eXene/examples/nbody. See the
2774 :     README file there for details.
2775 :    
2776 :     ----------------------------------------------------------------------
2777 :     Name: Matthias Blume
2778 : blume 844 Date: 2001/06/20 16:40:00 EDT
2779 :     Tag: blume-20010620-cml
2780 :     Description:
2781 :    
2782 :     CML now compiles and works again.
2783 :    
2784 :     Libraries (for version > 110.33):
2785 :    
2786 :     $cml/cml.cm Main CML library.
2787 :     $cml/basis.cm CML's version of $/basis.cm.
2788 :     $cml/cml-internal.cm Internal helper library.
2789 :     $cml/core-cml.cm Internal helper library.
2790 :     $cml-lib/trace-cml.cm Tracing facility.
2791 :     $cml-lib/smlnj-lib.cm CML's version of $/smlnj-lib.cm
2792 :    
2793 :     The installer (config/install.sh) has been taught how to properly
2794 :     install this stuff.
2795 :    
2796 :     ----------------------------------------------------------------------
2797 :     Name: Matthias Blume
2798 : blume 843 Date: 2001/06/19 17:55:00 EDT
2799 :     Tag: blume-20010619-instantiate
2800 :     Description:
2801 :    
2802 :     This un-breaks the fix for bug 1432.
2803 :     (The bug was originally fixed in 110.9 but I broke it again some
2804 :     time after that.)
2805 :    
2806 :     ----------------------------------------------------------------------
2807 :     Name: Matthias Blume
2808 : blume 842 Date: 2001/06/19 17:25:00 EDT
2809 :     Tag: blume-20010619-signals
2810 :     Description:
2811 :    
2812 :     This should (hopefully) fix the long-standing signal handling bug.
2813 :     (The runtime system was constructing a continuation record with an
2814 :     incorrect descriptor which would cause the GC to drop data on the floor...)
2815 :    
2816 :     ----------------------------------------------------------------------
2817 :     Name: Matthias Blume
2818 : blume 840 Date: 2001/06/15 15:05:00 EDT
2819 : blume 841 Tag: blume-20010615-moresparc
2820 :     Description:
2821 :    
2822 :     Here is a short late-hour update related to Sparc c-calls:
2823 :    
2824 :     -- made handling of double-word arguments a bit smarter
2825 :    
2826 :     -- instruction selection phase tries to collapse certain clumsily
2827 :     constructed ML-Trees; typical example:
2828 :    
2829 :     ADD(ty,ADD(_,e,LI d1),LI d2) -> ADD(ty,e,LI(d1+d2))
2830 :    
2831 :     This currently has no further impact on SML/NJ since mlriscGen does
2832 :     not seem to generate such patterns in the first place, and c-calls
2833 :     (which did generate them in the beginning) has meanwhile been fixed
2834 :     so as to avoid them as well.
2835 :    
2836 :     ----------------------------------------------------------------------
2837 :     Name: Matthias Blume
2838 :     Date: 2001/06/15 15:05:00 EDT
2839 : blume 840 Tag: blume-20010615-sparc
2840 :     Description:
2841 :    
2842 :     The purpose of this update is to provide an implementation of NLFFI
2843 :     on Sparc machines.
2844 :    
2845 :     Here are the changes in detail:
2846 :    
2847 :     * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
2848 :     the Sparc implementation of the c-calls API.
2849 :     * The Sparc backend of SML/NJ has been modified to uniformely use %fp
2850 :     for accessing the ML frame. Thus, we have a real frame pointer and
2851 :     can freely modify %sp without need for an omit-frame-ptr phase.
2852 :     The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
2853 :     this case.
2854 :     * ml-nlffigen has been taught to produce code for different architectures
2855 :     and calling conventions.
2856 :     * In a way similar to what was done in the x86 case, the Sparc
2857 :     backend uses its own specific extension to mltree. (For example,
2858 :     it needs to be able to generate UNIMP instructions which are part
2859 :     of the calling convention.)
2860 :     * ml-nlffi-lib was reorganized to make it more modular (in particular,
2861 :     to make it easier to plug in new machine- and os-dependent parts).
2862 :    
2863 :     There are some other fairly unrelated bug fixes and cleanups as well:
2864 :    
2865 :     * I further hacked the .cm files for MLRISC tools (like MDLGen) so
2866 :     that they properly share their libraries with existing SML/NJ libraries.
2867 :     * I fixed a minor cosmetic bug in CM, supressing certain spurious
2868 :     follow-up error messages.
2869 :     * Updates to CM/CMB documentation.
2870 :    
2871 :     TODO items:
2872 :    
2873 :     * MLRISC should use a different register as its asmTemp on the Sparc.
2874 :     (The current %o2 is a really bad choice because it is part of the
2875 :     calling conventions, so things might interfere in unexpected ways.)
2876 :    
2877 :     ----------------------------------------------------------------------
2878 :     Name: Matthias Blume
2879 : blume 839 Date: 2001/06/07
2880 :     Tag: blume-20010607-calls
2881 :     Description:
2882 :    
2883 :     A number of internal changes related to C calls and calling conventions:
2884 :    
2885 :     1. ML-Tree CALL statements now carry a "pops" field. It indicates the
2886 :     number of bytes popped implicitly (by the callee). In most cases
2887 :     this field is 0 but on x86/win32 it is some non-zero value. This
2888 :     is information provided for the benefit of the "omit-frameptr" pass.
2889 :     2. The CALL instruction on the x86 carries a similar "pops" field.
2890 :     The instruction selection phase copies its value from the ML-Tree
2891 :     CALL statement.
2892 :     3. On all other architectures, the instruction selection phase checks
2893 :     whether "pops=0" and complains if not.
2894 :     4. The c-calls implementation for x86 now accepts two calling conventions:
2895 :     "ccall" and "stdcall". When "ccall" is selected, the caller cleans
2896 :     up after the call and pops is set to 0. For "stdcall", the caller
2897 :     does nothing, leaving the cleanup to the callee; pops is set to
2898 :     the number of bytes that were pushed onto the stack.
2899 :     5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
2900 :     distinguish between "ccall" and "stdcall".
2901 :     6. The UNIMP instruction has been added to the supported Sparc instruction
2902 :     set. (This is needed for implementing the official C calling convention
2903 :     on this architecture.)
2904 :     7. I fixed some of the .cm files under src/MLRISC/Tools to make them
2905 :     work with the latest CM.
2906 :    
2907 :     ----------------------------------------------------------------------
2908 :     Name: Matthias Blume
2909 : blume 838 Date: 2001/06/05 15:10:00 EDT
2910 :     Tag: blume-20010605-cm-index
2911 :     Description:
2912 :    
2913 :     0. The "lambdasplit" parameter for class "sml" in CM has been documented.
2914 :    
2915 :     1. CM can now generate "index files". These are human-readable files
2916 :     that list on a per-.cm-file basis each toplevel symbol defined or
2917 :     imported. The location of the index file for
2918 :     <p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
2919 :     To enable index-file generation, set CM.Control.generate_index to true
2920 :     or export an environment-symbol: export CM_GENERATE_INDEX=true.
2921 :    
2922 :     The CM manual has been updated accordingly.
2923 :    
2924 :     2. I made some slight modifications to the c-calls API in MLRISC.
2925 :    
2926 :     a) There is now a callback to support saving/restoring of
2927 :     dedicated but caller-save registers around the actual call
2928 :     instruction.
2929 :     b) One can optionally specify a comment-annotation for the
2930 :     call instruction.
2931 :    
2932 :     3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
2933 :     (For example, the comment annotation shows the C prototype of
2934 :     the function being called.)
2935 :    
2936 :     ----------------------------------------------------------------------
2937 :     Name: Matthias Blume
2938 : blume 837 Date: 2001/06/01 13:30:00 EDT
2939 :     Tag: blume-20010601-nlffi-cleanup
2940 :     Description:
2941 :    
2942 :     This is mostly a cleanup of MLFFI stuff:
2943 :    
2944 :     - some signature files have been put into a more exposed place
2945 :     - the ugly 'f type parameter is gone (simplifies types tremendously!)
2946 :     - ml-nlffigen changed accordingly
2947 :     - tutorial updated
2948 :    
2949 :     Other changes:
2950 :    
2951 :     - author's affiliation in CM manual(s) updated
2952 :     - some more recognized keywords added to Allen's sml.sty
2953 :    
2954 :     ----------------------------------------------------------------------
2955 :     Name: Matthias Blume
2956 : blume 836 Date: 2001/05/25 15:30:00 EDT
2957 :     Tag: blume-20010525-iptr
2958 :     Description:
2959 :    
2960 :     - put the official 110.33-README (as it appears on the ftp server) under
2961 :     CVS
2962 :     - fixed a small bug related to incomplete pointer types in
2963 :     ml-nlffigen
2964 :     - small cosmetic change to the ml-nlffi-lib's "arr" type constructor
2965 :     (it does not need the 'f type parameter)
2966 :    
2967 :     ----------------------------------------------------------------------
2968 :     Name: Matthias Blume
2969 : blume 834 Date: 2001/05/23 14:30:00 EDT
2970 :     Tag: Release_110_33
2971 :     Description:
2972 :    
2973 :     New version number (110.33). New bootfiles.
