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 1009, Wed Jan 9 19:44:22 2002 UTC revision 1186, Fri Apr 12 17:54:31 2002 UTC
# Line 11  Line 11 
11  Date: yyyy/mm/dd  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: 2002/04/12 13:55:00 EDT
18    Tag: blume-20020412-assyntax
19    Description:
20    
21    1. Grabbed newer assyntax.h from the XFree86 project.
22    2. Fiddled with how to compile X86.prim.asm without warnings.
23    3. (Very) Minor cleanup in CM.
24    
25    ----------------------------------------------------------------------
26    Name: Matthias Blume
27    Date: 2002/04/01 (no joke!) 17:07:00 EST
28    Tag: blume-20020401-x86div
29    Description:
30    
31    Added full support for div/mod/rem/quot on the x86, using the machine
32    instruction's two results (without clumsily recomputing the remainder)
33    directly where appropriate.
34    
35    Some more extensive power-of-two support was added to the x86 instruction
36    selector (avoiding expensive divs, mods, and muls where they can be
37    replaced with cheaper shifts and masks).  However, this sort of thing
38    ought to be done earlier, e.g., within the CPS optimizer so that
39    all architectures benefit from it.
40    
41    The compiler compiles to a fixed point, but changes might be somewhat
42    fragile nevertheless.  Please, report any strange things that you might
43    see wrt. div/mod/quot/rem...
44    
45    ----------------------------------------------------------------------
46    Name: Matthias Blume
47    Date: 2002/03/29 17:22:00
48    Tag: blume-20020329-div
49    Description:
50    
51    Fixed my broken div/mod logic.  Unfortunately, this means that the
52    inline code for div/mod now has one more comparison than before.
53    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
54    The problem was with quotient = 0, because that alone does not tell
55    us which way the rounding went.  One then has to look at whether
56    remainder and divisor have the same sign...  :(
57    
58    Anyway, I replaced the bootfiles with fresh ones...
59    
60    ----------------------------------------------------------------------
61    Name: Matthias Blume
62    Date: 2002/03/29 14:10:00 EST
63    Tag: blume-20020329-inlprims
64    Description:
65    
66    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
67    
68    Primops have changed. This means that the bin/boot-file formats have
69    changed as well.
70    
71    To make sure that there is no confusion, I made a new version.
72    
73    
74    CHANGES:
75    
76    * removed REMT from mltree (remainder should never overflow).
77    
78    * added primops to deal with divisions of all flavors to the frontend
79    
80    * handled these primops all the way through so they map to their respective
81      MLRISC support
82    
83    * used these primops in the implementation of Int, Int32, Word, Word32
84    
85    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
86    
87    * parameterized INLMIN, INLMAX, and INLABS by a numkind
88    
89    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
90      floating point
91    
92    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
93      and Real (but Real.abs maps to a separate floating-point-only primop)
94    
95    
96    TODO items:
97    
98    * Hacked Alpha32 instruction selection, disabling the selection of REMx
99      instructions because the machine instruction encoder cannot handle
100      them.  (Hppa, PPC, and Sparc instruction selection did not handle
101      REM in the first place, and REM is supported by the x86 machine coder.)
102    
103    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
104      selection phase.  (The two can be streamlined because the hardware
105      delivers both quotient and remainder at the same time anyway.)
106    
107    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
108      (Currently the behavior is inconsistent both across architectures and
109      wrt. the draft Basis spec.)
110    
111    * Word8 should eventually be handled natively, too.
112    
113    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
114      as if there currently is no execution path that could trigger it in
115      SML/NJ.  (The assumptions underlying functions arith and promotable do not
116      hold for things like multiplication and division.)
117    
118    ----------------------------------------------------------------------
119    Name: Matthias Blume
120    Date: 2002/03/27 16:27:00 EST
121    Tag: blume-20020327-mlrisc-divisions
122    Description:
123    
124    Added support for all four division operations (ML's div, mod, quot,
125    and rem) to MLRISC.  In the course of doing so, I also rationalized
126    the naming (no more annoying switch-around of DIV and QUOT), by
127    parameterizing the operation by div_rounding_mode (which can be either
128    DIV_TO_ZERO or DIV_TO_NEGINF).
129    
130    The generic MLTreeGen functor takes care of compiling all four
131    operations down to only round-to-zero div.
132    
133    Missing pieces:
134    
135      * Doing something smarter than relying on MLTreeGen on architectures
136        like, e.g., the x86 where hardware division delivers both quotient and
137        remainder at the same time.  With this, the implementation of the
138        round-to-neginf operations could be further streamlined.
139    
140      * Remove inlining support for div/mod/rem from the frontend and replace it
141        with primops that get carried through to the backend.  Do this for all
142        int and word types.
143    
144    ----------------------------------------------------------------------
145    Name: Matthias Blume
146    Date: 2002/03/25 17:25:00 EST
147    Tag: blume-20020325-divmod
148    Description:
149    
150    I improved (hopefully without breaking them) the implementation of Int.div,
151    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
152    advantage of the following observations:
153    
154      Let  q = x quot y      r = x rem y
155           d = x div  y      m = x mod y
156    
157    where "quot" is the round-to-zero version of integer division that
158    hardware usually provides.  Then we have:
159    
160         r = x - q * y        where neither the * nor the - will overflow
161         d = if q >= 0 orelse x = q * y then q else q - 1
162                              where neither the * nor the - will overflow
163         m = if q >= 0 orelse r = 0 then r else r + y
164                              where the + will not overflow
165    
166    This results in substantial simplification of the generated code.
167    The following table shows the number of CFG nodes and edges generated
168    for
169            fun f (x, y) = x OPER y
170            (* with OPER \in div, mod, quot, rem *)
171    
172    
173        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
174        --------------------------------------------------------
175         div |         24 |         39 |         12 |         16
176         mod |         41 |         71 |         12 |         16
177        quot |          8 |         10 |          8 |         10
178         rem |         10 |         14 |          8 |         10
179    
180    
181    ----------------------------------------------------------------------
182    Name: Matthias Blume
183    Date: 2002/03/25 22:06:00 EST
184    Tag: blume-20020325-cprotobug
185    Description:
186    
187    Fixed a bug in cproto (c prototype decoder).
188    
189    ----------------------------------------------------------------------
190    Name: Matthias Blume
191    Date: 2002/03/25 16:00:00 EST
192    Tag: blume-20020325-raw-primops
193    Description:
194    
195    I did some cleanup to Allen's new primop code and
196    replaced yesterday's bootfiles with new ones.
197    (But they are stored in the same place.)
198    
199    ----------------------------------------------------------------------
200    Name: Matthias Blume
201    Date: 2002/03/24 22:40:00 EST
202    Tag: blume-20020324-bootfiles
203    Description:
204    
205    Made the bootfiles that Allen asked for.
206    
207    ----------------------------------------------------------------------
208    Name: Allen Leung
209    Date: 2002/03/23 15:50:00 EST
210    Tag: leunga-20020323-flint-cps-rcc-primops
211    Description:
212    
213      1. Changes to FLINT primops:
214    
215        (* make a call to a C-function;
216         * The primop carries C function prototype information and specifies
217         * which of its (ML-) arguments are floating point. C prototype
218         * information is for use by the backend, ML information is for
219         * use by the CPS converter. *)
220      | RAW_CCALL of { c_proto: CTypes.c_proto,
221                       ml_args: ccall_type list,
222                       ml_res_opt: ccall_type option,
223                       reentrant : bool
224                     } option
225       (* Allocate uninitialized storage on the heap.
