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 952, Tue Oct 9 14:00:40 2001 UTC revision 1294, Thu Dec 12 21:22:48 2002 UTC
# Line 8  Line 8 
8  The form of an entry should be:  The form of an entry should be:
9    
10  Name:  Name:
11  Date:  Date: yyyy/mm/dd
12  Tag: <post-commit CVS tag>  Tag: <post-commit CVS tag>
13  Description:  Description:
14    
15  ----------------------------------------------------------------------  ----------------------------------------------------------------------
16    Name: Matthias Blume
17    Date: 2002/12/12 16:25:00 EST
18    Tag: blume-20021212-risc-ra
19    Description:
20    
21    Fixed a serious bug in the rewrite code for FP spilling/reloading that
22    sent the RA into an infinite loop when floating point registers get
23    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
24    110.37 and 110.38.)
25    There was another set of potential problems related to the handling of
26    MLRISC annotations (but those did not yet cause real problems, apparently).
27    
28    ----------------------------------------------------------------------
29    Name: Matthias Blume
30    Date: 2002/12/06 22:40:00 EST
31    Tag: blume-20021206-cm-fileid
32    Description:
33    
34    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
35    This fixes the problem of CM getting confused by files that suddenly
36    change their identity (e.g., by getting unlinked and recreated by some
37    text editor such as vi).  There might be a better/cheaper/cleaner way
38    of doing this, but for now this will have to do.
39    
40    ----------------------------------------------------------------------
41    Name: Matthias Blume
42    Date: 2002/10/28 09:50:00 EST
43    Tag: blume-20021028-typecheck
44    Description:
45    
46    Exported structure Typecheck from $smlnj/viscomp/core.cm.
47    
48    ----------------------------------------------------------------------
49    Name: Matthias Blume
50    Date: 2002/10/17 09:10:00 EDT
51    Tag: Release_110_42
52    Description:
53    
54    In good old tradition, there has been a slight hiccup so that we have
55    to patch 110.42 after the fact.   The old release tag has been replaced
56    (see below).
57    
58    The change solves a problem with two competing approaches the
59    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
60    each other's way.
61    
62    This change only affects the runtime system code and the installer script.
63    (No new bootfiles.)
64    
65    ----------------------------------------------------------------------
66    Name: Matthias Blume
67    Date: 2002/10/16 12:00:00 EDT
68    Tag: Release_110_42_removed
69    Description:
70    
71    New working release.  New bootfiles.
72    
73    ----------------------------------------------------------------------
74    Name: Matthias Blume
75    Date: 2002/10/10 13:10:00 EDT
76    Tag: blume-20021010-ppc-divs
77    Description:
78    
79    The mltree operator DIVS must be implemented with an overflow check on
80    the PPC because the hardware indicates divide-by-zero using "overflow" as
81    well.
82    
83    ----------------------------------------------------------------------
84    Name: Matthias Blume
85    Date: 2002/07/23 11:45:00
86    Tag: blume-20020723-smlnj-home
87    Description:
88    
89    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
90    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
91    CM_PATHCONFIG is also set), the path configuration file is assumed
92    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
93    the entire tree to some other place and everything will "just work".
94    
95    (Companion commands such as ml-build and ml-makedepend also sense this
96    variable.)
97    
98    ----------------------------------------------------------------------
99    Name: Matthias Blume
100    Date: 2002/07/12 21:19:00 EDT
101    Tag: blume-20020712-liveness
102    Description:
103    
104    Exported two useful "step" functions from liveness module (MLRISC).
105    
106    ----------------------------------------------------------------------
107    Name: Matthias Blume
108    Date: 2002/07/05 16:00 EDT
109    Tag: Release_110_41
110    Description:
111    
112    New working release.  New bootfiles.
113    
114    ----------------------------------------------------------------------
115    Name: Matthias Blume
116    Date: 2002/07/05 10:25:00 EDT
117    Tag: blume-20020705-btimp
118    Description:
119    
120    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
121    other clients can set up backtracing support.
122    
123    ----------------------------------------------------------------------
124    Name: Matthias Blume
125    Date: 2002/06/25 17:23:00 EDT
126    Tag: blume-20020625-fpmax
127    Description:
128    
129    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
130    case.  (The sense of the isNaN test was reversed -- which made min and
131    max always return their first argument.)
132    
133    ----------------------------------------------------------------------
134    Name: Matthias Blume
135    Date: 2002/06/11
136    Tag: blume-20020611-unixpath
137    Description:
138    
139    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
140    
141    ----------------------------------------------------------------------
142    Name: Matthias Blume
143    Date: 2002/06/10 16:35:00 EDT
144    Tag: blume-20020610-ieeereal
145    Description:
146    
147    I back-ported my implementation of IEEEReal.fromString from the
148    idlbasis-devel branch so that we can test it.
149    
150    Another small change is that ppDec tries to give more information
151    than just "<sig>" in the case of functors.  However, this code is
152    broken in some mysterious way if the functor's body's signature
153    has not been declared by ascription but gets inferred from the
154    implementation.  This needs fixing...
155    
156    ----------------------------------------------------------------------
157    Name: Matthias Blume
158    Date: 2002/05/31
159    Tag: blume-20020531-btrace-mode
160    Description:
161    
162    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
163    the wayside when I switched over to using Controls everywhere.)
164    
165    ----------------------------------------------------------------------
166    Name: Lal George
167    Date: 2002/05/23 12:21:40 EDT
168    Tag: george-20020523-visual-labels
169    Description:
170    
171    Labels are now displayed in the graphical output to make
172    the fall-through and target blocks obvious.
173    
174    ----------------------------------------------------------------------
175    Name: Matthias Blume
176    Date: 2002/05/22 11:03:00 EDT
177    Tag: blume-20020522-shrink
178    Description:
179    
180    John tweaked yesterday's fix for 1131 to handle an out-of-memory
181    situation that comes up when allocating huge arrays.
182    
183    ----------------------------------------------------------------------
184    Name: Matthias Blume
185    Date: 2002/05/21 16:00:00 EDT
186    Tag: Release_110_40
187    Description:
188    
189    New working release (110.40).  New bootfiles.
190    
191    [Also: John Reppy fixed GC bug 1131.]
192    
193    ----------------------------------------------------------------------
194    Name: Matthias Blume
195    Date: 2002/05/21 12:35:00 EDT
196    Tag: blume-20020521-cmdoc
197    Description:
198    
199    CM documentation update.
200    
201    ----------------------------------------------------------------------
202    Name: Matthias Blume
203    Date: 2002/05/21 10:55:00 EDT
204    Tag: blume-20020521-misc
205    Description:
206    
207    - John tweaked runtime to be silent on heap export (except when
208      GC messages are on).
209    - I added a few more things (cross-compiling versions of CMB) to
210      config/preloads (as suggestions).
211    
212    ----------------------------------------------------------------------
213    Name: Matthias Blume
214    Date: 2002/05/20 22:25:00 EDT
215    Tag: blume-20020520-controls
216    Description:
217    
218    - Added ControlUtil structure to control-lib.cm.
219    - Use it throughout.
220    - Used Controls facility to define MLRISC controls (as opposed to
221      registering MLRISC control ref cells with Controls after the fact)
222    - Fixed messed-up controls priorities.
223    
224    * Removed again all the stuff from config/preloads that one wouldn't
225      be able to preload at the time the initial heap image is built.
226      (Many libraries, e.g., CML, do not exist yet at this time.  The
227       only libraries that can be preloaded via config/preloads are those
228       that come bundled with the bootfiles.)
229    
230    ----------------------------------------------------------------------
231    Name: Matthias Blume
232    Date: 2002/05/20 10:59:00 EDT
233    Tag: blume-20020520-preloads
234    Description:
235    
236    Added a lot of commented-out suggestions for things to be included
237    in config/preloads.
238    
239    ----------------------------------------------------------------------
240    Name: Allen Leung
241    Date: 2002/05/18 14:20:00 EDT
242    Tag: leunga-20020518-mdl
243    Description:
244    
245    o Made the mdl tool stuff compile and run again.
246    o I've disabled all the stuff that depends on RTL specifications; they
247      are all badly broken anyway.
248    
249    ----------------------------------------------------------------------
250    Name: Matthias Blume
251    Date: 2002/05/17 16:49:00 EDT
252    Tag: blume-20020517-controls
253    Description:
254    
255    0. John Reppy made several modifications to the SML/NJ library.
256       In particular, there is a shiny new controls-lib.cm.
257    
258    1. Pushed new controls interface through compiler so that everything
259       compiles again.
260    
261    2. Added FormatComb and FORMAT_COMB to the CML version of the
262       SML/NJ library (so that CML compiles again).
263    
264    3. Modified init scripts because XXX_DEFAULT environment variables
265       are no longer with us.  (Boot-time initialization is now done
266       using the same environment variables that are also used for
267       startup-time initialization of controls.)
268    
269    ----------------------------------------------------------------------
270    Name: Lal George
271    Date: 2002/05/15 09:20:10 EDT
272    Tag: george-20020515-pseudo-op-decls
273    Description:
274    
275            All pseudo-ops emitted before the first segment declaration
276    such as TEXT, DATA, and BSS directives are assumed to be global
277    declarations and are emitted first in the assembly file. This is
278    useful in a number of situations where one has pseudo-ops that are not
279    specific to any segment, and also works around the constraint that one
280    cannot have client pseudo-ops in the TEXT segment.
281    
282            Because no segment is associated with these declarations it is
283    an error to allocate any space or objects before the first segment
284    directive and an exception will be raised. However, we cannot make
285    this check for client pseudo-ops.