2974 :    
2975 :     ----------------------------------------------------------------------
2976 :     Name: Matthias Blume
2977 : blume 833 Date: 2001/05/22 18:06:00 EDT
2978 :     Tag: blume-20010522-targets
2979 :     Description:
2980 :    
2981 :     Made install.sh use file config/targets.customized if it exists, falling
2982 :     back to config/targets if it doesn't. This way one can have a customized
2983 :     version of the targets file without touching the "real thing", thus
2984 :     eliminating the constant fear of accidentally checking something bogus
2985 :     back into the CVS repository... (File config/targets.customized must
2986 :     not be added to the repository!)
2987 :    
2988 :     ----------------------------------------------------------------------
2989 :     Name: Matthias Blume
2990 : blume 832 Date: 2001/05/22 16:30:00 EDT
2991 :     Tag: blume-20010522-minitut
2992 :     Description:
2993 :    
2994 :     1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
2995 :     struct returns.
2996 :     2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt. This is some very
2997 :     incomplete, preliminary documentation for NLFFI.
2998 :    
2999 :     ----------------------------------------------------------------------
3000 :     Name: Matthias Blume
3001 : blume 829 Date: 2001/05/14 11:30:00 EDT
3002 :     Tag: blume-20010514-script
3003 :     Description:
3004 :    
3005 :     Some bugs in install script fixed.
3006 :    
3007 :     In addition to that I also made a slight change to the NLFFI API:
3008 :     Functors generated by ml-nlffigen now take the dynamic library as a
3009 :     straight functor argument, not as a suspended one. (The original
3010 :     functor code used to force the suspension right away anyway, so there
3011 :     was nothing gained by this complication of the interface.)
3012 :    
3013 :     ----------------------------------------------------------------------
3014 :     Name: Matthias Blume
3015 : blume 828 Date: 2001/05/11 14:35:00 EDT
3016 :     Tag: blume-20010511-ml-nlffi
3017 :     Description:
3018 :    
3019 :     I finally took the plunge and added my new FFI code to the main
3020 :     repository. For x86-linux it is now ready for prime-time.
3021 :    
3022 :     There are two new subdirectories of "src":
3023 :    
3024 :     - ml-nlffi-lib:
3025 :     The utility library for programs using the FFI interface.
3026 :     Here is the implementation of $/c.cm and its associated low-level
3027 :     partners $/c-int.cm and $/memory.cm.
3028 :     - ml-nlffigen:
3029 :     A stand-alone program for generating ML glue code from C source
3030 :     code.
3031 :    
3032 :     Building ml-nlffigen requires $/ckit-lib.cm.
3033 :    
3034 :     The config/install.sh script has been updates to do the Right Thing
3035 :     (hopefully).
3036 :    
3037 :     Notice that the source tree for the C-Kit will not be put under "src"
3038 :     but directly under the installation root directory. (This is the
3039 :     structure that currently exists on the CVS server when you check out
3040 :     module "sml".) Fortunately, config/install.sh knows about this oddity.
3041 :    
3042 :     Bugs: No documentation yet.
3043 :    
3044 :     ----------------------------------------------------------------------
3045 :     Name: Matthias Blume
3046 : blume 826 Date: 2001/05/09 16:35:00 EDT
3047 :     Tag: blume-20010509-cpscontract
3048 :     Description:
3049 :    
3050 :     Fixed a bug in the accounting code in cpsopt/contract.sml. (The
3051 :     wrapper/unwrapper elimination did not decrement usage counts and some
3052 :     dead variables got overlooked by the dead-up logic.)
3053 :    
3054 :     ----------------------------------------------------------------------
3055 : george 824 Name: Lal George
3056 :     Date: 2001/05/08 17:26:09 EDT
3057 :     Tag: george-20010508-omit-frameptr
3058 :     Description:
3059 :    
3060 :     Changes to implement the omit-frame-pointer optimization to support
3061 :     raw C calls. For now, there is only support on the Intel x86, but
3062 :     other architectures will follow as more experience is gained with this.
3063 :    
3064 :    
3065 :     ----------------------------------------------------------------------
3066 : blume 818 Name: Matthias Blume
3067 : blume 819 Date: 2001/05/07 14:40:00 EDT
3068 :     Tag: blume-20010507-proxies
3069 :     Description:
3070 :    
3071 :     I made into "proxy libraries" all libraries that qualify for such a
3072 :     change. (A qualifying library is a library that has another library or
3073 :     groups as its sole member and repeats that member's export list
3074 :     verbatim. A proxy library avoids this repetition by omitting its export
3075 :     list, effectively inheriting the list that its (only) member exports.
3076 :     See the CM manual for more explanation.)
3077 :     The main effect is that explicit export lists for these libraries
3078 :     do not have to be kepts in sync, making maintenance a bit easier.
3079 :    
3080 :     I also added copyright notices to many .cm-files.
3081 :    
3082 :     Last but not least, I made a new set of bootfiles.
3083 :    
3084 :     ----------------------------------------------------------------------
3085 :     Name: Matthias Blume
3086 : blume 818 Date: 2001/05/04 17:00:00 EDT
3087 :     Tag: blume-20010504-cm-lsplit
3088 :     Description:
3089 :    
3090 :     0. John merged pending changes to $/smlnj-lib.cm
3091 :    
3092 :     1. Allen's previous change accidentally backed out of one of Lal's
3093 :     earlier changes. I undid this mistake (re-introducing Lal's change).
3094 :    
3095 :     2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
3096 :     within the compiler where applicable. There is some code simplification
3097 :     because of that.
3098 :    
3099 :     3. The "split" phase (in FLINT) is now part of the default list of phases.
3100 :     Compiler.Control.LambdaSplitting.* can be used to globally control the
3101 :     lambda-splitting (cross-module-inlining) engine. In addition to that,
3102 :     it can now also be controlled on a per-source basis: CM has been taught
3103 :     a new tool parameter applicable to ML source files.
3104 :    
3105 :     - To turn lambda-splitting off completely:
3106 :     local open Compiler.Control.LambdaSplitting in
3107 :     val _ = set Off
3108 :     end
3109 :     - To make "no lambda-splitting" the global default (but allow per-source
3110 :     overriding); this is the initial setting:
3111 :     local open Compiler.Control.LambdaSplitting in
3112 :     val _ = set (Default NONE)
3113 :     end
3114 :     - To make "lambda-splitting with aggressiveness a" the global default
3115 :     (and allow per-source overriding):
3116 :     local open Compiler.Control.LambdaSplitting in
3117 :     val _ = set (Default (SOME a))
3118 :     end
3119 :    
3120 :     - To turn lambda-splitting off for a given ML souce file (say: a.sml)
3121 :     write (in the respective .cm-file):
3122 :     a.sml (lambdasplitting:off)
3123 :     - To turn lambda-splitting for a.sml on with minimal aggressiveness:
3124 :     a.sml (lambdasplitting:on)
3125 :     - To turn lambda-splitting for a.sml on with aggressiveness <a> (where
3126 :     <a> is a decimal non-negative integer):
3127 :     a.sml (lambdasplitting:<a>)
3128 :     - To turn lambda-splitting for a.sml on with maximal aggressiveness:
3129 :     a.sml (lambdasplitting:infinity)
3130 :     - To use the global default for a.sml:
3131 :     a.sml (lambdasplitting:default)
3132 :     or simply
3133 :     a.sml
3134 :    
3135 :     ----------------------------------------------------------------------
3136 : leunga 815 Name: Allen Leung
3137 :     Date: 2001/05/04 01:57:00 EDT
3138 :     Tag: leunga-20010504-sync
3139 :     Description:
3140 : leunga 788
3141 : leunga 815 MLRISC features.
3142 :    
3143 :     1. Fix to CMPXCHG instructions.
3144 :     2. Changed RA interface to allow annotations in callbacks.
3145 :     3. Added a new method to the stream interface to allow annotations updates.
3146 :    
3147 : george 749 ----------------------------------------------------------------------
3148 : blume 812 Name: Matthias Blume
3149 : blume 814 Date: 2001/05/01 11:45:00 EDT
3150 :     Tag: blume-20010501-pcedittmp
3151 :     Description:
3152 :    
3153 :     Changed install.sh to use the current working directory instead of
3154 :     /usr/tmp for a temporary file (pcedittmp). The previous choice
3155 :     of /usr/tmp caused trouble with MacOS X because of file premission
3156 :     problems.