226        * The record is meant to hold short-lived C objects, i.e., they
227        * are not ML pointers.  With the tag, the representation is
228        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
229        *)
230      | RAW_RECORD of {tag:bool,sz:int}
231      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
232    
233      2.  These CPS primops are now overloaded:
234    
235           rawload of {kind:numkind}
236           rawstore of {kind:numkind}
237    
238          The one argument form is:
239    
240             rawload {kind} address
241    
242          The two argument form is:
243    
244             rawload {kind} [ml object, byte-offset]
245    
246      3. RAW_CCALL/RCC now takes two extra arguments:
247    
248         a. The first is whether the C call is reentrant, i.e., whether
249            ML state should be saved and restored.
250         b. The second argument is a string argument specifying the name of
251            library and the C function.
252    
253         These things are currently not handled in the code generator, yet.
254    
255      4. In CProto,
256    
257         An encoding type of "bool" means "ml object" and is mapped into
258         C prototype of PTR.  Note that "bool" is different than "string",
259         even though "string" is also mapped into PTR, because "bool"
260         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
261    
262      5. Pickler/unpicker
263    
264         Changed to handle RAW_RECORD and newest RAW_CCALL
265    
266      6. MLRiscGen,
267    
268         1. Changed to handle the new rawload/rawstore/rawrecord operators.
269         2. Code for handling C Calls has been moved to a new module CPSCCalls,
270            in the file CodeGen/cpscompile/cps-c-calls.sml
271    
272      7. Added the conditional move operator
273    
274             condmove of branch
275    
276         to cps.  Generation of this is still buggy so it is currently
277         disabled.
278    
279    ----------------------------------------------------------------------
280    Name: Lal George
281    Date: 2002/03/22 14:18:25 EST
282    Tag: george-20020322-cps-branch-prob
283    Description:
284    
285    Implemented the Ball-Larus branch prediction-heuristics, and
286    incorporated graphical viewers for control flow graphs.
287    
288    Ball-Larus Heuristics:
289    ---------------------
290    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
291    
292    By design it uses the Dempster-Shafer theory for combining
293    probabilities.  For example, in the function:
294    
295        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
296    
297    the ball-larus heuristics predicts that the n=0 is unlikely
298    (OH-heuristic), and the 'then' branch is unlikely because of the
299    RH-heuristic -- giving the 'then' branch an even lower combined
300    probability using the Dempster-Shafer theory.
301    
302    Finally, John Reppy's loop analysis in MLRISC, further lowers the
303    probability of the 'then' branch because of the loop in the else
304    branch.
305    
306    
307    Graphical Viewing:
308    ------------------
309    I merely plugged in Allen's graphical viewers into the compiler. The
310    additional code is not much. At the top level, saying:
311    
312            Control.MLRISC.getFlag "cfg-graphical-view" := true;
313    
314    will display the graphical view of the control flow graph just before
315    back-patching.  daVinci must be in your path for this to work. If
316    daVinci is not available, then the default viewer can be changed
317    using:
318    
319            Control.MLRISC.getString "viewer"
320    
321    which can be set to "dot" or "vcg" for the corresponding viewers. Of
322    course, these viewers must be in your path.
323    
324    The above will display the compilation unit at the level of clusters,
325    many of which are small, boring, and un-interesting. Also setting:
326    
327            Control.MLRISC.getInt "cfg-graphical-view_size"
328    
329    will display clusters that are larger than the value set by the above.
330    
331    
332    ----------------------------------------------------------------------
333    Name: Matthias Blume
334    Date: 2002/03/21 22:20:00 EST
335    Tag: blume-20020321-kmp-bugfix
336    Description:
337    
338    Changed the interface to the KMP routine in PreString and fixed
339    a minor bug in one place where it was used.
340    
341    ----------------------------------------------------------------------
342    Name: Allen Leung
343    Date: 2002/03/21 20:30:00 EST
344    Tag: leunga-20020321-cfg
345    Description:
346    
347      Fixed a potential problem in cfg edge splitting.
348    
349    ----------------------------------------------------------------------
350    Name: Allen Leung
351    Date: 2002/03/21 17:15:00 EST
352    Tag: leunga-20020321-x86-fp-cfg
353    Description:
354    
355      1. Recoded the buggy parts of x86-fp.
356    
357         a. All the block reordering code has been removed.
358            We now depend on the block placement phases to do this work.
359    
360         b. Critical edge splitting code has been simplified and moved into the
361            CFG modules, as where they belong.
362    
363         Both of these were quite buggy and complex.  The code is now much, much
364         simpler.
365    
366      2. X86 backend.
367    
368         a. Added instructions for 64-bit support.  Instruction selection for
369            64-bit has not been committed, however, since that
370            requires changes to MLTREE which haven't been approved by
371            Lal and John.
372    
373         b. Added support for FUCOMI and FUCOMIP when generating code for
374            PentiumPro and above.  We only generate these instructions in
375            the fast-fp mode.
376    
377         c. Added cases for JP and JNP in X86FreqProps.
378    
379      3. CFG
380    
381         CFG now has a bunch of methods for edge splitting and merging.
382    
383      4. Machine description.
384    
385         John's simplification of MLTREE_BASIS.fcond broke a few machine
386         description things:
387    
388         rtl-build.{sig,sml} and hppa.mdl fixed.
389    
390         NOTE: the machine description stuff in the repository is still broken.
391               Again, I can't put my fixes in because that involves
392               changes to MLTREE.
393    
394    ----------------------------------------------------------------------
395    Name: Matthias Blume
396    Date: 2002/03/20 15:55:00 EST
397    Tag: blume-20020320-kmp
398    Description:
399    
400    Implemented Knuth-Morris-Pratt string matching in PreString and used
401    it for String.isSubstring, Substring.isSubstring, and
402    Substring.position.
403    
404    (Might need some stress-testing.  Simple examples worked fine.)
405    
406    ----------------------------------------------------------------------
407    Name: Matthias Blume
408    Date: 2002/03/19 16:37:00 EST
409    Tag: blume-20020319-witnesses
410    Description:
411    
412    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
413    
414    This implements a generic mechanism for changing constness qualifiers
415    anywhere within big C types without resorting to outright "casts".
416    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
417    modify the constness at the outermost level.)
418    The implementation of "convert" is based on the idea of "witness"
419    values -- values that are not used by the operation but whose types
420    "testify" to their applicability.  On the implementation side, "convert"
421    is simply a projection (returning its second curried argument).  With
422    cross-module inlining, it should not result in any machine code being
423    generated.
