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

SCM Repository

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

Diff of /sml/trunk/HISTORY

Parent Directory Parent Directory | Revision Log Revision Log | View Patch Patch

revision 577, Fri Mar 10 08:07:18 2000 UTC revision 678, Tue Jun 27 07:51:09 2000 UTC
# Line 13  Line 13 
13  Description:  Description:
14  ----------------------------------------------------------------------  ----------------------------------------------------------------------
15  Name: Matthias Blume  Name: Matthias Blume
16    Date: 2000/06/27 16:51:00 JST
17    Tag: blume-20000627-noweb
18    Description:
19    
20    Changes:
21    
22     1. Implemented "subdir" and "witness" options for noweb tool.
23        This caused some slight internal changes in CM's tool implementation.
24     2. Fixed bug in "tool plugin" mechanism.  This is essentially cleaning
25        some remaining issues from earlier path anchor changes.
26     3. Updated CM manual accordingly.
27    
28     4. Changed implementation of back-tracing so that I now consider it
29        ready for prime-time.
30    
31        In particular, you don't have to explicitly trigger the back-trace
32        anymore.  Instead, if you are running BTrace-instrumented code and
33        there is an uncaught exception (regardless of whether or not it was
34        raised in instrumented code), the top-level evalloop will print
35        the back-trace.
36    
37        Features:
38    
39          - Instrumented and uninstrumented code work together seemlessly.
40            (Of course, uninstrumented code is never mentioned in actual
41             back-traces.)
42    
43          - Asymptotic time- and space-complexity of instrumented code is
44            equal to that of uninstrumented code.  (This means that
45            tail-recursion is preserved by the instrumentation phase.)
46    
47          - Modules whose code has been instrumented in different sessions
48            work together without problem.
49    
50          - There is no penalty whatsoever on uninstrumented code.
51    
52          - There is no penalty on "raise" expressions, even in
53            instrumented code.
54    
55        A potential bug (or perhaps it is a feature, too):
56    
57          A back-trace reaches no further than the outermost instrumented
58          non-trivial "raise".  Here, a "trivial" raise is one that is the
59          sole RHS of a "handle" rule.  Thus, back-traces reach trough
60    
61               <exp> handle e => raise e
62    
63          and even
64    
65               <exp> handle Foo => raise Bar
66    
67          and, of course, through
68    
69               <exp> handle Foo => ...
70    
71         if the exception was not Foo.
72    
73         Back-traces always reach right through any un-instrumented code
74         including any of its "handle" expressions, trivial or not.
75    
76       To try this out, do the following:
77    
78         - Erase all existing binfiles for your program.
79           (You may keep binfiles for those modules where you think you
80            definitely don't need back-tracing.)
81         - Turn on back-trace instrumentation:
82              SMLofNJ.Internals.BTrace.mode (SOME true);
83         - Recompile your program.  (I.e., run "CM.make" or "use".)
84         - You may now turn instrumentation off again (if you want):
85              SMLofNJ.Internals.BTrace.mode (SOME false);
86         - Run your program as usual.  If it raises an exception that
87           reaches the interactive toplevel, then a back-trace will
88           automatically be printed.  After that, the toplevel loop
89           will print the exception history as usual.
90    
91    ----------------------------------------------------------------------
92    Name: Matthias Blume
93    Date: 2000/06/26 09:56:46 JST
94    Tag: blume-20000626-setup
95    Description:
96    
97    CM: - setup-parameter to "sml" added; this can be used to run arbitrary
98          ML code before and after compiling a file (e.g., to set compiler
99          flags)
100    
101    Compiler: - improved btrace API (in core.sml, internals.{sig,sml})
102              - associated changes to btrace.sml (BTrace instrumentation pass)
103              - cleaner implementation of btimp.sml (BTrace tracing and report
104                module)
105    
106    CM manual: * new path encoding documented
107               * description of setup-parameter to "sml" added
108    
109    The biggest user-visible change to back-tracing is that it is no
110    longer necessary to compile all traced modules within the same
111    session.  (This was a real limitation.)
112    
113    ----------------------------------------------------------------------
114    Name: Matthias Blume
115    Date: 2000/06/24 12:40:00 JST
116    Tag: blume-20000624-startup
117    Description:
118    
119    Fixes startup slowdown problem.  (I was calling SrcPath.sync a _tad_
120    bit too often -- to put it mildly. :)
121    
122    ----------------------------------------------------------------------
123    Name: Matthias Blume
124    Date: 2000/06/23 18:20:00 JST
125    Tag: blume-20000623-btrace
126    Description:
127    
128    This updates adds a backtrace facility to aid programmers in debugging
129    their programs.  This involves the following changes:
130    
131    1. Module system/smlnj/init/core.sml (structure _Core) now has hooks for
132       keeping track of the current call stack.  When programs are compiled
133       in a special mode, the compiler will insert calls to these hooks
134       into the user program.
135       "Hook" means that it is possible for different implementations of
136       back-tracing to register themselves (at different times).
137    
138    2. compiler/MiscUtil/profile/btrace.sml implements the annotation phase
139       as an Absyn.dec->Absyn.dec rewrite.  Normally this phase is turned off.
140       It can be turned on using this call:
141         SMLofNJ.Internals.BTrace.mode (SOME true);
142       Turning it off again:
143         SMLofNJ.Internals.BTrace.mode (SOME false);
144       Querying the current status:
145         SMLofNJ.Internals.BTrace.mode NONE;
146       Annotated programs are about twice as big as normal ones, and they
147       run a factor of 2 to 4 slower with a dummy back-trace plugin (one
148       where all hooks do nothing).  The slowdown with a plugin that is
149       actually useful (such as the one supplied by default) is even greater,
150       but in the case of the default plugin it is still only an constant
151       factor (amortized).
152    
153    3. system/Basis/Implementation/NJ/internals.{sig,sml} have been augmented
154       with a sub-structure BTrace for controlling back-tracing.  In particular,
155       the above-mentioned function "mode" controls whether the annotation
156       phase is invoked by the compiler.  Another important function is
157       "trigger": when called it aborts the current execution and causes
158       the top-level loop to print a full back-trace.
159    
160    4. compiler/MiscUtil/profile/btimp.sml is the current default plugin
161       for back-tracing.  It keeps track of the dynamic call stack and in
162       addition to that it keeps a partial history at each "level" of that
163       stack.  For example, if a tail-calls b, b tail-calls c, and c tail-calls
164       d and b (at separate times, dynamically), then the report will show:
165    
166       GOTO   d
167             /c
168       GOTO  \b
169       CALL   a
170    
171       This shows that there was an initial non-tail call of a, then a
172       tail-call to b or c, looping behavior in a cluster of functions that
173       consist of b and c, and then a goto from that cluster (i.e., either from
174       b or from c) to d.
175    
176       Note that (depending on the user program) the amount of information
177       that the back-trace module has to keep track of at each level is bounded
178       by a constant.  Thus, the whole implementation has the same asymptotical
179       complexity as the original program (both in space and in time).
180    
181    5. compiler/TopLevel/interact/evalloop.sml has been modified to
182       handle the special exception SMLofNJ.Internals.BTrace.BTrace
183       which is raised by the "trigger" function mentioned above.
184    
185    Notes on usage:
186    
187    - Annotated code works well together with unannotated code:
188    Unannotated calls simply do not show up at all in the backtrace.
189    
190    - It is not a good idea to let modules that were annotated during
191    different sessions run at the same time.  This is because the compiler
192    chooses small integers to identify individual functions, and there
193    will be clashes if different modules were compiled in separate sessions.
194    (Nothing will crash, and you will even be told about the clashes, but
195    back-trace information will in general not be useful.)
196    
197    - Back-tracing can be confused by callcc and capture.
198    
199    - The only way of getting a back-trace right now is to explicitly
200    invoke the "trigger" function from your user program.  Eventually, we
201    should make every exception carry back-trace information (if
202    available).  But since this creates more overhead at "raise"-time
203    (similar to the current exnHistory overhead), I have not yet
204    implemented this.  (The implementation will be rather easy.)  With
205    exceptions carrying back-trace information, this facility will be even
206    more useful because users don't need to modify their programs...
