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 1055, Thu Feb 7 20:38:59 2002 UTC revision 1132, Mon Mar 11 18:32:50 2002 UTC
# Line 13  Line 13 
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume
17    Date: 2002/03/11 13:30:00 EST
18    Tag: blume-20020311-mltreeeval
19    Description:
20    
21    A functor parameter was missing.
22    
23    ----------------------------------------------------------------------
24    Name: Allen Leung
25    Date: 2002/03/11 10:30:00 EST
26    Tag: leunga-20020310-runtime-string0
27    Description:
28    
29       The representation of the empty string now points to a
30    legal null terminated C string instead of unit.  It is now possible
31    to convert an ML string into C string with InlineT.CharVector.getData.
32    This compiles into one single machine instruction.
33    
34    ----------------------------------------------------------------------
35    Name: Allen Leung
36    Date: 2002/03/10 23:55:00 EST
37    Tag: leunga-20020310-x86-call
38    Description:
39    
40       Added machine generation for CALL instruction (relative displacement mode)
41    
42    ----------------------------------------------------------------------
43    Name: Matthias Blume
44    Date: 2002/03/08 16:05:00
45    Tag: blume-20020308-entrypoints
46    Description:
47    
48    Version number bumped to 110.39.1.  NEW BOOTFILES!
49    
50    Entrypoints: non-zero offset into a code object where execution should begin.
51    
52    - Added the notion of an entrypoint to CodeObj.
53    - Added reading/writing of entrypoint info to Binfile.
54    - Made runtime system bootloader aware of entrypoints.
55    - Use the address of the label of the first function given to mlriscGen
56      as the entrypoint.  This address is currently always 0, but it will
57      not be 0 once we turn on block placement.
58    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
59      with entry points) from mlriscGen.
60    
61    ----------------------------------------------------------------------
62    Name: Allen Leung
63    Date: 2002/03/07 20:45:00 EST
64    Tag: leunga-20020307-x86-cmov
65    Description:
66    
67       Bug fixes for CMOVcc on x86.
68    
69       1. Added machine code generation for CMOVcc
70       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
71       3. CMOVcc cannot have an immediate operand as argument.
72    
73    ----------------------------------------------------------------------
74    Name: Matthias Blume
75    Date: 2002/03/07 16:15:00 EST
76    Tag: blume-20020307-controls
77    Description:
78    
79    This is a very large but mostly boring patch which makes (almost)
80    every tuneable compiler knob (i.e., pretty much everything under
81    Control.* plus a few other things) configurable via both the command
82    line and environment variables in the style CM did its configuration
83    until now.
84    
85    Try starting sml with '-h' (or, if you are brave, '-H')
86    
87    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
88    implements the underlying generic mechanism.
89    
90    The interface to some of the existing such facilities has changed somewhat.
91    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
92    (The getFoo interface is still there for backward-compatibility, but its
93    use is deprecated.)
94    
95    The ml-build script passes -Cxxx=yyy command-line arguments through so
96    that one can now twiddle the compiler settings when using this "batch"
97    compiler.
98    
99    TODO items:
100    
101    We should go through and throw out all controls that are no longer
102    connected to anything.  Moreover, we should go through and provide
103    meaningful (and correct!) documentation strings for those controls
104    that still are connected.
105    
106    Currently, multiple calls to Controls.new are accepted (only the first
107    has any effect).  Eventually we should make sure that every control
108    is being made (via Controls.new) exactly once.  Future access can then
109    be done using Controls.acc.
110    
111    Finally, it would probably be a good idea to use the getter-setter
112    interface to controls rather than ref cells.  For the time being, both
113    styles are provided by the Controls module, but getter-setter pairs are
114    better if thread-safety is of any concern because they can be wrapped.
115    
116    *****************************************
117    
118    One bug fix: The function blockPlacement in three of the MLRISC
119    backpatch files used to be hard-wired to one of two possibilities at
120    link time (according to the value of the placementFlag).  But (I
121    think) it should rather sense the flag every time.