424    
425    ----------------------------------------------------------------------
426    Name: Matthias Blume
427    Date: 2002/03/15 16:40:00 EST
428    Tag: blume-20020315-basis
429    Description:
430    
431    Provided (preliminary?) implementations for
432    
433      {String,Substring}.{concatWith,isSuffix,isSubstring}
434    
435    and
436    
437      Substring.full
438    
439    Those are in the Basis spec but they were missing in SML/NJ.
440    
441    ----------------------------------------------------------------------
442    Name: Matthias Blume
443    Date: 2002/03/14 21:30:00 EST
444    Tag: blume-20020314-controls
445    Description:
446    
447    Controls:
448    ---------
449    
450    1. Factored out the recently-added Controls : CONTROLS stuff and put
451       it into its own library $/controls-lib.cm.  The source tree for
452       this is under src/smlnj-lib/Controls.
453    
454    2. Changed the names of types and functions in this interface, so they
455       make a bit more "sense":
456    
457          module -> registry
458          'a registry -> 'a group
459    
460    3. The interface now deals in ref cells only.  The getter/setter interface
461       is (mostly) gone.
462    
463    4. Added a function that lets one register an already-existing ref cell.
464    
465    5. Made the corresponding modifications to the rest of the code so that
466       everything compiles again.
467    
468    6. Changed the implementation of Controls.MLRISC back to something closer
469       to the original.  In particular, this module (and therefore MLRISC)
470       does not depend on Controls.  There now is some link-time code in
471       int-sys.sml that registers the MLRISC controls with the Controls
472       module.
473    
474    CM:
475    ---
476    
477      * One can now specify the lambda-split aggressiveness in init.cmi.
478    
479    ----------------------------------------------------------------------
480    Name: Allen Leung
481    Date: 2002/03/13 17:30:00 EST
482    Tag: leunga-20020313-x86-fp-unary
483    Description:
484    
485    Bug fix for:
486    
487    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
488    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
489    > - fun f(x,(y,z)) = Real.~ y;
490    > [autoloading]
491    > [autoloading done]
492    >       fchsl   (%eax), 184(%esp)
493    > Error: MLRisc bug: X86MCEmitter.emitInstr
494    >
495    > uncaught exception Error
496    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
497    
498    The problem was that the code generator did not generate any fp registers
499    in this case, and the ra didn't know that it needed to run the X86FP phase to
500    translate the pseudo fp instruction.   This only happened with unary fp
501    operators in certain situations.
502    
503    ----------------------------------------------------------------------
504    Name: Matthias Blume
505    Date: 2002/03/13 14:00:00 EST
506    Tag: blume-20020313-overload-etc
507    Description:
508    
509    1. Added _overload as a synonym for overload for backward compatibility.
510       (Control.overloadKW must be true for either version to be accepted.)
511    
512    2. Fixed bug in install script that caused more things to be installed
513       than what was requested in config/targets.
514    
515    3. Made CM aware of the (_)overload construct so that autoloading
516       works.
517    
518    ----------------------------------------------------------------------
519    Name: Matthias Blume
520    Date: 2002/03/12 22:03:00 EST
521    Tag: blume-20020312-url
522    Description:
523    
524    Forgot to update BOOT and srcarchiveurl.
525    
526    ----------------------------------------------------------------------
527    Name: Matthias Blume
528    Date: 2002/03/12 17:30:00 EST
529    Tag: blume-20020312-version110392
530    Description:
531    
532    Yet another version number bump (because of small changes to the
533    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
534    
535    Changes:
536    
537      The new pid generation scheme described a few weeks ago was overly
538      complicated.  I implemented a new mechanism that is simpler and
539      provides a bit more "stability":  Once CM has seen a compilation
540      unit, it keeps its identity constant (as long as you do not delete
541      those crucial CM/GUID/* files).  This means that when you change
542      an interface, compile, then go back to the old interface, and
543      compile again, you arrive at the original pid.
544    
545      There now also is a mechanism that instructs CM to use the plain
546      environment hash as a module's pid (effectively making its GUID
547      the empty string).  For this, "noguid" must be specified as an
548      option to the .sml file in question within its .cm file.
549      This is most useful for code that is being generated by tools such
550      as ml-nlffigen (because during development programmers tend to
551      erase the tool's entire output directory tree including CM's cached
552      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
553      revert to the old, broken behavior of SML/NJ, but in specific cases
554      where there is no danger of interface confusion, its use is ok
555      (I think).
556    
557      ml-nlffigen by default generates "noguid" annotations.  They can be
558      turned off by specifying -guid in its command line.
559    
560    ----------------------------------------------------------------------
561    Name: Lal George
562    Date: 2002/03/12 12 14:42:36 EST
563    Tag: george-20020312-frequency-computation
564    Description:
565    
566    Integrated jump chaining and static block frequency into the
567    compiler. More details and numbers later.
568    
569    ----------------------------------------------------------------------
570    Name: Lal George
571    Date: 2002/03/11 11 22:38:53 EST
572    Tag: george-20020311-jump-chain-elim
573    Description:
574    
575    Tested the jump chain elimination on all architectures (except the
576    hppa).  This is on by default right now and is profitable for the
577    alpha and x86, however, it may not be profitable for the sparc and ppc
578    when compiling the compiler.
579    
580    The gc test will typically jump to a label at the end of the cluster,
581    where there is another jump to an external cluster containing the actual
582    code to invoke gc. This is to allow factoring of common gc invocation
583    sequences. That is to say, we generate:
584    
585            f:
586               testgc
587               ja   L1      % jump if above to L1
588    
589            L1:
590               jmp L2
591    
592    
593    After jump chain elimination the 'ja L1' instructions is converted to
594    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
595    up being implemented in their long form (if L2 is far away) using:
596    
597            jbe     L3      % jump if below or equal to L3
598            jmp     L2
599         L3:
600            ...
601    
602    
603    For large compilation units L2  may be far away.
604    
605    
606    ----------------------------------------------------------------------
607    Name: Matthias Blume
608    Date: 2002/03/11 13:30:00 EST
609    Tag: blume-20020311-mltreeeval
610    Description:
611    
612    A functor parameter was missing.
613    
614    ----------------------------------------------------------------------
615    Name: Allen Leung
616    Date: 2002/03/11 10:30:00 EST
617    Tag: leunga-20020311-runtime-string0
618    Description:
619    
620       The representation of the empty string now points to a
621    legal null terminated C string instead of unit.  It is now possible
622    to convert an ML string into C string with InlineT.CharVector.getData.
623    This compiles into one single machine instruction.
624    
625    ----------------------------------------------------------------------
626    Name: Allen Leung
627    Date: 2002/03/10 23:55:00 EST
628    Tag: leunga-20020310-x86-call
629    Description:
630    
631       Added machine generation for CALL instruction (relative displacement mode)
632    
633    ----------------------------------------------------------------------
634    Name: Matthias Blume
635    Date: 2002/03/08 16:05:00
636    Tag: blume-20020308-entrypoints
637    Description:
638    
639    Version number bumped to 110.39.1.  NEW BOOTFILES!
640    
641    Entrypoints: non-zero offset into a code object where execution should begin.
642    
643    - Added the notion of an entrypoint to CodeObj.
