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 587, Thu Mar 30 09:01:52 2000 UTC revision 1332, Sun May 18 03:21:16 2003 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 (blume@tti-c.org)
17    Date: 2003/05/17 10:20:00 CDT
18    Tag: blume-20030517-absyn
19    Description:
20    
21    1. Added cases for IF, WHILE, ANDALSO, and ORELSE to Absyn.
22    
23       This mainly affects the quality of error messages.  However, some
24       of the code is now more straightforward than before. (Treatment of
25       the above four constructs in translate.sml is much simpler than
26       the "macro-expansion" that was going on before.  Plus, the mach-
27       compiler no longer gets invoked just to be able to compile an
28       if-expression.)
29    
30    2. The ErrorMsg.Error exception is now caught and absorbed by the
31       interactive loop.
32    
33    ----------------------------------------------------------------------
34    Name: Allen Leung
35    Date: 2003/05/16 13:05:00 CDT
36    Tag: leunga-20030516-cygwin-runtime
37    Description:
38    
39        Ported the runtime system to cygwin, which uses the unix
40    x86-unix bin files.  Missing/buggy features:
41    
42       o getnetbyname, getnetbyaddr: these functions seem to be missing in
43         the Cygwin library.
44       o Ctrl-C handling may be flaky.
45       o Windows system calls and Windows I/O are not supported.
46    
47       A new set of binfiles is located at:
48    
49         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
50    
51       This is only needed for bootstrapping the cygwin version of smlnj.
52       Other x86 versions can use the existing binfiles.
53    
54    ----------------------------------------------------------------------
55    Name: Matthias Blume
56    Date: 2003/04/08 15:42:00 CDT
57    Tag: blume-20030408-listpair
58    Description:
59    
60    1. Added a target 'mlrisc' to installer.
61    
62    2. Added missing elements to structure ListPair.
63    
64    ----------------------------------------------------------------------
65    Name: Allen Leung
66    Date: 2003/01/07 10:40:00 EST
67    Tag: leunga-20030107-int-rem
68    Description:
69    
70    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
71    The arguments to the SUBL instruction were swapped.
72    
73    ----------------------------------------------------------------------
74    Name: Matthias Blume
75    Date: 2002/12/12 16:25:00 EST
76    Tag: blume-20021212-risc-ra
77    Description:
78    
79    Fixed a serious bug in the rewrite code for FP spilling/reloading that
80    sent the RA into an infinite loop when floating point registers get
81    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
82    110.37 and 110.38.)
83    There was another set of potential problems related to the handling of
84    MLRISC annotations (but those did not yet cause real problems, apparently).
85    
86    ----------------------------------------------------------------------
87    Name: Matthias Blume
88    Date: 2002/12/06 22:40:00 EST
89    Tag: blume-20021206-cm-fileid
90    Description:
91    
92    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
93    This fixes the problem of CM getting confused by files that suddenly
94    change their identity (e.g., by getting unlinked and recreated by some
95    text editor such as vi).  There might be a better/cheaper/cleaner way
96    of doing this, but for now this will have to do.
97    
98    ----------------------------------------------------------------------
99    Name: Matthias Blume
100    Date: 2002/10/28 09:50:00 EST
101    Tag: blume-20021028-typecheck
102    Description:
103    
104    Exported structure Typecheck from $smlnj/viscomp/core.cm.
105    
106    ----------------------------------------------------------------------
107    Name: Matthias Blume
108    Date: 2002/10/17 09:10:00 EDT
109    Tag: Release_110_42
110    Description:
111    
112    In good old tradition, there has been a slight hiccup so that we have
113    to patch 110.42 after the fact.   The old release tag has been replaced
114    (see below).
115    
116    The change solves a problem with two competing approaches the
117    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
118    each other's way.
119    
120    This change only affects the runtime system code and the installer script.
121    (No new bootfiles.)
122    
123    ----------------------------------------------------------------------
124    Name: Matthias Blume
125    Date: 2002/10/16 12:00:00 EDT
126    Tag: Release_110_42_removed
127    Description:
128    
129    New working release.  New bootfiles.
130    
131    ----------------------------------------------------------------------
132    Name: Matthias Blume
133    Date: 2002/10/10 13:10:00 EDT
134    Tag: blume-20021010-ppc-divs
135    Description:
136    
137    The mltree operator DIVS must be implemented with an overflow check on
138    the PPC because the hardware indicates divide-by-zero using "overflow" as
139    well.
140    
141    ----------------------------------------------------------------------
142    Name: Matthias Blume
143    Date: 2002/07/23 11:45:00
144    Tag: blume-20020723-smlnj-home
145    Description:
146    
147    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
148    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
149    CM_PATHCONFIG is also set), the path configuration file is assumed
150    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
151    the entire tree to some other place and everything will "just work".
152    
153    (Companion commands such as ml-build and ml-makedepend also sense this
154    variable.)
155    
156    ----------------------------------------------------------------------
157    Name: Matthias Blume
158    Date: 2002/07/12 21:19:00 EDT
159    Tag: blume-20020712-liveness
160    Description:
161    
162    Exported two useful "step" functions from liveness module (MLRISC).
163    
164    ----------------------------------------------------------------------
165    Name: Matthias Blume
166    Date: 2002/07/05 16:00 EDT
167    Tag: Release_110_41
168    Description:
169    
170    New working release.  New bootfiles.
171    
172    ----------------------------------------------------------------------
173    Name: Matthias Blume
174    Date: 2002/07/05 10:25:00 EDT
175    Tag: blume-20020705-btimp
176    Description:
177    
178    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
179    other clients can set up backtracing support.
180    
181    ----------------------------------------------------------------------
182    Name: Matthias Blume
183    Date: 2002/06/25 17:23:00 EDT
184    Tag: blume-20020625-fpmax
185    Description:
186    
187    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
188    case.  (The sense of the isNaN test was reversed -- which made min and
189    max always return their first argument.)
190    
191    ----------------------------------------------------------------------
192    Name: Matthias Blume
193    Date: 2002/06/11
194    Tag: blume-20020611-unixpath
195    Description:
196    
197    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
198    
199    ----------------------------------------------------------------------
200    Name: Matthias Blume
201    Date: 2002/06/10 16:35:00 EDT
202    Tag: blume-20020610-ieeereal
203    Description:
204    
205    I back-ported my implementation of IEEEReal.fromString from the
206    idlbasis-devel branch so that we can test it.
207    
208    Another small change is that ppDec tries to give more information
209    than just "<sig>" in the case of functors.  However, this code is
210    broken in some mysterious way if the functor's body's signature
211    has not been declared by ascription but gets inferred from the
212    implementation.  This needs fixing...
213    
214    ----------------------------------------------------------------------
215    Name: Matthias Blume
216    Date: 2002/05/31
217    Tag: blume-20020531-btrace-mode
218    Description:
219    
220    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
221    the wayside when I switched over to using Controls everywhere.)
222    
223    ----------------------------------------------------------------------
224    Name: Lal George
225    Date: 2002/05/23 12:21:40 EDT
226    Tag: george-20020523-visual-labels
227    Description:
228    
229    Labels are now displayed in the graphical output to make
230    the fall-through and target blocks obvious.
231    
232    ----------------------------------------------------------------------
233    Name: Matthias Blume
234    Date: 2002/05/22 11:03:00 EDT
235    Tag: blume-20020522-shrink
236    Description:
237    
238    John tweaked yesterday's fix for 1131 to handle an out-of-memory
239    situation that comes up when allocating huge arrays.
240    
241    ----------------------------------------------------------------------
242    Name: Matthias Blume
243    Date: 2002/05/21 16:00:00 EDT
244    Tag: Release_110_40
245    Description:
246    
247    New working release (110.40).  New bootfiles.
248    
249    [Also: John Reppy fixed GC bug 1131.]
250    
251    ----------------------------------------------------------------------
252    Name: Matthias Blume
253    Date: 2002/05/21 12:35:00 EDT
254    Tag: blume-20020521-cmdoc
255    Description:
256    
257    CM documentation update.
258    
259    ----------------------------------------------------------------------
260    Name: Matthias Blume
261    Date: 2002/05/21 10:55:00 EDT
262    Tag: blume-20020521-misc
263    Description:
264    
265    - John tweaked runtime to be silent on heap export (except when
266      GC messages are on).
267    - I added a few more things (cross-compiling versions of CMB) to
268      config/preloads (as suggestions).
269    
270    ----------------------------------------------------------------------
271    Name: Matthias Blume
272    Date: 2002/05/20 22:25:00 EDT
273    Tag: blume-20020520-controls
274    Description:
275    
276    - Added ControlUtil structure to control-lib.cm.
277    - Use it throughout.
278    - Used Controls facility to define MLRISC controls (as opposed to
279      registering MLRISC control ref cells with Controls after the fact)
280    - Fixed messed-up controls priorities.
281    
282    * Removed again all the stuff from config/preloads that one wouldn't
283      be able to preload at the time the initial heap image is built.
284      (Many libraries, e.g., CML, do not exist yet at this time.  The
285       only libraries that can be preloaded via config/preloads are those
286       that come bundled with the bootfiles.)
287    
288    ----------------------------------------------------------------------
289    Name: Matthias Blume
290    Date: 2002/05/20 10:59:00 EDT
291    Tag: blume-20020520-preloads
292    Description:
293    
294    Added a lot of commented-out suggestions for things to be included
295    in config/preloads.
296    
297    ----------------------------------------------------------------------
298    Name: Allen Leung
299    Date: 2002/05/18 14:20:00 EDT
300    Tag: leunga-20020518-mdl
301    Description:
302    
303    o Made the mdl tool stuff compile and run again.
304    o I've disabled all the stuff that depends on RTL specifications; they
305      are all badly broken anyway.
306    
307    ----------------------------------------------------------------------
308    Name: Matthias Blume
309    Date: 2002/05/17 16:49:00 EDT
310    Tag: blume-20020517-controls
311    Description:
312    
313    0. John Reppy made several modifications to the SML/NJ library.
314       In particular, there is a shiny new controls-lib.cm.
315    
316    1. Pushed new controls interface through compiler so that everything
317       compiles again.
318    
319    2. Added FormatComb and FORMAT_COMB to the CML version of the
320       SML/NJ library (so that CML compiles again).
321    
322    3. Modified init scripts because XXX_DEFAULT environment variables
323       are no longer with us.  (Boot-time initialization is now done
324       using the same environment variables that are also used for
325       startup-time initialization of controls.)
326    
327    ----------------------------------------------------------------------
328    Name: Lal George
329    Date: 2002/05/15 09:20:10 EDT
330    Tag: george-20020515-pseudo-op-decls
331    Description:
332    
333            All pseudo-ops emitted before the first segment declaration
334    such as TEXT, DATA, and BSS directives are assumed to be global
335    declarations and are emitted first in the assembly file. This is
336    useful in a number of situations where one has pseudo-ops that are not
337    specific to any segment, and also works around the constraint that one
338    cannot have client pseudo-ops in the TEXT segment.
339    
340            Because no segment is associated with these declarations it is
341    an error to allocate any space or objects before the first segment
342    directive and an exception will be raised. However, we cannot make
343    this check for client pseudo-ops.
344    
345            These top level declarations are a field in the CFG graph_info.
346    In theory you can continue to add to this field after the CFG has been
347    built -- provided you know what you are doing;-)
348    
349    ----------------------------------------------------------------------
350    Name: Matthias Blume
351    Date: 2002/05/13 16:40:00 EDT
352    Tag: blume-20020513-pp-etc
353    Description:
354    
355    A few minor bugfixes:
356    
357      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
358        (Bogus printouts for redefined bindings still occur.  Compiler
359        bug should no longer occur now.  We need to redo the prettyprinter
360        from scratch.)
361    
362      - CM pathname printer now also adds escape sequences for ( and )
363    
364      - commend and docu fixes for ml-nlffi
365    
366    ----------------------------------------------------------------------
367    Name: Matthias Blume
368    Date: 2002/05/10 16:40:00 EDT
369    Tag: blume-20020510-erg-textio
370    Description:
371    
372    Applied the following bugfix provided by Emden Gansner:
373    
374        Output is corrupted when outputSubstr is used rather than output.
375    
376        The problem occurs when a substring
377    
378            ss = (s, dataStart, dataLen)
379    
380        where dataStart > 0, fills a stream buffer with avail bytes left.
381        avail bytes of s, starting at index dataStart, are copied into the
382        buffer, the buffer is flushed, and then the remaining dataLen-avail
383        bytes of ss are copied into the beginning of the buffer. Instead of
384        starting this copy at index dataStart+avail in s, the current code
385        starts the copy at index avail.
386    
387        Fix:
388        In text-io-fn.sml, change line 695 from
389             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
390        to
391             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
392    
393    ----------------------------------------------------------------------
394    Name: Matthias Blume
395    Date: 2002/04/12 13:55:00 EDT
396    Tag: blume-20020412-assyntax
397    Description:
398    
399    1. Grabbed newer assyntax.h from the XFree86 project.
400    2. Fiddled with how to compile X86.prim.asm without warnings.
401    3. (Very) Minor cleanup in CM.
402    
403    ----------------------------------------------------------------------
404    Name: Matthias Blume
405    Date: 2002/04/01 (no joke!) 17:07:00 EST
406    Tag: blume-20020401-x86div
407    Description:
408    
409    Added full support for div/mod/rem/quot on the x86, using the machine
410    instruction's two results (without clumsily recomputing the remainder)
411    directly where appropriate.
412    
413    Some more extensive power-of-two support was added to the x86 instruction
414    selector (avoiding expensive divs, mods, and muls where they can be
415    replaced with cheaper shifts and masks).  However, this sort of thing
416    ought to be done earlier, e.g., within the CPS optimizer so that
417    all architectures benefit from it.
418    
419    The compiler compiles to a fixed point, but changes might be somewhat
420    fragile nevertheless.  Please, report any strange things that you might
421    see wrt. div/mod/quot/rem...
422    
423    ----------------------------------------------------------------------
424    Name: Matthias Blume
425    Date: 2002/03/29 17:22:00
426    Tag: blume-20020329-div
427    Description:
428    
429    Fixed my broken div/mod logic.  Unfortunately, this means that the
430    inline code for div/mod now has one more comparison than before.
431    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
432    The problem was with quotient = 0, because that alone does not tell
433    us which way the rounding went.  One then has to look at whether
434    remainder and divisor have the same sign...  :(
435    
436    Anyway, I replaced the bootfiles with fresh ones...
437    
438    ----------------------------------------------------------------------
439    Name: Matthias Blume
440    Date: 2002/03/29 14:10:00 EST
441    Tag: blume-20020329-inlprims
442    Description:
443    
444    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
445    
446    Primops have changed. This means that the bin/boot-file formats have
447    changed as well.
448    
449    To make sure that there is no confusion, I made a new version.
450    
451    
452    CHANGES:
453    
454    * removed REMT from mltree (remainder should never overflow).
455    
456    * added primops to deal with divisions of all flavors to the frontend
457    
458    * handled these primops all the way through so they map to their respective
459      MLRISC support
460    
461    * used these primops in the implementation of Int, Int32, Word, Word32
462    
463    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
464    
465    * parameterized INLMIN, INLMAX, and INLABS by a numkind
466    
467    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
468      floating point
469    
470    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
471      and Real (but Real.abs maps to a separate floating-point-only primop)
472    
473    
474    TODO items:
475    
476    * Hacked Alpha32 instruction selection, disabling the selection of REMx
477      instructions because the machine instruction encoder cannot handle
478      them.  (Hppa, PPC, and Sparc instruction selection did not handle
479      REM in the first place, and REM is supported by the x86 machine coder.)
480    
481    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
482      selection phase.  (The two can be streamlined because the hardware
483      delivers both quotient and remainder at the same time anyway.)