286    
287            These top level declarations are a field in the CFG graph_info.
288    In theory you can continue to add to this field after the CFG has been
289    built -- provided you know what you are doing;-)
290    
291    ----------------------------------------------------------------------
292    Name: Matthias Blume
293    Date: 2002/05/13 16:40:00 EDT
294    Tag: blume-20020513-pp-etc
295    Description:
296    
297    A few minor bugfixes:
298    
299      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
300        (Bogus printouts for redefined bindings still occur.  Compiler
301        bug should no longer occur now.  We need to redo the prettyprinter
302        from scratch.)
303    
304      - CM pathname printer now also adds escape sequences for ( and )
305    
306      - commend and docu fixes for ml-nlffi
307    
308    ----------------------------------------------------------------------
309    Name: Matthias Blume
310    Date: 2002/05/10 16:40:00 EDT
311    Tag: blume-20020510-erg-textio
312    Description:
313    
314    Applied the following bugfix provided by Emden Gansner:
315    
316        Output is corrupted when outputSubstr is used rather than output.
317    
318        The problem occurs when a substring
319    
320            ss = (s, dataStart, dataLen)
321    
322        where dataStart > 0, fills a stream buffer with avail bytes left.
323        avail bytes of s, starting at index dataStart, are copied into the
324        buffer, the buffer is flushed, and then the remaining dataLen-avail
325        bytes of ss are copied into the beginning of the buffer. Instead of
326        starting this copy at index dataStart+avail in s, the current code
327        starts the copy at index avail.
328    
329        Fix:
330        In text-io-fn.sml, change line 695 from
331             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
332        to
333             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
334    
335    ----------------------------------------------------------------------
336    Name: Matthias Blume
337    Date: 2002/04/12 13:55:00 EDT
338    Tag: blume-20020412-assyntax
339    Description:
340    
341    1. Grabbed newer assyntax.h from the XFree86 project.
342    2. Fiddled with how to compile X86.prim.asm without warnings.
343    3. (Very) Minor cleanup in CM.
344    
345    ----------------------------------------------------------------------
346    Name: Matthias Blume
347    Date: 2002/04/01 (no joke!) 17:07:00 EST
348    Tag: blume-20020401-x86div
349    Description:
350    
351    Added full support for div/mod/rem/quot on the x86, using the machine
352    instruction's two results (without clumsily recomputing the remainder)
353    directly where appropriate.
354    
355    Some more extensive power-of-two support was added to the x86 instruction
356    selector (avoiding expensive divs, mods, and muls where they can be
357    replaced with cheaper shifts and masks).  However, this sort of thing
358    ought to be done earlier, e.g., within the CPS optimizer so that
359    all architectures benefit from it.
360    
361    The compiler compiles to a fixed point, but changes might be somewhat
362    fragile nevertheless.  Please, report any strange things that you might
363    see wrt. div/mod/quot/rem...
364    
365    ----------------------------------------------------------------------
366    Name: Matthias Blume
367    Date: 2002/03/29 17:22:00
368    Tag: blume-20020329-div
369    Description:
370    
371    Fixed my broken div/mod logic.  Unfortunately, this means that the
372    inline code for div/mod now has one more comparison than before.
373    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
374    The problem was with quotient = 0, because that alone does not tell
375    us which way the rounding went.  One then has to look at whether
376    remainder and divisor have the same sign...  :(
377    
378    Anyway, I replaced the bootfiles with fresh ones...
379    
380    ----------------------------------------------------------------------
381    Name: Matthias Blume
382    Date: 2002/03/29 14:10:00 EST
383    Tag: blume-20020329-inlprims
384    Description:
385    
386    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
387    
388    Primops have changed. This means that the bin/boot-file formats have
389    changed as well.
390    
391    To make sure that there is no confusion, I made a new version.
392    
393    
394    CHANGES:
395    
396    * removed REMT from mltree (remainder should never overflow).
397    
398    * added primops to deal with divisions of all flavors to the frontend
399    
400    * handled these primops all the way through so they map to their respective
401      MLRISC support
402    
403    * used these primops in the implementation of Int, Int32, Word, Word32
404    
405    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
406    
407    * parameterized INLMIN, INLMAX, and INLABS by a numkind
408    
409    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
410      floating point
411    
412    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
413      and Real (but Real.abs maps to a separate floating-point-only primop)
414    
415    
416    TODO items:
417    
418    * Hacked Alpha32 instruction selection, disabling the selection of REMx
419      instructions because the machine instruction encoder cannot handle
420      them.  (Hppa, PPC, and Sparc instruction selection did not handle
421      REM in the first place, and REM is supported by the x86 machine coder.)
422    
423    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
424      selection phase.  (The two can be streamlined because the hardware
425      delivers both quotient and remainder at the same time anyway.)
426    
427    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
428      (Currently the behavior is inconsistent both across architectures and
429      wrt. the draft Basis spec.)
430    
431    * Word8 should eventually be handled natively, too.
432    
433    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
434      as if there currently is no execution path that could trigger it in
435      SML/NJ.  (The assumptions underlying functions arith and promotable do not
436      hold for things like multiplication and division.)
437    
438    ----------------------------------------------------------------------
439    Name: Matthias Blume
440    Date: 2002/03/27 16:27:00 EST
441    Tag: blume-20020327-mlrisc-divisions
442    Description:
443    
444    Added support for all four division operations (ML's div, mod, quot,
445    and rem) to MLRISC.  In the course of doing so, I also rationalized
446    the naming (no more annoying switch-around of DIV and QUOT), by
447    parameterizing the operation by div_rounding_mode (which can be either
448    DIV_TO_ZERO or DIV_TO_NEGINF).
449    
450    The generic MLTreeGen functor takes care of compiling all four
451    operations down to only round-to-zero div.
452    
453    Missing pieces:
454    
455      * Doing something smarter than relying on MLTreeGen on architectures
456        like, e.g., the x86 where hardware division delivers both quotient and
457        remainder at the same time.  With this, the implementation of the
458        round-to-neginf operations could be further streamlined.
459    
460      * Remove inlining support for div/mod/rem from the frontend and replace it
461        with primops that get carried through to the backend.  Do this for all
462        int and word types.
463    
464    ----------------------------------------------------------------------
465    Name: Matthias Blume
466    Date: 2002/03/25 17:25:00 EST
467    Tag: blume-20020325-divmod
468    Description:
469    
470    I improved (hopefully without breaking them) the implementation of Int.div,
471    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
472    advantage of the following observations:
473    
474      Let  q = x quot y      r = x rem y
475           d = x div  y      m = x mod y
476    
477    where "quot" is the round-to-zero version of integer division that
478    hardware usually provides.  Then we have:
479    
480         r = x - q * y        where neither the * nor the - will overflow
481         d = if q >= 0 orelse x = q * y then q else q - 1
482                              where neither the * nor the - will overflow
483         m = if q >= 0 orelse r = 0 then r else r + y
484                              where the + will not overflow
485    
486    This results in substantial simplification of the generated code.
487    The following table shows the number of CFG nodes and edges generated
488    for
489            fun f (x, y) = x OPER y
490            (* with OPER \in div, mod, quot, rem *)
491    
492    
493        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
494        --------------------------------------------------------
495         div |         24 |         39 |         12 |         16
496         mod |         41 |         71 |         12 |         16
497        quot |          8 |         10 |          8 |         10
498         rem |         10 |         14 |          8 |         10
499    
500    
501    ----------------------------------------------------------------------
502    Name: Matthias Blume
503    Date: 2002/03/25 22:06:00 EST
504    Tag: blume-20020325-cprotobug
505    Description:
506    
507    Fixed a bug in cproto (c prototype decoder).
508    
509    ----------------------------------------------------------------------
510    Name: Matthias Blume
511    Date: 2002/03/25 16:00:00 EST
512    Tag: blume-20020325-raw-primops
513    Description:
514    
515    I did some cleanup to Allen's new primop code and
516    replaced yesterday's bootfiles with new ones.
517    (But they are stored in the same place.)
518    
519    ----------------------------------------------------------------------
520    Name: Matthias Blume
521    Date: 2002/03/24 22:40:00 EST
522    Tag: blume-20020324-bootfiles
523    Description:
524    
525    Made the bootfiles that Allen asked for.
526    
527    ----------------------------------------------------------------------
528    Name: Allen Leung
529    Date: 2002/03/23 15:50:00 EST
530    Tag: leunga-20020323-flint-cps-rcc-primops
531    Description:
532    
533      1. Changes to FLINT primops:
534    
535        (* make a call to a C-function;
536         * The primop carries C function prototype information and specifies
537         * which of its (ML-) arguments are floating point. C prototype
538         * information is for use by the backend, ML information is for
539         * use by the CPS converter. *)
540      | RAW_CCALL of { c_proto: CTypes.c_proto,
541                       ml_args: ccall_type list,
542                       ml_res_opt: ccall_type option,
543                       reentrant : bool
544                     } option
545       (* Allocate uninitialized storage on the heap.
546        * The record is meant to hold short-lived C objects, i.e., they
547        * are not ML pointers.  With the tag, the representation is
548        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
549        *)
550      | RAW_RECORD of {tag:bool,sz:int}
551      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
552    
553      2.  These CPS primops are now overloaded:
554    
555           rawload of {kind:numkind}
556           rawstore of {kind:numkind}
557    
558          The one argument form is:
559    
560             rawload {kind} address
561    
562          The two argument form is:
563    
564             rawload {kind} [ml object, byte-offset]
565    
566      3. RAW_CCALL/RCC now takes two extra arguments:
567    
568         a. The first is whether the C call is reentrant, i.e., whether
569            ML state should be saved and restored.
570         b. The second argument is a string argument specifying the name of
571            library and the C function.