644    - Added reading/writing of entrypoint info to Binfile.
645    - Made runtime system bootloader aware of entrypoints.
646    - Use the address of the label of the first function given to mlriscGen
647      as the entrypoint.  This address is currently always 0, but it will
648      not be 0 once we turn on block placement.
649    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
650      with entry points) from mlriscGen.
651    
652    ----------------------------------------------------------------------
653    Name: Allen Leung
654    Date: 2002/03/07 20:45:00 EST
655    Tag: leunga-20020307-x86-cmov
656    Description:
657    
658       Bug fixes for CMOVcc on x86.
659    
660       1. Added machine code generation for CMOVcc
661       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
662       3. CMOVcc cannot have an immediate operand as argument.
663    
664    ----------------------------------------------------------------------
665    Name: Matthias Blume
666    Date: 2002/03/07 16:15:00 EST
667    Tag: blume-20020307-controls
668    Description:
669    
670    This is a very large but mostly boring patch which makes (almost)
671    every tuneable compiler knob (i.e., pretty much everything under
672    Control.* plus a few other things) configurable via both the command
673    line and environment variables in the style CM did its configuration
674    until now.
675    
676    Try starting sml with '-h' (or, if you are brave, '-H')
677    
678    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
679    implements the underlying generic mechanism.
680    
681    The interface to some of the existing such facilities has changed somewhat.
682    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
683    (The getFoo interface is still there for backward-compatibility, but its
684    use is deprecated.)
685    
686    The ml-build script passes -Cxxx=yyy command-line arguments through so
687    that one can now twiddle the compiler settings when using this "batch"
688    compiler.
689    
690    TODO items:
691    
692    We should go through and throw out all controls that are no longer
693    connected to anything.  Moreover, we should go through and provide
694    meaningful (and correct!) documentation strings for those controls
695    that still are connected.
696    
697    Currently, multiple calls to Controls.new are accepted (only the first
698    has any effect).  Eventually we should make sure that every control
699    is being made (via Controls.new) exactly once.  Future access can then
700    be done using Controls.acc.
701    
702    Finally, it would probably be a good idea to use the getter-setter
703    interface to controls rather than ref cells.  For the time being, both
704    styles are provided by the Controls module, but getter-setter pairs are
705    better if thread-safety is of any concern because they can be wrapped.
706    
707    *****************************************
708    
709    One bug fix: The function blockPlacement in three of the MLRISC
710    backpatch files used to be hard-wired to one of two possibilities at
711    link time (according to the value of the placementFlag).  But (I
712    think) it should rather sense the flag every time.
713    
714    *****************************************
715    
716    Other assorted changes (by other people who did not supply a HISTORY entry):
717    
718    1. the cross-module inliner now works much better (Monnier)
719    2. representation of weights, frequencies, and probabilities in MLRISC
720       changed in preparation of using those for weighted block placement
721       (Reppy, George)
722    
723    ----------------------------------------------------------------------
724    Name: Lal George
725    Date: 2002/03/07 14:44:24 EST 2002
726    Tag: george-20020307-weighted-block-placement
727    
728    Tested the weighted block placement optimization on all architectures
729    (except the hppa) using AMPL to generate the block and edge frequencies.
730    Changes were required in the machine properties to correctly
731    categorize trap instructions. There is an MLRISC flag
732    "weighted-block-placement" that can be used to enable weighted block
733    placement, but this will be ineffective without block/edge
734    frequencies (coming soon).
735    
736    
737    ----------------------------------------------------------------------
738    Name: Lal George
739    Date: 2002/03/05 17:24:48 EST
740    Tag: george-20020305-linkage-cluster
741    
742    In order to support the block placement optimization, a new cluster
743    is generated as the very first cluster (called the linkage cluster).
744    It contains a single jump to the 'real' entry point for the compilation
745    unit. Block placement has no effect on the linkage cluster itself, but
746    all the other clusters  have full freedom in the manner in which they
747    reorder blocks or functions.
748    
749    On the x86 the typical linkage code that is generated is:
750       ----------------------
751            .align 2
752       L0:
753            addl    $L1-L0, 72(%esp)
754            jmp     L1
755    
756    
757            .align  2
758       L1:
759       ----------------------
760    
761    72(%esp) is the memory location for the stdlink register. This
762    must contain the address of the CPS function being called. In the
763    above example, it contains the address of  L0; before
764    calling L1 (the real entry point for the compilation unit), it
765    must contain the address for L1, and hence
766    
767            addl $L1-L0, 72(%esp)
768    
769    I have tested this on all architectures except the hppa.The increase
770    in code size is of course negligible
771    
772    ----------------------------------------------------------------------
773    Name: Allen Leung
774    Date: 2002/03/03 13:20:00 EST
775    Tag: leunga-20020303-mlrisc-tools
776    
777      Added #[ ... ] expressions to mlrisc tools
778    
779    ----------------------------------------------------------------------
780    Name: Matthias Blume
781    Date: 2002/02/27 12:29:00 EST
782    Tag: blume-20020227-cdebug
783    Description:
784    
785    - made types in structure C and C_Debug to be equal
786    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
787    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
788    
789    ----------------------------------------------------------------------
790    Name: Matthias Blume
791    Date: 2002/02/26 12:00:00 EST
792    Tag: blume-20020226-ffi
793    Description:
794    
795    1. Fixed a minor bug in CM's "noweb" tool:
796       If numbering is turned off, then truly don't number (i.e., do not
797       supply the -L option to noweb).  The previous behavior was to supply
798       -L'' -- which caused noweb to use the "default" line numbering scheme.
799       Thanks to Chris Richards for pointing this out (and supplying the fix).
800    
801    2. Once again, I reworked some aspects of the FFI:
802    
803       A. The incomplete/complete type business:
804    
805       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
806         gone!
807       - ML types representing an incomplete type are now *equal* to
808         ML types representing their corresponding complete types (just like
809         in C).  This is still safe because ml-nlffigen will not generate
810         RTTI for incomplete types, nor will it generate functions that
811         require access to such RTTI.   But when ML code generated from both
812         incomplete and complete versions of the C type meet, the ML types
813         are trivially interoperable.
814    
815         NOTE:  These changes restore the full generality of the translation
816         (which was previously lost when I eliminated functorization)!
817    
818       B. Enum types:
819    
820       - Structure C now has a type constructor "enum" that is similar to
821         how the "su" constructor works.  However, "enum" is not a phantom
822         type because each "T enum" has values (and is isomorphic to
823         MLRep.Signed.int).
824       - There are generic access operations for enum objects (using
825         MLRep.Signed.int).
826       - ml-nlffigen will generate a structure E_foo for each "enum foo".
827         * The structure contains the definition of type "mlrep" (the ML-side
828         representation type of the enum).  Normally, mlrep is the same
829         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
830         then mlrep will be defined as a datatype -- thus facilitating
831         pattern matching on mlrep values.
832         ("-ec" will be suppressed if there are duplicate values in an
833          enumeration.)
