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 959, Fri Oct 12 21:21:41 2001 UTC revision 1096, Tue Feb 26 16:59:02 2002 UTC
# Line 8  Line 8 
8  The form of an entry should be:  The form of an entry should be:
9    
10  Name:  Name:
11  Date:  Date: yyyy/mm/dd
12  Tag: <post-commit CVS tag>  Tag: <post-commit CVS tag>
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume
17    Date: 2002/02/26 12:00:00 EST
18    Tag: blume-20020226-ffi
19    Description:
20    
21    1. Fixed a minor bug in CM's "noweb" tool:
22       If numbering is turned off, then truly don't number (i.e., do not
23       supply the -L option to noweb).  The previous behavior was to supply
24       -L'' -- which caused noweb to use the "default" line numbering scheme.
25       Thanks to Chris Richards for pointing this out (and supplying the fix).
26    
27    2. Once again, I reworked some aspects of the FFI:
28    
29       A. The incomplete/complete type business:
30    
31       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
32         gone!
33       - ML types representing an incomplete type are now *equal* to
34         ML types representing their corresponding complete types (just like
35         in C).  This is still safe because ml-nlffigen will not generate
36         RTTI for incomplete types, nor will it generate functions that
37         require access to such RTTI.   But when ML code generated from both
38         incomplete and complete versions of the C type meet, the ML types
39         are trivially interoperable.
40    
41         NOTE:  These changes restore the full generality of the translation
42         (which was previously lost when I eliminated functorization)!
43    
44       B. Enum types:
45    
46       - Structure C now has a type constructor "enum" that is similar to
47         how the "su" constructor works.  However, "enum" is not a phantom
48         type because each "T enum" has values (and is isomorphic to
49         MLRep.Signed.int).
50       - There are generic access operations for enum objects (using
51         MLRep.Signed.int).
52       - ml-nlffigen will generate a structure E_foo for each "enum foo".
53         * The structure contains the definition of type "mlrep" (the ML-side
54         representation type of the enum).  Normally, mlrep is the same
55         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
56         then mlrep will be defined as a datatype -- thus facilitating
57         pattern matching on mlrep values.
58         ("-ec" will be suppressed if there are duplicate values in an
59          enumeration.)
60         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
61         will be generated for each C enum constant xxx.
62         * Conversion functions m2i and i2m convert between mlrep and
63         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
64         * Coversion functions c and ml convert between mlrep and "tag enum".
65         * Access functions (get/set) fetch and store mlrep values.
66       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
67         enumerations are merged into one single enumeration represented by
68         structure E_'.
69    
70    ----------------------------------------------------------------------
71    Name: Allen Leung
72    Date: 2002/02/25 04:45:00 EST
73    Tag: leunga-20020225-cps-spill
74    
75    This is a new implementation of the CPS spill phase.
76    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
77    In case of problems, replace it with the old file spill.sml
78    
79    The current compiler runs into some serious performance problems when
80    constructing a large record.  This can happen when we try to compile a
81    structure with many items.  Even a very simple structure like the following
82    makes the compiler slow down.
83    
84        structure Foo = struct
85           val x_1 = 0w1 : Word32.int
86           val x_2 = 0w2 : Word32.int
87           val x_3 = 0w3 : Word32.int
88           ...
89           val x_N = 0wN : Word32.int
90        end
91    
92    The following table shows the compile time, from N=1000 to N=4000,
93    with the old compiler:
94    
95    N
96    1000   CPS 100 spill                           0.04u  0.00s  0.00g
97           MLRISC ra                               0.06u  0.00s  0.05g
98              (spills = 0 reloads = 0)
99           TOTAL                                   0.63u  0.07s  0.21g
100    
101    1100   CPS 100 spill                           8.25u  0.32s  0.64g
102           MLRISC ra                               5.68u  0.59s  3.93g
103              (spills = 0 reloads = 0)
104           TOTAL                                   14.71u  0.99s  4.81g
105    
106    1500   CPS 100 spill                           58.55u  2.34s  1.74g
107           MLRISC ra                               5.54u  0.65s  3.91g
108              (spills = 543 reloads = 1082)
109           TOTAL                                   65.40u  3.13s  6.00g
110    
111    2000   CPS 100 spill                           126.69u  4.84s  3.08g
112           MLRISC ra                               0.80u  0.10s  0.55g
113              (spills = 42 reloads = 84)
114           TOTAL                                   129.42u  5.10s  4.13g
115    
116    3000   CPS 100 spill                           675.59u  19.03s  11.64g
117           MLRISC ra                               2.69u  0.27s  1.38g
118              (spills = 62 reloads = 124)
119           TOTAL                                   682.48u  19.61s  13.99g
120    
121    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
122           MLRISC ra                               4.96u  0.27s  2.72g
123              (spills = 85 reloads = 170)
124           TOTAL                                   2375.26u  57.21s  48.00g
125    
126    As you can see the old cps spill module suffers from some serious
127    performance problem but since I cannot decipher the old code fully,
128    innstead of patching the problems up, I'm reimplementing it
129    with a different algorithm.  The new code is more modular,
130    smaller when compiled, and substantially faster
131    (O(n log n) time and O(n) space).  Timing of the new spill module:
132    
133    4000  CPS 100 spill                           0.02u  0.00s  0.00g
134          MLRISC ra                               0.25u  0.02s  0.15g
135             (spills=1 reloads=3)
136          TOTAL                                   7.74u  0.34s  1.62g
137    
138    Implementation details:
139    
140    As far as I can tell, the purpose of the CPS spill module is to make sure the
141    number of live variables at any program point (the bandwidth)
142    does not exceed a certain limit, which is determined by the
143    size of the spill area.