572    
573         These things are currently not handled in the code generator, yet.
574    
575      4. In CProto,
576    
577         An encoding type of "bool" means "ml object" and is mapped into
578         C prototype of PTR.  Note that "bool" is different than "string",
579         even though "string" is also mapped into PTR, because "bool"
580         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
581    
582      5. Pickler/unpicker
583    
584         Changed to handle RAW_RECORD and newest RAW_CCALL
585    
586      6. MLRiscGen,
587    
588         1. Changed to handle the new rawload/rawstore/rawrecord operators.
589         2. Code for handling C Calls has been moved to a new module CPSCCalls,
590            in the file CodeGen/cpscompile/cps-c-calls.sml
591    
592      7. Added the conditional move operator
593    
594             condmove of branch
595    
596         to cps.  Generation of this is still buggy so it is currently
597         disabled.
598    
599    ----------------------------------------------------------------------
600    Name: Lal George
601    Date: 2002/03/22 14:18:25 EST
602    Tag: george-20020322-cps-branch-prob
603    Description:
604    
605    Implemented the Ball-Larus branch prediction-heuristics, and
606    incorporated graphical viewers for control flow graphs.
607    
608    Ball-Larus Heuristics:
609    ---------------------
610    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
611    
612    By design it uses the Dempster-Shafer theory for combining
613    probabilities.  For example, in the function:
614    
615        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
616    
617    the ball-larus heuristics predicts that the n=0 is unlikely
618    (OH-heuristic), and the 'then' branch is unlikely because of the
619    RH-heuristic -- giving the 'then' branch an even lower combined
620    probability using the Dempster-Shafer theory.
621    
622    Finally, John Reppy's loop analysis in MLRISC, further lowers the
623    probability of the 'then' branch because of the loop in the else
624    branch.
625    
626    
627    Graphical Viewing:
628    ------------------
629    I merely plugged in Allen's graphical viewers into the compiler. The
630    additional code is not much. At the top level, saying:
631    
632            Control.MLRISC.getFlag "cfg-graphical-view" := true;
633    
634    will display the graphical view of the control flow graph just before
635    back-patching.  daVinci must be in your path for this to work. If
636    daVinci is not available, then the default viewer can be changed
637    using:
638    
639            Control.MLRISC.getString "viewer"
640    
641    which can be set to "dot" or "vcg" for the corresponding viewers. Of
642    course, these viewers must be in your path.
643    
644    The above will display the compilation unit at the level of clusters,
645    many of which are small, boring, and un-interesting. Also setting:
646    
647            Control.MLRISC.getInt "cfg-graphical-view_size"
648    
649    will display clusters that are larger than the value set by the above.
650    
651    
652    ----------------------------------------------------------------------
653    Name: Matthias Blume
654    Date: 2002/03/21 22:20:00 EST
655    Tag: blume-20020321-kmp-bugfix
656    Description:
657    
658    Changed the interface to the KMP routine in PreString and fixed
659    a minor bug in one place where it was used.
660    
661    ----------------------------------------------------------------------
662    Name: Allen Leung
663    Date: 2002/03/21 20:30:00 EST
664    Tag: leunga-20020321-cfg
665    Description:
666    
667      Fixed a potential problem in cfg edge splitting.
668    
669    ----------------------------------------------------------------------
670    Name: Allen Leung
671    Date: 2002/03/21 17:15:00 EST
672    Tag: leunga-20020321-x86-fp-cfg
673    Description:
674    
675      1. Recoded the buggy parts of x86-fp.
676    
677         a. All the block reordering code has been removed.
678            We now depend on the block placement phases to do this work.
679    
680         b. Critical edge splitting code has been simplified and moved into the
681            CFG modules, as where they belong.
682    
683         Both of these were quite buggy and complex.  The code is now much, much
684         simpler.
685    
686      2. X86 backend.
687    
688         a. Added instructions for 64-bit support.  Instruction selection for
689            64-bit has not been committed, however, since that
690            requires changes to MLTREE which haven't been approved by
691            Lal and John.
692    
693         b. Added support for FUCOMI and FUCOMIP when generating code for
694            PentiumPro and above.  We only generate these instructions in
695            the fast-fp mode.
696    
697         c. Added cases for JP and JNP in X86FreqProps.
698    
699      3. CFG
700    
701         CFG now has a bunch of methods for edge splitting and merging.
702    
703      4. Machine description.
704    
705         John's simplification of MLTREE_BASIS.fcond broke a few machine
706         description things:
707    
708         rtl-build.{sig,sml} and hppa.mdl fixed.
709    
710         NOTE: the machine description stuff in the repository is still broken.
711               Again, I can't put my fixes in because that involves
712               changes to MLTREE.
713    
714    ----------------------------------------------------------------------
715    Name: Matthias Blume
716    Date: 2002/03/20 15:55:00 EST
717    Tag: blume-20020320-kmp
718    Description:
719    
720    Implemented Knuth-Morris-Pratt string matching in PreString and used
721    it for String.isSubstring, Substring.isSubstring, and
722    Substring.position.
723    
724    (Might need some stress-testing.  Simple examples worked fine.)
725    
726    ----------------------------------------------------------------------
727    Name: Matthias Blume
728    Date: 2002/03/19 16:37:00 EST
729    Tag: blume-20020319-witnesses
730    Description:
731    
732    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
733    
734    This implements a generic mechanism for changing constness qualifiers
735    anywhere within big C types without resorting to outright "casts".
736    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
737    modify the constness at the outermost level.)
738    The implementation of "convert" is based on the idea of "witness"
739    values -- values that are not used by the operation but whose types
740    "testify" to their applicability.  On the implementation side, "convert"
741    is simply a projection (returning its second curried argument).  With
742    cross-module inlining, it should not result in any machine code being
743    generated.
744    
745    ----------------------------------------------------------------------
746    Name: Matthias Blume
747    Date: 2002/03/15 16:40:00 EST
748    Tag: blume-20020315-basis
749    Description:
750    
751    Provided (preliminary?) implementations for
752    
753      {String,Substring}.{concatWith,isSuffix,isSubstring}
754    
755    and
756    
757      Substring.full
758    
759    Those are in the Basis spec but they were missing in SML/NJ.
760    
761    ----------------------------------------------------------------------
762    Name: Matthias Blume
763    Date: 2002/03/14 21:30:00 EST
764    Tag: blume-20020314-controls
765    Description:
766    
767    Controls:
768    ---------
769    
770    1. Factored out the recently-added Controls : CONTROLS stuff and put
771       it into its own library $/controls-lib.cm.  The source tree for
772       this is under src/smlnj-lib/Controls.
773    
774    2. Changed the names of types and functions in this interface, so they
775       make a bit more "sense":
776    
777          module -> registry
778          'a registry -> 'a group
779    
780    3. The interface now deals in ref cells only.  The getter/setter interface
781       is (mostly) gone.
782    
783    4. Added a function that lets one register an already-existing ref cell.
784    
785    5. Made the corresponding modifications to the rest of the code so that
786       everything compiles again.
787    
788    6. Changed the implementation of Controls.MLRISC back to something closer
789       to the original.  In particular, this module (and therefore MLRISC)
790       does not depend on Controls.  There now is some link-time code in
791       int-sys.sml that registers the MLRISC controls with the Controls
792       module.
793    
794    CM:
795    ---
796    
797      * One can now specify the lambda-split aggressiveness in init.cmi.
798    
799    ----------------------------------------------------------------------
800    Name: Allen Leung
801    Date: 2002/03/13 17:30:00 EST
802    Tag: leunga-20020313-x86-fp-unary
803    Description:
804    
805    Bug fix for:
806    
807    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
808    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
809    > - fun f(x,(y,z)) = Real.~ y;
810    > [autoloading]
811    > [autoloading done]
812    >       fchsl   (%eax), 184(%esp)
813    > Error: MLRisc bug: X86MCEmitter.emitInstr
814    >
815    > uncaught exception Error
816    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
817    
818    The problem was that the code generator did not generate any fp registers
819    in this case, and the ra didn't know that it needed to run the X86FP phase to
820    translate the pseudo fp instruction.   This only happened with unary fp
821    operators in certain situations.
822    
823    ----------------------------------------------------------------------
824    Name: Matthias Blume
825    Date: 2002/03/13 14:00:00 EST
826    Tag: blume-20020313-overload-etc
827    Description:
828    
829    1. Added _overload as a synonym for overload for backward compatibility.
830       (Control.overloadKW must be true for either version to be accepted.)
831    
832    2. Fixed bug in install script that caused more things to be installed
833       than what was requested in config/targets.
834    
835    3. Made CM aware of the (_)overload construct so that autoloading
836       works.
837    
838    ----------------------------------------------------------------------
839    Name: Matthias Blume
840    Date: 2002/03/12 22:03:00 EST
841    Tag: blume-20020312-url
842    Description:
843    
844    Forgot to update BOOT and srcarchiveurl.
845    
846    ----------------------------------------------------------------------
847    Name: Matthias Blume
848    Date: 2002/03/12 17:30:00 EST
849    Tag: blume-20020312-version110392
850    Description:
851    
852    Yet another version number bump (because of small changes to the
853    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
854    
855    Changes:
856    
857      The new pid generation scheme described a few weeks ago was overly
858      complicated.  I implemented a new mechanism that is simpler and
859      provides a bit more "stability":  Once CM has seen a compilation
860      unit, it keeps its identity constant (as long as you do not delete
861      those crucial CM/GUID/* files).  This means that when you change
862      an interface, compile, then go back to the old interface, and
863      compile again, you arrive at the original pid.
864    
865      There now also is a mechanism that instructs CM to use the plain
866      environment hash as a module's pid (effectively making its GUID
867      the empty string).  For this, "noguid" must be specified as an
868      option to the .sml file in question within its .cm file.