834         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
835         will be generated for each C enum constant xxx.
836         * Conversion functions m2i and i2m convert between mlrep and
837         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
838         * Coversion functions c and ml convert between mlrep and "tag enum".
839         * Access functions (get/set) fetch and store mlrep values.
840       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
841         enumerations are merged into one single enumeration represented by
842         structure E_'.
843    
844    ----------------------------------------------------------------------
845    Name: Allen Leung
846    Date: 2002/02/25 04:45:00 EST
847    Tag: leunga-20020225-cps-spill
848    
849    This is a new implementation of the CPS spill phase.
850    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
851    In case of problems, replace it with the old file spill.sml
852    
853    The current compiler runs into some serious performance problems when
854    constructing a large record.  This can happen when we try to compile a
855    structure with many items.  Even a very simple structure like the following
856    makes the compiler slow down.
857    
858        structure Foo = struct
859           val x_1 = 0w1 : Word32.int
860           val x_2 = 0w2 : Word32.int
861           val x_3 = 0w3 : Word32.int
862           ...
863           val x_N = 0wN : Word32.int
864        end
865    
866    The following table shows the compile time, from N=1000 to N=4000,
867    with the old compiler:
868    
869    N
870    1000   CPS 100 spill                           0.04u  0.00s  0.00g
871           MLRISC ra                               0.06u  0.00s  0.05g
872              (spills = 0 reloads = 0)
873           TOTAL                                   0.63u  0.07s  0.21g
874    
875    1100   CPS 100 spill                           8.25u  0.32s  0.64g
876           MLRISC ra                               5.68u  0.59s  3.93g
877              (spills = 0 reloads = 0)
878           TOTAL                                   14.71u  0.99s  4.81g
879    
880    1500   CPS 100 spill                           58.55u  2.34s  1.74g
881           MLRISC ra                               5.54u  0.65s  3.91g
882              (spills = 543 reloads = 1082)
883           TOTAL                                   65.40u  3.13s  6.00g
884    
885    2000   CPS 100 spill                           126.69u  4.84s  3.08g
886           MLRISC ra                               0.80u  0.10s  0.55g
887              (spills = 42 reloads = 84)
888           TOTAL                                   129.42u  5.10s  4.13g
889    
890    3000   CPS 100 spill                           675.59u  19.03s  11.64g
891           MLRISC ra                               2.69u  0.27s  1.38g
892              (spills = 62 reloads = 124)
893           TOTAL                                   682.48u  19.61s  13.99g
894    
895    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
896           MLRISC ra                               4.96u  0.27s  2.72g
897              (spills = 85 reloads = 170)
898           TOTAL                                   2375.26u  57.21s  48.00g
899    
900    As you can see the old cps spill module suffers from some serious
901    performance problem.  But since I cannot decipher the old code fully,
902    instead of patching the problems up, I'm reimplementing it
903    with a different algorithm.  The new code is more modular,
904    smaller when compiled, and substantially faster
905    (O(n log n) time and O(n) space).  Timing of the new spill module:
906    
907    4000  CPS 100 spill                           0.02u  0.00s  0.00g
908          MLRISC ra                               0.25u  0.02s  0.15g
909             (spills=1 reloads=3)
910          TOTAL                                   7.74u  0.34s  1.62g
911    
912    Implementation details:
913    
914    As far as I can tell, the purpose of the CPS spill module is to make sure the
915    number of live variables at any program point (the bandwidth)
916    does not exceed a certain limit, which is determined by the
917    size of the spill area.
918    
919    When the bandwidth is too large, we decrease the register pressure by
920    packing live variables into spill records.  How we achieve this is
921    completely different than what we did in the old code.
922    
923    First, there is something about the MLRiscGen code generator
924    that we should be aware of:
925    
926    o MLRiscGen performs code motion!
927    
928       In particular, it will move floating point computations and
929       address computations involving only the heap pointer to
930       their use sites (if there is only a single use).
931       What this means is that if we have a CPS record construction
932       statement
933    
934           RECORD(k,vl,w,e)
935    
936       we should never count the new record address w as live if w
937       has only one use (which is often the case).
938    
939       We should do something similar to floating point, but the transformation
940       there is much more complex, so I won't deal with that.
941    
942    Secondly, there are now two new cps primops at our disposal:
943    
944     1. rawrecord of record_kind option
945        This pure operator allocates some uninitialized storage from the heap.
946        There are two forms:
947    
948         rawrecord NONE [INT n]  allocates a tagless record of length n
949         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
950                                     and initializes the tag.
951    
952     2. rawupdate of cty
953          rawupdate cty (v,i,x)
954          Assigns to x to the ith component of record v.
955          The storelist is not updated.
956    
957    We use these new primops for both spilling and increment record construction.
958    
959     1. Spilling.
960    
961        This is implemented with a linear scan algorithm (but generalized
962        to trees).  The algorithm will create a single spill record at the
963        beginning of the cps function and use rawupdate to spill to it,
964        and SELECT or SELp to reload from it.  So both spills and reloads
965        are fine-grain operations.  In contrast, in the old algorithm
966        "spills" have to be bundled together in records.
967    
968        Ideally, we should sink the spill record construction to where
969        it is needed.  We can even split the spill record into multiple ones
970        at the places where they are needed.  But CPS is not a good
971        representation for global code motion, so I'll keep it simple and
972        am not attempting this.
973    
974     2. Incremental record construction (aka record splitting).
975    
976        Long records with many component values which are simulatenously live
977        (recall that single use record addresses are not considered to
978         be live) are constructed with rawrecord and rawupdate.
979        We allocate space on the heap with rawrecord first, then gradually
980        fill it in with rawupdate.  This is the technique suggested to me
981        by Matthias.
982    
983        Some restrictions on when this is applicable:
984        1. It is not a VECTOR record.  The code generator currently does not handle
985           this case. VECTOR record uses double indirection like arrays.
986        2. All the record component values are defined in the same "basic block"
987           as the record constructor.  This is to prevent speculative
988           record construction.
989    
990    ----------------------------------------------------------------------
991    Name: Allen Leung
992    Date: 2002/02/22 01:02:00 EST
993    Tag: leunga-20020222-mlrisc-tools
994    
995    Minor bug fixes in the parser and rewriter
996    
997    ----------------------------------------------------------------------
998    Name: Allen Leung
999    Date: 2002/02/21 20:20:00 EST
1000    Tag: leunga-20020221-peephole
1001    
1002    Regenerated the peephole files.  Some contained typos in the specification
1003    and some didn't compile because of pretty printing bugs in the old version
1004    of 'nowhere'.
1005    
1006    ----------------------------------------------------------------------
1007    Name: Allen Leung
1008    Date: 2002/02/19 20:20:00 EST
1009    Tag: leunga-20020219-mlrisc-tools
1010    Description:
1011    
1012       Minor bug fixes to the mlrisc-tools library:
1013    
1014       1.  Fixed up parsing colon suffixed keywords
1015       2.  Added the ability to shut the error messages up
1016       3.  Reimplemented the pretty printer and fixed up/improved
1017           the pretty printing of handle and -> types.