3157 :    
3158 :     ----------------------------------------------------------------------
3159 :     Name: Matthias Blume
3160 : blume 812 Date: 2001/04/20 11:10:00 EDT
3161 :     Tag: blume-20010420-inMLflag
3162 :     Description:
3163 :    
3164 :     - added vp_limitPtrMask to vproc-state.h
3165 :     (for use by the raw-C-calls mechanism to implement proper interrupt
3166 :     handling)
3167 :     - made the ML compiler aware of various data-structure offsets so it
3168 :     can generate code for accessing the vp_inML flag and vp_limitPtrMask
3169 :     - tweaked mlriscGen.sml to have it emit interrupt-handling code for
3170 :     raw C-calls
3171 :    
3172 :     ----------------------------------------------------------------------
3173 : george 810 Name: Lal George
3174 :     Date: 2001/04/20 09:15:28 EDT
3175 :     Tag: george-20010420-macosX
3176 :     Description:
3177 :    
3178 :     - Changes to port to Mac OS X; Darwin.
3179 :    
3180 :     - In the process I found that sqrt was broken on the PPC, because the
3181 :     fsqrt instruction is not implemented.
3182 :    
3183 :     ----------------------------------------------------------------------
3184 : blume 799 Name: Matthias Blume
3185 : blume 808 Date: 2001/04/18 12:45:00 EDT
3186 :     Tag: blume-20010418-ccalls
3187 :     Description:
3188 :    
3189 :     - fixed two off-by-4 errors in the x86-specific c-calls implementation
3190 :     (this bug prevented structure arguments containing pointers from being
3191 :     passed correctly)
3192 :     - changed the raw-C-call code in mlriscGen.sml in such a way that
3193 :     structure arguments are represented as a pointer to the beginning
3194 :     of the structure (instead of having a series of synthesized arguments,
3195 :     one for each structure member)
3196 :    
3197 :     - made makeml script's verbosity level configurable via environment
3198 :     variable (MAKEML_VERBOSITY)
3199 :    
3200 : blume 813 - eliminated placeholder implementations for f32l, w16s, i16s, and f32s
3201 : blume 808 in rawmem-x86.sml; we are now using the real thing
3202 :    
3203 :     ----------------------------------------------------------------------
3204 :     Name: Matthias Blume
3205 : blume 806 Date: 2001/03/22 16:25:00 EST
3206 :     Tag: blume-20010322-bootfiles
3207 :     Description:
3208 :    
3209 :     Created a new set of bootfiles (for your automatic installation convenience).
3210 :    
3211 :     ----------------------------------------------------------------------
3212 :     Name: Matthias Blume
3213 : blume 805 Date: 2001/03/22 15:10:00 EST
3214 :     Tag: blume-20010322-rawmem-parcm
3215 :     Description:
3216 :    
3217 :     1. All "raw memory access" primitives for the new FFI are implemented now
3218 :     (at least on the x86).
3219 :     2. Some further cleanup of CM's parallel make mechanism.
3220 :    
3221 :     ----------------------------------------------------------------------
3222 :     Name: Matthias Blume
3223 : blume 801 Date: 2001/03/19 17:53:00 EST
3224 :     Tag: blume-20010319-parallel
3225 :     Description:
3226 :    
3227 :     Parallel make (using compile servers) now works again.
3228 :    
3229 :     To this end, CM.stabilize and CMB.make have been modified to work in
3230 :     two passes when compile servers are attached:
3231 :     1. Compile everything, do not perform stabilization; this pass
3232 :     uses compile servers
3233 :     2. Stabilize everything; this pass does not use compile servers
3234 :     If there are no compile servers, the two passes are combined into one
3235 :     (as before). Splitting the passes increases the inherent parallelism
3236 :     in the dependency graph because the entire graph including all
3237 :     libraries is available at the same time. This, in turn, improves
3238 :     server utilization. The downside is that the master process will
3239 :     have to do some extra work after compilation is done (because for
3240 :     technical reasons it must re-read all the binfiles during stabilization).
3241 :    
3242 :     ----------------------------------------------------------------------
3243 :     Name: Matthias Blume
3244 : blume 800 Date: 2001/03/16 12:22:00 EST
3245 :     Tag: blume-20010316-bootfiles
3246 :     Description:
3247 :    
3248 :     Created a new set of bootfiles (for your automatic installation convenience).
3249 :    
3250 :     ----------------------------------------------------------------------
3251 :     Name: Matthias Blume
3252 : blume 799 Date: 2001/03/16 11:00:00 EST
3253 :     Tag: blume-20010316-MLTREE-fixup
3254 :     Description:
3255 :    
3256 :     This is a minor fixup for an (untagged) earlier commit by Allen.
3257 :     (A file was missing).
3258 :    
3259 :     ----------------------------------------------------------------------
3260 : leunga 788 Name: Allen Leung
3261 : leunga 796 Date: Mon Mar 5 18:54:57 EST 2001
3262 :     Tag: leunga-20010305-cut-support
3263 :    
3264 :     1. New support for alternative control-flow in MLTREE.
3265 :     Currently we support
3266 :    
3267 :     FLOW_TO(CALL ...., [k1,...,kn])
3268 :    
3269 :     This is needed for 'cuts to' in C-- and try/handle-like constructs
3270 :     in Moby
3271 :    
3272 :     New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
3273 :    
3274 :     2. Register Allocator
3275 :    
3276 :     Changes in interface [from Fermin, John]
3277 :    
3278 :     3. Alpha 8-bit SLL support [Fermin]
3279 :    
3280 :     4. All architectures
3281 :    
3282 :     A new module (ClusterExpandCopies) for expanding parallel copies.
3283 :    
3284 :     ----------------------------------------------------------------------
3285 :     Name: Allen Leung
3286 : leunga 788 Date: 2001/02/27 23:07:00 EST
3287 :     Tag: leunga-20010227-minor-stuff
3288 :    
3289 :     1. Alpha bug fix for CMOVNE
3290 :     2. Handle mltree COND(..,FCMP ...,...)
3291 :     3. Bug fix in simplifier
3292 :    
3293 :     ----------------------------------------------------------------------
3294 : blume 777 Name: Matthias Blume
3295 : blume 781 Date: 2001/01/30 17:50:00 EST
3296 :     Tag: blume-20010130-sync
3297 :     Description:
3298 :    
3299 :     This is just a minor update to sync my devel branch with the main brach.
3300 :     The only visible change is the addition of some README files.
3301 :    
3302 :     ----------------------------------------------------------------------
3303 :     Name: Matthias Blume
3304 : blume 778 Date: 2001/01/12 23:30:00 JST
3305 :     Tag: blume-20010112-bootfiles
3306 :     Description:
3307 :    
3308 :     Made a new set of bootfiles that goes with the current state of the
3309 :     repository.
3310 :    
3311 :     ----------------------------------------------------------------------
3312 :     Name: Matthias Blume
3313 : blume 777 Date: 2001/01/12 21:20:00 JST
3314 :     Tag: blume-20010112-sync
3315 :     Description:
3316 :    
3317 :     I am just flushing out some minor changes that had accumulated in
3318 :     my private branch in order to sync with the main tree. (This is
3319 :     mainly because I had CVS trouble when trying to merge _into_ my
3320 :     private branch.)
3321 :    
3322 :     Most people should be completely unaffected by this.
3323 :    
3324 :     ----------------------------------------------------------------------
3325 : leunga 775 Name: Allen Leung
3326 :     Date: Thu Jan 11 21:03:00 EST 2001
3327 :     Tag: leunga-20010111-labexp=mltree
3328 :     Description:
3329 :    
3330 :     1. Removed the type LabelExp and replace it by MLTree.
3331 :     2. Rewritten mltree-simplify with the pattern matcher tool.
3332 :     3. There were some bugs in alpha code generator which would break
3333 :     64-bit code generation.
3334 :     4. Redo the tools to generate code with the
3335 :     5. The CM files in MLRISC (and in src/system/smlnj/MLRISC)
3336 :     are now generated by perl scripts.
3337 :    
3338 :     ----------------------------------------------------------------------
3339 : blume 771 Name: Matthias Blume
3340 : blume 774 Date: 2001/01/10 21:55:00 JST
3341 :     Tag: blume-20010110-rcc
3342 :     Description:
3343 :    
3344 :     The RCC stuff now seems to work (but only on the x86).
3345 :     This required hacking of the c-calls interface (and -implementation) in
3346 :     MLRISC.
3347 :    
3348 :     Normal compiler users should be unaffected.
3349 :    
3350 :     ----------------------------------------------------------------------
3351 :     Name: Matthias Blume
3352 : blume 773 Date: 2001/01/09 01:20:00 JST
3353 :     Tag: blume-20010109-rcc
3354 :     Description:
3355 :    
3356 :     This is a fairly big patch, flushing out a large number of pending
3357 :     changes that I made to my development copy over the last couple of days.