484    
485    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
486      (Currently the behavior is inconsistent both across architectures and
487      wrt. the draft Basis spec.)
488    
489    * Word8 should eventually be handled natively, too.
490    
491    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
492      as if there currently is no execution path that could trigger it in
493      SML/NJ.  (The assumptions underlying functions arith and promotable do not
494      hold for things like multiplication and division.)
495    
496    ----------------------------------------------------------------------
497    Name: Matthias Blume
498    Date: 2002/03/27 16:27:00 EST
499    Tag: blume-20020327-mlrisc-divisions
500    Description:
501    
502    Added support for all four division operations (ML's div, mod, quot,
503    and rem) to MLRISC.  In the course of doing so, I also rationalized
504    the naming (no more annoying switch-around of DIV and QUOT), by
505    parameterizing the operation by div_rounding_mode (which can be either
506    DIV_TO_ZERO or DIV_TO_NEGINF).
507    
508    The generic MLTreeGen functor takes care of compiling all four
509    operations down to only round-to-zero div.
510    
511    Missing pieces:
512    
513      * Doing something smarter than relying on MLTreeGen on architectures
514        like, e.g., the x86 where hardware division delivers both quotient and
515        remainder at the same time.  With this, the implementation of the
516        round-to-neginf operations could be further streamlined.
517    
518      * Remove inlining support for div/mod/rem from the frontend and replace it
519        with primops that get carried through to the backend.  Do this for all
520        int and word types.
521    
522    ----------------------------------------------------------------------
523    Name: Matthias Blume
524    Date: 2002/03/25 17:25:00 EST
525    Tag: blume-20020325-divmod
526    Description:
527    
528    I improved (hopefully without breaking them) the implementation of Int.div,
529    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
530    advantage of the following observations:
531    
532      Let  q = x quot y      r = x rem y
533           d = x div  y      m = x mod y
534    
535    where "quot" is the round-to-zero version of integer division that
536    hardware usually provides.  Then we have:
537    
538         r = x - q * y        where neither the * nor the - will overflow
539         d = if q >= 0 orelse x = q * y then q else q - 1
540                              where neither the * nor the - will overflow
541         m = if q >= 0 orelse r = 0 then r else r + y
542                              where the + will not overflow
543    
544    This results in substantial simplification of the generated code.
545    The following table shows the number of CFG nodes and edges generated
546    for
547            fun f (x, y) = x OPER y
548            (* with OPER \in div, mod, quot, rem *)
549    
550    
551        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
552        --------------------------------------------------------
553         div |         24 |         39 |         12 |         16
554         mod |         41 |         71 |         12 |         16
555        quot |          8 |         10 |          8 |         10
556         rem |         10 |         14 |          8 |         10
557    
558    
559    ----------------------------------------------------------------------
560    Name: Matthias Blume
561    Date: 2002/03/25 22:06:00 EST
562    Tag: blume-20020325-cprotobug
563    Description:
564    
565    Fixed a bug in cproto (c prototype decoder).
566    
567    ----------------------------------------------------------------------
568    Name: Matthias Blume
569    Date: 2002/03/25 16:00:00 EST
570    Tag: blume-20020325-raw-primops
571    Description:
572    
573    I did some cleanup to Allen's new primop code and
574    replaced yesterday's bootfiles with new ones.
575    (But they are stored in the same place.)
576    
577    ----------------------------------------------------------------------
578    Name: Matthias Blume
579    Date: 2002/03/24 22:40:00 EST
580    Tag: blume-20020324-bootfiles
581    Description:
582    
583    Made the bootfiles that Allen asked for.
584    
585    ----------------------------------------------------------------------
586    Name: Allen Leung
587    Date: 2002/03/23 15:50:00 EST
588    Tag: leunga-20020323-flint-cps-rcc-primops
589    Description:
590    
591      1. Changes to FLINT primops:
592    
593        (* make a call to a C-function;
594         * The primop carries C function prototype information and specifies
595         * which of its (ML-) arguments are floating point. C prototype
596         * information is for use by the backend, ML information is for
597         * use by the CPS converter. *)
598      | RAW_CCALL of { c_proto: CTypes.c_proto,
599                       ml_args: ccall_type list,
600                       ml_res_opt: ccall_type option,
601                       reentrant : bool
602                     } option
603       (* Allocate uninitialized storage on the heap.
604        * The record is meant to hold short-lived C objects, i.e., they
605        * are not ML pointers.  With the tag, the representation is
606        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
607        *)
608      | RAW_RECORD of {tag:bool,sz:int}
609      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
610    
611      2.  These CPS primops are now overloaded:
612    
613           rawload of {kind:numkind}
614           rawstore of {kind:numkind}
615    
616          The one argument form is:
617    
618             rawload {kind} address
619    
620          The two argument form is:
621    
622             rawload {kind} [ml object, byte-offset]
623    
624      3. RAW_CCALL/RCC now takes two extra arguments:
625    
626         a. The first is whether the C call is reentrant, i.e., whether
627            ML state should be saved and restored.
628         b. The second argument is a string argument specifying the name of
629            library and the C function.
630    
631         These things are currently not handled in the code generator, yet.
632    
633      4. In CProto,
634    
635         An encoding type of "bool" means "ml object" and is mapped into
636         C prototype of PTR.  Note that "bool" is different than "string",
637         even though "string" is also mapped into PTR, because "bool"
638         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
639    
640      5. Pickler/unpicker
641    
642         Changed to handle RAW_RECORD and newest RAW_CCALL
643    
644      6. MLRiscGen,
645    
646         1. Changed to handle the new rawload/rawstore/rawrecord operators.
647         2. Code for handling C Calls has been moved to a new module CPSCCalls,
648            in the file CodeGen/cpscompile/cps-c-calls.sml
649    
650      7. Added the conditional move operator
651    
652             condmove of branch
653    
654         to cps.  Generation of this is still buggy so it is currently
655         disabled.
656    
657    ----------------------------------------------------------------------
658    Name: Lal George
659    Date: 2002/03/22 14:18:25 EST
660    Tag: george-20020322-cps-branch-prob
661    Description:
662    
663    Implemented the Ball-Larus branch prediction-heuristics, and
664    incorporated graphical viewers for control flow graphs.
665    
666    Ball-Larus Heuristics:
667    ---------------------
668    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
669    
670    By design it uses the Dempster-Shafer theory for combining
671    probabilities.  For example, in the function:
672    
673        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
674    
675    the ball-larus heuristics predicts that the n=0 is unlikely
676    (OH-heuristic), and the 'then' branch is unlikely because of the
677    RH-heuristic -- giving the 'then' branch an even lower combined
678    probability using the Dempster-Shafer theory.
679    
680    Finally, John Reppy's loop analysis in MLRISC, further lowers the
681    probability of the 'then' branch because of the loop in the else
682    branch.
683    
684    
685    Graphical Viewing:
686    ------------------
687    I merely plugged in Allen's graphical viewers into the compiler. The
688    additional code is not much. At the top level, saying:
689    
690            Control.MLRISC.getFlag "cfg-graphical-view" := true;
691    
692    will display the graphical view of the control flow graph just before
693    back-patching.  daVinci must be in your path for this to work. If
694    daVinci is not available, then the default viewer can be changed
695    using:
696    
697            Control.MLRISC.getString "viewer"
698    
699    which can be set to "dot" or "vcg" for the corresponding viewers. Of
700    course, these viewers must be in your path.
701    
702    The above will display the compilation unit at the level of clusters,
703    many of which are small, boring, and un-interesting. Also setting:
704    
705            Control.MLRISC.getInt "cfg-graphical-view_size"
706    
707    will display clusters that are larger than the value set by the above.
708    
709    
710    ----------------------------------------------------------------------
711    Name: Matthias Blume
712    Date: 2002/03/21 22:20:00 EST
713    Tag: blume-20020321-kmp-bugfix
714    Description:
715    
716    Changed the interface to the KMP routine in PreString and fixed
717    a minor bug in one place where it was used.
718    
719    ----------------------------------------------------------------------
720    Name: Allen Leung
721    Date: 2002/03/21 20:30:00 EST
722    Tag: leunga-20020321-cfg
723    Description:
724    
725      Fixed a potential problem in cfg edge splitting.
726    
727    ----------------------------------------------------------------------
728    Name: Allen Leung
729    Date: 2002/03/21 17:15:00 EST
730    Tag: leunga-20020321-x86-fp-cfg
731    Description:
732    
733      1. Recoded the buggy parts of x86-fp.
734    
735         a. All the block reordering code has been removed.
736            We now depend on the block placement phases to do this work.
737    
738         b. Critical edge splitting code has been simplified and moved into the
739            CFG modules, as where they belong.
740    
741         Both of these were quite buggy and complex.  The code is now much, much
742         simpler.
743    
744      2. X86 backend.
745    
746         a. Added instructions for 64-bit support.  Instruction selection for
747            64-bit has not been committed, however, since that
748            requires changes to MLTREE which haven't been approved by
749            Lal and John.
750    
751         b. Added support for FUCOMI and FUCOMIP when generating code for
752            PentiumPro and above.  We only generate these instructions in
753            the fast-fp mode.
754    
755         c. Added cases for JP and JNP in X86FreqProps.
756    
757      3. CFG
758    
759         CFG now has a bunch of methods for edge splitting and merging.
760    
761      4. Machine description.
762    
763         John's simplification of MLTREE_BASIS.fcond broke a few machine
764         description things:
765    
766         rtl-build.{sig,sml} and hppa.mdl fixed.
767    
768         NOTE: the machine description stuff in the repository is still broken.
769               Again, I can't put my fixes in because that involves
770               changes to MLTREE.
771    
772    ----------------------------------------------------------------------
773    Name: Matthias Blume
774    Date: 2002/03/20 15:55:00 EST
775    Tag: blume-20020320-kmp
776    Description:
777    
778    Implemented Knuth-Morris-Pratt string matching in PreString and used
779    it for String.isSubstring, Substring.isSubstring, and
780    Substring.position.
781    
782    (Might need some stress-testing.  Simple examples worked fine.)
783    
784    ----------------------------------------------------------------------
785    Name: Matthias Blume
786    Date: 2002/03/19 16:37:00 EST
787    Tag: blume-20020319-witnesses
788    Description:
789    
790    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
791    
792    This implements a generic mechanism for changing constness qualifiers
793    anywhere within big C types without resorting to outright "casts".
794    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
795    modify the constness at the outermost level.)
796    The implementation of "convert" is based on the idea of "witness"
797    values -- values that are not used by the operation but whose types
798    "testify" to their applicability.  On the implementation side, "convert"
799    is simply a projection (returning its second curried argument).  With
800    cross-module inlining, it should not result in any machine code being
801    generated.
802    
803    ----------------------------------------------------------------------
804    Name: Matthias Blume
805    Date: 2002/03/15 16:40:00 EST
806    Tag: blume-20020315-basis
807    Description:
808    
809    Provided (preliminary?) implementations for
810    
811      {String,Substring}.{concatWith,isSuffix,isSubstring}
812    
813    and
814    
815      Substring.full
816    
817    Those are in the Basis spec but they were missing in SML/NJ.
818    
819    ----------------------------------------------------------------------
820    Name: Matthias Blume
821    Date: 2002/03/14 21:30:00 EST
822    Tag: blume-20020314-controls
823    Description:
824    
825    Controls:
826    ---------
827    
828    1. Factored out the recently-added Controls : CONTROLS stuff and put
829       it into its own library $/controls-lib.cm.  The source tree for
830       this is under src/smlnj-lib/Controls.
831    
832    2. Changed the names of types and functions in this interface, so they
833       make a bit more "sense":
834    
835          module -> registry
836          'a registry -> 'a group
837    
838    3. The interface now deals in ref cells only.  The getter/setter interface
839       is (mostly) gone.
840    
841    4. Added a function that lets one register an already-existing ref cell.
842    
843    5. Made the corresponding modifications to the rest of the code so that
844       everything compiles again.
845    
846    6. Changed the implementation of Controls.MLRISC back to something closer
847       to the original.  In particular, this module (and therefore MLRISC)
848       does not depend on Controls.  There now is some link-time code in
849       int-sys.sml that registers the MLRISC controls with the Controls
850       module.
851    
852    CM:
853    ---
854    
855      * One can now specify the lambda-split aggressiveness in init.cmi.
856    
857    ----------------------------------------------------------------------
858    Name: Allen Leung
859    Date: 2002/03/13 17:30:00 EST
860    Tag: leunga-20020313-x86-fp-unary
861    Description:
862    
863    Bug fix for:
864    
865    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
866    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
867    > - fun f(x,(y,z)) = Real.~ y;
868    > [autoloading]
869    > [autoloading done]
870    >       fchsl   (%eax), 184(%esp)
871    > Error: MLRisc bug: X86MCEmitter.emitInstr
872    >
873    > uncaught exception Error
874    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
875    
876    The problem was that the code generator did not generate any fp registers
877    in this case, and the ra didn't know that it needed to run the X86FP phase to
878    translate the pseudo fp instruction.   This only happened with unary fp
879    operators in certain situations.
880    
881    ----------------------------------------------------------------------
882    Name: Matthias Blume
883    Date: 2002/03/13 14:00:00 EST
884    Tag: blume-20020313-overload-etc
885    Description:
886    
887    1. Added _overload as a synonym for overload for backward compatibility.
888       (Control.overloadKW must be true for either version to be accepted.)
889    
890    2. Fixed bug in install script that caused more things to be installed
891       than what was requested in config/targets.
892    
893    3. Made CM aware of the (_)overload construct so that autoloading
894       works.
895    
896    ----------------------------------------------------------------------
897    Name: Matthias Blume
898    Date: 2002/03/12 22:03:00 EST
899    Tag: blume-20020312-url
900    Description:
901    
902    Forgot to update BOOT and srcarchiveurl.
903    
904    ----------------------------------------------------------------------
905    Name: Matthias Blume
906    Date: 2002/03/12 17:30:00 EST
907    Tag: blume-20020312-version110392
908    Description:
909    
910    Yet another version number bump (because of small changes to the
911    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
912    
913    Changes:
914    
915      The new pid generation scheme described a few weeks ago was overly
916      complicated.  I implemented a new mechanism that is simpler and
917      provides a bit more "stability":  Once CM has seen a compilation
918      unit, it keeps its identity constant (as long as you do not delete
919      those crucial CM/GUID/* files).  This means that when you change
920      an interface, compile, then go back to the old interface, and
921      compile again, you arrive at the original pid.
922    
923      There now also is a mechanism that instructs CM to use the plain
924      environment hash as a module's pid (effectively making its GUID
925      the empty string).  For this, "noguid" must be specified as an
926      option to the .sml file in question within its .cm file.
927      This is most useful for code that is being generated by tools such
928      as ml-nlffigen (because during development programmers tend to
929      erase the tool's entire output directory tree including CM's cached
930      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
931      revert to the old, broken behavior of SML/NJ, but in specific cases
932      where there is no danger of interface confusion, its use is ok
933      (I think).
934    
935      ml-nlffigen by default generates "noguid" annotations.  They can be
936      turned off by specifying -guid in its command line.