869      This is most useful for code that is being generated by tools such
870      as ml-nlffigen (because during development programmers tend to
871      erase the tool's entire output directory tree including CM's cached
872      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
873      revert to the old, broken behavior of SML/NJ, but in specific cases
874      where there is no danger of interface confusion, its use is ok
875      (I think).
876    
877      ml-nlffigen by default generates "noguid" annotations.  They can be
878      turned off by specifying -guid in its command line.
879    
880    ----------------------------------------------------------------------
881    Name: Lal George
882    Date: 2002/03/12 12 14:42:36 EST
883    Tag: george-20020312-frequency-computation
884    Description:
885    
886    Integrated jump chaining and static block frequency into the
887    compiler. More details and numbers later.
888    
889    ----------------------------------------------------------------------
890    Name: Lal George
891    Date: 2002/03/11 11 22:38:53 EST
892    Tag: george-20020311-jump-chain-elim
893    Description:
894    
895    Tested the jump chain elimination on all architectures (except the
896    hppa).  This is on by default right now and is profitable for the
897    alpha and x86, however, it may not be profitable for the sparc and ppc
898    when compiling the compiler.
899    
900    The gc test will typically jump to a label at the end of the cluster,
901    where there is another jump to an external cluster containing the actual
902    code to invoke gc. This is to allow factoring of common gc invocation
903    sequences. That is to say, we generate:
904    
905            f:
906               testgc
907               ja   L1      % jump if above to L1
908    
909            L1:
910               jmp L2
911    
912    
913    After jump chain elimination the 'ja L1' instructions is converted to
914    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
915    up being implemented in their long form (if L2 is far away) using:
916    
917            jbe     L3      % jump if below or equal to L3
918            jmp     L2
919         L3:
920            ...
921    
922    
923    For large compilation units L2  may be far away.
924    
925    
926    ----------------------------------------------------------------------
927    Name: Matthias Blume
928    Date: 2002/03/11 13:30:00 EST
929    Tag: blume-20020311-mltreeeval
930    Description:
931    
932    A functor parameter was missing.
933    
934    ----------------------------------------------------------------------
935    Name: Allen Leung
936    Date: 2002/03/11 10:30:00 EST
937    Tag: leunga-20020311-runtime-string0
938    Description:
939    
940       The representation of the empty string now points to a
941    legal null terminated C string instead of unit.  It is now possible
942    to convert an ML string into C string with InlineT.CharVector.getData.
943    This compiles into one single machine instruction.
944    
945    ----------------------------------------------------------------------
946    Name: Allen Leung
947    Date: 2002/03/10 23:55:00 EST
948    Tag: leunga-20020310-x86-call
949    Description:
950    
951       Added machine generation for CALL instruction (relative displacement mode)
952    
953    ----------------------------------------------------------------------
954    Name: Matthias Blume
955    Date: 2002/03/08 16:05:00
956    Tag: blume-20020308-entrypoints
957    Description:
958    
959    Version number bumped to 110.39.1.  NEW BOOTFILES!
960    
961    Entrypoints: non-zero offset into a code object where execution should begin.
962    
963    - Added the notion of an entrypoint to CodeObj.
964    - Added reading/writing of entrypoint info to Binfile.
965    - Made runtime system bootloader aware of entrypoints.
966    - Use the address of the label of the first function given to mlriscGen
967      as the entrypoint.  This address is currently always 0, but it will
968      not be 0 once we turn on block placement.
969    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
970      with entry points) from mlriscGen.
971    
972    ----------------------------------------------------------------------
973    Name: Allen Leung
974    Date: 2002/03/07 20:45:00 EST
975    Tag: leunga-20020307-x86-cmov
976    Description:
977    
978       Bug fixes for CMOVcc on x86.
979    
980       1. Added machine code generation for CMOVcc
981       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
982       3. CMOVcc cannot have an immediate operand as argument.
983    
984    ----------------------------------------------------------------------
985    Name: Matthias Blume
986    Date: 2002/03/07 16:15:00 EST
987    Tag: blume-20020307-controls
988    Description:
989    
990    This is a very large but mostly boring patch which makes (almost)
991    every tuneable compiler knob (i.e., pretty much everything under
992    Control.* plus a few other things) configurable via both the command
993    line and environment variables in the style CM did its configuration
994    until now.
995    
996    Try starting sml with '-h' (or, if you are brave, '-H')
997    
998    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
999    implements the underlying generic mechanism.
1000    
1001    The interface to some of the existing such facilities has changed somewhat.
1002    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1003    (The getFoo interface is still there for backward-compatibility, but its
1004    use is deprecated.)
1005    
1006    The ml-build script passes -Cxxx=yyy command-line arguments through so
1007    that one can now twiddle the compiler settings when using this "batch"
1008    compiler.
1009    
1010    TODO items:
1011    
1012    We should go through and throw out all controls that are no longer
1013    connected to anything.  Moreover, we should go through and provide
1014    meaningful (and correct!) documentation strings for those controls
1015    that still are connected.
1016    
1017    Currently, multiple calls to Controls.new are accepted (only the first
1018    has any effect).  Eventually we should make sure that every control
1019    is being made (via Controls.new) exactly once.  Future access can then
1020    be done using Controls.acc.
1021    
1022    Finally, it would probably be a good idea to use the getter-setter
1023    interface to controls rather than ref cells.  For the time being, both
1024    styles are provided by the Controls module, but getter-setter pairs are
1025    better if thread-safety is of any concern because they can be wrapped.
1026    
1027    *****************************************
1028    
1029    One bug fix: The function blockPlacement in three of the MLRISC
1030    backpatch files used to be hard-wired to one of two possibilities at
1031    link time (according to the value of the placementFlag).  But (I
1032    think) it should rather sense the flag every time.
1033    
1034    *****************************************
1035    
1036    Other assorted changes (by other people who did not supply a HISTORY entry):
1037    
1038    1. the cross-module inliner now works much better (Monnier)
1039    2. representation of weights, frequencies, and probabilities in MLRISC
1040       changed in preparation of using those for weighted block placement
1041       (Reppy, George)
1042    
1043    ----------------------------------------------------------------------
1044    Name: Lal George
1045    Date: 2002/03/07 14:44:24 EST 2002
1046    Tag: george-20020307-weighted-block-placement
1047    
1048    Tested the weighted block placement optimization on all architectures
1049    (except the hppa) using AMPL to generate the block and edge frequencies.
1050    Changes were required in the machine properties to correctly
1051    categorize trap instructions. There is an MLRISC flag
1052    "weighted-block-placement" that can be used to enable weighted block
1053    placement, but this will be ineffective without block/edge
1054    frequencies (coming soon).
1055    
1056    
1057    ----------------------------------------------------------------------
1058    Name: Lal George
1059    Date: 2002/03/05 17:24:48 EST
1060    Tag: george-20020305-linkage-cluster
1061    
1062    In order to support the block placement optimization, a new cluster
1063    is generated as the very first cluster (called the linkage cluster).
1064    It contains a single jump to the 'real' entry point for the compilation
1065    unit. Block placement has no effect on the linkage cluster itself, but
1066    all the other clusters  have full freedom in the manner in which they
1067    reorder blocks or functions.
1068    
1069    On the x86 the typical linkage code that is generated is:
1070       ----------------------
1071            .align 2
1072       L0:
1073            addl    $L1-L0, 72(%esp)
1074            jmp     L1
1075    
1076    
1077            .align  2
1078       L1:
1079       ----------------------
1080    
1081    72(%esp) is the memory location for the stdlink register. This
1082    must contain the address of the CPS function being called. In the
1083    above example, it contains the address of  L0; before
1084    calling L1 (the real entry point for the compilation unit), it
1085    must contain the address for L1, and hence
1086    
1087            addl $L1-L0, 72(%esp)
1088    
1089    I have tested this on all architectures except the hppa.The increase
1090    in code size is of course negligible
1091    
1092    ----------------------------------------------------------------------
1093    Name: Allen Leung
1094    Date: 2002/03/03 13:20:00 EST
1095    Tag: leunga-20020303-mlrisc-tools
1096    
1097      Added #[ ... ] expressions to mlrisc tools
1098    
1099    ----------------------------------------------------------------------
1100    Name: Matthias Blume
1101    Date: 2002/02/27 12:29:00 EST
1102    Tag: blume-20020227-cdebug
1103    Description:
1104    
1105    - made types in structure C and C_Debug to be equal
1106    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1107    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1108    
1109    ----------------------------------------------------------------------
1110    Name: Matthias Blume
1111    Date: 2002/02/26 12:00:00 EST
1112    Tag: blume-20020226-ffi
1113    Description:
1114    
1115    1. Fixed a minor bug in CM's "noweb" tool:
1116       If numbering is turned off, then truly don't number (i.e., do not
1117       supply the -L option to noweb).  The previous behavior was to supply
1118       -L'' -- which caused noweb to use the "default" line numbering scheme.
1119       Thanks to Chris Richards for pointing this out (and supplying the fix).
1120    
1121    2. Once again, I reworked some aspects of the FFI:
1122    
1123       A. The incomplete/complete type business:
1124    
1125       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1126         gone!
1127       - ML types representing an incomplete type are now *equal* to
1128         ML types representing their corresponding complete types (just like
1129         in C).  This is still safe because ml-nlffigen will not generate
1130         RTTI for incomplete types, nor will it generate functions that
1131         require access to such RTTI.   But when ML code generated from both
1132         incomplete and complete versions of the C type meet, the ML types
1133         are trivially interoperable.
1134    
1135         NOTE:  These changes restore the full generality of the translation
1136         (which was previously lost when I eliminated functorization)!