3358 :    
3359 :     Of practical relevance at this moment is a workaround for a pickling
3360 :     bug that Allen ran into the other day. The cause of the bug itself is
3361 :     still unknown and it might be hard to fix it properly, but the
3362 :     workaround has some merits of its own (namely somewhat reducing pickling
3363 :     overhead for certain libraries). Therefore, I think this solution should
3364 :     be satisfactory at this time.
3365 :    
3366 :     The rest of the changes (i.e., the vast majority) has to do with my
3367 :     ongoing efforts of providing direct support for C function calls from
3368 :     ML. At the moment there is a new primop "RAW_CCALL", typing magic
3369 :     in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new
3370 :     case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to
3371 :     translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to
3372 :     handle RCC.
3373 :    
3374 :     The last part (the changes to mlriscGen.sml) are still known to be
3375 :     wrong on the x86 and not implemented on all other architectures. But
3376 :     the infrastructure is in place. I had to change a few functor
3377 :     signatures in the backend to be able to route the CCalls interface
3378 :     from MLRISC there, and I had to specialize the mltree type (on the
3379 :     x86) to include the necessary extensions. (The extensions themselves
3380 :     were already there and redy to go in MLRISC/x86).
3381 :    
3382 :     Everything should be very happy as soon as someone helps me with
3383 :     mlriscGen.sml...
3384 :    
3385 :     In any case, nothing of this should matter to anyone as long as the
3386 :     new primop is not being used (which is going to be the case unless you
3387 :     find it where I hid it :). The rest of the compiler is completely
3388 :     unaffected.
3389 :    
3390 :     ----------------------------------------------------------------------
3391 :     Name: Matthias Blume
3392 : blume 772 Date: 2001/01/05 00:30:00 JST
3393 :     Tag: blume-20010105-primops
3394 :     Description:
3395 :    
3396 :     Added some experimental support for work that I am doing right now.
3397 :     These changes mostly concern added primops, but there is also a new
3398 :     experimental C library in the runtime system (but currently not enabled
3399 :     anywhere except on Linux/X86).
3400 :    
3401 :     In the course of adding primops (and playing with them), I discovered that
3402 :     Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly
3403 :     broken. (Zhong was very right he labeled this stuff as "major gross hack".)
3404 :     To recover, I made type information in INL_PRIM mandatory and changed
3405 :     prim.sml as well as built-in.sml accordingly. The InLine structure now
3406 :     has complete, correct type information (i.e., no bottom types).
3407 :    
3408 :     Since all these changes mean that we need new binfiles, I also bumped the
3409 :     version number to 110.32.1.
3410 :    
3411 :     ----------------------------------------------------------------------
3412 :     Name: Matthias Blume
3413 : blume 771 Date: 2000/12/30 22:10:00 JST
3414 :     Tag: blume-20001230-various
3415 :     Description:
3416 :    
3417 :     Added proxy libraries for MLRISC and let MLRISC libraries refer
3418 :     to each other using path anchors. (See CM manual for explanation.)
3419 :    
3420 :     Updated CM documentation.
3421 :    
3422 :     Fixed some bugs in CM.
3423 :    
3424 :     Implemented "proxy" libraries (= syntactic sugar for CM).
3425 :    
3426 :     Added "-quiet" option to makeml and changed runtime system accordingly.
3427 :    
3428 :     Added cleanup handler for exportML to reset timers and compiler stats.
3429 :    
3430 :     ----------------------------------------------------------------------
3431 : george 761 Name: Lal George
3432 :     Date: 2000/12/22 22:22:58 EST 2000
3433 :     Tag: Release_110_32
3434 :     Description:
3435 :    
3436 :     Infinite precision used throughout MLRISC.
3437 :     see MLRISC/mltree/machine-int.sig
3438 :    
3439 :     ----------------------------------------------------------------------
3440 : blume 756 Name: Matthias Blume
3441 : blume 760 Date: 2000/12/22 23:16:00 JST
3442 :     Tag: blume-20001222-warn
3443 :     Description:
3444 : blume 759
3445 : blume 760 Corrected wording and formatting of some CM warning message which I
3446 :     broke in my previous patch.
3447 : blume 759
3448 :     ----------------------------------------------------------------------
3449 :     Name: Matthias Blume
3450 : blume 763 Date: 2000/12/22 21:20:00 JST
3451 :     Tag: blume-20001222-anchorenv
3452 :     Description:
3453 :    
3454 :     Fixed CM's handling of anchor environments in connection with CMB.make.
3455 :    
3456 :     ----------------------------------------------------------------------
3457 :     Name: Matthias Blume
3458 : blume 758 Date: 2000/12/22 13:15:00 JST
3459 :     Tag: blume-20001222-cleanup
3460 :     Description:
3461 :    
3462 :     Removed src/cm/ffi which does not (and did not) belong here.
3463 :    
3464 :     ----------------------------------------------------------------------
3465 :     Name: Matthias Blume
3466 : blume 757 Date: 2000/12/21 23:55:00 JST
3467 :     Tag: blume-20001221-exn
3468 :     Description:
3469 :    
3470 :     Probably most important: CM no longer silently swallows all exceptions
3471 :     in the compiler.
3472 :     Plus: some other minor CM changes. For example, CM now reports some
3473 :     sizes for generated binfiles (code, data, envpickle, lambdapickle).
3474 :    
3475 :     ----------------------------------------------------------------------
3476 :     Name: Matthias Blume
3477 : blume 756 Date: 2000/12/15 00:01:05 JST
3478 :     Tag: blume-20001215-dirtool
3479 :     Description:
3480 :    
3481 :     - "dir" tool added.
3482 :     - improvements and cleanup to Tools structure
3483 :     - documentation updates
3484 :    
3485 :     ----------------------------------------------------------------------
3486 : leunga 752 Name: Allen Leung
3487 : leunga 755 Date: Thu Dec 14 03:45:24 EST 2000
3488 :     Description:
3489 :     Tag: leunga-20001214-int-inf
3490 :     Description:
3491 :    
3492 :     In IntInf, added these standard functions, which are missing from our
3493 :     implementation:
3494 :    
3495 :     andb : int * int -> int
3496 :     xorb : int * int -> int
3497 :     orb : int * int -> int
3498 :     notb : int -> int
3499 :     << : int * word -> int
3500 :     ~>> : int * word -> int
3501 :    
3502 :     Not tested, I hope they are correct.
3503 :    
3504 :     ----------------------------------------------------------------------
3505 :     Name: Allen Leung
3506 : leunga 752 Date: Fri Dec 8 19:23:26 EST 2000
3507 :     Description:
3508 :     Tag: leunga-20001208-nowhere
3509 :     Description:
3510 :    
3511 :     Slight improvements to the 'nowhere' tool to handle OR-patterns,
3512 :     to generate better error messages etc. Plus a brief manual.
3513 :    
3514 :     ----------------------------------------------------------------------
3515 : george 749 Name: Lal George
3516 :     Date: 2000/12/08 09:54:02 EST 2000
3517 :     Tag: Release_110_31
3518 :     Description:
3519 : george 705
3520 : george 749 - Version 110.31
3521 : leunga 591 ----------------------------------------------------------------------
3522 : leunga 744 Name: Allen Leung
3523 :     Date: Thu Dec 7 22:01:04 EST 2000
3524 :     Tag: leunga-20001207-cell-monster-hack
3525 :     Description:
3526 :    
3527 :     Major MLRISC internal changes. Affect all clients.
3528 :     Summary:
3529 :    
3530 :     1. Type CELLS.cell = int is now replaced by a datatype.
3531 :     As a result, the old regmap is now gone. Almost all interfaces
3532 :     in MLRISC change as a consequence.
3533 :    
3534 :     2. A new brand version of machine description tool (v3.0) that generates
3535 :     modules expecting the new interface. The old version is removed.
3536 :    
3537 :     3. The RA interface has been further abstracted into two new functors.
3538 :     RISC_RA and X86RA. These functors have much simpler interfaces.
3539 :     [See also directory MLRISC/demo.]
3540 :    
3541 :     4. Some other new source->source code generation tools are available:
3542 :    
3543 :     a. MLRISC/Tools/RewriteGen -- generate rewriters from rules.
3544 :     b. MLRISC/Tools/WhereGen -- expands conditional pattern matching rules.
3545 :     I use this tool to generate the peephole optimizers---with the new
3546 :     cell type changes, peephole rules are becoming difficult to write
3547 :     without conditional pattern matching.
3548 :    
3549 :     5. More Intmap -> IntHashTable change. Previous changes by Matthias didn't
3550 :     cover the entire MLRISC source tree so many things broke.
3551 :    
3552 :     6. CM files have been moved to the subdirectory MLRISC/cm.
3553 :     They are moved because there are a lot of them and they clutter up the
3554 :     root dir.
3555 :    
3556 :     7. More detailed documentation to come...
3557 :    
3558 :     NOTE: To rebuild from 110.30 (ftp distribution), you'll have to do
3559 :     a makeml -rebuild first. This is because of other other
3560 :     changes that Matthias has made (see below).