937    
938    ----------------------------------------------------------------------
939    Name: Lal George
940    Date: 2002/03/12 12 14:42:36 EST
941    Tag: george-20020312-frequency-computation
942    Description:
943    
944    Integrated jump chaining and static block frequency into the
945    compiler. More details and numbers later.
946    
947    ----------------------------------------------------------------------
948    Name: Lal George
949    Date: 2002/03/11 11 22:38:53 EST
950    Tag: george-20020311-jump-chain-elim
951    Description:
952    
953    Tested the jump chain elimination on all architectures (except the
954    hppa).  This is on by default right now and is profitable for the
955    alpha and x86, however, it may not be profitable for the sparc and ppc
956    when compiling the compiler.
957    
958    The gc test will typically jump to a label at the end of the cluster,
959    where there is another jump to an external cluster containing the actual
960    code to invoke gc. This is to allow factoring of common gc invocation
961    sequences. That is to say, we generate:
962    
963            f:
964               testgc
965               ja   L1      % jump if above to L1
966    
967            L1:
968               jmp L2
969    
970    
971    After jump chain elimination the 'ja L1' instructions is converted to
972    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
973    up being implemented in their long form (if L2 is far away) using:
974    
975            jbe     L3      % jump if below or equal to L3
976            jmp     L2
977         L3:
978            ...
979    
980    
981    For large compilation units L2  may be far away.
982    
983    
984    ----------------------------------------------------------------------
985    Name: Matthias Blume
986    Date: 2002/03/11 13:30:00 EST
987    Tag: blume-20020311-mltreeeval
988    Description:
989    
990    A functor parameter was missing.
991    
992    ----------------------------------------------------------------------
993    Name: Allen Leung
994    Date: 2002/03/11 10:30:00 EST
995    Tag: leunga-20020311-runtime-string0
996    Description:
997    
998       The representation of the empty string now points to a
999    legal null terminated C string instead of unit.  It is now possible
1000    to convert an ML string into C string with InlineT.CharVector.getData.
1001    This compiles into one single machine instruction.
1002    
1003    ----------------------------------------------------------------------
1004    Name: Allen Leung
1005    Date: 2002/03/10 23:55:00 EST
1006    Tag: leunga-20020310-x86-call
1007    Description:
1008    
1009       Added machine generation for CALL instruction (relative displacement mode)
1010    
1011    ----------------------------------------------------------------------
1012    Name: Matthias Blume
1013    Date: 2002/03/08 16:05:00
1014    Tag: blume-20020308-entrypoints
1015    Description:
1016    
1017    Version number bumped to 110.39.1.  NEW BOOTFILES!
1018    
1019    Entrypoints: non-zero offset into a code object where execution should begin.
1020    
1021    - Added the notion of an entrypoint to CodeObj.
1022    - Added reading/writing of entrypoint info to Binfile.
1023    - Made runtime system bootloader aware of entrypoints.
1024    - Use the address of the label of the first function given to mlriscGen
1025      as the entrypoint.  This address is currently always 0, but it will
1026      not be 0 once we turn on block placement.
1027    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1028      with entry points) from mlriscGen.
1029    
1030    ----------------------------------------------------------------------
1031    Name: Allen Leung
1032    Date: 2002/03/07 20:45:00 EST
1033    Tag: leunga-20020307-x86-cmov
1034    Description:
1035    
1036       Bug fixes for CMOVcc on x86.
1037    
1038       1. Added machine code generation for CMOVcc
1039       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1040       3. CMOVcc cannot have an immediate operand as argument.
1041    
1042    ----------------------------------------------------------------------
1043    Name: Matthias Blume
1044    Date: 2002/03/07 16:15:00 EST
1045    Tag: blume-20020307-controls
1046    Description:
1047    
1048    This is a very large but mostly boring patch which makes (almost)
1049    every tuneable compiler knob (i.e., pretty much everything under
1050    Control.* plus a few other things) configurable via both the command
1051    line and environment variables in the style CM did its configuration
1052    until now.
1053    
1054    Try starting sml with '-h' (or, if you are brave, '-H')
1055    
1056    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1057    implements the underlying generic mechanism.
1058    
1059    The interface to some of the existing such facilities has changed somewhat.
1060    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1061    (The getFoo interface is still there for backward-compatibility, but its
1062    use is deprecated.)
1063    
1064    The ml-build script passes -Cxxx=yyy command-line arguments through so
1065    that one can now twiddle the compiler settings when using this "batch"
1066    compiler.
1067    
1068    TODO items:
1069    
1070    We should go through and throw out all controls that are no longer
1071    connected to anything.  Moreover, we should go through and provide
1072    meaningful (and correct!) documentation strings for those controls
1073    that still are connected.
1074    
1075    Currently, multiple calls to Controls.new are accepted (only the first
1076    has any effect).  Eventually we should make sure that every control
1077    is being made (via Controls.new) exactly once.  Future access can then
1078    be done using Controls.acc.
1079    
1080    Finally, it would probably be a good idea to use the getter-setter
1081    interface to controls rather than ref cells.  For the time being, both
1082    styles are provided by the Controls module, but getter-setter pairs are
1083    better if thread-safety is of any concern because they can be wrapped.
1084    
1085    *****************************************
1086    
1087    One bug fix: The function blockPlacement in three of the MLRISC
1088    backpatch files used to be hard-wired to one of two possibilities at
1089    link time (according to the value of the placementFlag).  But (I
1090    think) it should rather sense the flag every time.
1091    
1092    *****************************************
1093    
1094    Other assorted changes (by other people who did not supply a HISTORY entry):
1095    
1096    1. the cross-module inliner now works much better (Monnier)
1097    2. representation of weights, frequencies, and probabilities in MLRISC
1098       changed in preparation of using those for weighted block placement
1099       (Reppy, George)
1100    
1101    ----------------------------------------------------------------------
1102    Name: Lal George
1103    Date: 2002/03/07 14:44:24 EST 2002
1104    Tag: george-20020307-weighted-block-placement
1105    
1106    Tested the weighted block placement optimization on all architectures
1107    (except the hppa) using AMPL to generate the block and edge frequencies.
1108    Changes were required in the machine properties to correctly
1109    categorize trap instructions. There is an MLRISC flag
1110    "weighted-block-placement" that can be used to enable weighted block
1111    placement, but this will be ineffective without block/edge
1112    frequencies (coming soon).
1113    
1114    
1115    ----------------------------------------------------------------------
1116    Name: Lal George
1117    Date: 2002/03/05 17:24:48 EST
1118    Tag: george-20020305-linkage-cluster
1119    
1120    In order to support the block placement optimization, a new cluster
1121    is generated as the very first cluster (called the linkage cluster).
1122    It contains a single jump to the 'real' entry point for the compilation
1123    unit. Block placement has no effect on the linkage cluster itself, but
1124    all the other clusters  have full freedom in the manner in which they
1125    reorder blocks or functions.
1126    
1127    On the x86 the typical linkage code that is generated is:
1128       ----------------------
1129            .align 2
1130       L0:
1131            addl    $L1-L0, 72(%esp)
1132            jmp     L1
1133    
1134    
1135            .align  2
1136       L1:
1137       ----------------------
1138    
1139    72(%esp) is the memory location for the stdlink register. This
1140    must contain the address of the CPS function being called. In the
1141    above example, it contains the address of  L0; before
1142    calling L1 (the real entry point for the compilation unit), it
1143    must contain the address for L1, and hence
1144    
1145            addl $L1-L0, 72(%esp)
1146    
1147    I have tested this on all architectures except the hppa.The increase
1148    in code size is of course negligible
1149    
1150    ----------------------------------------------------------------------
1151    Name: Allen Leung
1152    Date: 2002/03/03 13:20:00 EST
1153    Tag: leunga-20020303-mlrisc-tools
1154    
1155      Added #[ ... ] expressions to mlrisc tools
1156    
1157    ----------------------------------------------------------------------
1158    Name: Matthias Blume
1159    Date: 2002/02/27 12:29:00 EST
1160    Tag: blume-20020227-cdebug
1161    Description:
1162    
1163    - made types in structure C and C_Debug to be equal
1164    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1165    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1166    
1167    ----------------------------------------------------------------------
1168    Name: Matthias Blume
1169    Date: 2002/02/26 12:00:00 EST
1170    Tag: blume-20020226-ffi
1171    Description:
1172    
1173    1. Fixed a minor bug in CM's "noweb" tool:
1174       If numbering is turned off, then truly don't number (i.e., do not
1175       supply the -L option to noweb).  The previous behavior was to supply
1176       -L'' -- which caused noweb to use the "default" line numbering scheme.
1177       Thanks to Chris Richards for pointing this out (and supplying the fix).
1178    
1179    2. Once again, I reworked some aspects of the FFI:
1180    
1181       A. The incomplete/complete type business:
1182    
1183       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1184         gone!
1185       - ML types representing an incomplete type are now *equal* to
1186         ML types representing their corresponding complete types (just like
1187         in C).  This is still safe because ml-nlffigen will not generate
1188         RTTI for incomplete types, nor will it generate functions that
1189         require access to such RTTI.   But when ML code generated from both
1190         incomplete and complete versions of the C type meet, the ML types
1191         are trivially interoperable.
1192    
1193         NOTE:  These changes restore the full generality of the translation
1194         (which was previously lost when I eliminated functorization)!
1195    
1196       B. Enum types:
1197    
1198       - Structure C now has a type constructor "enum" that is similar to
1199         how the "su" constructor works.  However, "enum" is not a phantom
1200         type because each "T enum" has values (and is isomorphic to
1201         MLRep.Signed.int).
1202       - There are generic access operations for enum objects (using
1203         MLRep.Signed.int).
1204       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1205         * The structure contains the definition of type "mlrep" (the ML-side
1206         representation type of the enum).  Normally, mlrep is the same
1207         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1208         then mlrep will be defined as a datatype -- thus facilitating
1209         pattern matching on mlrep values.
1210         ("-ec" will be suppressed if there are duplicate values in an
1211          enumeration.)
1212         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1213         will be generated for each C enum constant xxx.
1214         * Conversion functions m2i and i2m convert between mlrep and
1215         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1216         * Coversion functions c and ml convert between mlrep and "tag enum".
1217         * Access functions (get/set) fetch and store mlrep values.
1218       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1219         enumerations are merged into one single enumeration represented by
1220         structure E_'.
1221    
1222    ----------------------------------------------------------------------
1223    Name: Allen Leung
1224    Date: 2002/02/25 04:45:00 EST
1225    Tag: leunga-20020225-cps-spill
1226    
1227    This is a new implementation of the CPS spill phase.
1228    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1229    In case of problems, replace it with the old file spill.sml
1230    
1231    The current compiler runs into some serious performance problems when
1232    constructing a large record.  This can happen when we try to compile a
1233    structure with many items.  Even a very simple structure like the following
1234    makes the compiler slow down.
1235    
1236        structure Foo = struct
1237           val x_1 = 0w1 : Word32.int
1238           val x_2 = 0w2 : Word32.int
1239           val x_3 = 0w3 : Word32.int
1240           ...
1241           val x_N = 0wN : Word32.int
1242        end
1243    
1244    The following table shows the compile time, from N=1000 to N=4000,
1245    with the old compiler:
1246    
1247    N
1248    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1249           MLRISC ra                               0.06u  0.00s  0.05g
1250              (spills = 0 reloads = 0)
1251           TOTAL                                   0.63u  0.07s  0.21g
1252    
1253    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1254           MLRISC ra                               5.68u  0.59s  3.93g
1255              (spills = 0 reloads = 0)
1256           TOTAL                                   14.71u  0.99s  4.81g
1257    
1258    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1259           MLRISC ra                               5.54u  0.65s  3.91g
1260              (spills = 543 reloads = 1082)
1261           TOTAL                                   65.40u  3.13s  6.00g
1262    
1263    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1264           MLRISC ra                               0.80u  0.10s  0.55g
1265              (spills = 42 reloads = 84)
1266           TOTAL                                   129.42u  5.10s  4.13g
1267    
1268    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1269           MLRISC ra                               2.69u  0.27s  1.38g
1270              (spills = 62 reloads = 124)
1271           TOTAL                                   682.48u  19.61s  13.99g
1272    
1273    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1274           MLRISC ra                               4.96u  0.27s  2.72g
1275              (spills = 85 reloads = 170)
1276           TOTAL                                   2375.26u  57.21s  48.00g
1277    
1278    As you can see the old cps spill module suffers from some serious
1279    performance problem.  But since I cannot decipher the old code fully,
1280    instead of patching the problems up, I'm reimplementing it
1281    with a different algorithm.  The new code is more modular,
1282    smaller when compiled, and substantially faster
1283    (O(n log n) time and O(n) space).  Timing of the new spill module:
1284    
1285    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1286          MLRISC ra                               0.25u  0.02s  0.15g
1287             (spills=1 reloads=3)
1288          TOTAL                                   7.74u  0.34s  1.62g
1289    
1290    Implementation details:
1291    
1292    As far as I can tell, the purpose of the CPS spill module is to make sure the
1293    number of live variables at any program point (the bandwidth)
1294    does not exceed a certain limit, which is determined by the
1295    size of the spill area.
1296    
1297    When the bandwidth is too large, we decrease the register pressure by
1298    packing live variables into spill records.  How we achieve this is
1299    completely different than what we did in the old code.
1300    
1301    First, there is something about the MLRiscGen code generator
1302    that we should be aware of:
1303    
1304    o MLRiscGen performs code motion!
1305    
1306       In particular, it will move floating point computations and
1307       address computations involving only the heap pointer to
1308       their use sites (if there is only a single use).
1309       What this means is that if we have a CPS record construction
1310       statement
1311    
1312           RECORD(k,vl,w,e)
1313    
1314       we should never count the new record address w as live if w
1315       has only one use (which is often the case).
1316    
1317       We should do something similar to floating point, but the transformation
1318       there is much more complex, so I won't deal with that.
1319    
1320    Secondly, there are now two new cps primops at our disposal:
1321    
1322     1. rawrecord of record_kind option
1323        This pure operator allocates some uninitialized storage from the heap.
1324        There are two forms:
1325    
1326         rawrecord NONE [INT n]  allocates a tagless record of length n
1327         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1328                                     and initializes the tag.
1329    
1330     2. rawupdate of cty
1331          rawupdate cty (v,i,x)
1332          Assigns to x to the ith component of record v.
1333          The storelist is not updated.
1334    
1335    We use these new primops for both spilling and increment record construction.
1336    
1337     1. Spilling.
1338    
1339        This is implemented with a linear scan algorithm (but generalized
1340        to trees).  The algorithm will create a single spill record at the
1341        beginning of the cps function and use rawupdate to spill to it,
1342        and SELECT or SELp to reload from it.  So both spills and reloads
1343        are fine-grain operations.  In contrast, in the old algorithm
1344        "spills" have to be bundled together in records.
1345    
1346        Ideally, we should sink the spill record construction to where
1347        it is needed.  We can even split the spill record into multiple ones
1348        at the places where they are needed.  But CPS is not a good
1349        representation for global code motion, so I'll keep it simple and
1350        am not attempting this.
1351    
1352     2. Incremental record construction (aka record splitting).
1353    
1354        Long records with many component values which are simulatenously live
1355        (recall that single use record addresses are not considered to
1356         be live) are constructed with rawrecord and rawupdate.
1357        We allocate space on the heap with rawrecord first, then gradually
1358        fill it in with rawupdate.  This is the technique suggested to me
1359        by Matthias.
1360    
1361        Some restrictions on when this is applicable:
1362        1. It is not a VECTOR record.  The code generator currently does not handle
1363           this case. VECTOR record uses double indirection like arrays.