207    
208    - While it is possible to compile the compiler with back-trace
209    annotations turned on (I did it to get some confidence in
210    correctness), you must make absolutely sure that core.sml and
211    btimp.sml are compiled WITHOUT annotation!  (core.sml cannot actually
212    be compiled with annotation because there is no core access yet, but
213    if you compile btimp.sml with annotation, then the system will go into
214    an infinite recursion and crash.)
215    Since CM currently does not know about BTrace, the only way to turn
216    annotations on and off for different modules of the compiler is to
217    interrupt CMB.make, change the settings, and re-invoke it.  Of course,
218    this is awkward and clumsy.
219    
220    Sample sessions:
221    
222    Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
223    - SMLofNJ.Internals.BTrace.mode (SOME true);
224    [autoloading]
225    [autoloading done]
226    val it = false : bool
227    - structure X = struct
228    -     fun main n = let
229    -         fun a (x, 0) = d x
230    -           | a (x, n) = b (x, n - 1)
231    -         and b (x, n) = c (x, n)
232    -         and c (x, n) = a (x, n)
233    -         and d x = e (x, 3)
234    -         and e (x, 0) = f x
235    -           | e (x, n) = e (x, n - 1)
236    -         and f 0 = SMLofNJ.Internals.BTrace.trigger ()
237    -           | f n = n * g (n - 1)
238    -         and g n = a (n, 3)
239    -     in
240    -         f n
241    -     end
242    - end;
243    structure X : sig val main : int -> int end
244    - X.main 3;
245    *** BACK-TRACE ***
246    GOTO   stdIn:4.2-13.20: X.main[2].f
247    GOTO-( stdIn:4.2-13.20: X.main[2].e
248    GOTO   stdIn:4.2-13.20: X.main[2].d
249         / stdIn:4.2-13.20: X.main[2].a
250         | stdIn:4.2-13.20: X.main[2].b
251    GOTO-\ stdIn:4.2-13.20: X.main[2].c
252    CALL   stdIn:4.2-13.20: X.main[2].g
253    GOTO   stdIn:4.2-13.20: X.main[2].f
254    GOTO-( stdIn:4.2-13.20: X.main[2].e
255    GOTO   stdIn:4.2-13.20: X.main[2].d
256         / stdIn:4.2-13.20: X.main[2].a
257         | stdIn:4.2-13.20: X.main[2].b
258    GOTO-\ stdIn:4.2-13.20: X.main[2].c
259    CALL   stdIn:4.2-13.20: X.main[2].g
260    GOTO   stdIn:4.2-13.20: X.main[2].f
261    GOTO-( stdIn:4.2-13.20: X.main[2].e
262    GOTO   stdIn:4.2-13.20: X.main[2].d
263         / stdIn:4.2-13.20: X.main[2].a
264         | stdIn:4.2-13.20: X.main[2].b
265    GOTO-\ stdIn:4.2-13.20: X.main[2].c
266    CALL   stdIn:4.2-13.20: X.main[2].g
267    GOTO   stdIn:4.2-13.20: X.main[2].f
268    CALL   stdIn:2.15-17.4: X.main[2]
269    -
270    
271    (Note that because of a FLINt bug the above code currently does not
272    compile without BTrace turned on.)
273    
274    Here is another example, using my modified Tiger compiler:
275    
276    Standard ML of New Jersey v110.28.1 [FLINT v1.5], June 5, 2000
277    - SMLofNJ.Internals.BTrace.mode (SOME true);
278    [autoloading]
279    [autoloading done]
280    val it = false : bool
281    - CM.make "sources.cm";
282    [autoloading]
283    ...
284    [autoloading done]
285    [scanning sources.cm]
286    [parsing (sources.cm):parse.sml]
287    [creating directory CM/SKEL ...]
288    [parsing (sources.cm):tiger.lex.sml]
289    ...
290    [wrote CM/sparc-unix/semant.sml]
291    [compiling (sources.cm):main.sml]
292    [wrote CM/sparc-unix/main.sml]
293    [New bindings added.]
294    val it = true : bool
295    - Main.compile ("../testcases/merge.tig", "foo.out");
296    *** BACK-TRACE ***
297    CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trvar
298    CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
299    CALL   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
300    GOTO   lib/semant.sml:289.3-295.22: SemantFun[2].transExp.trexp.check[2]
301    CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
302    CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
303    CALL   lib/semant.sml:488.3-505.6: SemantFun[2].transDec.trdec[2].transBody[2]
304         / lib/semant.sml:411.65-543.8: SemantFun[2].transDec
305    CALL-\ lib/semant.sml:413.2-540.9: SemantFun[2].transDec.trdec[2]
306    CALL   lib/semant.sml:99.2-396.21: SemantFun[2].transExp.trexp
307    CALL   lib/semant.sml:8.52-558.4: SemantFun[2].transProg[2]
308    CALL   main.sml:1.18-118.4: Main.compile[2]
309    -
310    
311    ----------------------------------------------------------------------
312    Name: Matthias Blumen
313    Date: 2000/06/21 18:00:00 JST
314    Tag: blume-20000621-manual
315    Description:
316    
317    CM manual update: Path environments documented.
318    
319    ----------------------------------------------------------------------
320    Name: Matthias Blume
321    Date: 2000/06/19 13:40:00
322    Tag: blume-20000619-manual
323    Description:
324    
325    CM manual and system/README update.  This only covers the fact that
326    there are no more implicit anchors.  (Path environments and the "bind"
327    option to "cm" have yet to be documented.)
328    
329    ----------------------------------------------------------------------
330    Name: Matthias Blume
331    Date: 2000/06/19 11:05:00 JST
332    Tag: blume-20000619-chdir-bugfix
333    Description:
334    
335    Fixed a bug in new SrcPath module that sometimes led to a bad chDir call.
336    
337    ----------------------------------------------------------------------
338    Name: Matthias Blume
339    Date: 2000/06/18 22:00:10 JST
340    Tag: blume-20000618-implicit-anchors-really-gone
341    Description:
342    
343    I updates the previous HISTORY entry where I forgot to mention that
344    implicit anchors are no longer with us.
345    
346    The current update also gets rid of the (now useless) controller
347    CM.Control.implicit_anchors.
348    
349    ----------------------------------------------------------------------
350    Name: Matthias Blume
351    Date: 2000/06/16 17:30:00 JST
352    Tag: blume-20000616-anchorenv
353    Description:
354    
355    This patch implements the long anticipated (just kidding :) "anchor
356    environment" mechanism.  In the course of doing this, I also
357    re-implemented CM's internal "SrcPath" module from scratch.  The new
358    one should be more robust in certain boundary cases.  In any case, it
359    is a lot cleaner than its predecessor (IMHO).
360    
361    This time, although there is yet another boot file format change, I
362    kept the unpickler backward-compatible.  As a result, no new bootfiles
363    are necessary and bootstrapping is straightforward.  (You cannot read
364    new bootfiles into an old system, but the other way around is no
365    problem.)
366    
367    Visible changes:
368    
369    ** 0. Implicit path anchors (without the leading $-symbol) are no
370    longer recognized at all. This means that such path names are not
371    illegal either.  For example, the name basis.cm simply refers to a
372    local file called "basis.cm" (i.e, the name is an ordinary path
373    relative to .cm-files directory).  Or, to put it differently, only
374    names that start with $ are anchored paths.
375    
376    ** 1. The $<singlearc> abbreviation for $/<singlearc> has finally
377    vanished.
378    
379    John (Reppy) had critizised this as soon as I originally proposed and
380    implemented it, but at that time I did not really deeply believe
381    him. :) Now I came full-circle because I need the $<singlearc> syntax
382    in another place where it cannot be seen as an abbreviation for
383    $/<singlearc>.  To avoid the confusion, $<singlearc> now means what it
384    seems to mean (i.e., it "expands" into the corresponding anchor
385    value).
386    
387    However, when paths are used as members in CM description files, it
388    continues to be true that there must be at least another arc after the
389    anchor.  This is now enforced separately during semantic analysis
390    (i.e., from a lexical/syntactical point of view, the notation is ok.)
391    
392    ** 2. The "cm" class now accepts an option "bind".  The option's value
393    is a sub-option list of precisely two items -- one labeled "anchor"
394    and the other one labeled "value".  As you might expect, "anchor" is
395    used to specify an anchor name to be bound, and "value" specifies what
396    the anchor is being bound to.