3561 :    
3562 :    
3563 :     ----------------------------------------------------------------------
3564 : blume 733 Name: Matthias Blume
3565 : blume 742 Date: 2000/11/30 23:12:00 JST
3566 :     Tag: blume-20001130-filereorg
3567 :     Description:
3568 :    
3569 :     Some manual updates and some file reorganizations in CM.
3570 :    
3571 :     ----------------------------------------------------------------------
3572 :     Name: Matthias Blume
3573 : blume 737 Date: 2000/11/24 17:45:00 JST
3574 :     Tag: blume-20001124-link
3575 :     Description:
3576 :    
3577 :     Drastically improved link traversal code for the case that the dynamic
3578 :     value was already loaded at bootstrap time. As a result, CM and CMB
3579 :     now both load blazingly fast -- even on a very slow machine. Also,
3580 :     memory consumption has been further reduced by this.
3581 :    
3582 :     Warning: The format of the PIDMAP file has changed. THerefore, to
3583 :     bootstrap you have to do this:
3584 :    
3585 :     1. Run CMB.make
3586 :     2. Make a symbolic link for the boot directory:
3587 :     ln -s sml.boot.ARCH-OS xxx
3588 :     3. "Rebuild" the boot directory:
3589 :     ./makeml -boot xxx -rebuild sml ; rm xxx
3590 :     4. Boot normally:
3591 :     ./makeml
3592 :    
3593 :     ----------------------------------------------------------------------
3594 :     Name: Matthias Blume
3595 : blume 735 Date: 2000/11/21 21:20:00 JST
3596 :     Tag: blume-20001121-tools
3597 :     Description:
3598 :    
3599 :     Continued hacking on autoloading problem -- with success this time.
3600 :     Also changed tool-plugin mechanism. See new CM manual.
3601 :    
3602 :     ----------------------------------------------------------------------
3603 :     Name: Matthias Blume
3604 : blume 734 Date: 2000/11/19 14:30:00 JST
3605 :     Tag: blume-20001119-autoload
3606 :     Description:
3607 :    
3608 :     Some hacking to make autoloading faster. Success for CMB, no success
3609 :     so far for CM. There is a reduced structure CM' that autoloads faster.
3610 :     (This is a temporary, non-documented hack to be eliminated again when
3611 :     the general problem is solved.)
3612 :    
3613 :     ----------------------------------------------------------------------
3614 :     Name: Matthias Blume
3615 : blume 733 Date: 2000/11/17 14:10:00 JST
3616 :     Tag: blume-20001117-pickle-lib
3617 :     Description:
3618 :    
3619 :     1. Eliminated comp-lib.cm
3620 :     2. Made pickle-lib.cm
3621 :     3. Eliminated all uses of intset.sml (from comp-lib.cm)
3622 :     4. Replaced all uses of intmap.{sig,sml} (from comp-lib.cm) with
3623 :     equivalent constructs from smlnj-lib.cm (INtHashTable).
3624 :     5. Point 4. also goes for those uses of intmap.* in MLRISC.
3625 :     Duplicated intmap modules thrown out.
3626 :     6. Hunted down all duplicated SCC code and replaced it with
3627 :     equivalent stuff (GraphSCCFn from smlnj-lib.cm).
3628 :     7. Rewrote Feedback module.
3629 :     8. Moved sortedlist.sml into viscomp-lib.cm. Eventually it
3630 :     should be thrown out and equivalent modules from smlnj-lib.cm
3631 :     should be used (IntRedBlackSet, IntListSet, ...).
3632 :    
3633 :     Confirmed that compiler compiles to fixpoint.
3634 :    
3635 :     ----------------------------------------------------------------------
3636 : leunga 731 Name: Allen Leung
3637 :     Date: 2000/11/10 18:00:00
3638 :     Tag: leunga-20001110-new-x86-fp
3639 :    
3640 :     A new x86 floating point code generator has been added.
3641 :     By default this is turned off. To turn this on, do:
3642 :    
3643 :     CM.autoload "$smlnj/compiler.cm";
3644 :     Compiler.Control.MLRISC.getFlag "x86-fast-fp" := true;
3645 :    
3646 :     Changes:
3647 :    
3648 :     1. Changed FTAN to FPTAN so that the assembly output is correct.
3649 :     2. Changed the extension callback for FTANGENT to generate:
3650 :    
3651 :     fptan
3652 :     fstp %st(0)
3653 :     instead of
3654 :     fptan
3655 :     fstpl ftempmem
3656 :    
3657 :     3. Numerous assembly fixes for x86.
3658 :    
3659 :     5. Cleaned up the machine code output module x86/x86MC.sml and added
3660 :     support for a whole bunch of instructions and addressing modes:
3661 :    
3662 :     fadd/fsub/fsubr/fmul/fdiv/fdivr %st, %st(n)
3663 :     faddp/fsubp/fsubrp/fmulp/fdivp/fdivrp %st, %st(n)
3664 :     fadd/fsub/fsubr/fmul/fdiv/fdivr %st(n), %st
3665 :     fiadd/fisub/fisubr/fimul/fidiv/fidivr mem
3666 :     fxch %st(n)
3667 :     fld %st(n)
3668 :     fst %st(n)
3669 :     fst mem
3670 :     fstp %st(n)
3671 :     fucom %st(n)
3672 :     fucomp %st(n)
3673 :    
3674 :     All these are now generated when the fast fp mode is turned on.
3675 :    
3676 :     6. Removed the dedicated registers %st(0), ..., %st(7) from X86CpsRegs
3677 :    
3678 :     ----------------------------------------------------------------------
3679 : blume 729 Name: Matthias Blume
3680 :     Date: 2000/11/09 11:20:00 JST
3681 :     Tag: blume-20001109-scc
3682 :     Description:
3683 :    
3684 :     Eliminated some code duplication:
3685 :    
3686 :     1. Added "where" clause to GraphSCCFn in SML/NJ Library.
3687 :     (Otherwise the functor is useless.)
3688 :     2. Used GraphSCCFn where SCCUtilFun was used previously.
3689 :     3. Got rid of SCCUtilFun (in comp-lib.cm).
3690 :    
3691 :     ----------------------------------------------------------------------
3692 : george 721 Name: Lal George
3693 :     Date: 2000/11/06 09:02:21 EST 2000
3694 :     Tag: Release_110_30
3695 :     Description:
3696 :    
3697 :     - Version 110.30
3698 :     ----------------------------------------------------------------------
3699 : blume 715 Name: Matthias Blume
3700 : blume 716 Date: 2000/11/04 14:45:00
3701 :     Tag: blume-20001104-mlbuild
3702 :     Description:
3703 :    
3704 :     - Made ml-build faster on startup.
3705 :     - Documentation fixes.
3706 :    
3707 :     ----------------------------------------------------------------------
3708 :     Name: Matthias Blume
3709 : blume 715 Date: 2000/11/02 17:00:00 JST
3710 :     Tag: blume-20001102-condcomp
3711 :     Description:
3712 :    
3713 :     - Small tweaks to pickler -- new BOOTFILES!
3714 :     - Version bumped to 110.29.2.
3715 :     - Added conditional compilation facility to init.cmi (see comment there).
3716 :     ----------------------------------------------------------------------
3717 : leunga 713 Name: Allen Leung
3718 :     Date: 2000/10/23 19:31:00
3719 :     Tag: leunga-20001023-demo-ra
3720 :    
3721 :     1. Minor RA changes that improves spilling on x86 (affects Moby and C-- only)
3722 :     2. Test programs for the graph library updated
3723 :     3. Some new MLRISC demo programs added
3724 :    
3725 :     ----------------------------------------------------------------------
3726 : blume 710 Name: Matthias Blume
3727 : blume 711 Date: 2000/08/31 22:15:00 JST
3728 :     Tag: blume-20001017-errmsg
3729 :     Description:
3730 :    
3731 :     More error message grief: Where there used to be no messages, there
3732 :     now were some that had bogus error regions. Fixed.
3733 :    
3734 :     ----------------------------------------------------------------------
3735 :     Name: Matthias Blume
3736 : blume 710 Date: 2000/08/31 17:30:00 JST
3737 :     Tag: blume-20001017-v110p29p1
3738 :     Description:
3739 :    
3740 :     I made a version 110.29.1 with new bootfiles.
3741 :    
3742 :     Changes: Modified pickler/unpickler for faster and leaner unpickling.
3743 :     CM documentation changes and a small bugfix in CM's error reporting.