1137    
1138       B. Enum types:
1139    
1140       - Structure C now has a type constructor "enum" that is similar to
1141         how the "su" constructor works.  However, "enum" is not a phantom
1142         type because each "T enum" has values (and is isomorphic to
1143         MLRep.Signed.int).
1144       - There are generic access operations for enum objects (using
1145         MLRep.Signed.int).
1146       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1147         * The structure contains the definition of type "mlrep" (the ML-side
1148         representation type of the enum).  Normally, mlrep is the same
1149         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1150         then mlrep will be defined as a datatype -- thus facilitating
1151         pattern matching on mlrep values.
1152         ("-ec" will be suppressed if there are duplicate values in an
1153          enumeration.)
1154         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1155         will be generated for each C enum constant xxx.
1156         * Conversion functions m2i and i2m convert between mlrep and
1157         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1158         * Coversion functions c and ml convert between mlrep and "tag enum".
1159         * Access functions (get/set) fetch and store mlrep values.
1160       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1161         enumerations are merged into one single enumeration represented by
1162         structure E_'.
1163    
1164    ----------------------------------------------------------------------
1165    Name: Allen Leung
1166    Date: 2002/02/25 04:45:00 EST
1167    Tag: leunga-20020225-cps-spill
1168    
1169    This is a new implementation of the CPS spill phase.
1170    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1171    In case of problems, replace it with the old file spill.sml
1172    
1173    The current compiler runs into some serious performance problems when
1174    constructing a large record.  This can happen when we try to compile a
1175    structure with many items.  Even a very simple structure like the following
1176    makes the compiler slow down.
1177    
1178        structure Foo = struct
1179           val x_1 = 0w1 : Word32.int
1180           val x_2 = 0w2 : Word32.int
1181           val x_3 = 0w3 : Word32.int
1182           ...
1183           val x_N = 0wN : Word32.int
1184        end
1185    
1186    The following table shows the compile time, from N=1000 to N=4000,
1187    with the old compiler:
1188    
1189    N
1190    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1191           MLRISC ra                               0.06u  0.00s  0.05g
1192              (spills = 0 reloads = 0)
1193           TOTAL                                   0.63u  0.07s  0.21g
1194    
1195    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1196           MLRISC ra                               5.68u  0.59s  3.93g
1197              (spills = 0 reloads = 0)
1198           TOTAL                                   14.71u  0.99s  4.81g
1199    
1200    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1201           MLRISC ra                               5.54u  0.65s  3.91g
1202              (spills = 543 reloads = 1082)
1203           TOTAL                                   65.40u  3.13s  6.00g
1204    
1205    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1206           MLRISC ra                               0.80u  0.10s  0.55g
1207              (spills = 42 reloads = 84)
1208           TOTAL                                   129.42u  5.10s  4.13g
1209    
1210    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1211           MLRISC ra                               2.69u  0.27s  1.38g
1212              (spills = 62 reloads = 124)
1213           TOTAL                                   682.48u  19.61s  13.99g
1214    
1215    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1216           MLRISC ra                               4.96u  0.27s  2.72g
1217              (spills = 85 reloads = 170)
1218           TOTAL                                   2375.26u  57.21s  48.00g
1219    
1220    As you can see the old cps spill module suffers from some serious
1221    performance problem.  But since I cannot decipher the old code fully,
1222    instead of patching the problems up, I'm reimplementing it
1223    with a different algorithm.  The new code is more modular,
1224    smaller when compiled, and substantially faster
1225    (O(n log n) time and O(n) space).  Timing of the new spill module:
1226    
1227    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1228          MLRISC ra                               0.25u  0.02s  0.15g
1229             (spills=1 reloads=3)
1230          TOTAL                                   7.74u  0.34s  1.62g
1231    
1232    Implementation details:
1233    
1234    As far as I can tell, the purpose of the CPS spill module is to make sure the
1235    number of live variables at any program point (the bandwidth)
1236    does not exceed a certain limit, which is determined by the
1237    size of the spill area.
1238    
1239    When the bandwidth is too large, we decrease the register pressure by
1240    packing live variables into spill records.  How we achieve this is
1241    completely different than what we did in the old code.
1242    
1243    First, there is something about the MLRiscGen code generator
1244    that we should be aware of:
1245    
1246    o MLRiscGen performs code motion!
1247    
1248       In particular, it will move floating point computations and
1249       address computations involving only the heap pointer to
1250       their use sites (if there is only a single use).
1251       What this means is that if we have a CPS record construction
1252       statement
1253    
1254           RECORD(k,vl,w,e)
1255    
1256       we should never count the new record address w as live if w
1257       has only one use (which is often the case).
1258    
1259       We should do something similar to floating point, but the transformation
1260       there is much more complex, so I won't deal with that.
1261    
1262    Secondly, there are now two new cps primops at our disposal:
1263    
1264     1. rawrecord of record_kind option
1265        This pure operator allocates some uninitialized storage from the heap.
1266        There are two forms:
1267    
1268         rawrecord NONE [INT n]  allocates a tagless record of length n
1269         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1270                                     and initializes the tag.
1271    
1272     2. rawupdate of cty
1273          rawupdate cty (v,i,x)
1274          Assigns to x to the ith component of record v.
1275          The storelist is not updated.
1276    
1277    We use these new primops for both spilling and increment record construction.
1278    
1279     1. Spilling.
1280    
1281        This is implemented with a linear scan algorithm (but generalized
1282        to trees).  The algorithm will create a single spill record at the
1283        beginning of the cps function and use rawupdate to spill to it,
1284        and SELECT or SELp to reload from it.  So both spills and reloads
1285        are fine-grain operations.  In contrast, in the old algorithm
1286        "spills" have to be bundled together in records.
1287    
1288        Ideally, we should sink the spill record construction to where
1289        it is needed.  We can even split the spill record into multiple ones
1290        at the places where they are needed.  But CPS is not a good
1291        representation for global code motion, so I'll keep it simple and
1292        am not attempting this.
1293    
1294     2. Incremental record construction (aka record splitting).
1295    
1296        Long records with many component values which are simulatenously live
1297        (recall that single use record addresses are not considered to
1298         be live) are constructed with rawrecord and rawupdate.
1299        We allocate space on the heap with rawrecord first, then gradually
1300        fill it in with rawupdate.  This is the technique suggested to me
1301        by Matthias.
1302    
1303        Some restrictions on when this is applicable:
1304        1. It is not a VECTOR record.  The code generator currently does not handle
1305           this case. VECTOR record uses double indirection like arrays.
1306        2. All the record component values are defined in the same "basic block"
1307           as the record constructor.  This is to prevent speculative
1308           record construction.
1309    
1310    ----------------------------------------------------------------------
1311    Name: Allen Leung
1312    Date: 2002/02/22 01:02:00 EST
1313    Tag: leunga-20020222-mlrisc-tools
1314    
1315    Minor bug fixes in the parser and rewriter
1316    
1317    ----------------------------------------------------------------------
1318    Name: Allen Leung
1319    Date: 2002/02/21 20:20:00 EST
1320    Tag: leunga-20020221-peephole
1321    
1322    Regenerated the peephole files.  Some contained typos in the specification
1323    and some didn't compile because of pretty printing bugs in the old version
1324    of 'nowhere'.
1325    
1326    ----------------------------------------------------------------------
1327    Name: Allen Leung
1328    Date: 2002/02/19 20:20:00 EST
1329    Tag: leunga-20020219-mlrisc-tools
1330    Description:
1331    
1332       Minor bug fixes to the mlrisc-tools library:
1333    
1334       1.  Fixed up parsing colon suffixed keywords
1335       2.  Added the ability to shut the error messages up
1336       3.  Reimplemented the pretty printer and fixed up/improved
1337           the pretty printing of handle and -> types.
1338       4.  Fixed up generation of literal symbols in the nowhere tool.
1339       5.  Added some SML keywords to to sml.sty
1340    
1341    ----------------------------------------------------------------------
1342    Name: Matthias Blume
1343    Date: 2002/02/19 16:20:00 EST
1344    Tag: blume-20020219-cmffi
1345    Description:
1346    
1347    A wild mix of changes, some minor, some major:
1348    
1349    * All C FFI-related libraries are now anchored under $c:
1350        $/c.cm      --> $c/c.cm
1351        $/c-int.cm  --> $c/internals/c-int.cm
1352        $/memory.cm --> $c/memory/memory.cm
1353    
1354    * "make" tool (in CM) now treats its argument pathname slightly
1355      differently:
1356        1. If the native expansion is an absolute name, then before invoking
1357           the "make" command on it, CM will apply OS.Path.mkRelative
1358           (with relativeTo = OS.FileSys.getDir()) to it.
1359        2. The argument will be passed through to subsequent phases of CM
1360           processing without "going native".  In particular, if the argument
1361           was an anchored path, then "make" will not lose track of that anchor.
1362    
1363    * Compiler backends now "know" their respective C calling conventions
1364      instead of having to be told about it by ml-nlffigen.  This relieves
1365      ml-nlffigen from one of its burdens.
1366    
1367    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1368    
1369    * Export C_DEBUG and C_Debug from $c/c.cm.
1370    
1371    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1372      subtyping relationship between incomplete pointers and their complete
1373      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1374      with the convention of instantiating 'o with ('t, 'c) obj whenever
1375      the pointer target type is complete.  In the incomplete case, 'o
1376      will be instantiated with some "'c iobj" -- a type obtained by
1377      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1378    
1379      Operations that work on both incomplete and complete pointer types are
1380      typed as taking an 'o ptr while operations that require the target to
1381      be known are typed as taking some ('t, 'c) obj ptr.
1382    
1383      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1384      where void is an eqtype without any values.  This makes it possible
1385      to work on voidptr values using functions meant to operate on light
1386      incomplete pointers.
1387    
1388    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1389      been vastly simplified.