122    
123    *****************************************
124    
125    Other assorted changes (by other people who did not supply a HISTORY entry):
126    
127    1. the cross-module inliner now works much better (Monnier)
128    2. representation of weights, frequencies, and probabilities in MLRISC
129       changed in preparation of using those for weighted block placement
130       (Reppy, George)
131    
132    ----------------------------------------------------------------------
133    Name: Lal George
134    Date: 2002/03/07 14:44:24 EST 2002
135    Tag: george-20020307-weighted-block-placement
136    
137    Tested the weighted block placement optimization on all architectures
138    (except the hppa) using AMPL to generate the block and edge frequencies.
139    Changes were required in the machine properties to correctly
140    categorize trap instructions. There is an MLRISC flag
141    "weighted-block-placement" that can be used to enable weighted block
142    placement, but this will be ineffective without block/edge
143    frequencies (coming soon).
144    
145    
146    ----------------------------------------------------------------------
147    Name: Lal George
148    Date: 2002/03/05 17:24:48 EST
149    Tag: george-20020305-linkage-cluster
150    
151    In order to support the block placement optimization, a new cluster
152    is generated as the very first cluster (called the linkage cluster).
153    It contains a single jump to the 'real' entry point for the compilation
154    unit. Block placement has no effect on the linkage cluster itself, but
155    all the other clusters  have full freedom in the manner in which they
156    reorder blocks or functions.
157    
158    On the x86 the typical linkage code that is generated is:
159       ----------------------
160            .align 2
161       L0:
162            addl    $L1-L0, 72(%esp)
163            jmp     L1
164    
165    
166            .align  2
167       L1:
168       ----------------------
169    
170    72(%esp) is the memory location for the stdlink register. This
171    must contain the address of the CPS function being called. In the
172    above example, it contains the address of  L0; before
173    calling L1 (the real entry point for the compilation unit), it
174    must contain the address for L1, and hence
175    
176            addl $L1-L0, 72(%esp)
177    
178    I have tested this on all architectures except the hppa.The increase
179    in code size is of course negligible
180    
181    ----------------------------------------------------------------------
182    Name: Allen Leung
183    Date: 2002/03/03 13:20:00 EST
184    Tag: leunga-20020303-mlrisc-tools
185    
186      Added #[ ... ] expressions to mlrisc tools
187    
188    ----------------------------------------------------------------------
189    Name: Matthias Blume
190    Date: 2002/02/27 12:29:00 EST
191    Tag: blume-20020227-cdebug
192    Description:
193    
194    - made types in structure C and C_Debug to be equal
195    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
196    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
197    
198    ----------------------------------------------------------------------
199    Name: Matthias Blume
200    Date: 2002/02/26 12:00:00 EST
201    Tag: blume-20020226-ffi
202    Description:
203    
204    1. Fixed a minor bug in CM's "noweb" tool:
205       If numbering is turned off, then truly don't number (i.e., do not
206       supply the -L option to noweb).  The previous behavior was to supply
207       -L'' -- which caused noweb to use the "default" line numbering scheme.
208       Thanks to Chris Richards for pointing this out (and supplying the fix).
209    
210    2. Once again, I reworked some aspects of the FFI:
211    
212       A. The incomplete/complete type business:
213    
214       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
215         gone!
216       - ML types representing an incomplete type are now *equal* to
217         ML types representing their corresponding complete types (just like
218         in C).  This is still safe because ml-nlffigen will not generate
219         RTTI for incomplete types, nor will it generate functions that
220         require access to such RTTI.   But when ML code generated from both
221         incomplete and complete versions of the C type meet, the ML types
222         are trivially interoperable.
223    
224         NOTE:  These changes restore the full generality of the translation
225         (which was previously lost when I eliminated functorization)!
226    
227       B. Enum types:
228    
229       - Structure C now has a type constructor "enum" that is similar to
230         how the "su" constructor works.  However, "enum" is not a phantom
231         type because each "T enum" has values (and is isomorphic to
232         MLRep.Signed.int).
233       - There are generic access operations for enum objects (using
234         MLRep.Signed.int).
235       - ml-nlffigen will generate a structure E_foo for each "enum foo".
236         * The structure contains the definition of type "mlrep" (the ML-side
237         representation type of the enum).  Normally, mlrep is the same
238         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
239         then mlrep will be defined as a datatype -- thus facilitating
240         pattern matching on mlrep values.
241         ("-ec" will be suppressed if there are duplicate values in an
242          enumeration.)