3744 :    
3745 :     ----------------------------------------------------------------------
3746 : george 705 Name: Lal George
3747 :     Date: 2000/09/27 14:42:35 EDT
3748 :     Tag: george-20000927-nodestatus
3749 :     Description:
3750 :    
3751 :     Changed the type of the nodestatus, so that:
3752 :    
3753 :     SPILLED(~1) is now SPILLED
3754 :     SPILLED(m) where m>=0 is now MEMREG(m)
3755 :     SPILLED(s) where s<~1 is now SPILL_LOC(~s)
3756 :    
3757 :     ----------------------------------------------------------------------
3758 : blume 697 Name: Matthias Blume
3759 : blume 703 Date: 2000/09/07 14:45:00 JST
3760 :     Tag: blume-20000907-cmerrmsg
3761 :     Description:
3762 :    
3763 :     Small tweak to CM to avoid getting ML syntax error messages twice.
3764 :    
3765 :     ----------------------------------------------------------------------
3766 :     Name: Matthias Blume
3767 : blume 701 Date: 2000/08/31 18:00:00 JST
3768 :     Tag: blume-20000831-cvsbootfiles
3769 :     Description:
3770 :    
3771 :     New URL for boot files (because the 110.29 files on the BL server do
3772 :     now work correctly with my updated install scripts for yacc and lex).
3773 :    
3774 :     ----------------------------------------------------------------------
3775 :     Name: Matthias Blume
3776 : blume 697 Date: 2000/08/08 12:33:00 JST
3777 :     Tag: blume-20000808-manual
3778 :     Description:
3779 :    
3780 :     Tiny update to CM manual.
3781 :    
3782 :     ----------------------------------------------------------------------
3783 : leunga 695 Name: Allen Leung
3784 :     Date: 2000/08/7 19:31:00
3785 :     Tag: leunga-20000807-a-whole-bunch-of-stuff
3786 :    
3787 :     Moby, C--, SSA, x86, machine descriptions etc. Should only affect C--
3788 :     and Mobdy.
3789 :    
3790 :     1. x86
3791 :    
3792 :     a. Fixes to peephole module by John and Dan.
3793 :     b. Assembly fix to SETcc by Allen.
3794 :     c. Fix to c-call by John.
3795 :     d. Fix to spilling by John. (This one deals with the missing FSTPT case)
3796 :     e. Instruction selection optimization to SETcc as suggested by John.
3797 :    
3798 :     For example,
3799 :    
3800 :     MV(32, x, COND(32, CMP(32, LT, a, b), LI 1, LI 0))
3801 :    
3802 :     should generate:
3803 :    
3804 :     MOVL a, x
3805 :     SUBL b, x
3806 :     SHRL 31, x
3807 :    
3808 :     2. IR stuff
3809 :    
3810 :     A bunch of new DJ-graph related algorithms added. These
3811 :     speed up SSA construction.
3812 :    
3813 :     3. SSA + Scheduling
3814 :    
3815 :     Added code for SSA and scheduling to the repository
3816 :    
3817 :     ----------------------------------------------------------------------
3818 : george 694 Name: Lal George
3819 :     Date: 2000/07/27 11:53:14 EDT
3820 :    
3821 :     Tag: lal-20000727-linux-ppc
3822 :     Description:
3823 :    
3824 :     Made changes to support Linux PPC.
3825 :     p.s. I have confirmation that the 110.29 boot files work fine.
3826 :    
3827 :     ----------------------------------------------------------------------
3828 : blume 684 Name: Matthias Blume
3829 : blume 692 Date: 2000/07/27 17:40:00 JST
3830 :     Tag: blume-20000727-scripts
3831 :     Description:
3832 :    
3833 :     !!!! WARNING !!!!
3834 :     You must recompile the runtime system!
3835 :     !!!! WARNING !!!!
3836 :    
3837 :     This is basically another round of script-enhancements:
3838 :    
3839 :     1. sml, ml-build, and ml-makedepend accept options -D and -U to define
3840 :     and undefine CM preprocessor symbols.
3841 :    
3842 :     2. ml-build avoids generating a new heap image if it finds that the
3843 :     existing one is still ok. (The condition is that no ML file had to
3844 :     be recompiled and all ML files are found to be older that the heap
3845 :     file.)
3846 :    
3847 :     To make this work smoothly, I also hacked the runtime system as
3848 :     well as SMLofNJ.SysInfo to get access to the heap image suffix
3849 :     (.sparc-solaris, ...) that is currently being used.
3850 :    
3851 :     Moreover, the signature of CM.mk_standalone has changed. See the
3852 :     CM manual.
3853 :    
3854 :     3. ml-makedepend accepts additional options -n, -a, and -o. (See the
3855 :     CM manual for details.)
3856 :    
3857 :     4. More CM manual updates:
3858 :     - all of the above has been documented.
3859 :     - there is now a section describing the (CM-related) command line
3860 :     arguments that are accepted by the "sml" command
3861 :    
3862 :     ----------------------------------------------------------------------
3863 :     Name: Matthias Blume
3864 : blume 691 Date: 2000/07/25 16:20:00 JST
3865 :     Tag: blume-20000725-makedepend
3866 :     Description:
3867 :    
3868 :     Added a script called ml-makedepend. This can be used in makefiles
3869 :     for Unix' make in a way very similar to the "makedepend" command for
3870 :     C.
3871 :    
3872 :     The script internally uses function CM.sources.
3873 :    
3874 :     Synopsis:
3875 :    
3876 :     ml-makedepend [-f makefile] cmfile targetname
3877 :    
3878 :     The default for the makefile is "makefile" (or "Makefile" should
3879 :     "makefile" not exist).
3880 :    
3881 :     ml-makedepend adds a cmfile/targetname-specific section to this
3882 :     makefile (after removing the previous version of this section). The
3883 :     section contains a single dependency specification with targetname on
3884 :     the LHS (targetname is an arbitrary name), and a list of files derived
3885 :     from the cmfile on the RHS. Some of the files on the RHS are
3886 :     ARCH/OPSYS-specific. Therefore, ml-makedepend inserts references to
3887 :     "make" variables $(ARCH) and $(OPSYS) in place of the corresponding
3888 :     path names. The makefile writer is responsible for making sure that
3889 :     these variables have correct at the time "make" is invoked.
3890 :    
3891 :     ----------------------------------------------------------------------
3892 :     Name: Matthias Blume
3893 : blume 690 Date: 2000/07/22 23:30:00 JST
3894 :     Tag: blume-20000722-urlupdate
3895 :     Description:
3896 :    
3897 :     Changed BOOT and config/srcarchiveurl to point to BL server:
3898 :    
3899 :     ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.29/
3900 :    
3901 :     ----------------------------------------------------------------------
3902 :     Name: Matthias Blume
3903 : blume 686 Date: 2000/07/18 18:00:00 JST
3904 :     Tag: blume-20000718-Version_110_29
3905 :     Description:
3906 :    
3907 :     1. Updated src/compiler/TopLevel/main/version.sml to version 110.29
3908 :    
3909 :     2. Updated config/version to 110.29
3910 :    
3911 :     3. Updated config/srcarchiveurl
3912 :    
3913 :     3. New boot files!
3914 :     ftp://ftp.cs.princeton.edu/pub/people/blume/sml/110.29-autofetch
3915 :    
3916 :     ----------------------------------------------------------------------
3917 :     Name: Matthias Blume
3918 : blume 684 Date: 2000/07/11 13:58:00 JST
3919 :     Tag: blume-20000711-doctypo
3920 :     Description:
3921 :    
3922 :     Fixed a few typos in CM manual.
3923 :    
3924 :     ----------------------------------------------------------------------
3925 : leunga 683 Name: Allen Leung
3926 :     Date: 2000/06/15 00:38:00
3927 :     Tag: leunga-20000704-sparc-x86
3928 :    
3929 :     1. x86 peephole improvement sp += k; sp -= k => nop [from John]
3930 :     2. fix to x86 RET bug [found by Dan Grossman]
3931 :     3. sparc assembly bug fix for ticc instructions [found by Fermin]
3932 :    
3933 :     Affects c-- and moby only
3934 :    
3935 :     ----------------------------------------------------------------------
3936 : blume 680 Name: Matthias Blume
3937 : blume 682 Date: 2000/07/04 15:26:00
3938 :     Tag: blume-20000704-trigger
3939 :     Description:
3940 :    
3941 :     1. Improvements to CM manual.
3942 :     2. SMLofNJ.Internals.BTrace.trigger reinstated as an alternative way
3943 :     of getting a back-trace. The function, when called, raises an
3944 :     internal exception which explicitly carries the full back-trace history,
3945 :     so it is unaffected by any intervening handle-raise pairs ("trivial"
3946 :     or not). The interactive loop will print that history once it arrives
3947 :     at top level.
3948 :     Short of having all exceptions implicitly carry the full history, the
3949 :     recommended way of using this facility is:
3950 :     - compile your program with instrumentation "on"
3951 :     - run it, when it raises an exception, look at the history
3952 :     - if the history is "cut off" because of some handler, go and modify
3953 :     your program so that it explicitly calls BTrace.trigger
3954 :     - recompile (still instrumented), and rerun; look at the full history
3955 :    
3956 :     ----------------------------------------------------------------------
3957 :     Name: Matthias Blume
3958 :     Date: 2000/07/03 15:36:00 JST
3959 : blume 680 Tag: blume-20000702-manual
3960 :     Description:
3961 :    
3962 :     Small corrections and updates to CM manual.