144    
145    When the bandwidth is too large, we decrease the register pressure by
146    packing live variables into spill records.  How we achieve this is
147    completely different than what we did in the old code.
148    
149    First, there is something about the MLRiscGen code generator
150    that we should be aware of:
151    
152    o MLRiscGen performs code motion!
153    
154       In particular, it will move floating point computations and
155       address computations involving only the heap pointer to
156       their use sites (if there is only a single use).
157       What this means is that if we have a CPS record construction
158       statement
159    
160           RECORD(k,vl,w,e)
161    
162       we should never count the new record address w as live if w
163       has only one use (which is often the case).
164    
165       We should do something similar to floating point, but the transformation
166       there is much more complex, so I won't deal with that.
167    
168    Secondly, there are now two new cps primops at our disposal:
169    
170     1. rawrecord of record_kind option
171        This pure operator allocates some uninitialized storage from the heap.
172        There are two forms:
173    
174         rawrecord NONE [INT n]  allocates a tagless record of length n
175         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
176                                     and initializes the tag.
177    
178     2. rawupdate of cty
179          rawupdate cty (v,i,x)
180          Assigns to x to the ith component of record v.
181          The storelist is not updated.
182    
183    We use these new primops for both spilling and increment record construction.
184    
185     1. Spilling.
186    
187        This is implemented with a linear scan algorithm (but generalized
188        to trees).  The algorithm will create a single spill record at the
189        beginning of the cps function and use rawupdate to spill to it,
190        and SELECT or SELp to reload from it.  So both spills and reloads
191        are fine-grain operations.  In contrast, in the old algorithm
192        "spills" have to be bundled together in records.
193    
194        Ideally, we should sink the spill record construction to where
195        it is needed.  We can even split the spill record into multiple ones
196        at the places where they are needed.  But CPS is not a good
197        representation for global code motion, so I'll keep it simple and
198        am not attempting this.
199    
200     2. Incremental record construction (aka record splitting).
201    
202        Long records with many component values which are simulatenously live
203        (recall that single use record addresses are not considered to
204         be live) are constructed with rawrecord and rawupdate.
205        We allocate space on the heap with rawrecord first, then gradually
206        fill it in with rawupdate.  This is the technique suggested to me
207        by Matthias.
208    
209        Some restrictions on when this is applicable:
210        1. It is not a VECTOR record.  The code generator currently does not handle
211           this case. VECTOR record uses double indirection like arrays.
212        2. All the record component values are defined in the same "basic block"
213           as the record constructor.  This is to prevent speculative
214           record construction.
215    
216    ----------------------------------------------------------------------
217    Name: Allen Leung
218    Date: 2002/02/22 01:02:00 EST
219    Tag: leunga-20020222-mlrisc-tools
220    
221    Minor bug fixes in the parser and rewriter
222    
223    ----------------------------------------------------------------------
224    Name: Allen Leung
225    Date: 2002/02/21 20:20:00 EST
226    Tag: leunga-20020221-peephole
227    
228    Regenerated the peephole files.  Some contained typos in the specification
229    and some didn't compile because of pretty printing bugs in the old version
230    of 'nowhere'.
231    
232    ----------------------------------------------------------------------
233    Name: Allen Leung
234    Date: 2002/02/19 20:20:00 EST
235    Tag: leunga-20020219-mlrisc-tools
236    Description:
237    
238       Minor bug fixes to the mlrisc-tools library:
239    
240       1.  Fixed up parsing colon suffixed keywords
241       2.  Added the ability to shut the error messages up
242       3.  Reimplemented the pretty printer and fixed up/improved
243           the pretty printing of handle and -> types.
244       4.  Fixed up generation of literal symbols in the nowhere tool.
245       5.  Added some SML keywords to to sml.sty
246    
247    ----------------------------------------------------------------------
248    Name: Matthias Blume
249    Date: 2002/02/19 16:20:00 EST
250    Tag: blume-20020219-cmffi
251    Description:
252    
253    A wild mix of changes, some minor, some major:
254    
255    * All C FFI-related libraries are now anchored under $c:
256        $/c.cm      --> $c/c.cm
257        $/c-int.cm  --> $c/internals/c-int.cm
258        $/memory.cm --> $c/memory/memory.cm
259    
260    * "make" tool (in CM) now treats its argument pathname slightly
261      differently:
262        1. If the native expansion is an absolute name, then before invoking
263           the "make" command on it, CM will apply OS.Path.mkRelative
264           (with relativeTo = OS.FileSys.getDir()) to it.
