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 1078, Tue Feb 19 21:26:48 2002 UTC revision 1159, Fri Mar 22 03:18:24 2002 UTC
# Line 14  Line 14 
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16  Name: Matthias Blume  Name: Matthias Blume
17    Date: 2002/03/21 22:20:00 EST
18    Tag: blume-20020321-kmp-bugfix
19    Description:
20    
21    Changed the interface to the KMP routine in PreString and fixed
22    a minor bug in one place where it was used.
23    
24    ----------------------------------------------------------------------
25    Name: Allen Leung
26    Date: 2002/03/21 20:30:00 EST
27    Tag: leunga-20020321-cfg
28    Description:
29    
30      Fixed a potential problem in cfg edge splitting.
31    
32    ----------------------------------------------------------------------
33    Name: Allen Leung
34    Date: 2002/03/21 17:15:00 EST
35    Tag: leunga-20020321-x86-fp-cfg
36    Description:
37    
38      1. Recoded the buggy parts of x86-fp.
39    
40         a. All the block reordering code has been removed.
41            We now depend on the block placement phases to do this work.
42    
43         b. Critical edge splitting code has been simplified and moved into the
44            CFG modules, as where they belong.
45    
46         Both of these were quite buggy and complex.  The code is now much, much
47         simpler.
48    
49      2. X86 backend.
50    
51         a. Added instructions for 64-bit support.  Instruction selection for
52            64-bit has not been committed, however, since that
53            requires changes to MLTREE which haven't been approved by
54            Lal and John.
55    
56         b. Added support for FUCOMI and FUCOMIP when generating code for
57            PentiumPro and above.  We only generate these instructions in
58            the fast-fp mode.
59    
60         c. Added cases for JP and JNP in X86FreqProps.
61    
62      3. CFG
63    
64         CFG now has a bunch of methods for edge splitting and merging.
65    
66      4. Machine description.
67    
68         John's simplification of MLTREE_BASIS.fcond broke a few machine
69         description things:
70    
71         rtl-build.{sig,sml} and hppa.mdl fixed.
72    
73         NOTE: the machine description stuff in the repository is still broken.
74               Again, I can't put my fixes in because that involves
75               changes to MLTREE.
76    
77    ----------------------------------------------------------------------
78    Name: Matthias Blume
79    Date: 2002/03/20 15:55:00 EST
80    Tag: blume-20020320-kmp
81    Description:
82    
83    Implemented Knuth-Morris-Pratt string matching in PreString and used
84    it for String.isSubstring, Substring.isSubstring, and
85    Substring.position.
86    
87    (Might need some stress-testing.  Simple examples worked fine.)
88    
89    ----------------------------------------------------------------------
90    Name: Matthias Blume
91    Date: 2002/03/19 16:37:00 EST
92    Tag: blume-20020319-witnesses
93    Description:
94    
95    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
96    
97    This implements a generic mechanism for changing constness qualifiers
98    anywhere within big C types without resorting to outright "casts".
99    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
100    modify the constness at the outermost level.)
101    The implementation of "convert" is based on the idea of "witness"
102    values -- values that are not used by the operation but whose types
103    "testify" to their applicability.  On the implementation side, "convert"
104    is simply a projection (returning its second curried argument).  With
105    cross-module inlining, it should not result in any machine code being
106    generated.
107    
108    ----------------------------------------------------------------------
109    Name: Matthias Blume
110    Date: 2002/03/15 16:40:00 EST
111    Tag: blume-20020315-basis
112    Description:
113    
114    Provided (preliminary?) implementations for
115    
116      {String,Substring}.{concatWith,isSuffix,isSubstring}
117    
118    and
119    
120      Substring.full
121    
122    Those are in the Basis spec but they were missing in SML/NJ.
123    
124    ----------------------------------------------------------------------
125    Name: Matthias Blume
126    Date: 2002/03/14 21:30:00 EST
127    Tag: blume-20020314-controls
128    Description:
129    
130    Controls:
131    ---------
132    
133    1. Factored out the recently-added Controls : CONTROLS stuff and put
134       it into its own library $/controls-lib.cm.  The source tree for
135       this is under src/smlnj-lib/Controls.