3963 :    
3964 :     ----------------------------------------------------------------------
3965 : blume 682 Name: Matthias Blume
3966 : blume 679 Date: 2000/06/29 16:04:00 JST
3967 :     Tag: blume-20000629-yacctool
3968 :     Description:
3969 :    
3970 :     Changes:
3971 :    
3972 :     1. Class "mlyacc" now takes separate arguments to pass options to
3973 :     generated .sml- and .sig-files independently.
3974 :     2. Corresponding CM manual updates.
3975 :     3. BTrace module now also reports call sites. (However, for loop clusters
3976 :     it only shows from where the cluster was entered.) There are associated
3977 :     modifications to core.sml, internals.{sig,sml}, btrace.sml, and btimp.sml.
3978 :    
3979 :     ----------------------------------------------------------------------
3980 : blume 675 Name: Matthias Blume
3981 : blume 678 Date: 2000/06/27 16:51:00 JST
3982 :     Tag: blume-20000627-noweb
3983 :     Description:
3984 :    
3985 :     Changes:
3986 :    
3987 :     1. Implemented "subdir" and "witness" options for noweb tool.
3988 :     This caused some slight internal changes in CM's tool implementation.
3989 :     2. Fixed bug in "tool plugin" mechanism. This is essentially cleaning
3990 :     some remaining issues from earlier path anchor changes.
3991 :     3. Updated CM manual accordingly.
3992 :    
3993 :     4. Changed implementation of back-tracing so that I now consider it
3994 :     ready for prime-time.
3995 :    
3996 :     In particular, you don't have to explicitly trigger the back-trace
3997 :     anymore. Instead, if you are running BTrace-instrumented code and
3998 :     there is an uncaught exception (regardless of whether or not it was
3999 :     raised in instrumented code), the top-level evalloop will print
4000 :     the back-trace.
4001 :    
4002 :     Features:
4003 :    
4004 :     - Instrumented and uninstrumented code work together seemlessly.
4005 :     (Of course, uninstrumented code is never mentioned in actual
4006 :     back-traces.)
4007 :    
4008 :     - Asymptotic time- and space-complexity of instrumented code is
4009 :     equal to that of uninstrumented code. (This means that
4010 :     tail-recursion is preserved by the instrumentation phase.)
4011 :    
4012 :     - Modules whose code has been instrumented in different sessions
4013 :     work together without problem.
4014 :    
4015 :     - There is no penalty whatsoever on uninstrumented code.
4016 :    
4017 :     - There is no penalty on "raise" expressions, even in
4018 :     instrumented code.
4019 :    
4020 :     A potential bug (or perhaps it is a feature, too):
4021 :    
4022 :     A back-trace reaches no further than the outermost instrumented
4023 :     non-trivial "raise". Here, a "trivial" raise is one that is the
4024 :     sole RHS of a "handle" rule. Thus, back-traces reach trough
4025 :    
4026 :     <exp> handle e => raise e
4027 :    
4028 :     and even
4029 :    
4030 :     <exp> handle Foo => raise Bar
4031 :    
4032 :     and, of course, through
4033 :    
4034 :     <exp> handle Foo => ...
4035 :    
4036 :     if the exception was not Foo.
4037 :    
4038 :     Back-traces always reach right through any un-instrumented code
4039 :     including any of its "handle" expressions, trivial or not.
4040 :    
4041 :     To try this out, do the following:
4042 :    
4043 :     - Erase all existing binfiles for your program.
4044 :     (You may keep binfiles for those modules where you think you
4045 :     definitely don't need back-tracing.)
4046 :     - Turn on back-trace instrumentation:
4047 :     SMLofNJ.Internals.BTrace.mode (SOME true);
4048 :     - Recompile your program. (I.e., run "CM.make" or "use".)
4049 :     - You may now turn instrumentation off again (if you want):
4050 :     SMLofNJ.Internals.BTrace.mode (SOME false);
4051 :     - Run your program as usual. If it raises an exception that
4052 :     reaches the interactive toplevel, then a back-trace will
4053 :     automatically be printed. After that, the toplevel loop
4054 :     will print the exception history as usual.
4055 :    
4056 :     ----------------------------------------------------------------------
4057 :     Name: Matthias Blume
4058 : blume 677 Date: 2000/06/26 09:56:46 JST
4059 :     Tag: blume-20000626-setup
4060 :     Description:
4061 :    
4062 :     CM: - setup-parameter to "sml" added; this can be used to run arbitrary
4063 :     ML code before and after compiling a file (e.g., to set compiler
4064 :     flags)
4065 :    
4066 :     Compiler: - improved btrace API (in core.sml, internals.{sig,sml})
4067 :     - associated changes to btrace.sml (BTrace instrumentation pass)
4068 :     - cleaner implementation of btimp.sml (BTrace tracing and report
4069 :     module)
4070 :    
4071 :     CM manual: * new path encoding documented
4072 :     * description of setup-parameter to "sml" added
4073 :    
4074 :     The biggest user-visible change to back-tracing is that it is no
4075 :     longer necessary to compile all traced modules within the same
4076 :     session. (This was a real limitation.)
4077 :    
4078 :     ----------------------------------------------------------------------
4079 :     Name: Matthias Blume
4080 : blume 676 Date: 2000/06/24 12:40:00 JST
4081 :     Tag: blume-20000624-startup
4082 :     Description:
4083 :    
4084 :     Fixes startup slowdown problem. (I was calling SrcPath.sync a _tad_
4085 :     bit too often -- to put it mildly. :)
4086 :    
4087 :     ----------------------------------------------------------------------
4088 :     Name: Matthias Blume
4089 : blume 675 Date: 2000/06/23 18:20:00 JST
4090 :     Tag: blume-20000623-btrace
4091 :     Description:
4092 :    
4093 :     This updates adds a backtrace facility to aid programmers in debugging
4094 :     their programs. This involves the following changes:
4095 :    
4096 :     1. Module system/smlnj/init/core.sml (structure _Core) now has hooks for
4097 :     keeping track of the current call stack. When programs are compiled
4098 :     in a special mode, the compiler will insert calls to these hooks
4099 :     into the user program.
4100 :     "Hook" means that it is possible for different implementations of
4101 :     back-tracing to register themselves (at different times).
4102 :    
4103 :     2. compiler/MiscUtil/profile/btrace.sml implements the annotation phase
4104 :     as an Absyn.dec->Absyn.dec rewrite. Normally this phase is turned off.
4105 :     It can be turned on using this call:
4106 :     SMLofNJ.Internals.BTrace.mode (SOME true);
4107 :     Turning it off again:
4108 :     SMLofNJ.Internals.BTrace.mode (SOME false);
4109 :     Querying the current status:
4110 :     SMLofNJ.Internals.BTrace.mode NONE;
4111 :     Annotated programs are about twice as big as normal ones, and they
4112 :     run a factor of 2 to 4 slower with a dummy back-trace plugin (one
4113 :     where all hooks do nothing). The slowdown with a plugin that is
4114 :     actually useful (such as the one supplied by default) is even greater,
4115 :     but in the case of the default plugin it is still only an constant
4116 :     factor (amortized).
4117 :    
4118 :     3. system/Basis/Implementation/NJ/internals.{sig,sml} have been augmented
4119 :     with a sub-structure BTrace for controlling back-tracing. In particular,
4120 :     the above-mentioned function "mode" controls whether the annotation
4121 :     phase is invoked by the compiler. Another important function is
4122 :     "trigger": when called it aborts the current execution and causes
4123 :     the top-level loop to print a full back-trace.
4124 :    
4125 :     4. compiler/MiscUtil/profile/btimp.sml is the current default plugin
4126 :     for back-tracing. It keeps track of the dynamic call stack and in
4127 :     addition to that it keeps a partial history at each "level" of that
4128 :     stack. For example, if a tail-calls b, b tail-calls c, and c tail-calls
4129 :     d and b (at separate times, dynamically), then the report will show:
4130 :    
4131 :     GOTO d
4132 :     /c
4133 :     GOTO \b
4134 :     CALL a
4135 :    
4136 :     This shows that there was an initial non-tail call of a, then a
4137 :     tail-call to b or c, looping behavior in a cluster of functions that
4138 :     consist of b and c, and then a goto from that cluster (i.e., either from
4139 :     b or from c) to d.
4140 :    
4141 :     Note that (depending on the user program) the amount of information
4142 :     that the back-trace module has to keep track of at each level is bounded
4143 :     by a constant. Thus, the whole implementation has the same asymptotical
4144 :     complexity as the original program (both in space and in time).
4145 :    
4146 :     5. compiler/TopLevel/interact/evalloop.sml has been modified to
4147 :     handle the special exception SMLofNJ.Internals.BTrace.BTrace
4148 :     which is raised by the "trigger" function mentioned above.