265        2. The argument will be passed through to subsequent phases of CM
266           processing without "going native".  In particular, if the argument
267           was an anchored path, then "make" will not lose track of that anchor.
268    
269    * Compiler backends now "know" their respective C calling conventions
270      instead of having to be told about it by ml-nlffigen.  This relieves
271      ml-nlffigen from one of its burdens.
272    
273    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
274    
275    * Export C_DEBUG and C_Debug from $c/c.cm.
276    
277    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
278      subtyping relationship between incomplete pointers and their complete
279      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
280      with the convention of instantiating 'o with ('t, 'c) obj whenever
281      the pointer target type is complete.  In the incomplete case, 'o
282      will be instantiated with some "'c iobj" -- a type obtained by
283      using one of the functors PointerToIncompleteType or PointerToCompleteType.
284    
285      Operations that work on both incomplete and complete pointer types are
286      typed as taking an 'o ptr while operations that require the target to
287      be known are typed as taking some ('t, 'c) obj ptr.
288    
289      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
290      where void is an eqtype without any values.  This makes it possible
291      to work on voidptr values using functions meant to operate on light
292      incomplete pointers.
293    
294    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
295      been vastly simplified.
296    
297    ----------------------------------------------------------------------
298    Name: Matthias Blume
299    Date: 2002/02/19 10:48:00 EST
300    Tag: blume-20020219-pqfix
301    Description:
302    
303    Applied Chris Okasaki's bug fix for priority queues.
304    
305    ----------------------------------------------------------------------
306    Name: Matthias Blume
307    Date: 2002/02/15 17:05:00
308    Tag: Release_110_39
309    Description:
310    
311    Last-minute retagging is becoming a tradition... :-(
312    
313    This is the working release 110.39.
314    
315    ----------------------------------------------------------------------
316    Name: Matthias Blume
317    Date: 2002/02/15 16:00:00 EST
318    Tag: Release_110_39-orig
319    Description:
320    
321    Working release 110.39.  New bootfiles.
322    
323    (Update: There was a small bug in the installer so it wouldn't work
324    with all shells.  So I retagged. -Matthias)
325    
326    ----------------------------------------------------------------------
327    Name: Matthias Blume
328    Date: 2002/02/15 14:17:00 EST
329    Tag: blume-20020215-showbindings
330    Description:
331    
332    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
333    the latter can be useful for exploring what bindings are available at
334    the interactive prompt.  (The first function returns only the list
335    of symbols that are really bound, the second prints those but also the
336    ones that CM's autoloading mechanism knows about.)
337    
338    ----------------------------------------------------------------------
339    Name: Matthias Blume
340    Date: 2002/02/15 12:08:00 EST
341    Tag: blume-20020215-iptrs
342    Description:
343    
344    Two improvements to ml-nlffigen:
345    
346      1. Write files only if they do not exist or if their current contents
347         do not coincide with what's being written.  (That is, avoid messing
348         with the time stamps unless absolutely necessary.)
349    
350      2. Implement a "repository" mechanism for generated files related
351         to "incomplete pointer types".   See the README file for details.
352    
353    ----------------------------------------------------------------------
354    Name: Matthias Blume
355    Date: 2002/02/14 11:50:00 EST
356    Tag: blume-20020214-quote
357    Description:
358    
359    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
360    because of the new and improved tag generation scheme.  (Thanks to Allen
361    Leung for pointing it out.)
362    
363    ----------------------------------------------------------------------
364    Name: Lal George
365    Date: 2002/02/14 09:55:27 EST 2002
366    Tag: george-20020214-isabelle-bug
367    Description:
368    
369    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
370    of Isabelle on the x86.
371    
372    From Allen:
373    -----------
374     I've found the problem:
375    
376         in ra-core.sml, I use the counter "blocked" to keep track of the
377         true number of elements in the freeze queue.  When the counter goes
378         to zero, I skip examining the queue.  But I've messed up the
379         bookkeeping in combine():
380    
381             else ();
382             case !ucol of
383               PSEUDO => (if !cntv > 0 then
384                     (if !cntu > 0 then blocked := !blocked - 1 else ();
385                                        ^^^^^^^^^^^^^^^^^^^^^^^
386                      moveu := mergeMoveList(!movev, !moveu)
387                     )
388                  else ();
389    
390         combine() is called to coalesce two nodes u and v.
391         I think I was thinking that if the move counts of u and v are both
392         greater than zero then after they are coalesced then one node is
393         removed from the freeze queue.  Apparently I was thinking that
394         both u and v are of low degree, but that's clearly not necessarily true.