243         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
244         will be generated for each C enum constant xxx.
245         * Conversion functions m2i and i2m convert between mlrep and
246         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
247         * Coversion functions c and ml convert between mlrep and "tag enum".
248         * Access functions (get/set) fetch and store mlrep values.
249       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
250         enumerations are merged into one single enumeration represented by
251         structure E_'.
252    
253    ----------------------------------------------------------------------
254    Name: Allen Leung
255    Date: 2002/02/25 04:45:00 EST
256    Tag: leunga-20020225-cps-spill
257    
258    This is a new implementation of the CPS spill phase.
259    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
260    In case of problems, replace it with the old file spill.sml
261    
262    The current compiler runs into some serious performance problems when
263    constructing a large record.  This can happen when we try to compile a
264    structure with many items.  Even a very simple structure like the following
265    makes the compiler slow down.
266    
267        structure Foo = struct
268           val x_1 = 0w1 : Word32.int
269           val x_2 = 0w2 : Word32.int
270           val x_3 = 0w3 : Word32.int
271           ...
272           val x_N = 0wN : Word32.int
273        end
274    
275    The following table shows the compile time, from N=1000 to N=4000,
276    with the old compiler:
277    
278    N
279    1000   CPS 100 spill                           0.04u  0.00s  0.00g
280           MLRISC ra                               0.06u  0.00s  0.05g
281              (spills = 0 reloads = 0)
282           TOTAL                                   0.63u  0.07s  0.21g
283    
284    1100   CPS 100 spill                           8.25u  0.32s  0.64g
285           MLRISC ra                               5.68u  0.59s  3.93g
286              (spills = 0 reloads = 0)
287           TOTAL                                   14.71u  0.99s  4.81g
288    
289    1500   CPS 100 spill                           58.55u  2.34s  1.74g
290           MLRISC ra                               5.54u  0.65s  3.91g
291              (spills = 543 reloads = 1082)
292           TOTAL                                   65.40u  3.13s  6.00g
293    
294    2000   CPS 100 spill                           126.69u  4.84s  3.08g
295           MLRISC ra                               0.80u  0.10s  0.55g
296              (spills = 42 reloads = 84)
297           TOTAL                                   129.42u  5.10s  4.13g
298    
299    3000   CPS 100 spill                           675.59u  19.03s  11.64g
300           MLRISC ra                               2.69u  0.27s  1.38g
301              (spills = 62 reloads = 124)
302           TOTAL                                   682.48u  19.61s  13.99g
303    
304    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
305           MLRISC ra                               4.96u  0.27s  2.72g
306              (spills = 85 reloads = 170)
307           TOTAL                                   2375.26u  57.21s  48.00g
308    
309    As you can see the old cps spill module suffers from some serious
310    performance problem.  But since I cannot decipher the old code fully,
311    instead of patching the problems up, I'm reimplementing it
312    with a different algorithm.  The new code is more modular,
313    smaller when compiled, and substantially faster
314    (O(n log n) time and O(n) space).  Timing of the new spill module:
315    
316    4000  CPS 100 spill                           0.02u  0.00s  0.00g
317          MLRISC ra                               0.25u  0.02s  0.15g
318             (spills=1 reloads=3)
319          TOTAL                                   7.74u  0.34s  1.62g
320    
321    Implementation details:
322    
323    As far as I can tell, the purpose of the CPS spill module is to make sure the
324    number of live variables at any program point (the bandwidth)
325    does not exceed a certain limit, which is determined by the
326    size of the spill area.
327    
328    When the bandwidth is too large, we decrease the register pressure by
329    packing live variables into spill records.  How we achieve this is
330    completely different than what we did in the old code.
331    
332    First, there is something about the MLRiscGen code generator
333    that we should be aware of:
334    
335    o MLRiscGen performs code motion!
336    
337       In particular, it will move floating point computations and
338       address computations involving only the heap pointer to
339       their use sites (if there is only a single use).
340       What this means is that if we have a CPS record construction
341       statement
342    
343           RECORD(k,vl,w,e)
344    
345       we should never count the new record address w as live if w
346       has only one use (which is often the case).
347    
348       We should do something similar to floating point, but the transformation
349       there is much more complex, so I won't deal with that.