4149 :    
4150 :     Notes on usage:
4151 :    
4152 :     - Annotated code works well together with unannotated code:
4153 :     Unannotated calls simply do not show up at all in the backtrace.
4154 :    
4155 :     - It is not a good idea to let modules that were annotated during
4156 :     different sessions run at the same time. This is because the compiler
4157 :     chooses small integers to identify individual functions, and there
4158 :     will be clashes if different modules were compiled in separate sessions.
4159 :     (Nothing will crash, and you will even be told about the clashes, but
4160 :     back-trace information will in general not be useful.)
4161 :    
4162 :     - Back-tracing can be confused by callcc and capture.
4163 :    
4164 :     - The only way of getting a back-trace right now is to explicitly
4165 :     invoke the "trigger" function from your user program. Eventually, we
4166 :     should make every exception carry back-trace information (if
4167 :     available). But since this creates more overhead at "raise"-time
4168 :     (similar to the current exnHistory overhead), I have not yet
4169 :     implemented this. (The implementation will be rather easy.) With
4170 :     exceptions carrying back-trace information, this facility will be even
4171 :     more useful because users don't need to modify their programs...
4172 :    
4173 :     - While it is possible to compile the compiler with back-trace
4174 :     annotations turned on (I did it to get some confidence in
4175 :     correctness), you must make absolutely sure that core.sml and
4176 :     btimp.sml are compiled WITHOUT annotation! (core.sml cannot actually
4177 :     be compiled with annotation because there is no core access yet, but
4178 :     if you compile btimp.sml with annotation, then the system will go into
4179 :     an infinite recursion and crash.)
4180 :     Since CM currently does not know about BTrace, the only way to turn
4181 :     annotations on and off for different modules of the compiler is to
4182 :     interrupt CMB.make, change the settings, and re-invoke it. Of course,
4183 :     this is awkward and clumsy.
4184 :    
4185 :     Sample sessions:
4186 :    
4187 :     Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
4188 :     - SMLofNJ.Internals.BTrace.mode (SOME true);
4189 :     [autoloading]
4190 :     [autoloading done]
4191 :     val it = false : bool
4192 :     - structure X = struct
4193 :     - fun main n = let
4194 :     - fun a (x, 0) = d x
4195 :     - | a (x, n) = b (x, n - 1)
4196 :     - and b (x, n) = c (x, n)
4197 :     - and c (x, n) = a (x, n)
4198 :     - and d x = e (x, 3)
4199 :     - and e (x, 0) = f x
4200 :     - | e (x, n) = e (x, n - 1)
4201 :     - and f 0 = SMLofNJ.Internals.BTrace.trigger ()
4202 :     - | f n = n * g (n - 1)
4203 :     - and g n = a (n, 3)
4204 :     - in
4205 :     - f n
4206 :     - end
4207 :     - end;
4208 :     structure X : sig val main : int -> int end
4209 :     - X.main 3;
4210 :     *** BACK-TRACE ***
4211 :     GOTO stdIn:4.2-13.20: X.main[2].f
4212 :     GOTO-( stdIn:4.2-13.20: X.main[2].e
4213 :     GOTO stdIn:4.2-13.20: X.main[2].d
4214 :     / stdIn:4.2-13.20: X.main[2].a
4215 :     | stdIn:4.2-13.20: X.main[2].b
4216 :     GOTO-\ stdIn:4.2-13.20: X.main[2].c
4217 :     CALL stdIn:4.2-13.20: X.main[2].g
4218 :     GOTO stdIn:4.2-13.20: X.main[2].f
4219 :     GOTO-( stdIn:4.2-13.20: X.main[2].e
4220 :     GOTO stdIn:4.2-13.20: X.main[2].d
4221 :     / stdIn:4.2-13.20: X.main[2].a
4222 :     | stdIn:4.2-13.20: X.main[2].b
4223 :     GOTO-\ stdIn:4.2-13.20: X.main[2].c
4224 :     CALL stdIn:4.2-13.20: X.main[2].g
4225 :     GOTO stdIn:4.2-13.20: X.main[2].f
4226 :     GOTO-( stdIn:4.2-13.20: X.main[2].e
4227 :     GOTO stdIn:4.2-13.20: X.main[2].d
4228 :     / stdIn:4.2-13.20: X.main[2].a
4229 :     | stdIn:4.2-13.20: X.main[2].b
4230 :     GOTO-\ stdIn:4.2-13.20: X.main[2].c
4231 :     CALL stdIn:4.2-13.20: X.main[2].g
4232 :     GOTO stdIn:4.2-13.20: X.main[2].f
4233 :     CALL stdIn:2.15-17.4: X.main[2]
4234 :     -
4235 :    
4236 :     (Note that because of a FLINt bug the above code currently does not
4237 :     compile without BTrace turned on.)
4238 :    
4239 :     Here is another example, using my modified Tiger compiler:
4240 :    
4241 :     Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
4242 :     - SMLofNJ.Internals.BTrace.mode (SOME true);
4243 :     [autoloading]
4244 :     [autoloading done]
4245 :     val it = false : bool
4246 :     - CM.make "sources.cm";
4247 :     [autoloading]
4248 :     ...
4249 :     [autoloading done]
4250 :     [scanning sources.cm]
4251 :     [parsing (sources.cm):parse.sml]
4252 :     [creating directory CM/SKEL ...]
4253 :     [parsing (sources.cm):tiger.lex.sml]
4254 :     ...
4255 :     [wrote CM/sparc-unix/semant.sml]
4256 :     [compiling (sources.cm):main.sml]
4257 :     [wrote CM/sparc-unix/main.sml]
4258 :     [New bindings added.]
4259 :     val it = true : bool
4260 :     - Main.compile ("../testcases/merge.tig", "foo.out");
4261 :     *** BACK-TRACE ***
4262 :     CALL lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trvar
4263 :     CALL lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
4264 :     CALL lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
4265 :     GOTO lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
4266 :     CALL lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
4267 :     CALL lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
4268 :     CALL lib/semant.sml:488.3-505.6: SemantFun[2].transDec.trdec[2].transBody[2]
4269 :     / lib/semant.sml:411.65-543.8: SemantFun[2].transDec
4270 :     CALL-\ lib/semant.sml:413.2-540.9: SemantFun[2].transDec.trdec[2]
4271 :     CALL lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
4272 :     CALL lib/semant.sml:8.52-558.4: SemantFun[2].transProg[2]
4273 :     CALL main.sml:1.18-118.4: Main.compile[2]
4274 :     -
4275 :    
4276 :     ----------------------------------------------------------------------
4277 : blume 673 Name: Matthias Blumen
4278 :     Date: 2000/06/21 18:00:00 JST
4279 :     Tag: blume-20000621-manual
4280 :     Description:
4281 :    
4282 :     CM manual update: Path environments documented.
4283 :    
4284 :     ----------------------------------------------------------------------
4285 : blume 666 Name: Matthias Blume
4286 : blume 672 Date: 2000/06/19 13:40:00
4287 :     Tag: blume-20000619-manual
4288 :     Description:
4289 :    
4290 :     CM manual and system/README update. This only covers the fact that
4291 :     there are no more implicit anchors. (Path environments and the "bind"
4292 :     option to "cm" have yet to be documented.)
4293 :    
4294 :     ----------------------------------------------------------------------
4295 :     Name: Matthias Blume
4296 : blume 671 Date: 2000/06/19 11:05:00 JST
4297 :     Tag: blume-20000619-chdir-bugfix
4298 :     Description:
4299 :    
4300 :     Fixed a bug in new SrcPath module that sometimes led to a bad chDir call.
4301 :    
4302 :     ----------------------------------------------------------------------
4303 :     Name: Matthias Blume
4304 : blume 670 Date: 2000/06/18 22:00:10 JST
4305 :     Tag: blume-20000618-implicit-anchors-really-gone
4306 :     Description:
4307 :    
4308 :     I updates the previous HISTORY entry where I forgot to mention that
4309 :     implicit anchors are no longer with us.
4310 :    
4311 :     The current update also gets rid of the (now useless) controller
4312 :     CM.Control.implicit_anchors.
4313 :    
4314 :     ----------------------------------------------------------------------
4315 :     Name: Matthias Blume
4316 : blume 666 Date: 2000/06/16 17:30:00 JST
4317 :     Tag: blume-20000616-anchorenv
4318 :     Description:
4319 :    
4320 :     This patch implements the long anticipated (just kidding :) "anchor
4321 :     environment" mechanism. In the course of doing this, I also
4322 :     re-implemented CM's internal "SrcPath" module from scratch. The new
4323 :     one should be more robust in certain boundary cases. In any case, it
4324 :     is a lot cleaner than its predecessor (IMHO).
4325 :    
4326 :     This time, although there is yet another boot file format change, I
4327 :     kept th