395    
396    
397    02/12/2002:
398        Here's the patch.  HOL now compiles.
399    
400        I don't know how this impact on performance (compile
401        time or runtime).  This bug caused the RA (especially on the x86)
402        to go thru the potential spill phase when there are still nodes on the
403        freeze queue.
404    
405    
406    
407    
408    ----------------------------------------------------------------------
409    Name: Matthias Blume
410    Date: 2002/02/13 22:40:00 EST
411    Tag: blume-20020213-fptr-rtti
412    Description:
413    
414    Fixed a bug in ml-nlffigen that was introduced with one of the previous
415    updates.
416    
417    ----------------------------------------------------------------------
418    Name: Matthias Blume
419    Date: 2002/02/13 16:41:00 EST
420    Tag: blume-20020213-cmlpq
421    Description:
422    
423    Added new priority queue export symbols (which have just been added to
424    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
425    not compile and the installer would choke.)
426    
427    ----------------------------------------------------------------------
428    Name: Matthias Blume
429    Date: 2002/02/13 16:15:00 EST
430    Tag: blume-20020213-various
431    Description:
432    
433    1. More tweaks to ml-nlffigen:
434    
435       - better internal datastructures (resulting in slight speedup)
436       - "-match" option requires exact match
437       - "localized" gensym counters (untagged structs/unions nested within
438         other structs/unions or within typedefs get a fresh counter; their
439         tag will be prefixed by a concatenation of their parents' tags)
440       - bug fixes (related to calculation of transitive closure of types
441         to be included in the output)
442    
443    2. Minor Basis updates:
444    
445       - added implementations for List.collate and Option.app
446    
447    ----------------------------------------------------------------------
448    Name: Matthias Blume
449    Date: 2002/02/11 15:55:00 EST
450    Tag: blume-20020211-gensym
451    Description:
452    
453    Added a "-gensym" option to command line of ml-nlffigen.  This can be
454    used to specify a "stem" -- a string that is inserted in all "gensym'd"
455    names (ML structure names that correspond to unnamed C structs, unions,
456    and enums), so that separate runs of ml-nlffigen do not clash.
457    
458    ----------------------------------------------------------------------
459    Name: Matthias Blume
460    Date: 2002/02/11 12:05:00 EST
461    Tag: blume-20020211-gensml
462    Description:
463    
464    A quick fix for a problem with GenSML (in the pgraph-util library):
465    Make generation of toplevel "local" optional.  (Strictly speaking,
466    signature definitions within "local" are not legal SML.)
467    
468    Other than that: updates to INSTALL and cm/TODO.
469    
470    ----------------------------------------------------------------------
471    Name: Matthias Blume
472    Date: 2002/02/08 15:00:00 EST
473    Tag: blume-20020208-uniquepid
474    Description:
475    
476    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
477    
478    1. The installer (config/install.sh) has gotten smarter:
479    
480         - Configuration options are a bit easier to specify now
481           (in config/targets).
482         - Bug in recognizing .tar.bz2 files fixed.
483         - Installer automatically resolves dependencies between
484           configuration options (e.g., if you ask for eXene, you will
485           also get cml -- regardless whether you asked for it or not).
486         - Installer can run in "quieter mode" by setting the environment
487           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
488           "completely silent", though.
489         - Build HashCons library as part of smlnj-lib.
490    
491    2. A new scheme for assigning persistent identifiers to compilation
492       units (and, by extension, to types etc.) has been put into place.
493       This fixes a long-standing bug where types and even dynamic values
494       can get internally confused, thereby compromising type safety
495       (abstraction) and dynamic correctness.  See
496    
497         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
498    
499       for an example of how things could go wrong until now.
500    
501       The downside of the new scheme is that pids are not quite as
502       persistent as they used to be: CM will generate a fresh pid
503       for every compilation unit that it thinks it sees for the first
504       time.  That means that if you compile starting from a clean, fresh
505       source tree at two different times, you end up with different
506       binaries.
507    
508       Cutoff recompilation, however, has not been compromised because
509       CM keeps pid information in special caches between runs.
510    
511    ----------------------------------------------------------------------
512    Name: Lal George
513    Date: 2002/02/07 15:34:13 EST 2002
514    Tag: <none>
515    Description:
516    
517    Compilers that generate assembly code may produce  global labels
518    whose value is resolved at link time. The various peephole optimization
519    modules did not take this in account.
520    
521    TODO. The Labels.addrOf function should really return an option
522    type so that clients are forced to deal with this issue, rather
523    than an exception being raised.
524    
525    ----------------------------------------------------------------------
526    Name: Lal George
527    Date: 2002/02/06 13:55:02 EST
528    Tag: george-20020206-ra-breakup
529    Description:
530    
531    1. A bug fix from Allen.