136    
137    2. Changed the names of types and functions in this interface, so they
138       make a bit more "sense":
139    
140          module -> registry
141          'a registry -> 'a group
142    
143    3. The interface now deals in ref cells only.  The getter/setter interface
144       is (mostly) gone.
145    
146    4. Added a function that lets one register an already-existing ref cell.
147    
148    5. Made the corresponding modifications to the rest of the code so that
149       everything compiles again.
150    
151    6. Changed the implementation of Controls.MLRISC back to something closer
152       to the original.  In particular, this module (and therefore MLRISC)
153       does not depend on Controls.  There now is some link-time code in
154       int-sys.sml that registers the MLRISC controls with the Controls
155       module.
156    
157    CM:
158    ---
159    
160      * One can now specify the lambda-split aggressiveness in init.cmi.
161    
162    ----------------------------------------------------------------------
163    Name: Allen Leung
164    Date: 2002/03/13 17:30:00 EST
165    Tag: leunga-20020313-x86-fp-unary
166    Description:
167    
168    Bug fix for:
169    
170    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
171    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
172    > - fun f(x,(y,z)) = Real.~ y;
173    > [autoloading]
174    > [autoloading done]
175    >       fchsl   (%eax), 184(%esp)
176    > Error: MLRisc bug: X86MCEmitter.emitInstr
177    >
178    > uncaught exception Error
179    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
180    
181    The problem was that the code generator did not generate any fp registers
182    in this case, and the ra didn't know that it needed to run the X86FP phase to
183    translate the pseudo fp instruction.   This only happened with unary fp
184    operators in certain situations.
185    
186    ----------------------------------------------------------------------
187    Name: Matthias Blume
188    Date: 2002/03/13 14:00:00 EST
189    Tag: blume-20020313-overload-etc
190    Description:
191    
192    1. Added _overload as a synonym for overload for backward compatibility.
193       (Control.overloadKW must be true for either version to be accepted.)
194    
195    2. Fixed bug in install script that caused more things to be installed
196       than what was requested in config/targets.
197    
198    3. Made CM aware of the (_)overload construct so that autoloading
199       works.
200    
201    ----------------------------------------------------------------------
202    Name: Matthias Blume
203    Date: 2002/03/12 22:03:00 EST
204    Tag: blume-20020312-url
205    Description:
206    
207    Forgot to update BOOT and srcarchiveurl.
208    
209    ----------------------------------------------------------------------
210    Name: Matthias Blume
211    Date: 2002/03/12 17:30:00 EST
212    Tag: blume-20020312-version110392
213    Description:
214    
215    Yet another version number bump (because of small changes to the
216    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
217    
218    Changes:
219    
220      The new pid generation scheme described a few weeks ago was overly
221      complicated.  I implemented a new mechanism that is simpler and
222      provides a bit more "stability":  Once CM has seen a compilation
223      unit, it keeps its identity constant (as long as you do not delete
224      those crucial CM/GUID/* files).  This means that when you change
225      an interface, compile, then go back to the old interface, and
226      compile again, you arrive at the original pid.
227    
228      There now also is a mechanism that instructs CM to use the plain
229      environment hash as a module's pid (effectively making its GUID
230      the empty string).  For this, "noguid" must be specified as an
231      option to the .sml file in question within its .cm file.
232      This is most useful for code that is being generated by tools such
233      as ml-nlffigen (because during development programmers tend to
234      erase the tool's entire output directory tree including CM's cached
235      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
236      revert to the old, broken behavior of SML/NJ, but in specific cases
237      where there is no danger of interface confusion, its use is ok
238      (I think).
239    
240      ml-nlffigen by default generates "noguid" annotations.  They can be
241      turned off by specifying -guid in its command line.
242    
243    ----------------------------------------------------------------------
244    Name: Lal George
245    Date: 2002/03/12 12 14:42:36 EST
246    Tag: george-20020312-frequency-computation
247    Description:
248    
249    Integrated jump chaining and static block frequency into the
250    compiler. More details and numbers later.
251    
252    ----------------------------------------------------------------------
253    Name: Lal George
254    Date: 2002/03/11 11 22:38:53 EST
255    Tag: george-20020311-jump-chain-elim
256    Description:
257    
258    Tested the jump chain elimination on all architectures (except the
259    hppa).  This is on by default right now and is profitable for the
260    alpha and x86, however, it may not be profitable for the sparc and ppc
261    when compiling the compiler.