1018       4.  Fixed up generation of literal symbols in the nowhere tool.
1019       5.  Added some SML keywords to to sml.sty
1020    
1021    ----------------------------------------------------------------------
1022    Name: Matthias Blume
1023    Date: 2002/02/19 16:20:00 EST
1024    Tag: blume-20020219-cmffi
1025    Description:
1026    
1027    A wild mix of changes, some minor, some major:
1028    
1029    * All C FFI-related libraries are now anchored under $c:
1030        $/c.cm      --> $c/c.cm
1031        $/c-int.cm  --> $c/internals/c-int.cm
1032        $/memory.cm --> $c/memory/memory.cm
1033    
1034    * "make" tool (in CM) now treats its argument pathname slightly
1035      differently:
1036        1. If the native expansion is an absolute name, then before invoking
1037           the "make" command on it, CM will apply OS.Path.mkRelative
1038           (with relativeTo = OS.FileSys.getDir()) to it.
1039        2. The argument will be passed through to subsequent phases of CM
1040           processing without "going native".  In particular, if the argument
1041           was an anchored path, then "make" will not lose track of that anchor.
1042    
1043    * Compiler backends now "know" their respective C calling conventions
1044      instead of having to be told about it by ml-nlffigen.  This relieves
1045      ml-nlffigen from one of its burdens.
1046    
1047    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1048    
1049    * Export C_DEBUG and C_Debug from $c/c.cm.
1050    
1051    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1052      subtyping relationship between incomplete pointers and their complete
1053      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1054      with the convention of instantiating 'o with ('t, 'c) obj whenever
1055      the pointer target type is complete.  In the incomplete case, 'o
1056      will be instantiated with some "'c iobj" -- a type obtained by
1057      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1058    
1059      Operations that work on both incomplete and complete pointer types are
1060      typed as taking an 'o ptr while operations that require the target to
1061      be known are typed as taking some ('t, 'c) obj ptr.
1062    
1063      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1064      where void is an eqtype without any values.  This makes it possible
1065      to work on voidptr values using functions meant to operate on light
1066      incomplete pointers.
1067    
1068    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1069      been vastly simplified.
1070    
1071    ----------------------------------------------------------------------
1072    Name: Matthias Blume
1073    Date: 2002/02/19 10:48:00 EST
1074    Tag: blume-20020219-pqfix
1075    Description:
1076    
1077    Applied Chris Okasaki's bug fix for priority queues.
1078    
1079    ----------------------------------------------------------------------
1080    Name: Matthias Blume
1081    Date: 2002/02/15 17:05:00
1082    Tag: Release_110_39
1083    Description:
1084    
1085    Last-minute retagging is becoming a tradition... :-(
1086    
1087    This is the working release 110.39.
1088    
1089    ----------------------------------------------------------------------
1090    Name: Matthias Blume
1091    Date: 2002/02/15 16:00:00 EST
1092    Tag: Release_110_39-orig
1093    Description:
1094    
1095    Working release 110.39.  New bootfiles.
1096    
1097    (Update: There was a small bug in the installer so it wouldn't work
1098    with all shells.  So I retagged. -Matthias)
1099    
1100    ----------------------------------------------------------------------
1101    Name: Matthias Blume
1102    Date: 2002/02/15 14:17:00 EST
1103    Tag: blume-20020215-showbindings
1104    Description:
1105    
1106    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1107    the latter can be useful for exploring what bindings are available at
1108    the interactive prompt.  (The first function returns only the list
1109    of symbols that are really bound, the second prints those but also the
1110    ones that CM's autoloading mechanism knows about.)
1111    
1112    ----------------------------------------------------------------------
1113    Name: Matthias Blume
1114    Date: 2002/02/15 12:08:00 EST
1115    Tag: blume-20020215-iptrs
1116    Description:
1117    
1118    Two improvements to ml-nlffigen:
1119    
1120      1. Write files only if they do not exist or if their current contents
1121         do not coincide with what's being written.  (That is, avoid messing
1122         with the time stamps unless absolutely necessary.)
1123    
1124      2. Implement a "repository" mechanism for generated files related
1125         to "incomplete pointer types".   See the README file for details.
1126    
1127    ----------------------------------------------------------------------
1128    Name: Matthias Blume
1129    Date: 2002/02/14 11:50:00 EST
1130    Tag: blume-20020214-quote
1131    Description:
1132    
1133    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1134    because of the new and improved tag generation scheme.  (Thanks to Allen
1135    Leung for pointing it out.)
1136    
1137    ----------------------------------------------------------------------
1138    Name: Lal George
1139    Date: 2002/02/14 09:55:27 EST 2002
1140    Tag: george-20020214-isabelle-bug
1141    Description:
1142    
1143    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1144    of Isabelle on the x86.
1145    
1146    From Allen:
1147    -----------
1148     I've found the problem:
1149    
1150         in ra-core.sml, I use the counter "blocked" to keep track of the
1151         true number of elements in the freeze queue.  When the counter goes
1152         to zero, I skip examining the queue.  But I've messed up the
1153         bookkeeping in combine():
1154    
1155             else ();
1156             case !ucol of
1157               PSEUDO => (if !cntv > 0 then
1158                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1159                                        ^^^^^^^^^^^^^^^^^^^^^^^
1160                      moveu := mergeMoveList(!movev, !moveu)
1161                     )
1162                  else ();
1163    
1164         combine() is called to coalesce two nodes u and v.
1165         I think I was thinking that if the move counts of u and v are both
1166         greater than zero then after they are coalesced then one node is
1167         removed from the freeze queue.  Apparently I was thinking that
1168         both u and v are of low degree, but that's clearly not necessarily true.
1169    
1170    
1171    02/12/2002:
1172        Here's the patch.  HOL now compiles.
1173    
1174        I don't know how this impact on performance (compile
1175        time or runtime).  This bug caused the RA (especially on the x86)
1176        to go thru the potential spill phase when there are still nodes on the
1177        freeze queue.
1178    
1179    
1180    
1181    
1182    ----------------------------------------------------------------------
1183    Name: Matthias Blume
1184    Date: 2002/02/13 22:40:00 EST
1185    Tag: blume-20020213-fptr-rtti
1186    Description:
1187    
1188    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1189    updates.
1190    
1191    ----------------------------------------------------------------------
1192    Name: Matthias Blume
1193    Date: 2002/02/13 16:41:00 EST
1194    Tag: blume-20020213-cmlpq
1195    Description:
1196    
1197    Added new priority queue export symbols (which have just been added to
1198    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1199    not compile and the installer would choke.)