532    
533        A typo causes extra fstp %st(0)'s to be generated at compensation
534        edges, which might cause stack underflow traps at runtime.  This
535        occurs in fft where there are extraneous fstps right before the 'into'
536        trap instruction (in this case they are harmless since none of the
537        integers overflow.)
538    
539    2. Pulled out various utility modules that were embedded in the modules
540       of the register allocator. I need these modules for other purposes, but
541       they are not complete enough to put into a library (just yet).
542    ----------------------------------------------------------------------
543    Name: Matthias Blume
544    Date: 2002/01/31 16:05:00 EST
545    Tag: blume-20020131-sparc-ccalls
546    Description:
547    
548    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
549       of extra stack space by mistake.  Fixed.
550    
551    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
552    
553    ----------------------------------------------------------------------
554    Name: Allen Leung
555    Date: 2002/01/30
556    Tag: leunga-20020130-nowhere-bug-fix
557    Description:
558    
559       MLRISC bug fixes:
560       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
561       2. MachineInt.fromString was negating its value.
562    
563    ----------------------------------------------------------------------
564    Name: Matthias Blume
565    Date: 2002/01/29
566    Tag: blume-20020129-INSTALL
567    Description:
568    
569    - Added somewhat detailed installation instructions (file INSTALL).
570    - Fixed curl-detection bug in config/install.sh.
571    - It is now possible to select the URL getter using the URLGETTER
572      environment variable:
573    
574          not set / "unknown"      --> automatic detection (script tries wget,
575                                       curl, and lynx)
576          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
577                                       how to properly invoke them)
578          other                    --> use $URLGETTER directly, it must take
579                                       precisely two command-line arguments
580                                       (source URL and destination file name)
581    
582    ----------------------------------------------------------------------
583    Name: Matthias Blume
584    Date: 2002/01/28
585    Tag: blume-20020128-sparc-ccalls
586    Description:
587    
588    - Fixed problem with calculation of "used" registers in sparc-c-calls.
589    - Make use of the allocParam argument in sparc-c-calls.
590    
591    ----------------------------------------------------------------------
592    Name: Matthias Blume
593    Date: 2002/01/28
594    Tag: blume-20020128-allocParam
595    Description:
596    
597    John Reppy:  Changes c-calls API to accept client-callback for
598    allocating extra stack space.
599    me: Corresponding changes to mlriscGen (using a dummy argument that
600        does not change the current behavior).
601    
602    ----------------------------------------------------------------------
603    Name: Matthias Blume
604    Date: 2002/01/28 12:00:00
605    Tag: Release_110_38
606    Description:
607    
608    This time for real!!!
609    
610    ----------------------------------------------------------------------
611    Name: Matthias Blume
612    Date: 2002/01/28 10:56:00 EST
613    Tag: blume-20020128-retraction
614    Description:
615    
616    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
617       with blume-Release_110_38-retracted.)
618    
619    1. Fixed a problem with incorrect rounding modes in real64.sml.
620       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
621    
622    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
623       and enums fixed.  The naming of corresponding ML identifiers should
624       now be consistent again.
625    
626    ----------------------------------------------------------------------
627    Name: Allen Leung
628    Date: 2002/01/27
629    Tag: leunga-20020127-nowhere
630    Description:
631    
632       Added a target called nowhere in the configuration scripts.
633       Enabling this will build the MLRISC 'nowhere' tool (for translating
634       programs with where-clauses into legal SML code) during installation.
635    
636    ----------------------------------------------------------------------
637    Name: Matthias Blume
638    Date: 2002/01/25 21:27:00 EST
639    Tag: blume-Release_110_38-retracted
640    Description:
641    
642    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
643    
644    README will be added later.
645    
646    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
647    (Release_110_38) removed.  Reason: Last-minute bug fixes.
648    
649    ----------------------------------------------------------------------
650    Name: Matthias Blume
651    Date: 2002/01/25
652    Tag: blume-20020125-ffi
653    Description:
654    
655    A large number of tweaks and improvements to ml-nlffi-lib and
656    ml-nlffigen:
657    
658       - ML represenation types have been streamlined
659       - getter and setter functions work with concrete values, not abstract
660         ones where possible
661       - ml-nlffigen command line more flexible (see README file there)
662       - some bugs have been fixed (hopefully)
663    
664    ----------------------------------------------------------------------
665    Name: Lal George
666    Date: 2002/01/24
667    Tag: george-20020124-risc-ra-interface
668    Description:
669    
670       There is a dramatic simplification in the interface to the
671       register allocator for RISC architectures as a result of making
672       parallel copy instructions explicit.
673    
674    ----------------------------------------------------------------------
675    Name: Matthias Blume
676    Date: 2002/01/22
677    Tag: blume-20020122-x86-ccalls
678    Description:
679    
680    Bug fix for c-calls on x86 (having to do with how char- and
681    short-arguments are being handled).