1364        2. All the record component values are defined in the same "basic block"
1365           as the record constructor.  This is to prevent speculative
1366           record construction.
1367    
1368    ----------------------------------------------------------------------
1369    Name: Allen Leung
1370    Date: 2002/02/22 01:02:00 EST
1371    Tag: leunga-20020222-mlrisc-tools
1372    
1373    Minor bug fixes in the parser and rewriter
1374    
1375    ----------------------------------------------------------------------
1376    Name: Allen Leung
1377    Date: 2002/02/21 20:20:00 EST
1378    Tag: leunga-20020221-peephole
1379    
1380    Regenerated the peephole files.  Some contained typos in the specification
1381    and some didn't compile because of pretty printing bugs in the old version
1382    of 'nowhere'.
1383    
1384    ----------------------------------------------------------------------
1385    Name: Allen Leung
1386    Date: 2002/02/19 20:20:00 EST
1387    Tag: leunga-20020219-mlrisc-tools
1388    Description:
1389    
1390       Minor bug fixes to the mlrisc-tools library:
1391    
1392       1.  Fixed up parsing colon suffixed keywords
1393       2.  Added the ability to shut the error messages up
1394       3.  Reimplemented the pretty printer and fixed up/improved
1395           the pretty printing of handle and -> types.
1396       4.  Fixed up generation of literal symbols in the nowhere tool.
1397       5.  Added some SML keywords to to sml.sty
1398    
1399    ----------------------------------------------------------------------
1400    Name: Matthias Blume
1401    Date: 2002/02/19 16:20:00 EST
1402    Tag: blume-20020219-cmffi
1403    Description:
1404    
1405    A wild mix of changes, some minor, some major:
1406    
1407    * All C FFI-related libraries are now anchored under $c:
1408        $/c.cm      --> $c/c.cm
1409        $/c-int.cm  --> $c/internals/c-int.cm
1410        $/memory.cm --> $c/memory/memory.cm
1411    
1412    * "make" tool (in CM) now treats its argument pathname slightly
1413      differently:
1414        1. If the native expansion is an absolute name, then before invoking
1415           the "make" command on it, CM will apply OS.Path.mkRelative
1416           (with relativeTo = OS.FileSys.getDir()) to it.
1417        2. The argument will be passed through to subsequent phases of CM
1418           processing without "going native".  In particular, if the argument
1419           was an anchored path, then "make" will not lose track of that anchor.
1420    
1421    * Compiler backends now "know" their respective C calling conventions
1422      instead of having to be told about it by ml-nlffigen.  This relieves
1423      ml-nlffigen from one of its burdens.
1424    
1425    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1426    
1427    * Export C_DEBUG and C_Debug from $c/c.cm.
1428    
1429    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1430      subtyping relationship between incomplete pointers and their complete
1431      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1432      with the convention of instantiating 'o with ('t, 'c) obj whenever
1433      the pointer target type is complete.  In the incomplete case, 'o
1434      will be instantiated with some "'c iobj" -- a type obtained by
1435      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1436    
1437      Operations that work on both incomplete and complete pointer types are
1438      typed as taking an 'o ptr while operations that require the target to
1439      be known are typed as taking some ('t, 'c) obj ptr.
1440    
1441      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1442      where void is an eqtype without any values.  This makes it possible
1443      to work on voidptr values using functions meant to operate on light
1444      incomplete pointers.
1445    
1446    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1447      been vastly simplified.
1448    
1449    ----------------------------------------------------------------------
1450    Name: Matthias Blume
1451    Date: 2002/02/19 10:48:00 EST
1452    Tag: blume-20020219-pqfix
1453    Description:
1454    
1455    Applied Chris Okasaki's bug fix for priority queues.
1456    
1457    ----------------------------------------------------------------------
1458    Name: Matthias Blume
1459    Date: 2002/02/15 17:05:00
1460    Tag: Release_110_39
1461    Description:
1462    
1463    Last-minute retagging is becoming a tradition... :-(
1464    
1465    This is the working release 110.39.
1466    
1467    ----------------------------------------------------------------------
1468    Name: Matthias Blume
1469    Date: 2002/02/15 16:00:00 EST
1470    Tag: Release_110_39-orig
1471    Description:
1472    
1473    Working release 110.39.  New bootfiles.
1474    
1475    (Update: There was a small bug in the installer so it wouldn't work
1476    with all shells.  So I retagged. -Matthias)
1477    
1478    ----------------------------------------------------------------------
1479    Name: Matthias Blume
1480    Date: 2002/02/15 14:17:00 EST
1481    Tag: blume-20020215-showbindings
1482    Description:
1483    
1484    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1485    the latter can be useful for exploring what bindings are available at
1486    the interactive prompt.  (The first function returns only the list
1487    of symbols that are really bound, the second prints those but also the
1488    ones that CM's autoloading mechanism knows about.)
1489    
1490    ----------------------------------------------------------------------
1491    Name: Matthias Blume
1492    Date: 2002/02/15 12:08:00 EST
1493    Tag: blume-20020215-iptrs
1494    Description:
1495    
1496    Two improvements to ml-nlffigen:
1497    
1498      1. Write files only if they do not exist or if their current contents
1499         do not coincide with what's being written.  (That is, avoid messing
1500         with the time stamps unless absolutely necessary.)
1501    
1502      2. Implement a "repository" mechanism for generated files related
1503         to "incomplete pointer types".   See the README file for details.
1504    
1505    ----------------------------------------------------------------------
1506    Name: Matthias Blume
1507    Date: 2002/02/14 11:50:00 EST
1508    Tag: blume-20020214-quote
1509    Description:
1510    
1511    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1512    because of the new and improved tag generation scheme.  (Thanks to Allen
1513    Leung for pointing it out.)
1514    
1515    ----------------------------------------------------------------------
1516    Name: Lal George
1517    Date: 2002/02/14 09:55:27 EST 2002
1518    Tag: george-20020214-isabelle-bug
1519    Description:
1520    
1521    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1522    of Isabelle on the x86.
1523    
1524    From Allen:
1525    -----------
1526     I've found the problem:
1527    
1528         in ra-core.sml, I use the counter "blocked" to keep track of the
1529         true number of elements in the freeze queue.  When the counter goes
1530         to zero, I skip examining the queue.  But I've messed up the
1531         bookkeeping in combine():
1532    
1533             else ();
1534             case !ucol of
1535               PSEUDO => (if !cntv > 0 then
1536                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1537                                        ^^^^^^^^^^^^^^^^^^^^^^^
1538                      moveu := mergeMoveList(!movev, !moveu)
1539                     )
1540                  else ();
1541    
1542         combine() is called to coalesce two nodes u and v.
1543         I think I was thinking that if the move counts of u and v are both
1544         greater than zero then after they are coalesced then one node is
1545         removed from the freeze queue.  Apparently I was thinking that
1546         both u and v are of low degree, but that's clearly not necessarily true.
1547    
1548    
1549    02/12/2002:
1550        Here's the patch.  HOL now compiles.
1551    
1552        I don't know how this impact on performance (compile
1553        time or runtime).  This bug caused the RA (especially on the x86)
1554        to go thru the potential spill phase when there are still nodes on the
1555        freeze queue.
1556    
1557    
1558    
1559    
1560    ----------------------------------------------------------------------
1561    Name: Matthias Blume
1562    Date: 2002/02/13 22:40:00 EST
1563    Tag: blume-20020213-fptr-rtti
1564    Description:
1565    
1566    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1567    updates.
1568    
1569    ----------------------------------------------------------------------
1570    Name: Matthias Blume
1571    Date: 2002/02/13 16:41:00 EST
1572    Tag: blume-20020213-cmlpq
1573    Description:
1574    
1575    Added new priority queue export symbols (which have just been added to
1576    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1577    not compile and the installer would choke.)
1578    
1579    ----------------------------------------------------------------------
1580    Name: Matthias Blume
1581    Date: 2002/02/13 16:15:00 EST
1582    Tag: blume-20020213-various
1583    Description:
1584    
1585    1. More tweaks to ml-nlffigen:
1586    
1587       - better internal datastructures (resulting in slight speedup)
1588       - "-match" option requires exact match
1589       - "localized" gensym counters (untagged structs/unions nested within
1590         other structs/unions or within typedefs get a fresh counter; their
1591         tag will be prefixed by a concatenation of their parents' tags)
1592       - bug fixes (related to calculation of transitive closure of types
1593         to be included in the output)
1594    
1595    2. Minor Basis updates:
1596    
1597       - added implementations for List.collate and Option.app
1598    
1599    ----------------------------------------------------------------------
1600    Name: Matthias Blume
1601    Date: 2002/02/11 15:55:00 EST
1602    Tag: blume-20020211-gensym
1603    Description:
1604    
1605    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1606    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1607    names (ML structure names that correspond to unnamed C structs, unions,
1608    and enums), so that separate runs of ml-nlffigen do not clash.
1609    
1610    ----------------------------------------------------------------------
1611    Name: Matthias Blume
1612    Date: 2002/02/11 12:05:00 EST
1613    Tag: blume-20020211-gensml
1614    Description:
1615    
1616    A quick fix for a problem with GenSML (in the pgraph-util library):
1617    Make generation of toplevel "local" optional.  (Strictly speaking,
1618    signature definitions within "local" are not legal SML.)
1619    
1620    Other than that: updates to INSTALL and cm/TODO.
1621    
1622    ----------------------------------------------------------------------
1623    Name: Matthias Blume
1624    Date: 2002/02/08 15:00:00 EST
1625    Tag: blume-20020208-uniquepid
1626    Description:
1627    
1628    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1629    
1630    1. The installer (config/install.sh) has gotten smarter:
1631    
1632         - Configuration options are a bit easier to specify now
1633           (in config/targets).
1634         - Bug in recognizing .tar.bz2 files fixed.
1635         - Installer automatically resolves dependencies between
1636           configuration options (e.g., if you ask for eXene, you will
1637           also get cml -- regardless whether you asked for it or not).
1638         - Installer can run in "quieter mode" by setting the environment
1639           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1640           "completely silent", though.
1641         - Build HashCons library as part of smlnj-lib.
1642    
1643    2. A new scheme for assigning persistent identifiers to compilation
1644       units (and, by extension, to types etc.) has been put into place.
1645       This fixes a long-standing bug where types and even dynamic values
1646       can get internally confused, thereby compromising type safety
1647       (abstraction) and dynamic correctness.  See
1648    
1649         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1650    
1651       for an example of how things could go wrong until now.
1652    
1653       The downside of the new scheme is that pids are not quite as
1654       persistent as they used to be: CM will generate a fresh pid
1655       for every compilation unit that it thinks it sees for the first
1656       time.  That means that if you compile starting from a clean, fresh
1657       source tree at two different times, you end up with different
1658       binaries.
1659    
1660       Cutoff recompilation, however, has not been compromised because
1661       CM keeps pid information in special caches between runs.
1662    
1663    ----------------------------------------------------------------------
1664    Name: Lal George
1665    Date: 2002/02/07 15:34:13 EST 2002
1666    Tag: <none>
1667    Description:
1668    
1669    Compilers that generate assembly code may produce  global labels
1670    whose value is resolved at link time. The various peephole optimization
1671    modules did not take this in account.
1672    
1673    TODO. The Labels.addrOf function should really return an option
1674    type so that clients are forced to deal with this issue, rather
1675    than an exception being raised.
1676    
1677    ----------------------------------------------------------------------
1678    Name: Lal George
1679    Date: 2002/02/06 13:55:02 EST
1680    Tag: george-20020206-ra-breakup
1681    Description:
1682    
1683    1. A bug fix from Allen.
1684    
1685        A typo causes extra fstp %st(0)'s to be generated at compensation
1686        edges, which might cause stack underflow traps at runtime.  This
1687        occurs in fft where there are extraneous fstps right before the 'into'
1688        trap instruction (in this case they are harmless since none of the
1689        integers overflow.)
1690    
1691    2. Pulled out various utility modules that were embedded in the modules
1692       of the register allocator. I need these modules for other purposes, but
1693       they are not complete enough to put into a library (just yet).
1694    ----------------------------------------------------------------------
1695    Name: Matthias Blume
1696    Date: 2002/01/31 16:05:00 EST
1697    Tag: blume-20020131-sparc-ccalls
1698    Description:
1699    
1700    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1701       of extra stack space by mistake.  Fixed.
1702    
1703    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1704    
1705    ----------------------------------------------------------------------
1706    Name: Allen Leung
1707    Date: 2002/01/30
1708    Tag: leunga-20020130-nowhere-bug-fix
1709    Description:
1710    
1711       MLRISC bug fixes:
1712       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1713       2. MachineInt.fromString was negating its value.
1714    
1715    ----------------------------------------------------------------------
1716    Name: Matthias Blume
1717    Date: 2002/01/29
1718    Tag: blume-20020129-INSTALL
1719    Description:
1720    
1721    - Added somewhat detailed installation instructions (file INSTALL).
1722    - Fixed curl-detection bug in config/install.sh.
1723    - It is now possible to select the URL getter using the URLGETTER
1724      environment variable:
1725    
1726          not set / "unknown"      --> automatic detection (script tries wget,
1727                                       curl, and lynx)
1728          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1729                                       how to properly invoke them)
1730          other                    --> use $URLGETTER directly, it must take
1731                                       precisely two command-line arguments
1732                                       (source URL and destination file name)
1733    
1734    ----------------------------------------------------------------------
1735    Name: Matthias Blume
1736    Date: 2002/01/28
1737    Tag: blume-20020128-sparc-ccalls
1738    Description:
1739    
1740    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1741    - Make use of the allocParam argument in sparc-c-calls.
1742    
1743    ----------------------------------------------------------------------
1744    Name: Matthias Blume
1745    Date: 2002/01/28
1746    Tag: blume-20020128-allocParam
1747    Description:
1748    
1749    John Reppy:  Changes c-calls API to accept client-callback for
1750    allocating extra stack space.
1751    me: Corresponding changes to mlriscGen (using a dummy argument that
1752        does not change the current behavior).
1753    
1754    ----------------------------------------------------------------------
1755    Name: Matthias Blume
1756    Date: 2002/01/28 12:00:00
1757    Tag: Release_110_38
1758    Description:
1759    
1760    This time for real!!!
1761    
1762    ----------------------------------------------------------------------
1763    Name: Matthias Blume
1764    Date: 2002/01/28 10:56:00 EST
1765    Tag: blume-20020128-retraction
1766    Description:
1767    
1768    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1769       with blume-Release_110_38-retracted.)
1770    
1771    1. Fixed a problem with incorrect rounding modes in real64.sml.
1772       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1773    
1774    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1775       and enums fixed.  The naming of corresponding ML identifiers should
1776       now be consistent again.
1777    
1778    ----------------------------------------------------------------------
1779    Name: Allen Leung
1780    Date: 2002/01/27
1781    Tag: leunga-20020127-nowhere
1782    Description:
1783    
1784       Added a target called nowhere in the configuration scripts.
1785       Enabling this will build the MLRISC 'nowhere' tool (for translating
1786       programs with where-clauses into legal SML code) during installation.
1787    
1788    ----------------------------------------------------------------------
1789    Name: Matthias Blume
1790    Date: 2002/01/25 21:27:00 EST
1791    Tag: blume-Release_110_38-retracted
1792    Description:
1793    
1794    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1795    
1796    README will be added later.
1797    
1798    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1799    (Release_110_38) removed.  Reason: Last-minute bug fixes.