397    
398    The value must be a directory name and can be given in either standard
399    syntax (including the possibility that it is itself an anchored path)
400    or native syntax.
401    
402    Examples:
403    
404       foo.cm (bind:(anchor:bar value:$mystuff/bar))
405       lib.cm (bind:(anchor:a value:"H:\\x\\y\\z"))  (* only works under windows *)
406    
407    and so on.
408    
409    The meaning of this is that the .cm-file will be processed with an
410    augmented anchor environment where the given anchor(s) is/are bound to
411    the given values(s).
412    
413    The rationale for having this feature is this: Suppose you are trying
414    to use two different (already stable) libraries a.cm and b.cm (that
415    you perhaps didn't write yourself).  Further, suppose each of these
416    two libraries internally uses its own auxiliary library $aux/lib.cm.
417    Normally you would now have a problem because the anchor "lib" can not
418    be bound to more than one value globally.  Therefore, the project that
419    uses both a.cm and b.cm must locally redirect the anchor to some other
420    place:
421    
422       a.cm (bind:(anchor:lib value:/usr/lib/smlnj/a-stuff))
423       b.cm (bind:(anchor:lib value:/usr/lib/smlnj/b-stuff))
424    
425    This hard-wires $lib/aux.cm to /usr/lib/smlnj/a-stuff/aux.cm or
426    /usr/lib/smlnj/b-stuff/aux.cm, respectively.
427    
428    Hard-wiring path names is a bit inflexible (and CM will verbosely warn
429    you when you do so at the time of CM.stabilize).  Therefore, you can
430    also use an anchored path as the value:
431    
432      a.cm (bind:(anchor:lib value:$a-lib))
433      b.cm (bind:(anchor:lib value:$b-lib))
434    
435    Now you can globally configure (using the usual CM.Anchor.anchor or
436    pathconfig machinery) bindings for "a-lib" and "b-lib".  Since "lib"
437    itself is always locally bound, setting it globally is no longer
438    meaningful or necessary (but it does not hurt either).  In fact, "lib"
439    can still be used as a global anchor for separate purposes.  As a
440    matter of fact, one can locally define "lib" in terms of a global
441    "lib":
442    
443      a.cm (bind:(anchor:lib value:$lib/a))
444      b.cm (bind:(anchor:lib value:$lib/b))
445    
446    ** 3: The encoding of path names has changed.  This affects the way
447    path names are shown in CM's progress report and also the internal
448    protocol encoding used for parallel make.
449    
450    The encoding now uses one or more ':'-separated segments.  Each
451    segments corresponds to a file that has been specified relative to the
452    file given by its preceding segment.  The first segment is either
453    relative to the CWD, absolute, or anchored.  Each segment itself is
454    basically a Unix pathname; all segments but the first are relative.
455    
456    Example:
457    
458       $foo/bar/baz.cm:a/b/c.sml
459    
460    This path denotes the file bar/a/b/c.sml relative to the directory
461    denoted by anchor "foo".  Notice that the encoding also includes
462    baz.cm which is the .cm-file that listed a/b/c.sml.  As usual, such
463    paths are resolved relative to the .cm-files directory, so baz.cm must
464    be ignored to get the "real" pathname.
465    
466    To make this fact more obvious, CM puts the names of such "virtual
467    arcs" into parentheses when they appear in progress reports. (No
468    parentheses will appear in the internal protocol encoding.)  Thus,
469    what you really see is:
470    
471      $foo/bar/(baz.cm):a/b/c.sml
472    
473    I find this notation to be much more informative than before.
474    
475    Another new feature of the encoding is that special characters
476    including parentheses, colons, (back)slashes, and white space are
477    written as \ddd (where ddd is the decimal encoding of the character).
478    
479    *** The CM manual still needs to be updated.
480    
481    ----------------------------------------------------------------------
482    Name: Allen Leung
483    Date: 2000/06/15 00:38:00
484    Tag: leunga-20000615-x86-peephole
485    
486    x86 Peephole fix by Fermin.  Affects c-- and moby only.
487    
488    ----------------------------------------------------------------------
489    Name: Matthias Blume
490    Date: 2000/06/12 11:40:00
491    Tag: blume-20000612-parmakefix
492    Description:
493    
494    More cleanup after changing the file naming scheme: This time I
495    repaired the parallel make mechanism for CMB.make which I broke earlier.
496    
497    ----------------------------------------------------------------------
498    Name: Allen Leung
499    Date: 2000/06/09 01:25:00
500    Tag: leunga-20000609-various
501    
502    None of these things should affect normal SML/NJ operations
503    
504    1. Peephole improvements provided by Fermin (c--)
505    2. New annotation DEFUSE for adding extra dependence (moby)
506    3. New X86 LOCK instructions (moby)
507    4. New machine description language for reservation tables (scheduling)
508    5. Fixes to various optimization/analysis modules (branch chaining, dominator
509       trees etc.)
510    6. I've changed the CM files so that they can work with versions
511       110.0.6, 110.25 and 110.28
512    
513    ----------------------------------------------------------------------
514    Name: Matthias Blume
515    Date: 2000/06/09 12:40:00
516    Tag: blume-20000609-log
517    Description:
518    
519    - Removed all(?) remaining RCS Log entries from sources.
520    
521    - Fixed bug in ml-yacc and ml-lex sources (use explicit anchors for
522      anchored paths).
523    
524    ----------------------------------------------------------------------
525    Name: Matthias Blume
526    Date: 2000/06/07 17:00:00 JST
527    Tag: blume-20000607-no-implicit-anchors
528    Description:
529    
530    1. This update changes the default setting for
531    CM.Control.implicit_anchors from true to false.  This means that
532    implicit anchors are no longer permitted by default.  I also tried to
533    make sure that nothing else still relies on implicit anchors.
534    (This is the next step on the schedule towards a CM that does not even
535    have the notion of implicit anchors anymore.)
536    
537    2. More CM manual updates.
538    
539    3. I managed to track down and fix the pickling bug I mentioned last
540    time.  Because of the previously existing workaround, this entails no
541    immediate practical changes.
542    
543    ----------------------------------------------------------------------
544    Name: Matthias Blume
545    Date: 2000/06/06 11:15:00 JST
546    Tag: blume-20000606-lazierpickle
547    Description:
548    
549    !!!! NEW BOOT FILES !!!!
550    
551    * The main purpose of this update is to make library pickles lazier in
552    order to reduce the initial space penalty for autoloading a library.
553    As a result, it is now possible to have $smlnj/compiler.cm
554    pre-registered.  This should take care of the many complaints or
555    inquiries about missing structure Compiler.  This required changes to
556    CM's internal data structures and small tweaks to some algorithms.
557    
558    As a neat additional effect, it is no longer necessary (for the sake
559    of lean heap image files) to distinguish between a "minimal" CM and a
560    "full" CM.  Now, there is only one CM (i.e., the "full" version:
561    $smlnj/cm.cm aka $smlnj/cm/full.cm), and it is always available at the
562    interactive top level. ($smlnj/cm/minimal.cm is gone.)
563    
564    To make the life of compiler-hackers easier, "makeml" now also
565    pre-registers $smlnj/cmb.cm (aka $smlnj/cmb/current.cm).  In other
566    words, after you bootstrap a new sml for the first time, you will not
567    have to autoload $smlnj/cmb.cm again afterwards.  (The first time
568    around you will still have to do it, though.)
569    
570    * A second change consists of major updates to the CM manual.  There
571    are now several appendices with summary information and also a full
572    specification of the CM description file syntax.
573    
574    * In directory src/system I added the script "allcross".  This script
575    invokes sml and cross-compiles the compiler for all supported
576    architectures.  (Useful when providing a new set of boot files.)
577    
578    * There seems to be a latent bug in my "lazy pickles" mechanism.  I
579    added a small tweak to pickle-util.sml to work around this problem,
580    but it is not a proper fix yet.  I will investigate further.  (The
581    effect of the bug was an inflation of library pickle size.)
582    
583    * Version number increased to 110.28.1 (to avoid compatibility problems).