682    
683    ----------------------------------------------------------------------
684    Name: Matthias Blume
685    Date: 2002/01/21
686    Tag: blume-20020121-ff
687    Description:
688    
689    Another day of fiddling with the FFI...
690    
691    1. Bug fix/workaround:  CKIT does not complain about negative array
692       dimensions, so ml-nlffigen has to guard itself against this possibility.
693       (Otherwise a negative dimension would send it into an infinite loop.)
694    
695    2. Some of the abstract types (light objects, light pointers, most "base"
696       types) in structure C are now eqtypes.
697    
698    3. Added constructors and test functions for NULL function pointers.
699    
700    ----------------------------------------------------------------------
701    Name: Matthias Blume
702    Date: 2002/01/18
703    Tag: blume-20020118-ready-for-new-release
704    Description:
705    
706    Made config/srcarchiveurl point to a new place.  (Will provide boot
707    files shortly.)
708    
709    Maybe we christen this to be 110.38?
710    
711    ----------------------------------------------------------------------
712    Name: Matthias Blume
713    Date: 2002/01/18
714    Tag: blume-20020118-more-ffifiddle
715    Description:
716    
717    Today's FFI fiddling:
718    
719      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
720        C.Set.* that use concrete (MLRep.*) arguments and results instead
721        of abstract ones.
722    
723      - Provided word-style bit operations etc. for "int" representation
724        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
725        and Long.
726    
727    ----------------------------------------------------------------------
728    Name: Matthias Blume
729    Date: 2002/01/18
730    Tag: blume-20020118-use-x86-fp
731    Description:
732    
733    Now that x86-fast-fp seems to be working, I turned it back on again
734    by default.  (Seems to work fine now, even with the FFI.)
735    
736    Other than that, I added some documentation about the FFI to
737    src/ml-nlffigen/README and updated the FFI test examples in
738    src/ml-nlffi-lib/Tests/*.
739    
740    ----------------------------------------------------------------------
741    Name: Allen Leung
742    Date: 2002/01/17
743    Tag: leunga-20020117-x86-fast-fp-call
744    Description:
745    
746       1. Fixed a problem with handling return fp values when x86's fast fp
747          mode is turned on.
748    
749       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
750          of %f32.
751    
752       3. Added a constructor INT32lit to the ast of MLRISC tools.
753    
754    ----------------------------------------------------------------------
755    Name: Matthias Blume
756    Date: 2002/01/16
757    Tag: blume-20020116-ffifiddle
758    Description:
759    
760    More fiddling with the FFI interface:
761    
762     - Make constness 'c instead of rw wherever possible.  This eliminates
763       the need for certain explicit coercions.  (However, due to ML's
764       value polymorphism, there will still be many cases where explicit
765       coercions are necessary.  Phantom types are not the whole answer
766       to modeling a subtyping relationship in ML.)
767    
768     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
769    
770     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
771       clumsy workaround for varargs problem.)
772    
773    ----------------------------------------------------------------------
774    Name: Lal George
775    Date: 2002/01/15
776    Tag: <none>
777    Description:
778    
779    1. Since COPY instructions are no longer native to the architecture,
780       a generic functor can be used to implement the expandCopies function.
781    
782    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
783       TEXT segment.
784    
785    ----------------------------------------------------------------------
786    Name: Matthias Blume
787    Date: 2002/01/15
788    Tag: blume-20020115-ffiupdates
789    Description:
790    
791    1. Fix for bug resulting in single-precision float values being returned
792       incorrectly from FFI calls.
793    
794    2. Small modifications to C FFI API:
795    
796        - memory-allocation routines return straight objects (no options)
797          and raise an exception in out-of-memory situations
798        - unsafe extensions to cast between function pointers and pointers
799          from/to ints
800        - added structure C_Debug as an alternative to structure C where
801          pointer-dereferencing (|*| and |*!) always check for null-pointers
802        - added open_lib' to DynLinkage;  open_lib' works like open_lib
803          but also takes a (possibly empty) list of existing library handles
804          that the current library depends on
805    
806    ----------------------------------------------------------------------
807    Name: Matthias Blume
808    Date: 2002/01/10
809    Tag: blume-20020110-newffigen
810    Description:
811    
812    1. Updates to portable graph code.
813    
814    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
815       more scalable now so that even huge interfaces such as the one
816       for GTK compile in finite time and space. :-)
817       See src/ml-nlffigen/README for details on what's new.
818    
819    ----------------------------------------------------------------------
820    Name: Lal George
821    Date: 2001/01/09 14:31:35 EST 2002
822    Tag: george-20011206-rm-native-copy
823    Description:
824    
825            Removed the native COPY and FCOPY instructions
826            from all the architectures and replaced it with the
827            explicit COPY instruction from the previous commit.
828    
829            It is now possible to simplify many of the optimizations
830            modules that manipulate copies. This has not been
831            done in this change.