1390    
1391    ----------------------------------------------------------------------
1392    Name: Matthias Blume
1393    Date: 2002/02/19 10:48:00 EST
1394    Tag: blume-20020219-pqfix
1395    Description:
1396    
1397    Applied Chris Okasaki's bug fix for priority queues.
1398    
1399    ----------------------------------------------------------------------
1400    Name: Matthias Blume
1401    Date: 2002/02/15 17:05:00
1402    Tag: Release_110_39
1403    Description:
1404    
1405    Last-minute retagging is becoming a tradition... :-(
1406    
1407    This is the working release 110.39.
1408    
1409    ----------------------------------------------------------------------
1410    Name: Matthias Blume
1411    Date: 2002/02/15 16:00:00 EST
1412    Tag: Release_110_39-orig
1413    Description:
1414    
1415    Working release 110.39.  New bootfiles.
1416    
1417    (Update: There was a small bug in the installer so it wouldn't work
1418    with all shells.  So I retagged. -Matthias)
1419    
1420    ----------------------------------------------------------------------
1421    Name: Matthias Blume
1422    Date: 2002/02/15 14:17:00 EST
1423    Tag: blume-20020215-showbindings
1424    Description:
1425    
1426    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1427    the latter can be useful for exploring what bindings are available at
1428    the interactive prompt.  (The first function returns only the list
1429    of symbols that are really bound, the second prints those but also the
1430    ones that CM's autoloading mechanism knows about.)
1431    
1432    ----------------------------------------------------------------------
1433    Name: Matthias Blume
1434    Date: 2002/02/15 12:08:00 EST
1435    Tag: blume-20020215-iptrs
1436    Description:
1437    
1438    Two improvements to ml-nlffigen:
1439    
1440      1. Write files only if they do not exist or if their current contents
1441         do not coincide with what's being written.  (That is, avoid messing
1442         with the time stamps unless absolutely necessary.)
1443    
1444      2. Implement a "repository" mechanism for generated files related
1445         to "incomplete pointer types".   See the README file for details.
1446    
1447    ----------------------------------------------------------------------
1448    Name: Matthias Blume
1449    Date: 2002/02/14 11:50:00 EST
1450    Tag: blume-20020214-quote
1451    Description:
1452    
1453    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1454    because of the new and improved tag generation scheme.  (Thanks to Allen
1455    Leung for pointing it out.)
1456    
1457    ----------------------------------------------------------------------
1458    Name: Lal George
1459    Date: 2002/02/14 09:55:27 EST 2002
1460    Tag: george-20020214-isabelle-bug
1461    Description:
1462    
1463    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1464    of Isabelle on the x86.
1465    
1466    From Allen:
1467    -----------
1468     I've found the problem:
1469    
1470         in ra-core.sml, I use the counter "blocked" to keep track of the
1471         true number of elements in the freeze queue.  When the counter goes
1472         to zero, I skip examining the queue.  But I've messed up the
1473         bookkeeping in combine():
1474    
1475             else ();
1476             case !ucol of
1477               PSEUDO => (if !cntv > 0 then
1478                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1479                                        ^^^^^^^^^^^^^^^^^^^^^^^
1480                      moveu := mergeMoveList(!movev, !moveu)
1481                     )
1482                  else ();
1483    
1484         combine() is called to coalesce two nodes u and v.
1485         I think I was thinking that if the move counts of u and v are both
1486         greater than zero then after they are coalesced then one node is
1487         removed from the freeze queue.  Apparently I was thinking that
1488         both u and v are of low degree, but that's clearly not necessarily true.
1489    
1490    
1491    02/12/2002:
1492        Here's the patch.  HOL now compiles.
1493    
1494        I don't know how this impact on performance (compile
1495        time or runtime).  This bug caused the RA (especially on the x86)
1496        to go thru the potential spill phase when there are still nodes on the
1497        freeze queue.
1498    
1499    
1500    
1501    
1502    ----------------------------------------------------------------------
1503    Name: Matthias Blume
1504    Date: 2002/02/13 22:40:00 EST
1505    Tag: blume-20020213-fptr-rtti
1506    Description:
1507    
1508    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1509    updates.
1510    
1511    ----------------------------------------------------------------------
1512    Name: Matthias Blume
1513    Date: 2002/02/13 16:41:00 EST
1514    Tag: blume-20020213-cmlpq
1515    Description:
1516    
1517    Added new priority queue export symbols (which have just been added to
1518    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1519    not compile and the installer would choke.)
1520    
1521    ----------------------------------------------------------------------
1522    Name: Matthias Blume
1523    Date: 2002/02/13 16:15:00 EST
1524    Tag: blume-20020213-various
1525    Description:
1526    
1527    1. More tweaks to ml-nlffigen:
1528    
1529       - better internal datastructures (resulting in slight speedup)
1530       - "-match" option requires exact match
1531       - "localized" gensym counters (untagged structs/unions nested within
1532         other structs/unions or within typedefs get a fresh counter; their
1533         tag will be prefixed by a concatenation of their parents' tags)
1534       - bug fixes (related to calculation of transitive closure of types
1535         to be included in the output)
1536    
1537    2. Minor Basis updates:
1538    
1539       - added implementations for List.collate and Option.app
1540    
1541    ----------------------------------------------------------------------
1542    Name: Matthias Blume
1543    Date: 2002/02/11 15:55:00 EST
1544    Tag: blume-20020211-gensym
1545    Description:
1546    
1547    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1548    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1549    names (ML structure names that correspond to unnamed C structs, unions,
1550    and enums), so that separate runs of ml-nlffigen do not clash.
1551    
1552    ----------------------------------------------------------------------
1553    Name: Matthias Blume
1554    Date: 2002/02/11 12:05:00 EST
1555    Tag: blume-20020211-gensml
1556    Description:
1557    
1558    A quick fix for a problem with GenSML (in the pgraph-util library):
1559    Make generation of toplevel "local" optional.  (Strictly speaking,
1560    signature definitions within "local" are not legal SML.)
1561    
1562    Other than that: updates to INSTALL and cm/TODO.
1563    
1564    ----------------------------------------------------------------------
1565    Name: Matthias Blume
1566    Date: 2002/02/08 15:00:00 EST
1567    Tag: blume-20020208-uniquepid
1568    Description:
1569    
1570    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1571    
1572    1. The installer (config/install.sh) has gotten smarter:
1573    
1574         - Configuration options are a bit easier to specify now
1575           (in config/targets).
1576         - Bug in recognizing .tar.bz2 files fixed.
1577         - Installer automatically resolves dependencies between
1578           configuration options (e.g., if you ask for eXene, you will
1579           also get cml -- regardless whether you asked for it or not).
1580         - Installer can run in "quieter mode" by setting the environment
1581           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1582           "completely silent", though.
1583         - Build HashCons library as part of smlnj-lib.
1584    
1585    2. A new scheme for assigning persistent identifiers to compilation
1586       units (and, by extension, to types etc.) has been put into place.
1587       This fixes a long-standing bug where types and even dynamic values
1588       can get internally confused, thereby compromising type safety
1589       (abstraction) and dynamic correctness.  See
1590    
1591         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1592    
1593       for an example of how things could go wrong until now.
1594    
1595       The downside of the new scheme is that pids are not quite as
1596       persistent as they used to be: CM will generate a fresh pid
1597       for every compilation unit that it thinks it sees for the first
1598       time.  That means that if you compile starting from a clean, fresh
1599       source tree at two different times, you end up with different
1600       binaries.
1601    
1602       Cutoff recompilation, however, has not been compromised because
1603       CM keeps pid information in special caches between runs.
1604    
1605    ----------------------------------------------------------------------
1606    Name: Lal George
1607    Date: 2002/02/07 15:34:13 EST 2002
1608    Tag: <none>
1609    Description:
1610    
1611    Compilers that generate assembly code may produce  global labels
1612    whose value is resolved at link time. The various peephole optimization
1613    modules did not take this in account.
1614    
1615    TODO. The Labels.addrOf function should really return an option
1616    type so that clients are forced to deal with this issue, rather
1617    than an exception being raised.
1618    
1619    ----------------------------------------------------------------------
1620    Name: Lal George
1621    Date: 2002/02/06 13:55:02 EST
1622    Tag: george-20020206-ra-breakup
1623    Description:
1624    
1625    1. A bug fix from Allen.
1626    
1627        A typo causes extra fstp %st(0)'s to be generated at compensation
1628        edges, which might cause stack underflow traps at runtime.  This
1629        occurs in fft where there are extraneous fstps right before the 'into'
1630        trap instruction (in this case they are harmless since none of the
1631        integers overflow.)
1632    
1633    2. Pulled out various utility modules that were embedded in the modules
1634       of the register allocator. I need these modules for other purposes, but
1635       they are not complete enough to put into a library (just yet).
1636    ----------------------------------------------------------------------
1637    Name: Matthias Blume
1638    Date: 2002/01/31 16:05:00 EST
1639    Tag: blume-20020131-sparc-ccalls
1640    Description:
1641    
1642    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1643       of extra stack space by mistake.  Fixed.
1644    
1645    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1646    
1647    ----------------------------------------------------------------------
1648    Name: Allen Leung
1649    Date: 2002/01/30
1650    Tag: leunga-20020130-nowhere-bug-fix
1651    Description:
1652    
1653       MLRISC bug fixes:
1654       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1655       2. MachineInt.fromString was negating its value.
1656    
1657    ----------------------------------------------------------------------
1658    Name: Matthias Blume
1659    Date: 2002/01/29
1660    Tag: blume-20020129-INSTALL
1661    Description:
1662    
1663    - Added somewhat detailed installation instructions (file INSTALL).
1664    - Fixed curl-detection bug in config/install.sh.