1800    
1801    ----------------------------------------------------------------------
1802    Name: Matthias Blume
1803    Date: 2002/01/25
1804    Tag: blume-20020125-ffi
1805    Description:
1806    
1807    A large number of tweaks and improvements to ml-nlffi-lib and
1808    ml-nlffigen:
1809    
1810       - ML represenation types have been streamlined
1811       - getter and setter functions work with concrete values, not abstract
1812         ones where possible
1813       - ml-nlffigen command line more flexible (see README file there)
1814       - some bugs have been fixed (hopefully)
1815    
1816    ----------------------------------------------------------------------
1817    Name: Lal George
1818    Date: 2002/01/24
1819    Tag: george-20020124-risc-ra-interface
1820    Description:
1821    
1822       There is a dramatic simplification in the interface to the
1823       register allocator for RISC architectures as a result of making
1824       parallel copy instructions explicit.
1825    
1826    ----------------------------------------------------------------------
1827    Name: Matthias Blume
1828    Date: 2002/01/22
1829    Tag: blume-20020122-x86-ccalls
1830    Description:
1831    
1832    Bug fix for c-calls on x86 (having to do with how char- and
1833    short-arguments are being handled).
1834    
1835    ----------------------------------------------------------------------
1836    Name: Matthias Blume
1837    Date: 2002/01/21
1838    Tag: blume-20020121-ff
1839    Description:
1840    
1841    Another day of fiddling with the FFI...
1842    
1843    1. Bug fix/workaround:  CKIT does not complain about negative array
1844       dimensions, so ml-nlffigen has to guard itself against this possibility.
1845       (Otherwise a negative dimension would send it into an infinite loop.)
1846    
1847    2. Some of the abstract types (light objects, light pointers, most "base"
1848       types) in structure C are now eqtypes.
1849    
1850    3. Added constructors and test functions for NULL function pointers.
1851    
1852    ----------------------------------------------------------------------
1853    Name: Matthias Blume
1854    Date: 2002/01/18
1855    Tag: blume-20020118-ready-for-new-release
1856    Description:
1857    
1858    Made config/srcarchiveurl point to a new place.  (Will provide boot
1859    files shortly.)
1860    
1861    Maybe we christen this to be 110.38?
1862    
1863    ----------------------------------------------------------------------
1864    Name: Matthias Blume
1865    Date: 2002/01/18
1866    Tag: blume-20020118-more-ffifiddle
1867    Description:
1868    
1869    Today's FFI fiddling:
1870    
1871      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
1872        C.Set.* that use concrete (MLRep.*) arguments and results instead
1873        of abstract ones.
1874    
1875      - Provided word-style bit operations etc. for "int" representation
1876        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1877        and Long.
1878    
1879    ----------------------------------------------------------------------
1880    Name: Matthias Blume
1881    Date: 2002/01/18
1882    Tag: blume-20020118-use-x86-fp
1883    Description:
1884    
1885    Now that x86-fast-fp seems to be working, I turned it back on again
1886    by default.  (Seems to work fine now, even with the FFI.)
1887    
1888    Other than that, I added some documentation about the FFI to
1889    src/ml-nlffigen/README and updated the FFI test examples in
1890    src/ml-nlffi-lib/Tests/*.
1891    
1892    ----------------------------------------------------------------------
1893    Name: Allen Leung
1894    Date: 2002/01/17
1895    Tag: leunga-20020117-x86-fast-fp-call
1896    Description:
1897    
1898       1. Fixed a problem with handling return fp values when x86's fast fp
1899          mode is turned on.
1900    
1901       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
1902          of %f32.
1903    
1904       3. Added a constructor INT32lit to the ast of MLRISC tools.
1905    
1906    ----------------------------------------------------------------------
1907    Name: Matthias Blume
1908    Date: 2002/01/16
1909    Tag: blume-20020116-ffifiddle
1910    Description:
1911    
1912    More fiddling with the FFI interface:
1913    
1914     - Make constness 'c instead of rw wherever possible.  This eliminates
1915       the need for certain explicit coercions.  (However, due to ML's
1916       value polymorphism, there will still be many cases where explicit
1917       coercions are necessary.  Phantom types are not the whole answer
1918       to modeling a subtyping relationship in ML.)
1919    
1920     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
1921    
1922     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
1923       clumsy workaround for varargs problem.)
1924    
1925    ----------------------------------------------------------------------
1926    Name: Lal George
1927    Date: 2002/01/15
1928    Tag: <none>
1929    Description:
1930    
1931    1. Since COPY instructions are no longer native to the architecture,
1932       a generic functor can be used to implement the expandCopies function.
1933    
1934    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1935       TEXT segment.
1936    
1937    ----------------------------------------------------------------------
1938    Name: Matthias Blume
1939    Date: 2002/01/15
1940    Tag: blume-20020115-ffiupdates
1941    Description:
1942    
1943    1. Fix for bug resulting in single-precision float values being returned
1944       incorrectly from FFI calls.
1945    
1946    2. Small modifications to C FFI API:
1947    
1948        - memory-allocation routines return straight objects (no options)
1949          and raise an exception in out-of-memory situations
1950        - unsafe extensions to cast between function pointers and pointers
1951          from/to ints
1952        - added structure C_Debug as an alternative to structure C where
1953          pointer-dereferencing (|*| and |*!) always check for null-pointers
1954        - added open_lib' to DynLinkage;  open_lib' works like open_lib
1955          but also takes a (possibly empty) list of existing library handles
1956          that the current library depends on
1957    
1958    ----------------------------------------------------------------------
1959    Name: Matthias Blume
1960    Date: 2002/01/10
1961    Tag: blume-20020110-newffigen
1962    Description:
1963    
1964    1. Updates to portable graph code.
1965    
1966    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
1967       more scalable now so that even huge interfaces such as the one
1968       for GTK compile in finite time and space. :-)
1969       See src/ml-nlffigen/README for details on what's new.
1970    
1971    ----------------------------------------------------------------------
1972    Name: Lal George
1973    Date: 2001/01/09 14:31:35 EST 2002
1974    Tag: george-20011206-rm-native-copy
1975    Description:
1976    
1977            Removed the native COPY and FCOPY instructions
1978            from all the architectures and replaced it with the
1979            explicit COPY instruction from the previous commit.
1980    
1981            It is now possible to simplify many of the optimizations
1982            modules that manipulate copies. This has not been
1983            done in this change.
1984    
1985    ----------------------------------------------------------------------
1986    Name: Lal George
1987    Date: 2001/12/06 16:50:13 EST 2001
1988    Tag: george-20011206-mlrisc-instruction
1989    Description:
1990    
1991    Changed the representation of instructions from being fully abstract
1992    to being partially concrete. That is to say:
1993    
1994      from
1995            type instruction
1996    
1997      to
1998            type instr                              (* machine instruction *)
1999    
2000            datatype instruction =
2001                LIVE of {regs: C.cellset, spilled: C.cellset}
2002              | KILL of {regs: C.cellset, spilled: C.cellset}
2003              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
2004              | ANNOTATION of {i: instruction, a: Annotations.annotation}
2005              | INSTR of instr
2006    
2007    This makes the handling of certain special instructions that appear on
2008    all architectures easier and uniform.
2009    
2010    LIVE and KILL say that a list of registers are live or killed at the
2011    program point where they appear. No spill code is generated when an
2012    element of the 'regs' field is spilled, but the register is moved to
2013    the 'spilled' (which is present, more for debugging than anything else).
2014    
2015    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
2016    We used to generate:
2017    
2018            DEFFREG f1
2019            f1 := f2 + f3
2020            trapb
2021    
2022    but now generate:
2023    
2024            f1 := f2 + f3
2025            trapb
2026            LIVE {regs=[f1,f2,f3], spilled=[]}
2027    
2028    Furthermore, the DEFFREG (hack) required that all floating point instruction
2029    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
2030    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
2031    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
2032    this architecture).
2033    
2034    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
2035    all the architectures. This will result in further simplification of the
2036    register allocator that must be aware of them for coalescing purposes, and
2037    will also simplify certain aspects of the machine description that provides
2038    callbacks related to parallel copies.
2039    
2040    ANNOTATION should be obvious, and now INSTR represents the honest to God
2041    machine instruction set!
2042    
2043    The <arch>/instructions/<arch>Instr.sml files define certain utility
2044    functions for making porting easier -- essentially converting upper case
2045    to lower case. All machine instructions (of type instr) are in upper case,
2046    and the lower case form generates an MLRISC instruction. For example on
2047    the alpha we have:
2048    
2049      datatype instr =
2050         LDA of {r:cell, b:cell, d:operand}
2051       | ...
2052    
2053      val lda : {r:cell, b:cell, d:operand} -> instruction
2054        ...
2055    
2056    where lda is just (INSTR o LDA), etc.
2057    
2058    ----------------------------------------------------------------------
2059    Name: Matthias Blume
2060    Date: 2001/11/22 21:40:00 EST
2061    Tag: Release_110_37
2062    Description:
2063    
2064    Release 110.37.  This time for real.
2065    
2066    ----------------------------------------------------------------------
2067    Name: Matthias Blume
2068    Date: 2001/11/21 16:35:00 EST
2069    Tag: blume-20011121-foot-in-mouth
2070    Description:
2071    
2072    Removed the "Release_110_37" tag because of a serious bug.
2073    This will be re-tagged once the bug is fixed.
2074    
2075    ----------------------------------------------------------------------
2076    Name: Matthias Blume
2077    Date: 2001/11/21 16:14:00 EST
2078    Tag: blume-20011121-forgottenfile
2079    Description:
2080    
2081    Forgot to add a file.  (Just a .tex-file -- part of
2082    the CM manual source.)
2083    
2084    ----------------------------------------------------------------------
2085    Name: Matthias Blume
2086    Date: 2001/11/21 16:10:00 EST
2087    Tag: blume-20011121-invalid_110_37
2088    Description:
2089    
2090    Note: I removed the original tag "Release_110_37" from this commit
2091          because we found a serious bug in all non-x86 backends.
2092          - Matthias
2093    
2094    1. Modifications to the SML/NJ code generator and to the runtime system
2095       so that code object name strings are directly inserted into code
2096       objects at code generation time.  The only business the runtime system
2097       has with this is now to read the name strings on occasions.
2098       (The encoding of the name string has also changed somewhat.)
2099    
2100    2. CM now implements a simple "set calculus" for specifying export lists.
2101       In particular, it is now possible to refer to the export lists of
2102       other libraries/groups/sources and form unions as well as differences.
2103       See the latest CM manual for details.
2104    
2105    3. An separate notion of "proxy" libraries has again be eliminated from
2106       CM's model.  (Proxy libraries are now simply a special case of using
2107       the export list calculus.)
2108    
2109    4. Some of the existing libraries now take advantage of the new set
2110       calculus.
2111       (Notice that not all libraries have been converted because some
2112       of the existing .cm-files are supposed to be backward compatible
2113       with 110.0.x.)
2114    
2115    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2116       "General.exnMessage"!  The former relies on a certain hook to be
2117       initialized, and that often does not happen in the stand-alone case.)
2118    
2119    ----------------------------------------------------------------------
2120    Name: Lal George
2121    Date: 2001/11/21  13:56:18 EST
2122    Tag: george-2001121-pseudo-ops
2123    Description:
2124    
2125      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2126      ought to never be any question of incompatabilities with
2127      pseudo-op syntax expected by host assemblers.
2128    
2129      For now, only modules supporting GAS syntax are implemented
2130      but more should follow, such as MASM, and vendor assembler
2131      syntax, e.g. IBM as, Sun as, etc.
2132    
2133    ----------------------------------------------------------------------
2134    Name: Matthias Blume
2135    Date: 2001/11/14 11:52:00 EST
2136    Tag: blume-20011114-srcname
2137    Description:
2138    
2139    1. Routed the name of the current source file to mlriscgen where it
2140       should be directly emitted into the code object.  (This last part
2141       is yet to be done.)
2142    
2143    2. Some cleanup of the pgraph code to make it match the proposal that
2144       I put out the other day.  (The proposal notwithstanding, things are
2145       still in flux here.)
2146    
2147    ----------------------------------------------------------------------
2148    Name: Lal George
2149    Date: 2001/11/14 09:44:04 EST
2150    Tag:
2151    Description:
2152    
2153      Fix for a backpatching bug reported by Allen.
2154    
2155      Because the boundary between short and long span-dependent
2156      instructions is +/- 128, there are an astounding number of
2157      span-dependent instructions whose size is over estimated.
2158    
2159      Allen came up with the idea of letting the size of span
2160      dependent instructions be non-monotonic, for a maxIter
2161      number of times, after which the size must be monotonically
2162      increasing.
2163    
2164      This table shows the number of span-dependent instructions
2165      whose size was over-estimated as a function of maxIter, for the
2166      file Parse/parse/ml.grm.sml:
2167    
2168         maxIter            # of instructions:
2169            10                      687
2170            20                      438
2171            30                      198
2172            40                        0
2173    
2174      In compiling the compiler, there is no significant difference in
2175      compilation speed between maxIter=10 and maxIter=40. Actually,
2176      my measurements showed that maxIter=40 was a tad faster than
2177      maxIter=10! Also 96% of the  files in the compiler reach a fix
2178      point within 13 iterations, so fixing maxIter at 40, while high,
2179      is okay.
2180    
2181    ----------------------------------------------------------------------
2182    Name: Matthias Blume
2183    Date: 2001/10/31 15:25:00 EST
2184    Tag: blume-20011031-pgraph
2185    Description:
2186    
2187    CKIT:
2188    * Changed the "Function" constructor of type Ast.ctype to carry optional
2189      argument identifiers.
2190    * Changed the return type of TypeUtil.getFunction accordingly.
2191    * Type equality ignores the argument names.
2192    * TypeUtil.composite tries to preserve argument names but gives up quickly
2193      if there is a mismatch.
2194    
2195    installation script:
2196    * attempts to use "curl" if available (unless "wget" is available as well)
2197    
2198    CM:
2199    * has an experimental implementation of "portable graphs" which I will
2200      soon propose as an implementation-independent library format
2201    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2202    
2203    NLFFI-LIB:
2204    * some cleanup (all cosmetic)
2205    
2206    NLFFIGEN:
2207    * temporarily disabled the mechanism that suppresses ML output for
2208      C definitions whose identifiers start with an underscore character
2209    * generate val bindings for enum constants
2210    * user can request that only one style (light or heavy) is being used;
2211      default is to use both (command-line arguments: -heavy and -light)
2212    * fixed bug in handling of function types involving incomplete pointers
2213    * generate ML entry points that take record arguments (i.e., using
2214      named arguments) for C functions that have a prototype with named
2215      arguments
2216      (see changes to CKIT)
2217    
2218    ----------------------------------------------------------------------
2219    Name: Allen Leung
2220    Date: 2001/10/27 20:34:00 EDT
2221    Tag: leunga-20011027-x86-fast-fp-call
2222    Description:
2223    
2224       Fixed the bug described in blume-20010920-slowfp.
2225    
2226       The fix involves
2227          1. generating FCOPYs in FSTP in ia32-svid
2228          2. marking a CALL with the appropriate annotation
2229    
2230    ----------------------------------------------------------------------
2231    Name: Matthias Blume
2232    Date: 2001/10/16 11:32:00 EDT
2233    Tag: blume-20011016-netbsd
2234    Description:
2235    
2236    Underscore patch from Chris Richards (fixing problem with compiling
2237    runtime system under recent NetBSD).