832    
833    ----------------------------------------------------------------------
834    Name: Lal George
835    Date: 2001/12/06 16:50:13 EST 2001
836    Tag: george-20011206-mlrisc-instruction
837    Description:
838    
839    Changed the representation of instructions from being fully abstract
840    to being partially concrete. That is to say:
841    
842      from
843            type instruction
844    
845      to
846            type instr                              (* machine instruction *)
847    
848            datatype instruction =
849                LIVE of {regs: C.cellset, spilled: C.cellset}
850              | KILL of {regs: C.cellset, spilled: C.cellset}
851              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
852              | ANNOTATION of {i: instruction, a: Annotations.annotation}
853              | INSTR of instr
854    
855    This makes the handling of certain special instructions that appear on
856    all architectures easier and uniform.
857    
858    LIVE and KILL say that a list of registers are live or killed at the
859    program point where they appear. No spill code is generated when an
860    element of the 'regs' field is spilled, but the register is moved to
861    the 'spilled' (which is present, more for debugging than anything else).
862    
863    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
864    We used to generate:
865    
866            DEFFREG f1
867            f1 := f2 + f3
868            trapb
869    
870    but now generate:
871    
872            f1 := f2 + f3
873            trapb
874            LIVE {regs=[f1,f2,f3], spilled=[]}
875    
876    Furthermore, the DEFFREG (hack) required that all floating point instruction
877    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
878    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
879    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
880    this architecture).
881    
882    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
883    all the architectures. This will result in further simplification of the
884    register allocator that must be aware of them for coalescing purposes, and
885    will also simplify certain aspects of the machine description that provides
886    callbacks related to parallel copies.
887    
888    ANNOTATION should be obvious, and now INSTR represents the honest to God
889    machine instruction set!
890    
891    The <arch>/instructions/<arch>Instr.sml files define certain utility
892    functions for making porting easier -- essentially converting upper case
893    to lower case. All machine instructions (of type instr) are in upper case,
894    and the lower case form generates an MLRISC instruction. For example on
895    the alpha we have:
896    
897      datatype instr =
898         LDA of {r:cell, b:cell, d:operand}
899       | ...
900    
901      val lda : {r:cell, b:cell, d:operand} -> instruction
902        ...
903    
904    where lda is just (INSTR o LDA), etc.
905    
906    ----------------------------------------------------------------------
907    Name: Matthias Blume
908    Date: 2001/11/22 21:40:00 EST
909    Tag: Release_110_37
910    Description:
911    
912    Release 110.37.  This time for real.
913    
914    ----------------------------------------------------------------------
915    Name: Matthias Blume
916    Date: 2001/11/21 16:35:00 EST
917    Tag: blume-20011121-foot-in-mouth
918    Description:
919    
920    Removed the "Release_110_37" tag because of a serious bug.
921    This will be re-tagged once the bug is fixed.
922    
923    ----------------------------------------------------------------------
924    Name: Matthias Blume
925    Date: 2001/11/21 16:14:00 EST
926    Tag: blume-20011121-forgottenfile
927    Description:
928    
929    Forgot to add a file.  (Just a .tex-file -- part of
930    the CM manual source.)
931    
932    ----------------------------------------------------------------------
933    Name: Matthias Blume
934    Date: 2001/11/21 16:10:00 EST
935    Tag: blume-20011121-invalid_110_37
936    Description:
937    
938    Note: I removed the original tag "Release_110_37" from this commit
939          because we found a serious bug in all non-x86 backends.
940          - Matthias
941    
942    1. Modifications to the SML/NJ code generator and to the runtime system
943       so that code object name strings are directly inserted into code
944       objects at code generation time.  The only business the runtime system
945       has with this is now to read the name strings on occasions.
946       (The encoding of the name string has also changed somewhat.)
947    
948    2. CM now implements a simple "set calculus" for specifying export lists.
949       In particular, it is now possible to refer to the export lists of
950       other libraries/groups/sources and form unions as well as differences.
951       See the latest CM manual for details.
952    
953    3. An separate notion of "proxy" libraries has again be eliminated from
954       CM's model.  (Proxy libraries are now simply a special case of using
955       the export list calculus.)
956    
957    4. Some of the existing libraries now take advantage of the new set
958       calculus.
959       (Notice that not all libraries have been converted because some
960       of the existing .cm-files are supposed to be backward compatible
961       with 110.0.x.)
962    
963    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
964       "General.exnMessage"!  The former relies on a certain hook to be
965       initialized, and that often does not happen in the stand-alone case.)
966    
967    ----------------------------------------------------------------------
968    Name: Lal George
969    Date: 2001/11/21  13:56:18 EST
970    Tag: george-2001121-pseudo-ops
971    Description:
972    
973      Implemented a complete redesign of MLRISC pseudo-ops. Now there
974      ought to never be any question of incompatabilities with
975      pseudo-op syntax expected by host assemblers.
976    
977      For now, only modules supporting GAS syntax are implemented
978      but more should follow, such as MASM, and vendor assembler
979      syntax, e.g. IBM as, Sun as, etc.