584    
585    ----------------------------------------------------------------------
586    Name: Allen Leung
587    Date: 2000/05/25 17:28 EDT
588    Tag: leunga-20000525-ra
589    Description:
590    
591      Fixed a bug in freezing phase of the register allocator.
592    
593    ----------------------------------------------------------------------
594    Name: Allen Leung
595    Date: 2000/05/15 22:53 EDT
596    Tag: leunga-20000515-alpha-x86-ra
597    Description:
598    
599      1. Alpha
600    
601          Slight cleanup.  Removed the instruction SGNXL
602    
603      2. X86
604    
605          Added the following instructions to the instruction set:
606    
607            ROLx, RORx,
608            BTx, BTSx, BTLx, BTRx,
609            XCHGx, and variants with the LOCK prefix
610    
611      3. Register Allocation
612    
613          The module ra-rewrite-with-renaming has been improved.
614    
615      These have no effect on SML/NJ.
616    
617    ----------------------------------------------------------------------
618    Name: Matthias Blume
619    Date: 2000/05/15 16:20:00 JST
620    Tag: blume-20000515-lightrebuild
621    Description:
622    
623    1. I added an alternative to "-rebuild" to "makeml".  The difference is
624       that prior to calling CMB.make' the CM-variable "LIGHT" will be
625       defined.  In effect, the command will not build any cross-compiler
626       backends and therefore finish more quickly.
627    
628       The "fixpt" script also takes a "-light" switch to be able to use
629       this new facility while compiling for a fixpoint.
630    
631    2. I replaced all mentions of anchored paths in group owner specifications
632       with simple relative paths (usually starting with "..").
633       The rationale is that a library's internal workings should not be
634       compromised by the lack of some anchor.  (An anchor is necessary
635       for someone who wants to refer to the library by an anchored path,
636       but it should not be necessary to build the same library in the first
637       place.)
638    
639    3. I changed the way CM's tool mechanism determines the shell command
640       string used for things like ml-yacc etc. so that it does not break
641       when CM.Control.implicit_anchors is turned off.
642    
643    ----------------------------------------------------------------------
644    Name: Matthias Blume
645    Date: 2000/05/12 18:20:00 JST
646    Tag: blume-20000512-ml-build
647    Description:
648    
649    Fixed a bug in config/_ml-build that prevented ml-yacc and ml-lex from
650    getting installed properly (by config/install.sh).
651    
652    ----------------------------------------------------------------------
653    Name: Matthias Blume
654    Date: 2000/05/12 17:30:00 JST
655    Tag: blume-20000512-anchors
656    Description:
657    
658    !!! NEW BOOT FILES !!!
659    
660    This change is in preparation of fading out support for "implicitly
661    anchored path names".  I went through all sources and used the
662    explicit (and relatively new) $-notation.  See system/README and the
663    CM manual for more info on this.
664    
665    I also modified the anchoring scheme for some things such as "smlnj",
666    "MLRISC", "cm", etc. to take advantage of the fact that explicit
667    anchors are more expressive: anchor name and first arc do not have to
668    coincide.  This entails the following user-visible change:
669    
670    You have to write $smlnj/foo/bar instead of smlnj/foo/bar.  In
671    particular, when you fire up sml with a command-line argument, say,
672    e.g.:
673    
674       sml '$smlnj/cmb.cm'
675    
676    At the ML toplevel prompt:
677    
678       CM.autoload "$smlnj/cmb.cm";
679    
680    There is also a new controller in CM.Control that can be used to turn
681    off all remaining support for implicit anchors by saying:
682    
683        CM.autoload "$smlnj/
684        #set CM.Control.implicit_anchors false;
685    
686    This causes CM to reject implicitly anchored paths.  This is (for the
687    time being) less permissive than the "final" version where there will
688    be no more such implicit anchors and relative paths will be just that:
689    relative.
690    
691    The next step (version after next version?) will be to make the
692    default for CM.Control.implicit_anchors false.  After the dust has
693    settled, I can then produce the "final" version of this...
694    
695    Note: Since bootstrapping is a bit tricky, I provided new boot files.
696    
697    ----------------------------------------------------------------------
698    Name: Matthias Blume
699    Date: 2000/05/11 16:30:00 JST
700    Tag: blume-20000511-sources
701    Description:
702    
703    The main change is that I added function CM.sources as a generalized
704    version of the earlier CM.makedepend.  This entails the following
705    additional changes:
706    
707      - CM.makedepend has been dropped.
708    
709      - CM manual has been updated.
710    
711      - TOOLS signature and API have been changed.
712    
713    ----------------------------------------------------------------------
714    Name: Allen Leung
715    Date: 2000/05/10 21:17 EDT
716    Tag: leunga-20000510-moby-c--ssa
717    Description:
718    
719      Various bug fixes and new features for C--, Moby and MLRISC optimizations.
720    None of these affect SML/NJ.
721    
722    1. Register Allocation
723    
724        a. A new ra spilling module (ra/ra-spill-with-renaming) is implemented.
725           This module tries to remove local (i.e. basic block level) redundancies
726           during spilling.
727    
728        b. A new framework for performing region based register allocation.
729           Not yet entirely functional.
730    
731    2. X86
732    
733       a. DefUse for POP was missing the stack pointer [found by Lal]
734       b. Reload for CALL was incorrect in X86Spill [found by John]
735       c. Various fixes in X86Spill so that it can be used correctly for
736          the new spilling module.
737    
738    3. SSA/IR
739    
740       a. New module ir/dj-dataflow.sml implements elimination based
741          data flow analysis.
742    
743    4. MLRiscGen
744    
745       a. Fix for gc type annotation
746    
747    5. MDGen
748    
749       Various fixes for machine description -> ml code translation.  For ssa
750       only.
751    
752    ----------------------------------------------------------------------
753    Name: Allen Leung
754    Date: 2000/05/08 22:17 EDT
755    Tag: leunga-20000508-labexp
756    Description:
757    
758      Fermin has found a few assembly problems with constant expressions
759      generated in LabelExp.  Mostly, the problems involve extra parentheses,
760      which choke on dumb assemblers.  This is his fix.
761    
762    ----------------------------------------------------------------------
763    Name: Dave MacQueen
764    Date: 2000/04/09 14:00 EDT
765    Tag: dbm-20000502-Version_110_28
766    Description:
767    
768    1. Updated src/compiler/TopLevel/main/version.sml to version 110.28
769    
770    2. Updated config/version to 110.28
771    
772    3. Updated config/srcarchiveurl
773    
774    3. New boot files!
775       ftp://ftp.research.bell-labs.com/dist/smlnj/working/110.28/
776    
777    ----------------------------------------------------------------------
778    Name: Matthias Blume
779    Date: 2000/05/01 19:05:00 JST
780    Tag: blume-20000501-noweb
781    Description:
782    
783    A new noweb tool has been added.  The existing system is entirely
784    unaffected by this, but some CM users have asked for renewed noweb
785    support.  Everything is documented in the CM manual.
786    
787    New (plugin) libraries:
788    
789       noweb-tool.cm
790       nw-ext.cm
791    
792    ----------------------------------------------------------------------
793    Name: Dave MacQueen
794    Date: 2000/04/30 12:40PM EDT
795    Tag: dbm-20000430-bug_fixes
796    Description:
797    
798    1. Fix for bug 1498
799       smlnj/src/system/Basis/Implementation/Unsafe/object.sig
800       smlnj/src/system/Basis/Implementation/Unsafe/object.sml
801         added toRealArray function
802       smlnj/src/compiler/MiscUtil/print/ppobj.sml
803         added check for tag Obj.RealArray to array printing case in ppObj
804    
805    2. Fix for bug 1510
806       smlnj/src/compiler/Semant/types/typesutil.sml
807         fixed definition of dummyargs (used by equalTycon) so that
808         dummy args are distinct types
809    
810    ----------------------------------------------------------------------
811    Name: Matthias Blume
812    Date: 2000/04/30 01:00:00 JST
813    Tag: blume-20000430-versions
814    Description:
815    
816    1. CM version numbering added.  This is an implementation of Lal's
817       proposal for adding version numbers and version checking to .cm
818       files.  Lal said that his proposal was just that -- a proposal.
819       For the time being I went ahead and implemented it so that people
820       can comment on it.  Everything is completely backward-compatible
821       (except for the stable library format, i.e., new bootfiles!).