2238    
2239    ----------------------------------------------------------------------
2240    Name: Allen Leung
2241    Date: 2001/10/12 17:18:32 EDT 2001
2242    Tag: leung-20011012-x86-printflowgraph
2243    Description:
2244    
2245    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2246    
2247    ----------------------------------------------------------------------
2248    Name: Lal George
2249    Date: 2001/10/11 23:51:34 EDT
2250    Tag: george-20011011-too-many-instrs
2251    Description:
2252    
2253    The representation of a program point never expected to see more
2254    than 65536 instructions in a basic block!
2255    
2256    ----------------------------------------------------------------------
2257    Name: Lal George
2258    Date: 2001/10/09 09:41:37 EDT
2259    Tag: george-20011008-mlrisc-labels
2260    Description:
2261    
2262    Changed the machine description files to support printing of
2263    local and global labels in assembly code, based on host assembler
2264    conventions.
2265    
2266    ----------------------------------------------------------------------
2267    Name: Matthias Blume
2268    Date: 2001/09/25 15:25:00 EDT
2269    Tag: blume-20010925-exninfo
2270    Description:
2271    
2272    I provided a non-hook implementation of exnName (at the toplevel) and
2273    made the "dummy" implementation of exnMessage (at the toplevel) more
2274    useful: if nothing gets "hooked in", then at least you are going to
2275    see the exception name and a message indicating why you don't see more.
2276    
2277    [For the time being, programs that need exnMessage and want to use
2278    ml-build should either use General.exnMessage (strongly recommended) or
2279    refer to structure General at some other point so that CM sees a
2280    static dependency.]
2281    
2282    [Similar remarks go for "print" and "use":  If you want to use their
2283    functionality in stand-alone programs generated by ml-build, then use
2284    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2285    
2286    ----------------------------------------------------------------------
2287    Name: Matthias Blume
2288    Date: 2001/09/20 17:28:00 EDT
2289    Tag: blume-20010920-slowfp
2290    Description:
2291    
2292    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2293    
2294    ----------------------------------------------------------------------
2295    Name: Matthias Blume
2296    Date: 2001/09/20 17:20:00 EDT
2297    Tag: blume-20010920-canonicalpaths
2298    Description:
2299    
2300    0. Updated the BOOT file (something that I forgot to do earlier).
2301    
2302    1. Small internal change to CM so that it avoids "/../" in filenames
2303       as much as possible (but only where it is safe).
2304    
2305    2. Changed config/_run-sml (resulting in a changed bin/.run-sml) so
2306       that arguments that contain delimiters are passed through correctly.
2307       This change also means that all "special" arguments of the form
2308       @SMLxxx... must come first.
2309    
2310    3. Changed install script to put relative anchor names for tool commands
2311       into pathconfig.
2312    
2313    ----------------------------------------------------------------------
2314    Name: Matthias Blume
2315    Date: 2001/09/18 15:35:00 EDT
2316    Tag: blume-20010918-readme11036
2317    Description:
2318    
2319    Added README files.
2320    
2321    ----------------------------------------------------------------------
2322    Name: Matthias Blume
2323    Date: 2001/09/18 11:45:00 EDT
2324    Tag: Release_110_36 (retag)
2325    Description:
2326    
2327    Fixed mistake in config/preloads. Retagged as 110.36.
2328    
2329    ----------------------------------------------------------------------
2330    Name: Matthias Blume
2331    Date: 2001/09/18 09:40:00 EDT
2332    Tag: Release_110_36_orig (tag changed)
2333    Description:
2334    
2335    New version (110.36).  New bootfiles.
2336    
2337    ----------------------------------------------------------------------
2338    Name: Matthias Blume
2339    Date: 2001/09/14 16:15:00 EDT
2340    Tag: blume-20010914-x86fastfp
2341    Description:
2342    
2343    John committed some changes that Allen made, in particular a (hopefully)
2344    correctly working version of the x86-fp module.
2345    
2346    I changed the default setting of the Control.MLRISC.getFlag "x86-fast-fp"
2347    flag to "true".  Everything seems to compile to a fixpoint ok, and
2348    "mandelbrot" speeds up by about 15%.
2349    
2350    ----------------------------------------------------------------------
2351    Name: Matthias Blume
2352    Date: 2001/09/13 11:20:00 EDT
2353    Tag: blume-20010913-minimal
2354    Description:
2355    
2356    1. Stefan Monnier's patch to fix a miscompilation problem that
2357       was brought to light by John Reppy's work on Moby.
2358    
2359    2. Implemented a minimal "structure Compiler" that contains just
2360       "version" and "architecture".  The minimal version will be
2361       available when the full version is not.  This is for backward-
2362       compatibility with code that wants to test Compiler.version.
2363    
2364    ----------------------------------------------------------------------
2365    Name: Matthias Blume
2366    Date: 2001/08/28 14:03:00 EDT
2367    Tag: blume-20010828-ml-lex
2368    Description:
2369    
2370    Fix for bug 1581, received from Neophytos Michael.
2371    
2372    ----------------------------------------------------------------------
2373    Name: Matthias Blume
2374    Date: 2001/08/27 11:20:00 EDT
2375    Tag: blume-20010827-readme11035
2376    Description:
2377    
2378    Fleshed out the README file for 110.35.
2379    
2380    ----------------------------------------------------------------------
2381    Name: Matthias Blume
2382    Date: 2001/08/24 17:10:00 EDT
2383    Tag: Release_110_35
2384    Description:
2385    
2386    New version number (110.35).  New bootfiles.
2387    
2388    ----------------------------------------------------------------------
2389    Name: Lal George
2390    Date: 2001/08/24 13:47:18 EDT 2001
2391    Tag: george-20010824-MLRISC-graphs
2392    Description:
2393    
2394     removed clusters from MLRISC completely and replaced with graphs.
2395    
2396    ----------------------------------------------------------------------
2397    Name: Matthias Blume
2398    Date: 2001/08/23 17:50:00 EDT
2399    Tag: blume-20010823-toplevel
2400    Description:
2401    
2402    - some reorganization of the code that implements various kinds of
2403      environments in the compiler (static, dynamic, symbolic, combined)
2404    - re-implemented the EnvRef module so that evalStream works properly
2405      (if the stream contains references to "use", "CM.make", etc.)
2406    - cleaned up evalloop.sml and interact.sml (but they need more cleaning)
2407    
2408    ----------------------------------------------------------------------
2409    Name: Matthias Blume
2410    Date: 2001/08/20 15:50 EDT
2411    Tag: blume20010820-slipup
2412    Description:
2413    
2414    I forgot to commit a few files.  Here they are...
2415    
2416    ----------------------------------------------------------------------
2417    Name: Matthias Blume
2418    Date: 2001/08/20 15:35:00 EDT
2419    Tag: blume-20010820-debugprof
2420    Description:
2421    
2422    !!!! NEW BOOTFILES !!!!
2423    
2424    This is another round of reorganizing the compiler sources.  This
2425    time the main goal was to factor out all the "instrumentation"
2426    passes (for profiling and backtracing) into their own library.
2427    The difficulty was to do it in such a way that it does not depend
2428    on elaborate.cm but only on elabdata.cm.
2429    
2430    Therefore there have been further changes to both elaborate.cm and
2431    elabdata.cm -- more "generic" things have been moved from the former
2432    to the latter.  As a result, I was forced to split the assignment
2433    of numbers indicating "primtyc"s into two portions: SML-generic and
2434    SML/NJ-specific.  Since it would have been awkward to maintain,
2435    I bit the bullet and actually _changed_ the mapping between these
2436    numbers and primtycs.  The bottom line of this is that you need
2437    a new set of bin- and bootfiles.
2438    
2439    I have built new bootfiles for all architectures, so doing a fresh
2440    checkout and config/install.sh should be all you need.
2441    
2442    The newly created library's name is
2443    
2444        $smlnj/viscomp/debugprof.cm
2445    
2446    and its sources live under
2447    
2448        src/compiler/DebugProf
2449    
2450    ----------------------------------------------------------------------
2451    Name: Matthias Blume
2452    Date: 2001/08/15 17:15:00 EDT
2453    Tag: blume-20010815-compreorg
2454    Description:
2455    
2456    This is a first cut at reorganizing the CM libraries that make up the
2457    core of the compiler.  The idea is to separate out pieces that could
2458    be used independently by tools, e.g., the parser, the typechecker, etc.
2459    
2460    The current status is a step in this direction, but it is not quite
2461    satisfactory yet.  Expect more changes in the future.
2462    
2463    Here is the current (new) organization...
2464    
2465        What used to be $smlnj/viscomp/core.cm is now divided into
2466        six CM libraries:
2467    
2468             $smlnj/viscomp/basics.cm
2469                           /parser.cm
2470                           /elabdata.cm
2471                           /elaborate.cm
2472                           /execute.cm
2473                           /core.cm
2474    
2475        The CM files for these libraries live under src/system/smlnj/viscomp.
2476        All these libraries are proxy libraries that contain precisely
2477        one CM library component.  Here are the locations of the components
2478        (all within the src/compiler tree):
2479    
2480             Basics/basics.cm
2481             Parse/parser.cm
2482             ElabData/elabdata.cm
2483             Elaborator/elaborate.cm
2484             Execution/execute.cm
2485             core.cm
2486    
2487         [This organization is the same that has been used already
2488         for a while for the architecture-specific parts of the visible
2489         compiler and for the old version of core.cm.]
2490    
2491         As you will notice, many source files have been moved from their
2492         respective original locations to a new home in one of the above
2493         subtrees.
2494    
2495         The division of labor between the new libraries is the following:
2496    
2497             basics.cm:
2498                - Simple, basic definitions that pertain to many (or all) of
2499                  the other libraries.
2500             parser.cm:
2501                - The SML parser, producing output of type Ast.dec.
2502                - The type family for Ast is also defined and exported here.
2503             elabdata.cm:
2504                - The datatypes that describe input and output of the elaborator.
2505                  This includes types, absyn, and static environments.
2506             elaborator.cm:
2507                - The SML/NJ type checker and elaborator.
2508                  This maps an Ast.dec (with a given static environment) to
2509                  an Absyn.dec (with a new static environment).
2510                - This libraries implements certain modules that used to be
2511                  structures as functors (to remove dependencies on FLINT).
2512             execute.cm:
2513                - Everything having to do with executing binary code objects.
2514                - Dynamic environments.
2515             core.cm:
2516                - SML/NJ-specific instantiations of the elaborator and MLRISC.
2517                - Top-level modules.
2518                - FLINT (this should eventually become its own library)
2519    
2520    Notes:
2521    
2522    I am not 100% happy with the way I separated the elaborator (and its
2523    data structures) from FLINT.  Two instances of the same problem:
2524    
2525        1. Data structures contain certain fields that carry FLINT-specific
2526           information.  I hacked around this using exn and the property list
2527           module from smlnj-lib.  But the fact that there are middle-end
2528           specific fields around at all is a bit annoying.
2529    
2530        2. The elaborator calculates certain FLINT-related information.  I tried
2531           to make this as abstract as I could using functorization, but, again,
2532           the fact that the elaborator has to perform calculations on behalf
2533           of the middle-end at all is not nice.
2534    
2535        3. Having to used exn and property lists is unfortunate because it
2536           weakens type checking.  The other alternative (parameterizing
2537           nearly *everything*) is not appealing, though.
2538    
2539    I removed the "rebinding =" warning hack because due to the new organization
2540    it was awkward to maintain it.  As a result, the compiler now issues some of
2541    these warnings when compiling init.cmi during bootstrap compilation. On
2542    the plus side, you also get a warning when you do, for example:
2543       val op = = Int32.+
2544    which was not the case up to now.
2545    
2546    I placed "assign" and "deref" into the _Core structure so that the
2547    code that deals with the "lazy" keyword can find them there.  This
2548    removes the need for having access to the primitive environment
2549    during elaboration.
2550    
2551    ----------------------------------------------------------------------
2552    Name: Matthias Blume
2553    Date: 2001/08/13
2554    Tag: blume-20010813-closures
2555    Description:
2556    
2557    This fix was sent to us by Zhong Shao.  It is supposed to improve the
2558    performance of certain loops by avoiding needless closure allocation.
2559    
2560    ----------------------------------------------------------------------
2561    Name: Lal George
2562    Date: 2001/07/31 10:03:23 EDT 2001
2563    Tag: george-20010731-x86-fmalloc
2564    Description: Fixed bug in x86 calls
2565    
2566        There was a bug where call instructions would mysteriously
2567        vanish. The call instruction had to be one that returned
2568        a floating point value.
2569    
2570    ----------------------------------------------------------------------
2571    Name: Lal George
2572    Date: 2001/07/19 16:36:29 EDT 2001
2573    Tag: george-20010719-simple-cells
2574    Description:
2575    
2576    I have dramatically simplified the interface for CELLS in MLRISC.
2577    
2578    In summary, the cells interface is broken up into three parts:
2579    
2580      1. CellsBasis : CELLS_BASIS
2581    
2582            CellsBasis is a top level structure and common for all
2583            architectures.  it contains the definitions of basic datatypes
2584            and utility  functions over these types.
2585    
2586      2. functor Cells() : CELLS
2587    
2588            Cells generates an interface for CELLS that incorporates the
2589            specific resources on the target architecture, such as the
2590            presence of special register classes, their number and size,
2591            and various useful substructures.
2592    
2593      3. <ARCH>CELLS
2594    
2595            e.g. SparcCells: SPARCCELLS
2596    
2597            <ARCH>CELLS usually contains additional bindings for special
2598            registers  on the architecture, such as:
2599    
2600                    val r0 : cell           (* register zero *)
2601                    val y : cell            (* Y register *)
2602                    val psr : cell          (* processor status register *)
2603                    ...
2604    
2605            The structure returned by applying the Cells functor is opened
2606            in this interface.
2607    
2608    The main implication of all this is that the datatypes for cells is
2609    split between CellsBasis and CELLS -- a fairly simple change for user
2610    code.
2611    
2612    In the old scheme the CELLS interface had a definitional binding of
2613    the form:
2614    
2615            signature CELLS = sig
2616    
2617               structure CellsBasis = CellsBasis
2618    
2619               ...
2620    
2621            end
2622    
2623    With all the sharing constraints that goes on in MLRISC, this old
2624    design  quickly leads to errors such as:
2625    
2626            "structure definition spec inside of sharing ... "
2627    
2628    
2629    and appears to require an unacceptable amount of sharing and where
2630    constraint hackery.
2631    
2632    I think this error message (the interaction of definitional specs and
2633    sharing) requires more explanation on our web page.
2634    
2635    ----------------------------------------------------------------------
2636    Name: Matthias Blume
2637    Date: 2001/07/19 15:00:00 EDT
2638    Tag: blume-20010719-libreorg
2639    Description:
2640    
2641    This update puts together a fairly extensive but straightforward change
2642    to the way the libraries that implement the interactive system are
2643    organized:
2644    
2645       The biggest change is the elimination of structure Compiler.  As a
2646       replacement for this structure, there is now a CM library
2647       (known as $smlnj/compiler.cm or $smlnj/compiler/current.cm)
2648       that exports all the substructures of the original structure Compiler
2649       directly.  So instead of saying Compiler.Foo.bar one now simply
2650       says Foo.bar.  (The CM libraries actually export a collection of
2651       structures that is richer than the collection of substructures of
2652       structure Compiler.)
2653    
2654       To make the transition smooth, there is a separate library called
2655       $smlnj/compiler/compiler.cm which puts together and exports the
2656       original structure Compiler (or at least something very close to it).