1665    - It is now possible to select the URL getter using the URLGETTER
1666      environment variable:
1667    
1668          not set / "unknown"      --> automatic detection (script tries wget,
1669                                       curl, and lynx)
1670          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1671                                       how to properly invoke them)
1672          other                    --> use $URLGETTER directly, it must take
1673                                       precisely two command-line arguments
1674                                       (source URL and destination file name)
1675    
1676    ----------------------------------------------------------------------
1677    Name: Matthias Blume
1678    Date: 2002/01/28
1679    Tag: blume-20020128-sparc-ccalls
1680    Description:
1681    
1682    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1683    - Make use of the allocParam argument in sparc-c-calls.
1684    
1685    ----------------------------------------------------------------------
1686    Name: Matthias Blume
1687    Date: 2002/01/28
1688    Tag: blume-20020128-allocParam
1689    Description:
1690    
1691    John Reppy:  Changes c-calls API to accept client-callback for
1692    allocating extra stack space.
1693    me: Corresponding changes to mlriscGen (using a dummy argument that
1694        does not change the current behavior).
1695    
1696    ----------------------------------------------------------------------
1697    Name: Matthias Blume
1698    Date: 2002/01/28 12:00:00
1699    Tag: Release_110_38
1700    Description:
1701    
1702    This time for real!!!
1703    
1704    ----------------------------------------------------------------------
1705    Name: Matthias Blume
1706    Date: 2002/01/28 10:56:00 EST
1707    Tag: blume-20020128-retraction
1708    Description:
1709    
1710    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1711       with blume-Release_110_38-retracted.)
1712    
1713    1. Fixed a problem with incorrect rounding modes in real64.sml.
1714       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1715    
1716    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1717       and enums fixed.  The naming of corresponding ML identifiers should
1718       now be consistent again.
1719    
1720    ----------------------------------------------------------------------
1721    Name: Allen Leung
1722    Date: 2002/01/27
1723    Tag: leunga-20020127-nowhere
1724    Description:
1725    
1726       Added a target called nowhere in the configuration scripts.
1727       Enabling this will build the MLRISC 'nowhere' tool (for translating
1728       programs with where-clauses into legal SML code) during installation.
1729    
1730    ----------------------------------------------------------------------
1731    Name: Matthias Blume
1732    Date: 2002/01/25 21:27:00 EST
1733    Tag: blume-Release_110_38-retracted
1734    Description:
1735    
1736    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1737    
1738    README will be added later.
1739    
1740    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1741    (Release_110_38) removed.  Reason: Last-minute bug fixes.
1742    
1743    ----------------------------------------------------------------------
1744    Name: Matthias Blume
1745    Date: 2002/01/25
1746    Tag: blume-20020125-ffi
1747    Description:
1748    
1749    A large number of tweaks and improvements to ml-nlffi-lib and
1750    ml-nlffigen:
1751    
1752       - ML represenation types have been streamlined
1753       - getter and setter functions work with concrete values, not abstract
1754         ones where possible
1755       - ml-nlffigen command line more flexible (see README file there)
1756       - some bugs have been fixed (hopefully)
1757    
1758    ----------------------------------------------------------------------
1759    Name: Lal George
1760    Date: 2002/01/24
1761    Tag: george-20020124-risc-ra-interface
1762    Description:
1763    
1764       There is a dramatic simplification in the interface to the
1765       register allocator for RISC architectures as a result of making
1766       parallel copy instructions explicit.
1767    
1768    ----------------------------------------------------------------------
1769    Name: Matthias Blume
1770    Date: 2002/01/22
1771    Tag: blume-20020122-x86-ccalls
1772    Description:
1773    
1774    Bug fix for c-calls on x86 (having to do with how char- and
1775    short-arguments are being handled).
1776    
1777    ----------------------------------------------------------------------
1778    Name: Matthias Blume
1779    Date: 2002/01/21
1780    Tag: blume-20020121-ff
1781    Description:
1782    
1783    Another day of fiddling with the FFI...
1784    
1785    1. Bug fix/workaround:  CKIT does not complain about negative array
1786       dimensions, so ml-nlffigen has to guard itself against this possibility.
1787       (Otherwise a negative dimension would send it into an infinite loop.)
1788    
1789    2. Some of the abstract types (light objects, light pointers, most "base"
1790       types) in structure C are now eqtypes.
1791    
1792    3. Added constructors and test functions for NULL function pointers.
1793    
1794    ----------------------------------------------------------------------
1795    Name: Matthias Blume
1796    Date: 2002/01/18
1797    Tag: blume-20020118-ready-for-new-release
1798    Description:
1799    
1800    Made config/srcarchiveurl point to a new place.  (Will provide boot
1801    files shortly.)
1802    
1803    Maybe we christen this to be 110.38?
1804    
1805    ----------------------------------------------------------------------
1806    Name: Matthias Blume
1807    Date: 2002/01/18
1808    Tag: blume-20020118-more-ffifiddle
1809    Description:
1810    
1811    Today's FFI fiddling:
1812    
1813      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
1814        C.Set.* that use concrete (MLRep.*) arguments and results instead
1815        of abstract ones.
1816    
1817      - Provided word-style bit operations etc. for "int" representation
1818        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1819        and Long.
1820    
1821    ----------------------------------------------------------------------
1822    Name: Matthias Blume
1823    Date: 2002/01/18
1824    Tag: blume-20020118-use-x86-fp
1825    Description:
1826    
1827    Now that x86-fast-fp seems to be working, I turned it back on again
1828    by default.  (Seems to work fine now, even with the FFI.)
1829    
1830    Other than that, I added some documentation about the FFI to
1831    src/ml-nlffigen/README and updated the FFI test examples in
1832    src/ml-nlffi-lib/Tests/*.
1833    
1834    ----------------------------------------------------------------------
1835    Name: Allen Leung
1836    Date: 2002/01/17
1837    Tag: leunga-20020117-x86-fast-fp-call
1838    Description:
1839    
1840       1. Fixed a problem with handling return fp values when x86's fast fp
1841          mode is turned on.
1842    
1843       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
1844          of %f32.
1845    
1846       3. Added a constructor INT32lit to the ast of MLRISC tools.
1847    
1848    ----------------------------------------------------------------------
1849    Name: Matthias Blume
1850    Date: 2002/01/16
1851    Tag: blume-20020116-ffifiddle
1852    Description:
1853    
1854    More fiddling with the FFI interface:
1855    
1856     - Make constness 'c instead of rw wherever possible.  This eliminates
1857       the need for certain explicit coercions.  (However, due to ML's
1858       value polymorphism, there will still be many cases where explicit
1859       coercions are necessary.  Phantom types are not the whole answer
1860       to modeling a subtyping relationship in ML.)
1861    
1862     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
1863    
1864     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
1865       clumsy workaround for varargs problem.)
1866    
1867    ----------------------------------------------------------------------
1868    Name: Lal George
1869    Date: 2002/01/15
1870    Tag: <none>
1871    Description:
1872    
1873    1. Since COPY instructions are no longer native to the architecture,
1874       a generic functor can be used to implement the expandCopies function.
1875    
1876    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1877       TEXT segment.
1878    
1879    ----------------------------------------------------------------------
1880    Name: Matthias Blume
1881    Date: 2002/01/15
1882    Tag: blume-20020115-ffiupdates
1883    Description:
1884    
1885    1. Fix for bug resulting in single-precision float values being returned
1886       incorrectly from FFI calls.
1887    
1888    2. Small modifications to C FFI API:
1889    
1890        - memory-allocation routines return straight objects (no options)
1891          and raise an exception in out-of-memory situations
1892        - unsafe extensions to cast between function pointers and pointers
1893          from/to ints
1894        - added structure C_Debug as an alternative to structure C where
1895          pointer-dereferencing (|*| and |*!) always check for null-pointers
1896        - added open_lib' to DynLinkage;  open_lib' works like open_lib
1897          but also takes a (possibly empty) list of existing library handles
1898          that the current library depends on
1899    
1900    ----------------------------------------------------------------------
1901    Name: Matthias Blume
1902    Date: 2002/01/10
1903    Tag: blume-20020110-newffigen
1904    Description:
1905    
1906    1. Updates to portable graph code.
1907    
1908    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
1909       more scalable now so that even huge interfaces such as the one
1910       for GTK compile in finite time and space. :-)
1911       See src/ml-nlffigen/README for details on what's new.
1912    
1913    ----------------------------------------------------------------------
1914    Name: Lal George
1915    Date: 2001/01/09 14:31:35 EST 2002
1916    Tag: george-20011206-rm-native-copy
1917    Description:
1918    
1919            Removed the native COPY and FCOPY instructions
1920            from all the architectures and replaced it with the
1921            explicit COPY instruction from the previous commit.
1922    
1923            It is now possible to simplify many of the optimizations
1924            modules that manipulate copies. This has not been
1925            done in this change.
1926    
1927    ----------------------------------------------------------------------
1928    Name: Lal George
1929    Date: 2001/12/06 16:50:13 EST 2001
1930    Tag: george-20011206-mlrisc-instruction
1931    Description:
1932    
1933    Changed the representation of instructions from being fully abstract
1934    to being partially concrete. That is to say:
1935    
1936      from
1937            type instruction
1938    
1939      to
1940            type instr                              (* machine instruction *)
1941    
1942            datatype instruction =
1943                LIVE of {regs: C.cellset, spilled: C.cellset}
1944              | KILL of {regs: C.cellset, spilled: C.cellset}
1945              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
1946              | ANNOTATION of {i: instruction, a: Annotations.annotation}
1947              | INSTR of instr
1948    
1949    This makes the handling of certain special instructions that appear on
1950    all architectures easier and uniform.