822    
823       As usual, see the CM manual for details.
824    
825    2. An alternative syntax for anchored paths has been implemented.
826       Dave has recently voiced the same concerns that I had when I did
827       this, so there should be some support.  My take is that eventually
828       I will let support for the current syntax (where anchors are
829       "implicit") fade out in favor of the new, explicit syntax.
830       In order to be backward-compatible, both old and new syntax are
831       currently supported.
832    
833       Again, see the CM manual for details.
834    
835    3. Parallel make is trying to be slightly smarter:  When the master
836       process finds a "bottleneck", i.e., when there is only one
837       compilation unit that can be compiled and everybody else is
838       waiting on it, then it will simply compile it directly instead
839       of clumsily telling one of the slaves to do it.
840    
841    4. Support for "unsharing" added.  This is necessary in order to be
842       able to have two different versions of the same library running
843       at the same time (e.g., for trying out a new MLRISC while still
844       having the old MLRISC linked into the current compiler, etc.)
845       See the CM manual.
846    
847    5. Simple "makedepend" functionality added for generating Makefile
848       dependency information.  (This is rather crude at the moment.
849       Expect some changes here in the future.)
850    
851    6. ".fun" added as a recognized suffix for ML files. Also documented
852       explicitly in the manual that the fallback behavior (unknown suffix
853       -> ML file) is not an official feature!
854    
855    7. Small changes to the pickler for stable libraries.
856    
857    8. Several internal changes to CM (for cleanup/improvement).
858    
859    
860    !!!! NEW BINFILES !!!!
861    
862    ----------------------------------------------------------------------
863    Name: Matthias Blume
864    Date: 2000/04/28 17:30:00 JST
865    Tag: blume-20000428-pathconfig
866    Description:
867    
868    1. I changed config/install.sh to remove duplicate entries from the
869       lib/pathconfig file at the end.  Moreover, the final version of
870       lib/pathconfig is sorted alphabetically.  The same (sorting) is done
871       in src/system/installml.
872    
873    2. The config/install.sh script now consistently uses relative
874       pathnames in lib/pathconfig whenever the anchor is in the lib
875       directory.  (So far this was true for the libraries that come
876       pre-compiled and bundled as part of the bootfiles but not for
877       libraries that are compiled by the script itself.)
878    
879    ----------------------------------------------------------------------
880    Name: Matthias Blume
881    Date: 2000/04/26 13:10:00 JST
882    Tag: blume-20000426-fun_suffix
883    Description:
884    
885    Added ".fun" as a recognized file name suffix (for ML code).
886    
887    ----------------------------------------------------------------------
888    Name: Allen Leung
889    Date: 2000/04/25 17:00:00 EST
890    Tag: leunga-20000425-alpha-ra
891    Description:
892    
893    1. Alpha
894    
895        PSEUDOARITH was missing in AlphaRewrite.  This causes an endless loop
896    in C--.
897    
898    2. RA
899    
900       Added a flag "ra-dump-size" to print out the size of the flowgraph
901       and the interference graph.
902    
903    ----------------------------------------------------------------------
904    Name: Dave MacQueen
905    Date: 2000/04/25/
906    Tag: dbm-20000425-mlyacc_doc_examples
907    Description:
908      Updated mlyacc.tex sections 5 and 7 for SML '97 and CM.
909      Updated all three examples in src/ml-yacc/examples to run
910      under 110.* using CM.make.
911    
912    ----------------------------------------------------------------------
913    Name: Allen Leung
914    Date: 2000/04/20 23:04:00 EST
915    Tag: leunga-20000420-ssa-c---stuff
916    Description:
917    
918      This update synchronizes my repository with Yale's.  Most of these
919    changes, however, do not affect SML/NJ at all (the RA is an exception).
920    
921    1. Register Allocator
922    
923       a. An improvement in the interference graph construction:
924          Given a copy
925    
926                s <- t
927    
928          no interference edge between s and t is added for this definition of s.
929    
930       b. I've added two new spill heuristic modules that Fermin and I developed
931          (in the new library RA.cm). These are unused in SML/NJ but maybe
932          useful for others (Moby?)
933    
934    2. X86
935    
936       a. Various fixes in the backend provided by Fermin [C--] and Lal.
937    
938    3. Alpha
939    
940       a. Added the BSR instruction and code generation that goes with it [C--]
941       b. Other fixes too numerous to recount provided by Fermin [C--]
942    
943    4. Regmaps
944    
945       a. The regmaps are not initialized with the identity physical bindings
946          at creation time.  This is unneeded.
947    
948    5. MLRISC Optimizations
949    
950       a. The DJ-Graph module can now compute the iterated dominance frontiers
951          intersects with liveness incrementally in linear time! Woohoo!
952          This is now used in my new SSA construction algorithm.
953    
954       b. THe branch reorganization module is now smarter about linear chains of
955          basic blocks.
956    
957    
958    ----------------------------------------------------------------------
959    Name: Matthias Blume
960    Date: 2000/04/12 13:52:00 JST
961    Tag: blume_main_v110p27_1
962    Description:
963    
964    Changed install.sh script to handle archive files without version number
965    and to use "boot.<arch>-<os>" instead of "sml.boot.<arch>-<os>" for the
966    name of the boot file archive.
967    
968    ----------------------------------------------------------------------
969    Name: Dave MacQueen
970    Date: 2000/04/09 14:00 EDT
971    Tag: dbm-20000410-Version_110_27
972    Description:
973    
974    1. Updated src/compiler/TopLevel/main/version.sml to version 110.27
975    
976    2. Updated src/config/version to 110.27
977    
978    3. New boot files!
979    
980    ----------------------------------------------------------------------
981    Name: Allen Leung
982    Date: 2000/04/09 19:09:00 EST
983    Tag: leunga-20000409-misc
984    Description:
985    
986    1.  Yet another fix for x86 assembly for idivl, imull, mull and friends.
987    
988    2.  Miscellaneous improvements to MLRISC (unused in sml/nj)
989    
990    ----------------------------------------------------------------------
991    Name: Stefan
992    Date: 2000/04/07 10:00:00 EDT
993    Tag: monnier-20000406-branch-handling
994    Description:
995    
996    Improved handling of branches (mostly those generated from
997    polymorphic equality), removed switchoff and changed the
998    default optimization settings (more cpsopt and less flintopt).
999    
1000    ----------------------------------------------------------------------
1001    Name: Allen Leung
1002    Date: 2000/04/06 01:30:00 EST
1003    Tag: leunga-20000406-peephole-x86-SSA-2
1004    Description:
1005    
1006       Forgot a few files.
1007    
1008    ----------------------------------------------------------------------
1009    Name: Allen Leung
1010    Date: 2000/04/06 00:36:00 EST
1011    Tag: leunga-20000406-peephole-x86-SSA
1012    Description:
1013    
1014    1.  New Peephole code
1015    
1016    2.  Minor improvement to X86 instruction selection
1017    
1018    3.  Various fixes to SSA and machine description -> code translator
1019    
1020    ----------------------------------------------------------------------
1021    Name: Matthias Blume
1022    Date: 2000/04/05 12:30:00 JST
1023    Tag: blume_main_v110p26p2_3
1024    Description:
1025    
1026    This update just merges three minor cosmetic updates to CM's sources
1027    to get ready for the 110.27 code freeze on Friday.  No functionality
1028    has changed.
1029    
1030    ----------------------------------------------------------------------
1031    Name: Allen Leung
1032    Date: 2000/04/04 19:39:00 EST
1033    Tag: leunga-20000404-x86-asm
1034    Description:
1035    
1036    1.  Fixed a problem in X86 assembly.
1037    
1038        Things like
1039    
1040           jmp %eax
1041           jmp (%eax)
1042    
1043        should be output as
1044    
1045           jmp *%eax
1046           jmp *(%eax)
1047    
1048    2.  Assembly output
1049    
1050          Added a new flag
1051    
1052              "asm-indent-copies" (default to false)
1053    
1054          When this flag is on, parallel copies will be indented an extra level.
1055    
1056    ----------------------------------------------------------------------
1057    Name: Allen Leung
1058    Date: 2000/04/04 03:18:00 EST
1059    Tag: leunga-20000404-C--Moby
1060    Description:
1061    
1062        All of these fixes are related to C--, Moby, and my own optimization
1063        stuff; so they shouldn't affect SML/NJ.