350    
351    Secondly, there are now two new cps primops at our disposal:
352    
353     1. rawrecord of record_kind option
354        This pure operator allocates some uninitialized storage from the heap.
355        There are two forms:
356    
357         rawrecord NONE [INT n]  allocates a tagless record of length n
358         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
359                                     and initializes the tag.
360    
361     2. rawupdate of cty
362          rawupdate cty (v,i,x)
363          Assigns to x to the ith component of record v.
364          The storelist is not updated.
365    
366    We use these new primops for both spilling and increment record construction.
367    
368     1. Spilling.
369    
370        This is implemented with a linear scan algorithm (but generalized
371        to trees).  The algorithm will create a single spill record at the
372        beginning of the cps function and use rawupdate to spill to it,
373        and SELECT or SELp to reload from it.  So both spills and reloads
374        are fine-grain operations.  In contrast, in the old algorithm
375        "spills" have to be bundled together in records.
376    
377        Ideally, we should sink the spill record construction to where
378        it is needed.  We can even split the spill record into multiple ones
379        at the places where they are needed.  But CPS is not a good
380        representation for global code motion, so I'll keep it simple and
381        am not attempting this.
382    
383     2. Incremental record construction (aka record splitting).
384    
385        Long records with many component values which are simulatenously live
386        (recall that single use record addresses are not considered to
387         be live) are constructed with rawrecord and rawupdate.
388        We allocate space on the heap with rawrecord first, then gradually
389        fill it in with rawupdate.  This is the technique suggested to me
390        by Matthias.
391    
392        Some restrictions on when this is applicable:
393        1. It is not a VECTOR record.  The code generator currently does not handle
394           this case. VECTOR record uses double indirection like arrays.
395        2. All the record component values are defined in the same "basic block"
396           as the record constructor.  This is to prevent speculative
397           record construction.
398    
399    ----------------------------------------------------------------------
400    Name: Allen Leung
401    Date: 2002/02/22 01:02:00 EST
402    Tag: leunga-20020222-mlrisc-tools
403    
404    Minor bug fixes in the parser and rewriter
405    
406    ----------------------------------------------------------------------
407    Name: Allen Leung
408    Date: 2002/02/21 20:20:00 EST
409    Tag: leunga-20020221-peephole
410    
411    Regenerated the peephole files.  Some contained typos in the specification
412    and some didn't compile because of pretty printing bugs in the old version
413    of 'nowhere'.
414    
415    ----------------------------------------------------------------------
416    Name: Allen Leung
417    Date: 2002/02/19 20:20:00 EST
418    Tag: leunga-20020219-mlrisc-tools
419    Description:
420    
421       Minor bug fixes to the mlrisc-tools library:
422    
423       1.  Fixed up parsing colon suffixed keywords
424       2.  Added the ability to shut the error messages up
425       3.  Reimplemented the pretty printer and fixed up/improved
426           the pretty printing of handle and -> types.
427       4.  Fixed up generation of literal symbols in the nowhere tool.
428       5.  Added some SML keywords to to sml.sty
429    
430    ----------------------------------------------------------------------
431    Name: Matthias Blume
432    Date: 2002/02/19 16:20:00 EST
433    Tag: blume-20020219-cmffi
434    Description:
435    
436    A wild mix of changes, some minor, some major:
437    
438    * All C FFI-related libraries are now anchored under $c:
439        $/c.cm      --> $c/c.cm
440        $/c-int.cm  --> $c/internals/c-int.cm
441        $/memory.cm --> $c/memory/memory.cm
442    
443    * "make" tool (in CM) now treats its argument pathname slightly
444      differently:
445        1. If the native expansion is an absolute name, then before invoking
446           the "make" command on it, CM will apply OS.Path.mkRelative
447           (with relativeTo = OS.FileSys.getDir()) to it.
448        2. The argument will be passed through to subsequent phases of CM
449           processing without "going native".  In particular, if the argument
450           was an anchored path, then "make" will not lose track of that anchor.
451    
452    * Compiler backends now "know" their respective C calling conventions
453      instead of having to be told about it by ml-nlffigen.  This relieves
454      ml-nlffigen from one of its burdens.
455    
456    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
457    
458    * Export C_DEBUG and C_Debug from $c/c.cm.