1200    
1201    ----------------------------------------------------------------------
1202    Name: Matthias Blume
1203    Date: 2002/02/13 16:15:00 EST
1204    Tag: blume-20020213-various
1205    Description:
1206    
1207    1. More tweaks to ml-nlffigen:
1208    
1209       - better internal datastructures (resulting in slight speedup)
1210       - "-match" option requires exact match
1211       - "localized" gensym counters (untagged structs/unions nested within
1212         other structs/unions or within typedefs get a fresh counter; their
1213         tag will be prefixed by a concatenation of their parents' tags)
1214       - bug fixes (related to calculation of transitive closure of types
1215         to be included in the output)
1216    
1217    2. Minor Basis updates:
1218    
1219       - added implementations for List.collate and Option.app
1220    
1221    ----------------------------------------------------------------------
1222    Name: Matthias Blume
1223    Date: 2002/02/11 15:55:00 EST
1224    Tag: blume-20020211-gensym
1225    Description:
1226    
1227    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1228    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1229    names (ML structure names that correspond to unnamed C structs, unions,
1230    and enums), so that separate runs of ml-nlffigen do not clash.
1231    
1232    ----------------------------------------------------------------------
1233    Name: Matthias Blume
1234    Date: 2002/02/11 12:05:00 EST
1235    Tag: blume-20020211-gensml
1236    Description:
1237    
1238    A quick fix for a problem with GenSML (in the pgraph-util library):
1239    Make generation of toplevel "local" optional.  (Strictly speaking,
1240    signature definitions within "local" are not legal SML.)
1241    
1242    Other than that: updates to INSTALL and cm/TODO.
1243    
1244    ----------------------------------------------------------------------
1245    Name: Matthias Blume
1246    Date: 2002/02/08 15:00:00 EST
1247    Tag: blume-20020208-uniquepid
1248    Description:
1249    
1250    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1251    
1252    1. The installer (config/install.sh) has gotten smarter:
1253    
1254         - Configuration options are a bit easier to specify now
1255           (in config/targets).
1256         - Bug in recognizing .tar.bz2 files fixed.
1257         - Installer automatically resolves dependencies between
1258           configuration options (e.g., if you ask for eXene, you will
1259           also get cml -- regardless whether you asked for it or not).
1260         - Installer can run in "quieter mode" by setting the environment
1261           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1262           "completely silent", though.
1263         - Build HashCons library as part of smlnj-lib.
1264    
1265    2. A new scheme for assigning persistent identifiers to compilation
1266       units (and, by extension, to types etc.) has been put into place.
1267       This fixes a long-standing bug where types and even dynamic values
1268       can get internally confused, thereby compromising type safety
1269       (abstraction) and dynamic correctness.  See
1270    
1271         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1272    
1273       for an example of how things could go wrong until now.
1274    
1275       The downside of the new scheme is that pids are not quite as
1276       persistent as they used to be: CM will generate a fresh pid
1277       for every compilation unit that it thinks it sees for the first
1278       time.  That means that if you compile starting from a clean, fresh
1279       source tree at two different times, you end up with different
1280       binaries.
1281    
1282       Cutoff recompilation, however, has not been compromised because
1283       CM keeps pid information in special caches between runs.
1284    
1285    ----------------------------------------------------------------------
1286    Name: Lal George
1287    Date: 2002/02/07 15:34:13 EST 2002
1288    Tag: <none>
1289    Description:
1290    
1291    Compilers that generate assembly code may produce  global labels
1292    whose value is resolved at link time. The various peephole optimization
1293    modules did not take this in account.
1294    
1295    TODO. The Labels.addrOf function should really return an option
1296    type so that clients are forced to deal with this issue, rather
1297    than an exception being raised.
1298    
1299    ----------------------------------------------------------------------
1300    Name: Lal George
1301    Date: 2002/02/06 13:55:02 EST
1302    Tag: george-20020206-ra-breakup
1303    Description:
1304    
1305    1. A bug fix from Allen.
1306    
1307        A typo causes extra fstp %st(0)'s to be generated at compensation
1308        edges, which might cause stack underflow traps at runtime.  This
1309        occurs in fft where there are extraneous fstps right before the 'into'
1310        trap instruction (in this case they are harmless since none of the
1311        integers overflow.)
1312    
1313    2. Pulled out various utility modules that were embedded in the modules
1314       of the register allocator. I need these modules for other purposes, but
1315       they are not complete enough to put into a library (just yet).
1316    ----------------------------------------------------------------------
1317    Name: Matthias Blume
1318    Date: 2002/01/31 16:05:00 EST
1319    Tag: blume-20020131-sparc-ccalls
1320    Description:
1321    
1322    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1323       of extra stack space by mistake.  Fixed.
1324    
1325    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1326    
1327    ----------------------------------------------------------------------
1328    Name: Allen Leung
1329    Date: 2002/01/30
1330    Tag: leunga-20020130-nowhere-bug-fix
1331    Description:
1332    
1333       MLRISC bug fixes:
1334       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1335       2. MachineInt.fromString was negating its value.
1336    
1337    ----------------------------------------------------------------------
1338    Name: Matthias Blume
1339    Date: 2002/01/29
1340    Tag: blume-20020129-INSTALL
1341    Description:
1342    
1343    - Added somewhat detailed installation instructions (file INSTALL).
1344    - Fixed curl-detection bug in config/install.sh.
1345    - It is now possible to select the URL getter using the URLGETTER
1346      environment variable:
1347    
1348          not set / "unknown"      --> automatic detection (script tries wget,
1349                                       curl, and lynx)
1350          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1351                                       how to properly invoke them)
1352          other                    --> use $URLGETTER directly, it must take
1353                                       precisely two command-line arguments
1354                                       (source URL and destination file name)
1355    
1356    ----------------------------------------------------------------------
1357    Name: Matthias Blume
1358    Date: 2002/01/28
1359    Tag: blume-20020128-sparc-ccalls
1360    Description:
1361    
1362    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1363    - Make use of the allocParam argument in sparc-c-calls.
1364    
1365    ----------------------------------------------------------------------
1366    Name: Matthias Blume
1367    Date: 2002/01/28
1368    Tag: blume-20020128-allocParam
1369    Description:
1370    
1371    John Reppy:  Changes c-calls API to accept client-callback for
1372    allocating extra stack space.
1373    me: Corresponding changes to mlriscGen (using a dummy argument that
1374        does not change the current behavior).
1375    
1376    ----------------------------------------------------------------------
1377    Name: Matthias Blume
1378    Date: 2002/01/28 12:00:00
1379    Tag: Release_110_38
1380    Description:
1381    
1382    This time for real!!!
1383    
1384    ----------------------------------------------------------------------
1385    Name: Matthias Blume
1386    Date: 2002/01/28 10:56:00 EST
1387    Tag: blume-20020128-retraction
1388    Description:
1389    
1390    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1391       with blume-Release_110_38-retracted.)
1392    
1393    1. Fixed a problem with incorrect rounding modes in real64.sml.
1394       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1395    
1396    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1397       and enums fixed.  The naming of corresponding ML identifiers should
1398       now be consistent again.
1399    
1400    ----------------------------------------------------------------------
1401    Name: Allen Leung
1402    Date: 2002/01/27
1403    Tag: leunga-20020127-nowhere
1404    Description:
1405    
1406       Added a target called nowhere in the configuration scripts.
1407       Enabling this will build the MLRISC 'nowhere' tool (for translating
1408       programs with where-clauses into legal SML code) during installation.