1064    
1065    1.  X86
1066    
1067        Various fixes related floating point, and extensions.
1068    
1069    2.  Alpha
1070    
1071        Some extra patterns related to loads with signed/zero extension
1072        provided by Fermin.
1073    
1074    3.  Assembly
1075    
1076        When generating assembly, resolve the value of client defined constants,
1077        instead of generating symbolic values.  This is controlled by the
1078        new flag "asm-resolve-constants", which is default to true.
1079    
1080    4.  Machine Descriptions
1081    
1082        a. The precedence parser was slightly broken when parsing infixr symbols.
1083        b. The type generalizing code had the bound variables reversed, resulting
1084           in a problem during arity raising.
1085        c. Various fixes in machine descriptions.
1086    
1087    ----------------------------------------------------------------------
1088    Name: Matthias Blume
1089    Date: 2000/04/03 16:05:00 JST
1090    Tag: blume_main_v110p26p2_2
1091    Description:
1092    
1093    I eliminated coreEnv from compInfo.  Access to the "Core" structure is
1094    now done via the ordinary static environment that is context to each
1095    compilation unit.
1096    
1097    To this end, I arranged that instead of "structure Core" as "structure
1098    _Core" is bound in the pervasive environment.  Core access is done via
1099    _Core (which can never be accidentally rebound because _Core is not a
1100    legal surface-syntax symbol).
1101    
1102    The current solution is much cleaner because the core environment is
1103    now simply part of the pervasive environment which is part of every
1104    compilation unit's context anyway.  In particular, this eliminates all
1105    special-case handling that was necessary until now in order to deal
1106    with dynamic and symbolic parts of the core environment.
1107    
1108    Remaining hackery (to bind the "magic" symbol _Core) is localized in the
1109    compilation manager's bootstrap compiler (actually: in the "init group"
1110    handling).  See the comments in src/system/smlnj/init/init.cmi for
1111    more details.
1112    
1113    I also tried to track down all mentions of "Core" (as string argument
1114    to Symbol.strSymbol) in the compiler and replaced them with a
1115    reference to the new CoreSym.coreSym.  Seems cleaner since the actual
1116    name appears in one place only.
1117    
1118    Binfile and bootfile format have not changed, but the switchover from
1119    the old "init.cmi" to the new one is a bit tricky, so I supplied new
1120    bootfiles anyway.
1121    
1122    ----------------------------------------------------------------------
1123    Name: Allen Leung
1124    Date: 2000/04/02 21:17:00 EST
1125    Tag: leunga-20000402-mltree
1126    Description:
1127    
1128       1. Renamed the constructor CALL in MLTREE by popular demand.
1129       2. Added a bunch of files from my repository.  These are currently
1130          used by other non-SMLNJ backends.
1131    
1132    ----------------------------------------------------------------------
1133    Name: Allen Leung
1134    Date: 2000/03/31 21:15:00 EST
1135    Tag: leunga-20000331-aliasing
1136    Description:
1137    
1138    This update contains a rewritten (and hopefully more correct) module
1139    for extracting aliasing information from CPS.
1140    
1141       To turn on this feature:
1142    
1143            Compiler.Control.CG.memDisambiguate := true
1144    
1145       To pretty print the region information with assembly
1146    
1147           Compiler.Control.MLRISC.getFlag "asm-show-region" := true;
1148    
1149       To control how many levels of aliasing information are printed, use:
1150    
1151           Compiler.Control.MLRISC.getInt "points-to-show-level" := n
1152    
1153       The default of n is 3.
1154    
1155    ----------------------------------------------------------------------
1156    Name: David MacQueen
1157    Date: 2000/03/31 11:15:00 EST
1158    Tag: dbm-20000331-runtime_fix
1159    Description:
1160    
1161    This update contains:
1162    
1163    1. runtime/c-lib/c-libraries.c
1164       includes added in revision 1.2 caused compilation errors on hppa-hpux
1165    
1166    2. fix for bug 1556
1167       system/Basis/Implementation/NJ/internal-signals.sml
1168    
1169    ----------------------------------------------------------------------
1170    Name: Matthias Blume
1171    Date: 2000/03/31 18:00:00 JST
1172    Tag: blume_main_v110p26p2_1
1173    Description:
1174    
1175    This update contains:
1176    
1177    1. A small change to CM's handling of stable libraries:
1178       CM now maintains one "global" modmap that is used for all stable
1179       libraries.  The use of such a global modmap maximizes sharing and
1180       minimizes the need for re-traversing parts of environments during
1181       modmap construction.  (However, this has minor impact since modmap
1182       construction seems to account for just one percent or less of total
1183       compile time.)
1184    
1185    2. I added a "genmap" phase to the statistics.  This is where I got the
1186       "one percent" number (see above).
1187    
1188    3. CM's new tool parameter mechanism just became _even_ better. :)
1189       - The parser understands named parameters and recursive options.
1190       - The "make" and "shell" tools use these new features.
1191         (This makes it a lot easier to cascade these tools.)
1192       - There is a small syntax change: named parameters use a
1193    
1194           <name> : ( <option> ... )            or
1195           <name> : <string>
1196    
1197         syntax.  Previously, named parameters were implemented in an
1198         ad-hoc fashion by each tool individually (by parsing strings)
1199         and had the form
1200    
1201           <name>=<string>
1202    
1203       See the CM manual for a full description of these issues.
1204    
1205    ----------------------------------------------------------------------
1206    Name: Matthias Blume
1207    Date: 2000/03/30 18:00:00 JST
1208    Tag: blume_main_v110p26p2_0
1209    Description:
1210    
1211    !!!!! WARNING !!!!!!
1212    !!  New binfiles  !!
1213    !!!!!!!!!!!!!!!!!!!!
1214    
1215    This update contains:
1216    
1217    1. Moderate changes to CM:
1218    
1219       - Changes to CM's tools mechanism.  In particular, it is now possible
1220       to have tools that accept additional "command line" parameters
1221       (specified in the .cm file at each instance where the tool's class is
1222       used).
1223    
1224       This was done to accommodate the new "make" and "shell" tools which
1225       facilitate fairly seamless hookup to portions of code managed using
1226       Makefiles or Shell scripts.
1227    
1228       There are no classes "shared" or "private" anymore.  Instead, the
1229       sharing annotation is now a parameter to the "sml" class.
1230    
1231       There is a bit of generic machinery for implementing one's own
1232       tools that accept command-line parameters.  However, I am not yet fully
1233       satisfied with that part, so expect changes here in the future.
1234    
1235       All existing tools are described in the CM manual.
1236    
1237       - Slightly better error handling.  (CM now suppresses many followup
1238       error messages that tended to be more annoying than helpful.)
1239    
1240    2. Major changes to the compiler's static environment data structures.
1241    
1242       - no CMStaticEnv anymore.
1243            - no CMEnv, no "BareEnvironment" (actually, _only_ BareEnvironment,
1244              but it is called Environment), no conversions between different
1245              kinds of static environments
1246    
1247       - There is still a notion of a "modmap", but such modmaps are generated
1248         on demand at the time when they are needed.  This sounds slow, but I
1249         sped up the code that generates modmaps enough for this not to lead to
1250         a slowdown of the compiler (at least I didn't detect any).
1251    
1252       - To facilitate rapid modmap generation, static environments now
1253         contain an (optional) "modtree" structure.  Modtree annotations are
1254         constructed by the unpickler during unpickling.  (This means that
1255         the elaborator does not have to worry about modtrees at all.)
1256         Modtrees have the advantage that they are compositional in the same
1257         way as the environment data structure itself is compositional.
1258         As a result, modtrees never hang on to parts of an environment that
1259         has already been rendered "stale" by filtering or rebinding.
1260    
1261       - I went through many, many trials and errors before arriving at the
1262         current solution.  (The initial idea of "linkpaths" did not work.)
1263         But the result of all this is that I have touched a lot of files that
1264         depend on the "modules" and "types" data structures (most of the
1265         elaborator). There were a lot of changes during my "linkpath" trials
1266         that could have been reverted to their original state but weren't.