262    
263    The gc test will typically jump to a label at the end of the cluster,
264    where there is another jump to an external cluster containing the actual
265    code to invoke gc. This is to allow factoring of common gc invocation
266    sequences. That is to say, we generate:
267    
268            f:
269               testgc
270               ja   L1      % jump if above to L1
271    
272            L1:
273               jmp L2
274    
275    
276    After jump chain elimination the 'ja L1' instructions is converted to
277    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
278    up being implemented in their long form (if L2 is far away) using:
279    
280            jbe     L3      % jump if below or equal to L3
281            jmp     L2
282         L3:
283            ...
284    
285    
286    For large compilation units L2  may be far away.
287    
288    
289    ----------------------------------------------------------------------
290    Name: Matthias Blume
291    Date: 2002/03/11 13:30:00 EST
292    Tag: blume-20020311-mltreeeval
293    Description:
294    
295    A functor parameter was missing.
296    
297    ----------------------------------------------------------------------
298    Name: Allen Leung
299    Date: 2002/03/11 10:30:00 EST
300    Tag: leunga-20020311-runtime-string0
301    Description:
302    
303       The representation of the empty string now points to a
304    legal null terminated C string instead of unit.  It is now possible
305    to convert an ML string into C string with InlineT.CharVector.getData.
306    This compiles into one single machine instruction.
307    
308    ----------------------------------------------------------------------
309    Name: Allen Leung
310    Date: 2002/03/10 23:55:00 EST
311    Tag: leunga-20020310-x86-call
312    Description:
313    
314       Added machine generation for CALL instruction (relative displacement mode)
315    
316    ----------------------------------------------------------------------
317    Name: Matthias Blume
318    Date: 2002/03/08 16:05:00
319    Tag: blume-20020308-entrypoints
320    Description:
321    
322    Version number bumped to 110.39.1.  NEW BOOTFILES!
323    
324    Entrypoints: non-zero offset into a code object where execution should begin.
325    
326    - Added the notion of an entrypoint to CodeObj.
327    - Added reading/writing of entrypoint info to Binfile.
328    - Made runtime system bootloader aware of entrypoints.
329    - Use the address of the label of the first function given to mlriscGen
330      as the entrypoint.  This address is currently always 0, but it will
331      not be 0 once we turn on block placement.
332    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
333      with entry points) from mlriscGen.
334    
335    ----------------------------------------------------------------------
336    Name: Allen Leung
337    Date: 2002/03/07 20:45:00 EST
338    Tag: leunga-20020307-x86-cmov
339    Description:
340    
341       Bug fixes for CMOVcc on x86.
342    
343       1. Added machine code generation for CMOVcc
344       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
345       3. CMOVcc cannot have an immediate operand as argument.
346    
347    ----------------------------------------------------------------------
348    Name: Matthias Blume
349    Date: 2002/03/07 16:15:00 EST
350    Tag: blume-20020307-controls
351    Description:
352    
353    This is a very large but mostly boring patch which makes (almost)
354    every tuneable compiler knob (i.e., pretty much everything under
355    Control.* plus a few other things) configurable via both the command
356    line and environment variables in the style CM did its configuration
357    until now.
358    
359    Try starting sml with '-h' (or, if you are brave, '-H')
360    
361    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
362    implements the underlying generic mechanism.
363    
364    The interface to some of the existing such facilities has changed somewhat.
365    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
366    (The getFoo interface is still there for backward-compatibility, but its
367    use is deprecated.)
368    
369    The ml-build script passes -Cxxx=yyy command-line arguments through so
370    that one can now twiddle the compiler settings when using this "batch"
371    compiler.
372    
373    TODO items:
374    
375    We should go through and throw out all controls that are no longer
376    connected to anything.  Moreover, we should go through and provide
377    meaningful (and correct!) documentation strings for those controls
378    that still are connected.
379    
380    Currently, multiple calls to Controls.new are accepted (only the first
381    has any effect).  Eventually we should make sure that every control
382    is being made (via Controls.new) exactly once.  Future access can then
383    be done using Controls.acc.