2657    
2658       There are five members of the original structure Compiler
2659       that are not exported directly but which instead became members
2660       of a new structure Backend (described by signature BACKEND).  These are:
2661       structure Profile (: PROFILE), structure Compile (: COMPILE), structure
2662       Interact (: INTERACT), structure Machine (: MACHINE), and val
2663       architecture (: string).
2664    
2665       Structure Compiler.Version has become structure CompilerVersion.
2666    
2667       Cross-compilers for alpha32, hppa, ppc, sparc, and x86 are provided
2668       by $smlnj/compiler/<arch>.cm where <arch> is alpha32, hppa, ppc, sparc,
2669       or x86, respectively.
2670       Each of these exports the same frontend structures that
2671       $smlnj/compiler.cm exports.  But they do not have a structure Backend
2672       and instead export some structure <Arch>Backend where <Arch> is Alpha32,
2673       Hppa, PPC, Sparc, or X86, respectively.
2674    
2675       Library $smlnj/compiler/all.cm exports the union of the exports of
2676       $smlnj/compiler/<arch>.cm
2677    
2678       There are no structures <Arch>Compiler anymore, use
2679       $smlnj/compiler/<arch>.cm instead.
2680    
2681       Library host-compiler-0.cm is gone.  Instead, the internal library
2682       that instantiates CM is now called cm0.cm.  Selection of the host
2683       compiler (backend) is no longer done here but. (Responsibility for it
2684       now lies with $smlnj/compiler/current.cm.  This seems to be more
2685       logical.)
2686    
2687       Many individual files have been moved or renamed.  Some files have
2688       been split into multiple files, and some "dead" files have been deleted.
2689    
2690    Aside from these changes to library organization, there are also changes
2691    to the way the code itself is organized:
2692    
2693       Structure Binfile has been re-implemented in such a way that it no
2694       longer needs any knowledge of the compiler.  It exclusively deals
2695       with the details of binfile layout.  It no longer invokes the
2696       compiler (for the purpose of creating new prospective binfile
2697       content), and it no longer has any knowledge of how to interpret
2698       pickles.
2699    
2700       Structure Compile (: COMPILE) has been stripped down to the bare
2701       essentials of compilation.  It no longer deals with linking/execution.
2702       The interface has been cleaned up considerably.
2703    
2704       Utility routines for dealing with linking and execution have been
2705       moved into their own substructures.
2706    
2707       (The ultimate goal of these changes is to provide a light-weight
2708       binfile loader/linker (at least for, e.g., stable libraries) that
2709       does not require CM or the compiler to be present.)
2710    
2711    CM documentation has been updated to reflect the changes to library
2712    organization.
2713    
2714    ----------------------------------------------------------------------
2715    Name: Matthias Blume
2716    Date: 2001/07/10 17:30:00 EDT
2717    Tag: Release_110_34
2718    Description:
2719    
2720    Minor tweak to 110.34 (re-tagged):
2721    
2722      - README.html file added to CVS repository
2723      - runtime compiles properly under FreeBSD 3.X and 4.X
2724    
2725    ----------------------------------------------------------------------
2726    Name: Matthias Blume
2727    Date: 2001/07/10 17:30:00 EDT
2728    Tag: Release_110_34
2729    Description:
2730    
2731    New version number (110.34). New bootfiles.
2732    
2733    ----------------------------------------------------------------------
2734    Name: Matthias Blume
2735    Date: 2001/07/09 16:00:00 EDT
2736    Tag: blume-20010709-more-varargs
2737    Description:
2738    
2739    I changed the handling of varargs in ml-nlffigen again:
2740    The ellipsis ... will now simply be ignored (with an accompanying warning).
2741    
2742    The immediate effect is that you can actually call a varargs function
2743    from ML -- but you can't actually supply any arguments beyond the ones
2744    specified explicitly.  (For example, you can call printf with its format
2745    string, but you cannot pass additional arguments.)
2746    
2747    This behavior is only marginally more useful than the one before, but
2748    it has the advantage that a function or, more importantly, a function
2749    type never gets dropped on the floor, thus avoiding follow-up problems with
2750    other types that refer to the offending one.
2751    
2752    ----------------------------------------------------------------------
2753    Name: Matthias Blume
2754    Date: 2001/07/09 11:25:00 EDT
2755    Tag: blume-20010709-varargs
2756    Description:
2757    
2758    1. ckit-lib.cm now exports structure Error
2759    2. ml-nlffigen reports occurences of "..." (i.e., varargs function types)
2760       with a warning accompanied by a source location.  Moreover, it
2761       merely skips the offending function or type and proceeds with the
2762       rest of its work.u  As a result, one can safely feed C code containing
2763       "..." to ml-nlffigen.
2764    3. There are some internal improvements to CM, providing slightly
2765       more general string substitutions in the tools subsystem.
2766    
2767    ----------------------------------------------------------------------
2768    Name: Matthias Blume
2769    Date: 2001/06/27 15:10:00 EDT
2770    Tag: blume-20010627-concur
2771    Description:
2772    
2773    Fixed a small bug in CM's handling of parallel compilation.
2774    (You could observe the bug by Control-C-interrupting an ordinary
2775    CMB.make or CM.stabilize and then attaching some compile servers.
2776    The result was that all of a sudden the previously interrupted
2777    compilation would continue on its own.  This was because of
2778    an over-optimization: CM did not bother to clean out certain queues
2779    when no servers were attached "anyway", resulting in the contents
2780    of these queues to grab control when new servers did get attached.)
2781    
2782    There is also another minor update to the CM manual.
2783    
2784    ----------------------------------------------------------------------
2785    Name: Matthias Blume
2786    Date: 2001/06/26 16:15:00 EDT
2787    Tag: blume-20010626-cmdoc
2788    Description:
2789    
2790    Minor typo fixed in CM manual (syntax diagram for libraries).
2791    
2792    ----------------------------------------------------------------------
2793    Name: Matthias Blume
2794    Date: 2001/06/25 22:55:00 EDT
2795    Tag: blume-20010625-x86pc
2796    Description:
2797    
2798    Fixed a nasty bug in the X86 assembly code that caused signal
2799    handlers to fail (crash) randomly.
2800    
2801    ----------------------------------------------------------------------
2802    Name: Matthias Blume
2803    Date: 2001/06/25 12:05:00 EDT
2804    Tag: blume-20010625-nlffigen
2805    Description:
2806    
2807    This update fixes a number of minor bugs in ml-nlffigen as reported by
2808    Nick Carter <nbc@andrew.cmu.edu>.
2809    
2810      1. Silly but ok typedefs of the form "typedef void myvoid;" are now accepted.
2811      2. Default names for generated files are now derived from the name of
2812         the C file *without its directory*.  In particular, this causes generated
2813         files to be placed locally even if the C file is in some system directory.
2814      3. Default names for generated signatures and structures are also derived
2815         from the C file name without its directory.  This avoids silly things
2816         like "structure GL/GL".
2817         (Other silly names are still possible because ml-nlffigen does not do
2818          a thorough check of whether generated names are legal ML identifiers.
2819          When in doubt, use command line arguments to force particular names.)
2820    
2821    ----------------------------------------------------------------------
2822    Name: Matthias Blume
2823    Date: 2001/06/21 12:25:00 EDT
2824    Tag: blume-20010621-eXene
2825    Description:
2826    
2827    eXene now compiles and (sort of) works again.
2828    
2829    The library name (for version > 110.33) is $/eXene.cm.
2830    
2831    I also added an new example in src/eXene/examples/nbody.  See the
2832    README file there for details.
2833    
2834    ----------------------------------------------------------------------
2835    Name: Matthias Blume
2836    Date: 2001/06/20 16:40:00 EDT
2837    Tag: blume-20010620-cml
2838    Description:
2839    
2840    CML now compiles and works again.
2841    
2842    Libraries (for version > 110.33):
2843    
2844      $cml/cml.cm            Main CML library.
2845      $cml/basis.cm          CML's version of $/basis.cm.
2846      $cml/cml-internal.cm   Internal helper library.
2847      $cml/core-cml.cm       Internal helper library.
2848      $cml-lib/trace-cml.cm  Tracing facility.
2849      $cml-lib/smlnj-lib.cm  CML's version of $/smlnj-lib.cm
2850    
2851    The installer (config/install.sh) has been taught how to properly
2852    install this stuff.
2853    
2854    ----------------------------------------------------------------------
2855    Name: Matthias Blume
2856    Date: 2001/06/19 17:55:00 EDT
2857    Tag: blume-20010619-instantiate
2858    Description:
2859    
2860    This un-breaks the fix for bug 1432.
2861    (The bug was originally fixed in 110.9 but I broke it again some
2862    time after that.)
2863    
2864    ----------------------------------------------------------------------
2865    Name: Matthias Blume
2866    Date: 2001/06/19 17:25:00 EDT
2867    Tag: blume-20010619-signals
2868    Description:
2869    
2870    This should (hopefully) fix the long-standing signal handling bug.
2871    (The runtime system was constructing a continuation record with an
2872    incorrect descriptor which would cause the GC to drop data on the floor...)
2873    
2874    ----------------------------------------------------------------------
2875    Name: Matthias Blume
2876    Date: 2001/06/15 15:05:00 EDT
2877    Tag: blume-20010615-moresparc
2878    Description:
2879    
2880    Here is a short late-hour update related to Sparc c-calls:
2881    
2882     -- made handling of double-word arguments a bit smarter
2883    
2884     -- instruction selection phase tries to collapse certain clumsily
2885        constructed ML-Trees; typical example:
2886    
2887            ADD(ty,ADD(_,e,LI d1),LI d2)  ->  ADD(ty,e,LI(d1+d2))
2888    
2889        This currently has no further impact on SML/NJ since mlriscGen does
2890        not seem to generate such patterns in the first place, and c-calls
2891        (which did generate them in the beginning) has meanwhile been fixed
2892        so as to avoid them as well.
2893    
2894    ----------------------------------------------------------------------
2895    Name: Matthias Blume
2896    Date: 2001/06/15 15:05:00 EDT
2897    Tag: blume-20010615-sparc
2898    Description:
2899    
2900    The purpose of this update is to provide an implementation of NLFFI
2901    on Sparc machines.
2902    
2903    Here are the changes in detail:
2904    
2905       * src/MLRISC/sparc/c-calls/sparc-c-calls.sml is a new file containing
2906       the Sparc implementation of the c-calls API.
2907       * The Sparc backend of SML/NJ has been modified to uniformely use %fp
2908       for accessing the ML frame.  Thus, we have a real frame pointer and
2909       can freely modify %sp without need for an omit-frame-ptr phase.
2910       The vfp logic in src/compiler/CodeGen/* has been changed to accomodate
2911       this case.
2912       * ml-nlffigen has been taught to produce code for different architectures
2913       and calling conventions.
2914       * In a way similar to what was done in the x86 case, the Sparc
2915       backend uses its own specific extension to mltree.  (For example,
2916       it needs to be able to generate UNIMP instructions which are part
2917       of the calling convention.)
2918       * ml-nlffi-lib was reorganized to make it more modular (in particular,
2919       to make it easier to plug in new machine- and os-dependent parts).
2920    
2921    There are some other fairly unrelated bug fixes and cleanups as well:
2922    
2923       * I further hacked the .cm files for MLRISC tools (like MDLGen) so
2924       that they properly share their libraries with existing SML/NJ libraries.
2925       * I fixed a minor cosmetic bug in CM, supressing certain spurious
2926       follow-up error messages.
2927       * Updates to CM/CMB documentation.
2928    
2929    TODO items:
2930    
2931       * MLRISC should use a different register as its asmTemp on the Sparc.
2932         (The current %o2 is a really bad choice because it is part of the
2933         calling conventions, so things might interfere in unexpected ways.)
2934    
2935    ----------------------------------------------------------------------
2936    Name: Matthias Blume
2937    Date: 2001/06/07
2938    Tag: blume-20010607-calls
2939    Description:
2940    
2941    A number of internal changes related to C calls and calling conventions:
2942    
2943    1. ML-Tree CALL statements now carry a "pops" field.  It indicates the
2944       number of bytes popped implicitly (by the callee).  In most cases
2945       this field is 0 but on x86/win32 it is some non-zero value.  This
2946       is information provided for the benefit of the "omit-frameptr" pass.
2947    2. The CALL instruction on the x86 carries a similar "pops" field.
2948       The instruction selection phase copies its value from the ML-Tree
2949       CALL statement.
2950    3. On all other architectures, the instruction selection phase checks
2951       whether "pops=0" and complains if not.
2952    4. The c-calls implementation for x86 now accepts two calling conventions:
2953       "ccall" and "stdcall".  When "ccall" is selected, the caller cleans
2954       up after the call and pops is set to 0.  For "stdcall", the caller
2955       does nothing, leaving the cleanup to the callee; pops is set to
2956       the number of bytes that were pushed onto the stack.
2957    5. The cproto decoder (compiler/Semant/types/cproto.sml) now can
2958       distinguish between "ccall" and "stdcall".
2959    6. The UNIMP instruction has been added to the supported Sparc instruction
2960       set. (This is needed for implementing the official C calling convention
2961       on this architecture.)
2962    7. I fixed some of the .cm files under src/MLRISC/Tools to make them
2963       work with the latest CM.
2964    
2965    ----------------------------------------------------------------------
2966    Name: Matthias Blume
2967    Date: 2001/06/05 15:10:00 EDT
2968    Tag: blume-20010605-cm-index
2969    Description:
2970    
2971    0. The "lambdasplit" parameter for class "sml" in CM has been documented.
2972    
2973    1. CM can now generate "index files".  These are human-readable files
2974       that list on a per-.cm-file basis each toplevel symbol defined or
2975       imported.  The location of the index file for
2976       <p>/<d>.cm is <p>/CM/INDEX/<d>.cm.
2977       To enable index-file generation, set CM.Control.generate_index to true
2978       or export an environment-symbol: export CM_GENERATE_INDEX=true.
2979    
2980       The CM manual has been updated accordingly.
2981    
2982    2. I made some slight modifications to the c-calls API in MLRISC.
2983    
2984         a) There is now a callback to support saving/restoring of
2985            dedicated but caller-save registers around the actual call
2986            instruction.
2987         b) One can optionally specify a comment-annotation for the
2988            call instruction.
2989    
2990    3. SML/NJ (mlriscGen.sml) uses this new API for the rawccall primop.
2991       (For example, the comment annotation shows the C prototype of
2992        the function being called.)
2993    
2994    ----------------------------------------------------------------------
2995    Name: Matthias Blume
2996    Date: 2001/06/01 13:30:00 EDT
2997    Tag: blume-20010601-nlffi-cleanup
2998    Description:
2999    
3000    This is mostly a cleanup of MLFFI stuff:
3001    
3002       - some signature files have been put into a more exposed place
3003       - the ugly 'f type parameter is gone (simplifies types tremendously!)