459    
460    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
461      subtyping relationship between incomplete pointers and their complete
462      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
463      with the convention of instantiating 'o with ('t, 'c) obj whenever
464      the pointer target type is complete.  In the incomplete case, 'o
465      will be instantiated with some "'c iobj" -- a type obtained by
466      using one of the functors PointerToIncompleteType or PointerToCompleteType.
467    
468      Operations that work on both incomplete and complete pointer types are
469      typed as taking an 'o ptr while operations that require the target to
470      be known are typed as taking some ('t, 'c) obj ptr.
471    
472      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
473      where void is an eqtype without any values.  This makes it possible
474      to work on voidptr values using functions meant to operate on light
475      incomplete pointers.
476    
477    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
478      been vastly simplified.
479    
480    ----------------------------------------------------------------------
481    Name: Matthias Blume
482    Date: 2002/02/19 10:48:00 EST
483    Tag: blume-20020219-pqfix
484    Description:
485    
486    Applied Chris Okasaki's bug fix for priority queues.
487    
488    ----------------------------------------------------------------------
489    Name: Matthias Blume
490    Date: 2002/02/15 17:05:00
491    Tag: Release_110_39
492    Description:
493    
494    Last-minute retagging is becoming a tradition... :-(
495    
496    This is the working release 110.39.
497    
498    ----------------------------------------------------------------------
499    Name: Matthias Blume
500    Date: 2002/02/15 16:00:00 EST
501    Tag: Release_110_39-orig
502    Description:
503    
504    Working release 110.39.  New bootfiles.
505    
506    (Update: There was a small bug in the installer so it wouldn't work
507    with all shells.  So I retagged. -Matthias)
508    
509    ----------------------------------------------------------------------
510    Name: Matthias Blume
511    Date: 2002/02/15 14:17:00 EST
512    Tag: blume-20020215-showbindings
513    Description:
514    
515    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
516    the latter can be useful for exploring what bindings are available at
517    the interactive prompt.  (The first function returns only the list
518    of symbols that are really bound, the second prints those but also the
519    ones that CM's autoloading mechanism knows about.)
520    
521    ----------------------------------------------------------------------
522    Name: Matthias Blume
523    Date: 2002/02/15 12:08:00 EST
524    Tag: blume-20020215-iptrs
525    Description:
526    
527    Two improvements to ml-nlffigen:
528    
529      1. Write files only if they do not exist or if their current contents
530         do not coincide with what's being written.  (That is, avoid messing
531         with the time stamps unless absolutely necessary.)
532    
533      2. Implement a "repository" mechanism for generated files related
534         to "incomplete pointer types".   See the README file for details.
535    
536    ----------------------------------------------------------------------
537    Name: Matthias Blume
538    Date: 2002/02/14 11:50:00 EST
539    Tag: blume-20020214-quote
540    Description:
541    
542    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
543    because of the new and improved tag generation scheme.  (Thanks to Allen
544    Leung for pointing it out.)
545    
546    ----------------------------------------------------------------------
547    Name: Lal George
548    Date: 2002/02/14 09:55:27 EST 2002
549    Tag: george-20020214-isabelle-bug
550    Description:
551    
552    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
553    of Isabelle on the x86.
554    
555    From Allen:
556    -----------
557     I've found the problem:
558    
559         in ra-core.sml, I use the counter "blocked" to keep track of the
560         true number of elements in the freeze queue.  When the counter goes
561         to zero, I skip examining the queue.  But I've messed up the
562         bookkeeping in combine():
563    
564             else ();
565             case !ucol of
566               PSEUDO => (if !cntv > 0 then
567                     (if !cntu > 0 then blocked := !blocked - 1 else ();
568                                        ^^^^^^^^^^^^^^^^^^^^^^^
569                      moveu := mergeMoveList(!movev, !moveu)
570                     )
571                  else ();
572    
573         combine() is called to coalesce two nodes u and v.
574         I think I was thinking that if the move counts of u and v are both
575         greater than zero then after they are coalesced then one node is
576         removed from the freeze queue.  Apparently I was thinking that
577         both u and v are of low degree, but that's clearly not necessarily true.
578    
579    
580    02/12/2002:
581        Here's the patch.  HOL now compiles.
582    
583        I don't know how this impact on performance (compile
584        time or runtime).  This bug caused the RA (especially on the x86)
585        to go thru the potential spill phase when there are still nodes on the
586        freeze queue.