1267         Please, don't be too harsh on me for messing with this code a bit more
1268         than what was strictly necessary...  (I _did_ resist the tempation
1269         of doing any "global reformatting" to avoid an untimely death at
1270         Dave's hands. :)
1271    
1272       - One positive aspect of the previous point:  At least I made sure that
1273         all files that I touched now compile without warnings (other than
1274         "polyEqual").
1275    
1276       - compiler now tends to run "leaner" (i.e., ties up less memory in
1277         redundant modmaps)
1278    
1279    ----------------------------------------------------------------------
1280    Name: Allen Leung
1281    Date: 2000/03/29 18:00:00
1282    Tag: leunga-20000327-mlriscGen_hppa_alpha_x86
1283    Boot files (optional): ftp://react-ilp.cs.nyu.edu/leunga/110.26.1-sml.boot.x86-unix-20000330.tar.gz
1284    Description:
1285    
1286       This update contains *MAJOR* changes to the way code is generated from CPS
1287    in the module mlriscGen, and in various backend modules.
1288    
1289    CHANGES
1290    =======
1291    
1292    1. MLRiscGen: forward propagation fix.
1293    
1294       There was a bug in forward propagation introduced at about the same time
1295       as the MLRISC x86 backend, which prohibits coalescing to be
1296       performed effectively in loops.
1297    
1298       Effect: speed up of loops in RISC architectures.
1299               By itself, this actually slowed down certain benchmarks on the x86.
1300    
1301    2. MLRiscGen:  forward propagating addresses from consing.
1302    
1303       I've changed the way consing code is generated.  Basically I separated
1304       out the initialization part:
1305    
1306            store tag,   offset(allocptr)
1307            store elem1, offset+4(allocptr)
1308            store elem2, offset+8(allocptr)
1309            ...
1310            store elemn, offset+4n(allocptr)
1311    
1312       and the address computation part:
1313    
1314            celladdr <- offset+4+alloctpr
1315    
1316       and move the address computation part
1317    
1318       Effect:  register pressure is generally lower as a result.  This
1319                makes compilation of certain expressions much faster, such as
1320                long lists with non-trivial elements.
1321    
1322                 [(0,0), (0,0), .... (0,0)]
1323    
1324    3. MLRiscGen: base pointer elimination.
1325    
1326        As part of the linkage mechanism, we generate the sequence:
1327    
1328         L:  ...  <- start of the code fragment
1329    
1330         L1:
1331             base pointer <- linkreg - L1 + L
1332    
1333         The base pointer was then used for computing relocatable addresses
1334       in the code fragment.  Frequently (such as in lots of continuations)
1335       this is not needed.  We now eliminate this sequence whenever possible.
1336    
1337         For compile time efficiency, I'm using a very stupid local heuristic.
1338       But in general, this should be done as a control flow analysis.
1339    
1340       Effect:  Smaller code size.  Speed up of most programs.
1341    
1342    4. Hppa back end
1343    
1344         Long jumps in span dependence resolution used to depend on the existence
1345      of the base pointer.
1346    
1347         A jump to a long label L was expanded into the following sequence:
1348    
1349          LDIL %hi(L-8192), %r29
1350          LDO  %lo(L-8192)(%r29), %r29
1351          ADD  %r29, baseptr, %r29
1352          BV,n %r0(%r29)
1353    
1354         In the presence of change (3) above, this will not work.  I've changed
1355       it so that the following sequence of instructions are generated, which
1356       doesn't mention the base pointer at all:
1357    
1358             BL,n  L', %r29           /* branch and link, L' + 4 -> %r29 */
1359        L':  ADDIL L-(L'+4), %r29     /* Compute address of L */
1360             BV,n  %r0(%r29)          /* Jump */
1361    
1362    5. Alpha back end
1363    
1364          New alpha instructions LDB/LDW have been added, as per Fermin's
1365       suggestions.   This is unrelated to all other changes.
1366    
1367    6. X86 back end
1368    
1369         I've changed andl to testl in the floating point test sequence
1370         whenever appropriate.  The Intel optimization guide states that
1371         testl is preferable to andl.
1372    
1373    7. RA (x86 only)
1374    
1375         I've improved the spill propagation algorithm, using an approximation
1376       of maximal weighted independent sets.   This seems to be necessary to
1377       alleviate the negative effect in light of the slow down in (1).
1378    
1379         I'll write down the algorithm one of these days.
1380    
1381    8. MLRiscGen: frequencies
1382    
1383         I've added an annotation that states that all call gc blocks have zero
1384       execution frequencies.  This improves register allocation on the x86.
1385    
1386    BENCHMARKS
1387    ==========
1388    
1389       I've only perform the comparison on 110.25.
1390    
1391       The platforms are:
1392    
1393        HPPA  A four processor HP machine (E9000) with 5G of memory.
1394        X86   A 300Hhz Pentium II with 128M of memory, and
1395        SPARC An Ultra sparc 2 with 512M of memory.
1396    
1397       I used the following parameters for the SML benchmarks:
1398    
1399                 @SMLalloc
1400         HPPA    256k
1401         SPARC   512k
1402         X86     256k
1403    
1404    COMPILATION TIME
1405    ----------------
1406       Here are the numbers comparing the compilation times of the compilers.
1407       I've only compared 110.25 compiling the new sources versus
1408       a fixpoint version of the new compiler compiling the same.
1409    
1410                     110.25                                  New
1411               Total  Time in RA  Spill+Reload   Total  Time In RA Spill+Reload
1412         HPPA   627s    116s        2684+3584     599s    95s       1003+1879
1413         SPARC  892s    173s        2891+3870     708s    116s      1004+1880
1414         X86    999s    315s       94006+130691   987s    296s    108877+141957
1415    
1416                   110.25         New
1417                Code Size      Code Size
1418         HPPA   8596736         8561421
1419         SPARC  8974299         8785143
1420         X86    9029180         8716783
1421    
1422       So in summary, things are at least as good as before.   Dramatic
1423       reduction in compilation is obtained on the Sparc; I can't explain it,
1424       but it is reproducible.  Perhaps someone should try to reproduce this
1425       on their own machines.