384    
385    Finally, it would probably be a good idea to use the getter-setter
386    interface to controls rather than ref cells.  For the time being, both
387    styles are provided by the Controls module, but getter-setter pairs are
388    better if thread-safety is of any concern because they can be wrapped.
389    
390    *****************************************
391    
392    One bug fix: The function blockPlacement in three of the MLRISC
393    backpatch files used to be hard-wired to one of two possibilities at
394    link time (according to the value of the placementFlag).  But (I
395    think) it should rather sense the flag every time.
396    
397    *****************************************
398    
399    Other assorted changes (by other people who did not supply a HISTORY entry):
400    
401    1. the cross-module inliner now works much better (Monnier)
402    2. representation of weights, frequencies, and probabilities in MLRISC
403       changed in preparation of using those for weighted block placement
404       (Reppy, George)
405    
406    ----------------------------------------------------------------------
407    Name: Lal George
408    Date: 2002/03/07 14:44:24 EST 2002
409    Tag: george-20020307-weighted-block-placement
410    
411    Tested the weighted block placement optimization on all architectures
412    (except the hppa) using AMPL to generate the block and edge frequencies.
413    Changes were required in the machine properties to correctly
414    categorize trap instructions. There is an MLRISC flag
415    "weighted-block-placement" that can be used to enable weighted block
416    placement, but this will be ineffective without block/edge
417    frequencies (coming soon).
418    
419    
420    ----------------------------------------------------------------------
421    Name: Lal George
422    Date: 2002/03/05 17:24:48 EST
423    Tag: george-20020305-linkage-cluster
424    
425    In order to support the block placement optimization, a new cluster
426    is generated as the very first cluster (called the linkage cluster).
427    It contains a single jump to the 'real' entry point for the compilation
428    unit. Block placement has no effect on the linkage cluster itself, but
429    all the other clusters  have full freedom in the manner in which they
430    reorder blocks or functions.
431    
432    On the x86 the typical linkage code that is generated is:
433       ----------------------
434            .align 2
435       L0:
436            addl    $L1-L0, 72(%esp)
437            jmp     L1
438    
439    
440            .align  2
441       L1:
442       ----------------------
443    
444    72(%esp) is the memory location for the stdlink register. This
445    must contain the address of the CPS function being called. In the
446    above example, it contains the address of  L0; before
447    calling L1 (the real entry point for the compilation unit), it
448    must contain the address for L1, and hence
449    
450            addl $L1-L0, 72(%esp)
451    
452    I have tested this on all architectures except the hppa.The increase
453    in code size is of course negligible
454    
455    ----------------------------------------------------------------------
456    Name: Allen Leung
457    Date: 2002/03/03 13:20:00 EST
458    Tag: leunga-20020303-mlrisc-tools
459    
460      Added #[ ... ] expressions to mlrisc tools
461    
462    ----------------------------------------------------------------------
463    Name: Matthias Blume
464    Date: 2002/02/27 12:29:00 EST
465    Tag: blume-20020227-cdebug
466    Description:
467    
468    - made types in structure C and C_Debug to be equal
469    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
470    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
471    
472    ----------------------------------------------------------------------
473    Name: Matthias Blume
474    Date: 2002/02/26 12:00:00 EST
475    Tag: blume-20020226-ffi
476    Description:
477    
478    1. Fixed a minor bug in CM's "noweb" tool:
479       If numbering is turned off, then truly don't number (i.e., do not
480       supply the -L option to noweb).  The previous behavior was to supply
481       -L'' -- which caused noweb to use the "default" line numbering scheme.
482       Thanks to Chris Richards for pointing this out (and supplying the fix).
483    
484    2. Once again, I reworked some aspects of the FFI:
485    
486       A. The incomplete/complete type business:
487    
488       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
489         gone!
490       - ML types representing an incomplete type are now *equal* to
491         ML types representing their corresponding complete types (just like
492         in C).  This is still safe because ml-nlffigen will not generate
493         RTTI for incomplete types, nor will it generate functions that
494         require access to such RTTI.   But when ML code generated from both
495         incomplete and complete versions of the C type meet, the ML types
496         are trivially interoperable.