1409    
1410    ----------------------------------------------------------------------
1411    Name: Matthias Blume
1412    Date: 2002/01/25 21:27:00 EST
1413    Tag: blume-Release_110_38-retracted
1414    Description:
1415    
1416    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1417    
1418    README will be added later.
1419    
1420    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1421    (Release_110_38) removed.  Reason: Last-minute bug fixes.
1422    
1423    ----------------------------------------------------------------------
1424    Name: Matthias Blume
1425    Date: 2002/01/25
1426    Tag: blume-20020125-ffi
1427    Description:
1428    
1429    A large number of tweaks and improvements to ml-nlffi-lib and
1430    ml-nlffigen:
1431    
1432       - ML represenation types have been streamlined
1433       - getter and setter functions work with concrete values, not abstract
1434         ones where possible
1435       - ml-nlffigen command line more flexible (see README file there)
1436       - some bugs have been fixed (hopefully)
1437    
1438    ----------------------------------------------------------------------
1439    Name: Lal George
1440    Date: 2002/01/24
1441    Tag: george-20020124-risc-ra-interface
1442    Description:
1443    
1444       There is a dramatic simplification in the interface to the
1445       register allocator for RISC architectures as a result of making
1446       parallel copy instructions explicit.
1447    
1448    ----------------------------------------------------------------------
1449    Name: Matthias Blume
1450    Date: 2002/01/22
1451    Tag: blume-20020122-x86-ccalls
1452    Description:
1453    
1454    Bug fix for c-calls on x86 (having to do with how char- and
1455    short-arguments are being handled).
1456    
1457    ----------------------------------------------------------------------
1458    Name: Matthias Blume
1459    Date: 2002/01/21
1460    Tag: blume-20020121-ff
1461    Description:
1462    
1463    Another day of fiddling with the FFI...
1464    
1465    1. Bug fix/workaround:  CKIT does not complain about negative array
1466       dimensions, so ml-nlffigen has to guard itself against this possibility.
1467       (Otherwise a negative dimension would send it into an infinite loop.)
1468    
1469    2. Some of the abstract types (light objects, light pointers, most "base"
1470       types) in structure C are now eqtypes.
1471    
1472    3. Added constructors and test functions for NULL function pointers.
1473    
1474    ----------------------------------------------------------------------
1475    Name: Matthias Blume
1476    Date: 2002/01/18
1477    Tag: blume-20020118-ready-for-new-release
1478    Description:
1479    
1480    Made config/srcarchiveurl point to a new place.  (Will provide boot
1481    files shortly.)
1482    
1483    Maybe we christen this to be 110.38?
1484    
1485    ----------------------------------------------------------------------
1486    Name: Matthias Blume
1487    Date: 2002/01/18
1488    Tag: blume-20020118-more-ffifiddle
1489    Description:
1490    
1491    Today's FFI fiddling:
1492    
1493      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
1494        C.Set.* that use concrete (MLRep.*) arguments and results instead
1495        of abstract ones.
1496    
1497      - Provided word-style bit operations etc. for "int" representation
1498        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1499        and Long.
1500    
1501    ----------------------------------------------------------------------
1502    Name: Matthias Blume
1503    Date: 2002/01/18
1504    Tag: blume-20020118-use-x86-fp
1505    Description:
1506    
1507    Now that x86-fast-fp seems to be working, I turned it back on again
1508    by default.  (Seems to work fine now, even with the FFI.)
1509    
1510    Other than that, I added some documentation about the FFI to
1511    src/ml-nlffigen/README and updated the FFI test examples in
1512    src/ml-nlffi-lib/Tests/*.
1513    
1514    ----------------------------------------------------------------------
1515    Name: Allen Leung
1516    Date: 2002/01/17
1517    Tag: leunga-20020117-x86-fast-fp-call
1518    Description:
1519    
1520       1. Fixed a problem with handling return fp values when x86's fast fp
1521          mode is turned on.
1522    
1523       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
1524          of %f32.
1525    
1526       3. Added a constructor INT32lit to the ast of MLRISC tools.
1527    
1528    ----------------------------------------------------------------------
1529    Name: Matthias Blume
1530    Date: 2002/01/16
1531    Tag: blume-20020116-ffifiddle
1532    Description:
1533    
1534    More fiddling with the FFI interface:
1535    
1536     - Make constness 'c instead of rw wherever possible.  This eliminates
1537       the need for certain explicit coercions.  (However, due to ML's
1538       value polymorphism, there will still be many cases where explicit
1539       coercions are necessary.  Phantom types are not the whole answer
1540       to modeling a subtyping relationship in ML.)
1541    
1542     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
1543    
1544     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
1545       clumsy workaround for varargs problem.)
1546    
1547    ----------------------------------------------------------------------
1548    Name: Lal George
1549    Date: 2002/01/15
1550    Tag: <none>
1551    Description:
1552    
1553    1. Since COPY instructions are no longer native to the architecture,
1554       a generic functor can be used to implement the expandCopies function.
1555    
1556    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1557       TEXT segment.
1558    
1559    ----------------------------------------------------------------------
1560    Name: Matthias Blume
1561    Date: 2002/01/15
1562    Tag: blume-20020115-ffiupdates
1563    Description:
1564    
1565    1. Fix for bug resulting in single-precision float values being returned
1566       incorrectly from FFI calls.
1567    
1568    2. Small modifications to C FFI API:
1569    
1570        - memory-allocation routines return straight objects (no options)
1571          and raise an exception in out-of-memory situations
1572        - unsafe extensions to cast between function pointers and pointers
1573          from/to ints
1574        - added structure C_Debug as an alternative to structure C where
1575          pointer-dereferencing (|*| and |*!) always check for null-pointers
1576        - added open_lib' to DynLinkage;  open_lib' works like open_lib
1577          but also takes a (possibly empty) list of existing library handles
1578          that the current library depends on
1579    
1580    ----------------------------------------------------------------------
1581    Name: Matthias Blume
1582    Date: 2002/01/10
1583    Tag: blume-20020110-newffigen
1584    Description:
1585    
1586    1. Updates to portable graph code.
1587    
1588    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
1589       more scalable now so that even huge interfaces such as the one
1590       for GTK compile in finite time and space. :-)
1591       See src/ml-nlffigen/README for details on what's new.
1592    
1593  ----------------------------------------------------------------------  ----------------------------------------------------------------------
1594  Name: Lal George  Name: Lal George
1595  Date: 2001/01/09 14:31:35 EST 2002  Date: 2001/01/09 14:31:35 EST 2002
# Line 355  Line 1934 
1934    
1935  ----------------------------------------------------------------------  ----------------------------------------------------------------------
1936  Name: Matthias Blume  Name: Matthias Blume
 >>>>>>> 1.169  
1937  Date: 2001/09/18 15:35:00 EDT  Date: 2001/09/18 15:35:00 EDT
1938  Tag: blume-20010918-readme11036  Tag: blume-20010918-readme11036
1939  Description:  Description:

Legend:
Removed from v.1009  
changed lines
  Added in v.1186

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