1426    
1427    SML BENCHMARKS
1428    --------------
1429    
1430        On the average, all benchmarks perform at least as well as before.
1431    
1432          HPPA         Compilation Time     Spill+Reload      Run Time
1433                     110.25  New            110.25    New   110.25  New
1434    
1435          barnesHut  3.158  3.015  4.75%    1+1       0+0   2.980  2.922   2.00%
1436              boyer  6.152  5.708  7.77%    0+0       0+0   0.218  0.213   2.34%
1437       count-graphs  1.168  1.120  4.32%    0+0       0+0  22.705 23.073  -1.60%
1438                fft  0.877  0.792 10.74%    1+3       1+3   0.602  0.587   2.56%
1439        knuthBendix  3.180  2.857 11.32%    0+0       0+0   0.675  0.662   2.02%
1440             lexgen  6.190  5.290 17.01%    0+0       0+0   0.913  0.788  15.86%
1441               life  0.803  0.703 14.22%   25+25      0+0   0.153  0.140   9.52%
1442              logic  2.048  2.007  2.08%    6+6       1+1   4.133  4.008   3.12%
1443         mandelbrot  0.077  0.080 -4.17%    0+0       0+0   0.765  0.712   7.49%
1444             mlyacc 22.932 20.937  9.53%  154+181    32+57  0.468  0.430   8.91%
1445            nucleic  5.183  5.060  2.44%    2+2       0+0   0.125  0.120   4.17%
1446      ratio-regions  3.357  3.142  6.84%    0+0       0+0  116.225 113.173 2.70%
1447                ray  1.283  1.290 -0.52%    0+0       0+0   2.887  2.855   1.11%
1448             simple  6.307  6.032  4.56%   28+30      5+7   3.705  3.658   1.28%
1449                tsp  0.888  0.862  3.09%    0+0       0+0   7.040  6.893   2.13%
1450               vliw 24.378 23.455  3.94%  106+127    25+45  2.758  2.707   1.91%
1451      --------------------------------------------------------------------------
1452       Average                     6.12%                                   4.09%
1453    
1454          SPARC        Compilation Time     Spill+Reload      Run Time
1455                     110.25  New            110.25    New   110.25  New
1456    
1457          barnesHut  3.778  3.592  5.20%    2+2       0+0   3.648  3.453    5.65%
1458              boyer  6.632  6.110  8.54%    0+0       0+0   0.258  0.242    6.90%
1459       count-graphs  1.435  1.325  8.30%    0+0       0+0  33.672 34.737   -3.07%
1460                fft  0.980  0.940  4.26%    3+9       2+6   0.838  0.827    1.41%
1461        knuthBendix  3.590  3.138 14.39%    0+0       0+0   0.962  0.967   -0.52%
1462             lexgen  6.593  6.072  8.59%    1+1       0+0   1.077  1.078   -0.15%
1463               life  0.972  0.868 11.90%   26+26      0+0   0.143  0.140    2.38%
1464              logic  2.525  2.387  5.80%    7+7       1+1   5.625  5.158    9.05%
1465         mandelbrot  0.090  0.093 -3.57%    0+0       0+0   0.855  0.728   17.39%
1466             mlyacc 26.732 23.827 12.19%  162+189    32+57  0.550  0.560   -1.79%
1467            nucleic  6.233  6.197  0.59%    3+3       0+0   0.163  0.173   -5.77%
1468      ratio-regions  3.780  3.507  7.79%    0+0       0+0 133.993 131.035   2.26%
1469                ray  1.595  1.550  2.90%    1+1       0+0   3.440  3.418    0.63%
1470             simple  6.972  6.487  7.48%   29+32      5+7   3.523  3.525   -0.05%
1471                tsp  1.115  1.063  4.86%    0+0       0+0   7.393  7.265    1.77%
1472               vliw 27.765 24.818 11.87%  110+135    25+45  2.265  2.135    6.09%
1473      ----------------------------------------------------------------------------
1474       Average                     6.94%                                    2.64%
1475    
1476          X86          Compilation Time     Spill+Reload      Run Time
1477                     110.25  New            110.25    New   110.25  New
1478    
1479          barnesHut  5.530  5.420  2.03%  593+893   597+915   3.532  3.440   2.66%
1480              boyer  8.768  7.747 13.19%  493+199   301+289   0.327  0.297  10.11%
1481       count-graphs  2.040  2.010  1.49%  298+394   315+457  26.578 28.660  -7.26%
1482                fft  1.327  1.302  1.92%  112+209   115+210   1.055  0.962   9.71%
1483        knuthBendix  5.218  5.475 -4.69%  451+598   510+650   0.928  0.932  -0.36%
1484             lexgen  9.970  9.623  3.60% 1014+841  1157+885   0.947  0.928   1.97%
1485               life  1.183  1.183  0.00%  162+182   145+148   0.127  0.103  22.58%
1486              logic  3.285  3.512 -6.45%  514+684   591+836   5.682  5.577   1.88%
1487         mandelbrot  0.147  0.143  2.33%   38+41     33+54    0.703  0.690   1.93%
1488             mlyacc 35.457 32.763  8.22% 3496+4564 3611+4860  0.552  0.550   0.30%
1489            nucleic  7.100  6.888  3.07%  239+168   201+158   0.175  0.173   0.96%
1490      ratio-regions  6.388  6.843 -6.65% 1182+257   981+300  120.142 120.345 -0.17%
1491                ray  2.332  2.338 -0.29%  346+398   402+494   3.593  3.540   1.51%
1492             simple  9.912  9.903  0.08% 1475+941  1579+1168  3.057  3.178  -3.83%
1493                tsp  1.623  1.532  5.98%  266+200   250+211   8.045  7.878   2.12%
1494               vliw 33.947 35.470 -4.29% 2629+2774 2877+3171  2.072  1.890   9.61%
1495      ----------------------------------------------------------------------------
1496       Average                     1.22%                                     3.36%
1497    
1498    ----------------------------------------------------------------------
1499    Name: Allen Leung
1500    Date: 2000/03/23 16:25:00
1501    Tag: leunga-20000323-fix_x86_alpha
1502    Description:
1503    
1504    1. X86 fixes/changes
1505    
1506       a.  The old code generated for SETcc was completely wrong.
1507           The Intel optimization guide is VERY misleading.
1508    
1509    2. ALPHA fixes/changes
1510    
1511       a.  Added the instructions LDBU, LDWU, STB, STW as per Fermin's suggestion.
1512       b.  Added a new mode byteWordLoadStores to the functor parameter to Alpha()
1513       c.  Added reassociation code for address computation.
1514    
1515    ----------------------------------------------------------------------
1516    Name: Allen Leung
1517    Date: 2000/03/22 01:23:00
1518    Tag: leunga-20000322-fix_x86_hppa_ra
1519    Description:
1520    
1521    1. X86 fixes/changes
1522    
1523       a.  x86Rewrite bug with MUL3 (found by Lal)
1524       b.  Added the instructions FSTS, FSTL
1525    
1526    2. PA-RISC fixes/changes
1527    
1528       a.  B label should not be a delay slot candidate!  Why did this work?
1529       b.  ADDT(32, REG(32, r), LI n) now generates one instruction instead of two,
1530           as it should be.
1531       c.  The assembly syntax for fstds and fstdd was wrong.
1532       d.  Added the composite instruction COMICLR/LDO, which is the immediate
1533           operand variant of COMCLR/LDO.
1534    
1535    3. Generic MLRISC
1536    
1537       a.  shuffle.sml rewritten to be slightly more efficient
1538       b.  DIV bug in mltree-simplify fixed (found by Fermin)
1539    
1540    4. Register Allocator
1541    
1542       a.  I now release the interference graph earlier during spilling.
1543           May improve memory usage.
1544    
1545    ----------------------------------------------------------------------
1546    Name: Matthias Blume
1547    Date: 2000/03/14 14:15:32
1548    Tag: blume_main_v110p26p1_2
1549    Description:
1550    
1551    1. Tools.registerStdShellCmdTool (from smlnj/cm/tool.cm) takes an
1552    additional argument called "template" which is an optional string that
1553    specifies the layout of the tool command line.  See the CM manual for
1554    explanation.
1555    
1556    2. A special-purpose tool can be "registered" by simply dropping the
1557    corresponding <...>-tool.cm (and/or <...>-ext.cm) into the same
1558    directory where the .cm file lives that uses this tool.  (The
1559    behavior/misfeature until now was to look for the tool description
1560    files in the current working directory.)  As before, tool description
1561    files could also be anchored -- in which case they can live anywhere
1562    they like.  Following the recent e-mail discussion, this change should
1563    make it easier to have special-purpose tools that are shipped together
1564    with the sources of the program that uses them.
1565    
1566    ----------------------------------------------------------------------
1567    Name: Matthias Blume
1568  Date: 2000/03/10 07:48:34  Date: 2000/03/10 07:48:34
1569  Tag: blume_main_v110p26p1_1  Tag: blume_main_v110p26p1_1
1570  Description:  Description:
# Line 45  Line 1597 
1597  rebuild.  Having sets of the form <base><k>.{bin,boot}.<arch>-unix for  rebuild.  Having sets of the form <base><k>.{bin,boot}.<arch>-unix for
1598  <k>=1,2,... is normally not a good idea when invoking fixpt.  However,  <k>=1,2,... is normally not a good idea when invoking fixpt.  However,
1599  they might be the result of an earlier partial run of fixpt (which  they might be the result of an earlier partial run of fixpt (which
1600  perhaps got accidentially killed).  In this case, fixpt will quickly  perhaps got accidentally killed).  In this case, fixpt will quickly
1601  move through what exists before continuing where it left off earlier,  move through what exists before continuing where it left off earlier,
1602  and, thus, saves a lot of time.  and, thus, saves a lot of time.
1603    
# Line 95  Line 1647 
1647    it from that remote directory.    it from that remote directory.
1648    This should simplify installation further:  For machines that have    This should simplify installation further:  For machines that have
1649    access to the internet, just fetch <version>-config.tgz, unpack it,    access to the internet, just fetch <version>-config.tgz, unpack it,
1650    edit config/targets, and go (run config/install.sh).  The scipt will    edit config/targets, and go (run config/install.sh).  The script will
1651    fetch everything else that it might need all by itself.    fetch everything else that it might need all by itself.
1652    
1653    For CVS users, this mechanism is not relevant for source archives, but    For CVS users, this mechanism is not relevant for source archives, but

Legend:
Removed from v.577  
changed lines
  Added in v.678

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