980    
981    ----------------------------------------------------------------------
982    Name: Matthias Blume
983    Date: 2001/11/14 11:52:00 EST
984    Tag: blume-20011114-srcname
985    Description:
986    
987    1. Routed the name of the current source file to mlriscgen where it
988       should be directly emitted into the code object.  (This last part
989       is yet to be done.)
990    
991    2. Some cleanup of the pgraph code to make it match the proposal that
992       I put out the other day.  (The proposal notwithstanding, things are
993       still in flux here.)
994    
995    ----------------------------------------------------------------------
996    Name: Lal George
997    Date: 2001/11/14 09:44:04 EST
998    Tag:
999    Description:
1000    
1001      Fix for a backpatching bug reported by Allen.
1002    
1003      Because the boundary between short and long span-dependent
1004      instructions is +/- 128, there are an astounding number of
1005      span-dependent instructions whose size is over estimated.
1006    
1007      Allen came up with the idea of letting the size of span
1008      dependent instructions be non-monotonic, for a maxIter
1009      number of times, after which the size must be monotonically
1010      increasing.
1011    
1012      This table shows the number of span-dependent instructions
1013      whose size was over-estimated as a function of maxIter, for the
1014      file Parse/parse/ml.grm.sml:
1015    
1016         maxIter            # of instructions:
1017            10                      687
1018            20                      438
1019            30                      198
1020            40                        0
1021    
1022      In compiling the compiler, there is no significant difference in
1023      compilation speed between maxIter=10 and maxIter=40. Actually,
1024      my measurements showed that maxIter=40 was a tad faster than
1025      maxIter=10! Also 96% of the  files in the compiler reach a fix
1026      point within 13 iterations, so fixing maxIter at 40, while high,
1027      is okay.
1028    
1029    ----------------------------------------------------------------------
1030    Name: Matthias Blume
1031    Date: 2001/10/31 15:25:00 EST
1032    Tag: blume-20011031-pgraph
1033    Description:
1034    
1035    CKIT:
1036    * Changed the "Function" constructor of type Ast.ctype to carry optional
1037      argument identifiers.
1038    * Changed the return type of TypeUtil.getFunction accordingly.
1039    * Type equality ignores the argument names.
1040    * TypeUtil.composite tries to preserve argument names but gives up quickly
1041      if there is a mismatch.
1042    
1043    installation script:
1044    * attempts to use "curl" if available (unless "wget" is available as well)
1045    
1046    CM:
1047    * has an experimental implementation of "portable graphs" which I will
1048      soon propose as an implementation-independent library format
1049    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
1050    
1051    NLFFI-LIB:
1052    * some cleanup (all cosmetic)
1053    
1054    NLFFIGEN:
1055    * temporarily disabled the mechanism that suppresses ML output for
1056      C definitions whose identifiers start with an underscore character
1057    * generate val bindings for enum constants
1058    * user can request that only one style (light or heavy) is being used;
1059      default is to use both (command-line arguments: -heavy and -light)
1060    * fixed bug in handling of function types involving incomplete pointers
1061    * generate ML entry points that take record arguments (i.e., using
1062      named arguments) for C functions that have a prototype with named
1063      arguments
1064      (see changes to CKIT)
1065    
1066    ----------------------------------------------------------------------
1067    Name: Allen Leung
1068    Date: 2001/10/27 20:34:00 EDT
1069    Tag: leunga-20011027-x86-fast-fp-call
1070    Description:
1071    
1072       Fixed the bug described in blume-20010920-slowfp.
1073    
1074       The fix involves
1075          1. generating FCOPYs in FSTP in ia32-svid
1076          2. marking a CALL with the appropriate annotation
1077    
1078    ----------------------------------------------------------------------
1079    Name: Matthias Blume
1080    Date: 2001/10/16 11:32:00 EDT
1081    Tag: blume-20011016-netbsd
1082    Description:
1083    
1084    Underscore patch from Chris Richards (fixing problem with compiling
1085    runtime system under recent NetBSD).
1086    
1087    ----------------------------------------------------------------------
1088  Name: Allen Leung  Name: Allen Leung
1089  Date: 2001/10/12 17:18:32 EDT 2001  Date: 2001/10/12 17:18:32 EDT 2001
1090  Tag: leung-20011012-x86-printflowgraph  Tag: leung-20011012-x86-printflowgraph
# Line 88  Line 1160 
1160    
1161  ----------------------------------------------------------------------  ----------------------------------------------------------------------
1162  Name: Matthias Blume  Name: Matthias Blume
 >>>>>>> 1.169  
1163  Date: 2001/09/18 15:35:00 EDT  Date: 2001/09/18 15:35:00 EDT
1164  Tag: blume-20010918-readme11036  Tag: blume-20010918-readme11036
1165  Description:  Description:

Legend:
Removed from v.959  
changed lines
  Added in v.1096

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