587    
588    
589    
590    
591    ----------------------------------------------------------------------
592    Name: Matthias Blume
593    Date: 2002/02/13 22:40:00 EST
594    Tag: blume-20020213-fptr-rtti
595    Description:
596    
597    Fixed a bug in ml-nlffigen that was introduced with one of the previous
598    updates.
599    
600    ----------------------------------------------------------------------
601    Name: Matthias Blume
602    Date: 2002/02/13 16:41:00 EST
603    Tag: blume-20020213-cmlpq
604    Description:
605    
606    Added new priority queue export symbols (which have just been added to
607    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
608    not compile and the installer would choke.)
609    
610    ----------------------------------------------------------------------
611    Name: Matthias Blume
612    Date: 2002/02/13 16:15:00 EST
613    Tag: blume-20020213-various
614    Description:
615    
616    1. More tweaks to ml-nlffigen:
617    
618       - better internal datastructures (resulting in slight speedup)
619       - "-match" option requires exact match
620       - "localized" gensym counters (untagged structs/unions nested within
621         other structs/unions or within typedefs get a fresh counter; their
622         tag will be prefixed by a concatenation of their parents' tags)
623       - bug fixes (related to calculation of transitive closure of types
624         to be included in the output)
625    
626    2. Minor Basis updates:
627    
628       - added implementations for List.collate and Option.app
629    
630    ----------------------------------------------------------------------
631    Name: Matthias Blume
632    Date: 2002/02/11 15:55:00 EST
633    Tag: blume-20020211-gensym
634    Description:
635    
636    Added a "-gensym" option to command line of ml-nlffigen.  This can be
637    used to specify a "stem" -- a string that is inserted in all "gensym'd"
638    names (ML structure names that correspond to unnamed C structs, unions,
639    and enums), so that separate runs of ml-nlffigen do not clash.
640    
641    ----------------------------------------------------------------------
642    Name: Matthias Blume
643    Date: 2002/02/11 12:05:00 EST
644    Tag: blume-20020211-gensml
645    Description:
646    
647    A quick fix for a problem with GenSML (in the pgraph-util library):
648    Make generation of toplevel "local" optional.  (Strictly speaking,
649    signature definitions within "local" are not legal SML.)
650    
651    Other than that: updates to INSTALL and cm/TODO.
652    
653    ----------------------------------------------------------------------
654    Name: Matthias Blume
655    Date: 2002/02/08 15:00:00 EST
656    Tag: blume-20020208-uniquepid
657    Description:
658    
659    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
660    
661    1. The installer (config/install.sh) has gotten smarter:
662    
663         - Configuration options are a bit easier to specify now
664           (in config/targets).
665         - Bug in recognizing .tar.bz2 files fixed.
666         - Installer automatically resolves dependencies between
667           configuration options (e.g., if you ask for eXene, you will
668           also get cml -- regardless whether you asked for it or not).
669         - Installer can run in "quieter mode" by setting the environment
670           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
671           "completely silent", though.
672         - Build HashCons library as part of smlnj-lib.
673    
674    2. A new scheme for assigning persistent identifiers to compilation
675       units (and, by extension, to types etc.) has been put into place.
676       This fixes a long-standing bug where types and even dynamic values
677       can get internally confused, thereby compromising type safety
678       (abstraction) and dynamic correctness.  See
679    
680         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
681    
682       for an example of how things could go wrong until now.
683    
684       The downside of the new scheme is that pids are not quite as
685       persistent as they used to be: CM will generate a fresh pid
686       for every compilation unit that it thinks it sees for the first
687       time.  That means that if you compile starting from a clean, fresh
688       source tree at two different times, you end up with different
689       binaries.
690    
691       Cutoff recompilation, however, has not been compromised because
692       CM keeps pid information in special caches between runs.
693    
694    ----------------------------------------------------------------------
695  Name: Lal George  Name: Lal George
696  Date: 2002/02/07 15:34:13 EST 2002  Date: 2002/02/07 15:34:13 EST 2002
697  Tag: <none>  Tag: <none>

Legend:
Removed from v.1055  
changed lines
  Added in v.1132

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