497    
498         NOTE:  These changes restore the full generality of the translation
499         (which was previously lost when I eliminated functorization)!
500    
501       B. Enum types:
502    
503       - Structure C now has a type constructor "enum" that is similar to
504         how the "su" constructor works.  However, "enum" is not a phantom
505         type because each "T enum" has values (and is isomorphic to
506         MLRep.Signed.int).
507       - There are generic access operations for enum objects (using
508         MLRep.Signed.int).
509       - ml-nlffigen will generate a structure E_foo for each "enum foo".
510         * The structure contains the definition of type "mlrep" (the ML-side
511         representation type of the enum).  Normally, mlrep is the same
512         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
513         then mlrep will be defined as a datatype -- thus facilitating
514         pattern matching on mlrep values.
515         ("-ec" will be suppressed if there are duplicate values in an
516          enumeration.)
517         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
518         will be generated for each C enum constant xxx.
519         * Conversion functions m2i and i2m convert between mlrep and
520         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
521         * Coversion functions c and ml convert between mlrep and "tag enum".
522         * Access functions (get/set) fetch and store mlrep values.
523       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
524         enumerations are merged into one single enumeration represented by
525         structure E_'.
526    
527    ----------------------------------------------------------------------
528    Name: Allen Leung
529    Date: 2002/02/25 04:45:00 EST
530    Tag: leunga-20020225-cps-spill
531    
532    This is a new implementation of the CPS spill phase.
533    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
534    In case of problems, replace it with the old file spill.sml
535    
536    The current compiler runs into some serious performance problems when
537    constructing a large record.  This can happen when we try to compile a
538    structure with many items.  Even a very simple structure like the following
539    makes the compiler slow down.
540    
541        structure Foo = struct
542           val x_1 = 0w1 : Word32.int
543           val x_2 = 0w2 : Word32.int
544           val x_3 = 0w3 : Word32.int
545           ...
546           val x_N = 0wN : Word32.int
547        end
548    
549    The following table shows the compile time, from N=1000 to N=4000,
550    with the old compiler:
551    
552    N
553    1000   CPS 100 spill                           0.04u  0.00s  0.00g
554           MLRISC ra                               0.06u  0.00s  0.05g
555              (spills = 0 reloads = 0)
556           TOTAL                                   0.63u  0.07s  0.21g
557    
558    1100   CPS 100 spill                           8.25u  0.32s  0.64g
559           MLRISC ra                               5.68u  0.59s  3.93g
560              (spills = 0 reloads = 0)
561           TOTAL                                   14.71u  0.99s  4.81g
562    
563    1500   CPS 100 spill                           58.55u  2.34s  1.74g
564           MLRISC ra                               5.54u  0.65s  3.91g
565              (spills = 543 reloads = 1082)
566           TOTAL                                   65.40u  3.13s  6.00g
567    
568    2000   CPS 100 spill                           126.69u  4.84s  3.08g
569           MLRISC ra                               0.80u  0.10s  0.55g
570              (spills = 42 reloads = 84)
571           TOTAL                                   129.42u  5.10s  4.13g
572    
573    3000   CPS 100 spill                           675.59u  19.03s  11.64g
574           MLRISC ra                               2.69u  0.27s  1.38g
575              (spills = 62 reloads = 124)
576           TOTAL                                   682.48u  19.61s  13.99g
577    
578    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
579           MLRISC ra                               4.96u  0.27s  2.72g
580              (spills = 85 reloads = 170)
581           TOTAL                                   2375.26u  57.21s  48.00g
582    
583    As you can see the old cps spill module suffers from some serious
584    performance problem.  But since I cannot decipher the old code fully,
585    instead of patching the problems up, I'm reimplementing it
586    with a different algorithm.  The new code is more modular,
587    smaller when compiled, and substantially faster
588    (O(n log n) time and O(n) space).  Timing of the new spill module:
589    
590    4000  CPS 100 spill                           0.02u  0.00s  0.00g
591          MLRISC ra                               0.25u  0.02s  0.15g
592             (spills=1 reloads=3)
593          TOTAL                                   7.74u  0.34s  1.62g
594    
595    Implementation details:
596    
597    As far as I can tell, the purpose of the CPS spill module is to make sure the
598    number of live variables at any program point (the bandwidth)
599    does not exceed a certain limit, which is determined by the
600    size of the spill area.