1951    
1952    LIVE and KILL say that a list of registers are live or killed at the
1953    program point where they appear. No spill code is generated when an
1954    element of the 'regs' field is spilled, but the register is moved to
1955    the 'spilled' (which is present, more for debugging than anything else).
1956    
1957    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
1958    We used to generate:
1959    
1960            DEFFREG f1
1961            f1 := f2 + f3
1962            trapb
1963    
1964    but now generate:
1965    
1966            f1 := f2 + f3
1967            trapb
1968            LIVE {regs=[f1,f2,f3], spilled=[]}
1969    
1970    Furthermore, the DEFFREG (hack) required that all floating point instruction
1971    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
1972    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
1973    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
1974    this architecture).
1975    
1976    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
1977    all the architectures. This will result in further simplification of the
1978    register allocator that must be aware of them for coalescing purposes, and
1979    will also simplify certain aspects of the machine description that provides
1980    callbacks related to parallel copies.
1981    
1982    ANNOTATION should be obvious, and now INSTR represents the honest to God
1983    machine instruction set!
1984    
1985    The <arch>/instructions/<arch>Instr.sml files define certain utility
1986    functions for making porting easier -- essentially converting upper case
1987    to lower case. All machine instructions (of type instr) are in upper case,
1988    and the lower case form generates an MLRISC instruction. For example on
1989    the alpha we have:
1990    
1991      datatype instr =
1992         LDA of {r:cell, b:cell, d:operand}
1993       | ...
1994    
1995      val lda : {r:cell, b:cell, d:operand} -> instruction
1996        ...
1997    
1998    where lda is just (INSTR o LDA), etc.
1999    
2000    ----------------------------------------------------------------------
2001    Name: Matthias Blume
2002    Date: 2001/11/22 21:40:00 EST
2003    Tag: Release_110_37
2004    Description:
2005    
2006    Release 110.37.  This time for real.
2007    
2008    ----------------------------------------------------------------------
2009    Name: Matthias Blume
2010    Date: 2001/11/21 16:35:00 EST
2011    Tag: blume-20011121-foot-in-mouth
2012    Description:
2013    
2014    Removed the "Release_110_37" tag because of a serious bug.
2015    This will be re-tagged once the bug is fixed.
2016    
2017    ----------------------------------------------------------------------
2018    Name: Matthias Blume
2019    Date: 2001/11/21 16:14:00 EST
2020    Tag: blume-20011121-forgottenfile
2021    Description:
2022    
2023    Forgot to add a file.  (Just a .tex-file -- part of
2024    the CM manual source.)
2025    
2026    ----------------------------------------------------------------------
2027    Name: Matthias Blume
2028    Date: 2001/11/21 16:10:00 EST
2029    Tag: blume-20011121-invalid_110_37
2030    Description:
2031    
2032    Note: I removed the original tag "Release_110_37" from this commit
2033          because we found a serious bug in all non-x86 backends.
2034          - Matthias
2035    
2036    1. Modifications to the SML/NJ code generator and to the runtime system
2037       so that code object name strings are directly inserted into code
2038       objects at code generation time.  The only business the runtime system
2039       has with this is now to read the name strings on occasions.
2040       (The encoding of the name string has also changed somewhat.)
2041    
2042    2. CM now implements a simple "set calculus" for specifying export lists.
2043       In particular, it is now possible to refer to the export lists of
2044       other libraries/groups/sources and form unions as well as differences.
2045       See the latest CM manual for details.
2046    
2047    3. An separate notion of "proxy" libraries has again be eliminated from
2048       CM's model.  (Proxy libraries are now simply a special case of using
2049       the export list calculus.)
2050    
2051    4. Some of the existing libraries now take advantage of the new set
2052       calculus.
2053       (Notice that not all libraries have been converted because some
2054       of the existing .cm-files are supposed to be backward compatible
2055       with 110.0.x.)
2056    
2057    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2058       "General.exnMessage"!  The former relies on a certain hook to be
2059       initialized, and that often does not happen in the stand-alone case.)
2060    
2061    ----------------------------------------------------------------------
2062    Name: Lal George
2063    Date: 2001/11/21  13:56:18 EST
2064    Tag: george-2001121-pseudo-ops
2065    Description:
2066    
2067      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2068      ought to never be any question of incompatabilities with
2069      pseudo-op syntax expected by host assemblers.
2070    
2071      For now, only modules supporting GAS syntax are implemented
2072      but more should follow, such as MASM, and vendor assembler
2073      syntax, e.g. IBM as, Sun as, etc.
2074    
2075    ----------------------------------------------------------------------
2076    Name: Matthias Blume
2077    Date: 2001/11/14 11:52:00 EST
2078    Tag: blume-20011114-srcname
2079    Description:
2080    
2081    1. Routed the name of the current source file to mlriscgen where it
2082       should be directly emitted into the code object.  (This last part
2083       is yet to be done.)
2084    
2085    2. Some cleanup of the pgraph code to make it match the proposal that
2086       I put out the other day.  (The proposal notwithstanding, things are
2087       still in flux here.)
2088    
2089    ----------------------------------------------------------------------
2090    Name: Lal George
2091    Date: 2001/11/14 09:44:04 EST
2092    Tag:
2093    Description:
2094    
2095      Fix for a backpatching bug reported by Allen.
2096    
2097      Because the boundary between short and long span-dependent
2098      instructions is +/- 128, there are an astounding number of
2099      span-dependent instructions whose size is over estimated.
2100    
2101      Allen came up with the idea of letting the size of span
2102      dependent instructions be non-monotonic, for a maxIter
2103      number of times, after which the size must be monotonically
2104      increasing.
2105    
2106      This table shows the number of span-dependent instructions
2107      whose size was over-estimated as a function of maxIter, for the
2108      file Parse/parse/ml.grm.sml:
2109    
2110         maxIter            # of instructions:
2111            10                      687
2112            20                      438
2113            30                      198
2114            40                        0
2115    
2116      In compiling the compiler, there is no significant difference in
2117      compilation speed between maxIter=10 and maxIter=40. Actually,
2118      my measurements showed that maxIter=40 was a tad faster than
2119      maxIter=10! Also 96% of the  files in the compiler reach a fix
2120      point within 13 iterations, so fixing maxIter at 40, while high,
2121      is okay.
2122    
2123    ----------------------------------------------------------------------
2124    Name: Matthias Blume
2125    Date: 2001/10/31 15:25:00 EST
2126    Tag: blume-20011031-pgraph
2127    Description:
2128    
2129    CKIT:
2130    * Changed the "Function" constructor of type Ast.ctype to carry optional
2131      argument identifiers.
2132    * Changed the return type of TypeUtil.getFunction accordingly.
2133    * Type equality ignores the argument names.
2134    * TypeUtil.composite tries to preserve argument names but gives up quickly
2135      if there is a mismatch.
2136    
2137    installation script:
2138    * attempts to use "curl" if available (unless "wget" is available as well)
2139    
2140    CM:
2141    * has an experimental implementation of "portable graphs" which I will
2142      soon propose as an implementation-independent library format
2143    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2144    
2145    NLFFI-LIB:
2146    * some cleanup (all cosmetic)
2147    
2148    NLFFIGEN:
2149    * temporarily disabled the mechanism that suppresses ML output for
2150      C definitions whose identifiers start with an underscore character
2151    * generate val bindings for enum constants
2152    * user can request that only one style (light or heavy) is being used;
2153      default is to use both (command-line arguments: -heavy and -light)
2154    * fixed bug in handling of function types involving incomplete pointers
2155    * generate ML entry points that take record arguments (i.e., using
2156      named arguments) for C functions that have a prototype with named
2157      arguments
2158      (see changes to CKIT)
2159    
2160    ----------------------------------------------------------------------
2161    Name: Allen Leung
2162    Date: 2001/10/27 20:34:00 EDT
2163    Tag: leunga-20011027-x86-fast-fp-call
2164    Description:
2165    
2166       Fixed the bug described in blume-20010920-slowfp.
2167    
2168       The fix involves
2169          1. generating FCOPYs in FSTP in ia32-svid
2170          2. marking a CALL with the appropriate annotation
2171    
2172    ----------------------------------------------------------------------
2173    Name: Matthias Blume
2174    Date: 2001/10/16 11:32:00 EDT
2175    Tag: blume-20011016-netbsd
2176    Description:
2177    
2178    Underscore patch from Chris Richards (fixing problem with compiling
2179    runtime system under recent NetBSD).
2180    
2181    ----------------------------------------------------------------------
2182    Name: Allen Leung
2183    Date: 2001/10/12 17:18:32 EDT 2001
2184    Tag: leung-20011012-x86-printflowgraph
2185    Description:
2186    
2187    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2188    
2189    ----------------------------------------------------------------------
2190    Name: Lal George
2191    Date: 2001/10/11 23:51:34 EDT
2192    Tag: george-20011011-too-many-instrs
2193    Description:
2194    
2195    The representation of a program point never expected to see more
2196    than 65536 instructions in a basic block!
2197    
2198    ----------------------------------------------------------------------
2199  Name: Lal George  Name: Lal George
2200  Date: 2001/10/09 09:41:37 EDT  Date: 2001/10/09 09:41:37 EDT
2201  Tag: george-20011008-mlrisc-labels  Tag: george-20011008-mlrisc-labels
# Line 71  Line 2254 
2254    
2255  ----------------------------------------------------------------------  ----------------------------------------------------------------------
2256  Name: Matthias Blume  Name: Matthias Blume
 >>>>>>> 1.169  
2257  Date: 2001/09/18 15:35:00 EDT  Date: 2001/09/18 15:35:00 EDT
2258  Tag: blume-20010918-readme11036  Tag: blume-20010918-readme11036
2259  Description:  Description:

Legend:
Removed from v.952  
changed lines
  Added in v.1294

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