601    
602    When the bandwidth is too large, we decrease the register pressure by
603    packing live variables into spill records.  How we achieve this is
604    completely different than what we did in the old code.
605    
606    First, there is something about the MLRiscGen code generator
607    that we should be aware of:
608    
609    o MLRiscGen performs code motion!
610    
611       In particular, it will move floating point computations and
612       address computations involving only the heap pointer to
613       their use sites (if there is only a single use).
614       What this means is that if we have a CPS record construction
615       statement
616    
617           RECORD(k,vl,w,e)
618    
619       we should never count the new record address w as live if w
620       has only one use (which is often the case).
621    
622       We should do something similar to floating point, but the transformation
623       there is much more complex, so I won't deal with that.
624    
625    Secondly, there are now two new cps primops at our disposal:
626    
627     1. rawrecord of record_kind option
628        This pure operator allocates some uninitialized storage from the heap.
629        There are two forms:
630    
631         rawrecord NONE [INT n]  allocates a tagless record of length n
632         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
633                                     and initializes the tag.
634    
635     2. rawupdate of cty
636          rawupdate cty (v,i,x)
637          Assigns to x to the ith component of record v.
638          The storelist is not updated.
639    
640    We use these new primops for both spilling and increment record construction.
641    
642     1. Spilling.
643    
644        This is implemented with a linear scan algorithm (but generalized
645        to trees).  The algorithm will create a single spill record at the
646        beginning of the cps function and use rawupdate to spill to it,
647        and SELECT or SELp to reload from it.  So both spills and reloads
648        are fine-grain operations.  In contrast, in the old algorithm
649        "spills" have to be bundled together in records.
650    
651        Ideally, we should sink the spill record construction to where
652        it is needed.  We can even split the spill record into multiple ones
653        at the places where they are needed.  But CPS is not a good
654        representation for global code motion, so I'll keep it simple and
655        am not attempting this.
656    
657     2. Incremental record construction (aka record splitting).
658    
659        Long records with many component values which are simulatenously live
660        (recall that single use record addresses are not considered to
661         be live) are constructed with rawrecord and rawupdate.
662        We allocate space on the heap with rawrecord first, then gradually
663        fill it in with rawupdate.  This is the technique suggested to me
664        by Matthias.
665    
666        Some restrictions on when this is applicable:
667        1. It is not a VECTOR record.  The code generator currently does not handle
668           this case. VECTOR record uses double indirection like arrays.
669        2. All the record component values are defined in the same "basic block"
670           as the record constructor.  This is to prevent speculative
671           record construction.
672    
673    ----------------------------------------------------------------------
674    Name: Allen Leung
675    Date: 2002/02/22 01:02:00 EST
676    Tag: leunga-20020222-mlrisc-tools
677    
678    Minor bug fixes in the parser and rewriter
679    
680    ----------------------------------------------------------------------
681    Name: Allen Leung
682    Date: 2002/02/21 20:20:00 EST
683    Tag: leunga-20020221-peephole
684    
685    Regenerated the peephole files.  Some contained typos in the specification
686    and some didn't compile because of pretty printing bugs in the old version
687    of 'nowhere'.
688    
689    ----------------------------------------------------------------------
690    Name: Allen Leung
691    Date: 2002/02/19 20:20:00 EST
692    Tag: leunga-20020219-mlrisc-tools
693    Description:
694    
695       Minor bug fixes to the mlrisc-tools library:
696    
697       1.  Fixed up parsing colon suffixed keywords
698       2.  Added the ability to shut the error messages up
699       3.  Reimplemented the pretty printer and fixed up/improved
700           the pretty printing of handle and -> types.
701       4.  Fixed up generation of literal symbols in the nowhere tool.
702       5.  Added some SML keywords to to sml.sty
703    
704    ----------------------------------------------------------------------
705    Name: Matthias Blume
706  Date: 2002/02/19 16:20:00 EST  Date: 2002/02/19 16:20:00 EST
707  Tag: blume-20020219-cmffi  Tag: blume-20020219-cmffi
708  Description:  Description:

Legend:
Removed from v.1078  
changed lines
  Added in v.1159

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