3004       - ml-nlffigen changed accordingly
3005       - tutorial updated
3006    
3007    Other changes:
3008    
3009       - author's affiliation in CM manual(s) updated
3010       - some more recognized keywords added to Allen's sml.sty
3011    
3012    ----------------------------------------------------------------------
3013    Name: Matthias Blume
3014    Date: 2001/05/25 15:30:00 EDT
3015    Tag: blume-20010525-iptr
3016    Description:
3017    
3018      - put the official 110.33-README (as it appears on the ftp server) under
3019        CVS
3020      - fixed a small bug related to incomplete pointer types in
3021        ml-nlffigen
3022      - small cosmetic change to the ml-nlffi-lib's "arr" type constructor
3023        (it does not need the 'f type parameter)
3024    
3025    ----------------------------------------------------------------------
3026    Name: Matthias Blume
3027    Date: 2001/05/23 14:30:00 EDT
3028    Tag: Release_110_33
3029    Description:
3030    
3031    New version number (110.33).  New bootfiles.
3032    
3033    ----------------------------------------------------------------------
3034    Name: Matthias Blume
3035    Date: 2001/05/22 18:06:00 EDT
3036    Tag: blume-20010522-targets
3037    Description:
3038    
3039    Made install.sh use file config/targets.customized if it exists, falling
3040    back to config/targets if it doesn't.  This way one can have a customized
3041    version of the targets file without touching the "real thing", thus
3042    eliminating the constant fear of accidentally checking something bogus
3043    back into the CVS repository...  (File config/targets.customized must
3044    not be added to the repository!)
3045    
3046    ----------------------------------------------------------------------
3047    Name: Matthias Blume
3048    Date: 2001/05/22 16:30:00 EDT
3049    Tag: blume-20010522-minitut
3050    Description:
3051    
3052    1. Bug fix in ml-nlffigen; now (hopefully) correctly handling
3053       struct returns.
3054    2. Added src/ml-nlffi-lib/Doc/mini-tutorial.txt.  This is some very
3055       incomplete, preliminary documentation for NLFFI.
3056    
3057    ----------------------------------------------------------------------
3058    Name: Matthias Blume
3059    Date: 2001/05/14 11:30:00 EDT
3060    Tag: blume-20010514-script
3061    Description:
3062    
3063    Some bugs in install script fixed.
3064    
3065    In addition to that I also made a slight change to the NLFFI API:
3066    Functors generated by ml-nlffigen now take the dynamic library as a
3067    straight functor argument, not as a suspended one.  (The original
3068    functor code used to force the suspension right away anyway, so there
3069    was nothing gained by this complication of the interface.)
3070    
3071    ----------------------------------------------------------------------
3072    Name: Matthias Blume
3073    Date: 2001/05/11 14:35:00 EDT
3074    Tag: blume-20010511-ml-nlffi
3075    Description:
3076    
3077    I finally took the plunge and added my new FFI code to the main
3078    repository.  For x86-linux it is now ready for prime-time.
3079    
3080    There are two new subdirectories of "src":
3081    
3082      - ml-nlffi-lib:
3083           The utility library for programs using the FFI interface.
3084           Here is the implementation of $/c.cm and its associated low-level
3085           partners $/c-int.cm and $/memory.cm.
3086      - ml-nlffigen:
3087           A stand-alone program for generating ML glue code from C source
3088           code.
3089    
3090    Building ml-nlffigen requires $/ckit-lib.cm.
3091    
3092    The config/install.sh script has been updates to do the Right Thing
3093    (hopefully).
3094    
3095    Notice that the source tree for the C-Kit will not be put under "src"
3096    but directly under the installation root directory.  (This is the
3097    structure that currently exists on the CVS server when you check out
3098    module "sml".)  Fortunately, config/install.sh knows about this oddity.
3099    
3100    Bugs: No documentation yet.
3101    
3102    ----------------------------------------------------------------------
3103    Name: Matthias Blume
3104    Date: 2001/05/09 16:35:00 EDT
3105    Tag: blume-20010509-cpscontract
3106    Description:
3107    
3108    Fixed a bug in the accounting code in cpsopt/contract.sml.  (The
3109    wrapper/unwrapper elimination did not decrement usage counts and some
3110    dead variables got overlooked by the dead-up logic.)
3111    
3112    ----------------------------------------------------------------------
3113    Name: Lal George
3114    Date: 2001/05/08  17:26:09 EDT
3115    Tag: george-20010508-omit-frameptr
3116    Description:
3117    
3118    Changes to implement the omit-frame-pointer optimization to support
3119    raw C calls. For now, there is only support on the Intel x86, but
3120    other architectures will follow as more experience is gained with this.
3121    
3122    
3123    ----------------------------------------------------------------------
3124    Name: Matthias Blume
3125    Date: 2001/05/07 14:40:00 EDT
3126    Tag: blume-20010507-proxies
3127    Description:
3128    
3129    I made into "proxy libraries" all libraries that qualify for such a
3130    change.  (A qualifying library is a library that has another library or
3131    groups as its sole member and repeats that member's export list
3132    verbatim.  A proxy library avoids this repetition by omitting its export
3133    list, effectively inheriting the list that its (only) member exports.
3134    See the CM manual for more explanation.)
3135    The main effect is that explicit export lists for these libraries
3136    do not have to be kepts in sync, making maintenance a bit easier.
3137    
3138    I also added copyright notices to many .cm-files.
3139    
3140    Last but not least, I made a new set of bootfiles.
3141    
3142    ----------------------------------------------------------------------
3143    Name: Matthias Blume
3144    Date: 2001/05/04 17:00:00 EDT
3145    Tag: blume-20010504-cm-lsplit
3146    Description:
3147    
3148    0. John merged pending changes to $/smlnj-lib.cm
3149    
3150    1. Allen's previous change accidentally backed out of one of Lal's
3151       earlier changes.  I undid this mistake (re-introducing Lal's change).
3152    
3153    2. I used the new topOrder' function from graph-scc.sml (from $/smlnj-lib.cm)
3154       within the compiler where applicable.  There is some code simplification
3155       because of that.
3156    
3157    3. The "split" phase (in FLINT) is now part of the default list of phases.
3158       Compiler.Control.LambdaSplitting.* can be used to globally control the
3159       lambda-splitting (cross-module-inlining) engine.  In addition to that,
3160       it can now also be controlled on a per-source basis: CM has been taught
3161       a new tool parameter applicable to ML source files.
3162    
3163       - To turn lambda-splitting off completely:
3164            local open Compiler.Control.LambdaSplitting in
3165                val _ = set Off
3166            end
3167       - To make "no lambda-splitting" the global default (but allow per-source
3168         overriding); this is the initial setting:
3169            local open Compiler.Control.LambdaSplitting in
3170                val _ = set (Default NONE)
3171            end
3172       - To make "lambda-splitting with aggressiveness a" the global default
3173         (and allow per-source overriding):
3174            local open Compiler.Control.LambdaSplitting in
3175                val _ = set (Default (SOME a))
3176            end
3177    
3178       - To turn lambda-splitting off for a given ML souce file (say: a.sml)
3179         write (in the respective .cm-file):
3180            a.sml (lambdasplitting:off)
3181       - To turn lambda-splitting for a.sml on with minimal aggressiveness:
3182            a.sml (lambdasplitting:on)
3183       - To turn lambda-splitting for a.sml on with aggressiveness <a> (where
3184         <a> is a decimal non-negative integer):
3185            a.sml (lambdasplitting:<a>)
3186       - To turn lambda-splitting for a.sml on with maximal aggressiveness:
3187            a.sml (lambdasplitting:infinity)
3188       - To use the global default for a.sml:
3189            a.sml (lambdasplitting:default)
3190         or simply
3191            a.sml
3192    
3193    ----------------------------------------------------------------------
3194    Name: Allen Leung
3195    Date: 2001/05/04 01:57:00 EDT
3196    Tag: leunga-20010504-sync
3197    Description:
3198    
3199      MLRISC features.
3200    
3201      1. Fix to CMPXCHG instructions.
3202      2. Changed RA interface to allow annotations in callbacks.
3203      3. Added a new method to the stream interface to allow annotations updates.
3204    
3205    ----------------------------------------------------------------------
3206    Name: Matthias Blume
3207    Date: 2001/05/01 11:45:00 EDT
3208    Tag: blume-20010501-pcedittmp
3209    Description:
3210    
3211    Changed install.sh to use the current working directory instead of
3212    /usr/tmp for a temporary file (pcedittmp).  The previous choice
3213    of /usr/tmp caused trouble with MacOS X because of file premission
3214    problems.
3215    
3216    ----------------------------------------------------------------------
3217    Name: Matthias Blume
3218    Date: 2001/04/20 11:10:00 EDT
3219    Tag: blume-20010420-inMLflag
3220    Description:
3221    
3222     - added vp_limitPtrMask to vproc-state.h
3223       (for use by the raw-C-calls mechanism to implement proper interrupt
3224        handling)
3225     - made the ML compiler aware of various data-structure offsets so it
3226       can generate code for accessing the vp_inML flag and vp_limitPtrMask
3227     - tweaked mlriscGen.sml to have it emit interrupt-handling code for
3228       raw C-calls
3229    
3230    ----------------------------------------------------------------------
3231    Name: Lal George
3232    Date: 2001/04/20 09:15:28 EDT
3233    Tag: george-20010420-macosX
3234    Description:
3235    
3236     - Changes to port to Mac OS X; Darwin.
3237    
3238     - In the process I found that sqrt was broken on the PPC, because the
3239       fsqrt instruction is not implemented.
3240    
3241    ----------------------------------------------------------------------
3242    Name: Matthias Blume
3243    Date: 2001/04/18 12:45:00 EDT
3244    Tag: blume-20010418-ccalls
3245    Description:
3246    
3247     - fixed two off-by-4 errors in the x86-specific c-calls implementation
3248       (this bug prevented structure arguments containing pointers from being
3249        passed correctly)
3250     - changed the raw-C-call code in mlriscGen.sml in such a way that
3251       structure arguments are represented as a pointer to the beginning
3252       of the structure (instead of having a series of synthesized arguments,
3253       one for each structure member)
3254    
3255     - made makeml script's verbosity level configurable via environment
3256       variable (MAKEML_VERBOSITY)
3257    
3258     - eliminated placeholder implementations for f32l, w16s, i16s, and f32s
3259       in rawmem-x86.sml; we are now using the real thing
3260    
3261    ----------------------------------------------------------------------
3262    Name: Matthias Blume
3263    Date: 2001/03/22 16:25:00 EST
3264    Tag: blume-20010322-bootfiles
3265    Description:
3266    
3267    Created a new set of bootfiles (for your automatic installation convenience).
3268    
3269    ----------------------------------------------------------------------
3270    Name: Matthias Blume
3271    Date: 2001/03/22 15:10:00 EST
3272    Tag: blume-20010322-rawmem-parcm
3273    Description:
3274    
3275    1. All "raw memory access" primitives for the new FFI are implemented now
3276       (at least on the x86).
3277    2. Some further cleanup of CM's parallel make mechanism.
3278    
3279    ----------------------------------------------------------------------
3280    Name: Matthias Blume
3281    Date: 2001/03/19 17:53:00 EST
3282    Tag: blume-20010319-parallel
3283    Description:
3284    
3285    Parallel make (using compile servers) now works again.
3286    
3287    To this end, CM.stabilize and CMB.make have been modified to work in
3288    two passes when compile servers are attached:
3289       1. Compile everything, do not perform stabilization; this pass
3290          uses compile servers
3291       2. Stabilize everything; this pass does not use compile servers
3292    If there are no compile servers, the two passes are combined into one
3293    (as before).  Splitting the passes increases the inherent parallelism
3294    in the dependency graph because the entire graph including all
3295    libraries is available at the same time.  This, in turn, improves
3296    server utilization.  The downside is that the master process will
3297    have to do some extra work after compilation is done (because for
3298    technical reasons it must re-read all the binfiles during stabilization).
3299    
3300    ----------------------------------------------------------------------
3301    Name: Matthias Blume
3302    Date: 2001/03/16 12:22:00 EST
3303    Tag: blume-20010316-bootfiles
3304    Description:
3305    
3306    Created a new set of bootfiles (for your automatic installation convenience).
3307    
3308    ----------------------------------------------------------------------
3309    Name: Matthias Blume
3310    Date: 2001/03/16 11:00:00 EST
3311    Tag: blume-20010316-MLTREE-fixup
3312    Description:
3313    
3314    This is a minor fixup for an (untagged) earlier commit by Allen.
3315    (A file was missing).
3316    
3317    ----------------------------------------------------------------------
3318    Name: Allen Leung
3319    Date: Mon Mar  5 18:54:57 EST 2001
3320    Tag: leunga-20010305-cut-support
3321    
3322    1. New support for alternative control-flow in MLTREE.
3323       Currently we support
3324    
3325          FLOW_TO(CALL ...., [k1,...,kn])
3326    
3327       This is needed for 'cuts to' in C-- and try/handle-like constructs
3328       in Moby
3329    
3330       New assembler flag "asm-show-cutsto" to turn on control-flow debugging.
3331    
3332    2. Register Allocator
3333    
3334       Changes in interface [from Fermin, John]
3335    
3336    3. Alpha 8-bit SLL support [Fermin]
3337    
3338    4. All architectures
3339    
3340       A new module (ClusterExpandCopies) for expanding parallel copies.
3341    
3342    ----------------------------------------------------------------------
3343    Name: Allen Leung
3344    Date: 2001/02/27 23:07:00 EST
3345    Tag: leunga-20010227-minor-stuff
3346    
3347    1. Alpha bug fix for CMOVNE
3348    2. Handle mltree COND(..,FCMP ...,...)
3349    3. Bug fix in simplifier
3350    
3351    ----------------------------------------------------------------------
3352    Name: Matthias Blume
3353    Date: 2001/01/30 17:50:00 EST
3354    Tag: blume-20010130-sync
3355    Description:
3356    
3357    This is just a minor update to sync my devel branch with the main brach.
3358    The only visible change is the addition of some README files.
3359    
3360    ----------------------------------------------------------------------
3361    Name: Matthias Blume
3362    Date: 2001/01/12 23:30:00 JST
3363    Tag: blume-20010112-bootfiles
3364    Description:
3365    
3366    Made a new set of bootfiles that goes with the current state of the
3367    repository.
3368    
3369    ----------------------------------------------------------------------
3370    Name: Matthias Blume
3371    Date: 2001/01/12 21:20:00 JST
3372    Tag: blume-20010112-sync
3373    Description:
3374    
3375    I am just flushing out some minor changes that had accumulated in
3376    my private branch in order to sync with the main tree.  (This is
3377    mainly because I had CVS trouble when trying to merge _into_ my
3378    private branch.)
3379    
3380    Most people should be completely unaffected by this.
3381    
3382    ----------------------------------------------------------------------
3383    Name: Allen Leung
3384    Date: Thu Jan 11 21:03:00 EST 2001
3385    Tag: leunga-20010111-labexp=mltree
3386    Description:
3387    
3388    1.  Removed the type LabelExp and replace it by MLTree.
3389    2.  Rewritten mltree-simplify with the pattern matcher tool.
3390    3.  There were some bugs in alpha code generator which would break
3391        64-bit code generation.
3392    4.  Redo the tools to generate code with the
3393    5.  The CM files in MLRISC (and in src/system/smlnj/MLRISC)
3394        are now generated by perl scripts.
3395    
3396    ----------------------------------------------------------------------
3397    Name: Matthias Blume
3398    Date: 2001/01/10 21:55:00 JST
3399    Tag: blume-20010110-rcc
3400    Description:
3401    
3402    The RCC stuff now seems to work (but only on the x86).
3403    This required hacking of the c-calls interface (and -implementation) in
3404    MLRISC.
3405    
3406    Normal compiler users should be unaffected.
3407    
3408    ----------------------------------------------------------------------
3409    Name: Matthias Blume
3410    Date: 2001/01/09 01:20:00 JST
3411    Tag: blume-20010109-rcc
3412    Description:
3413    
3414    This is a fairly big patch, flushing out a large number of pending
3415    changes that I made to my development copy over the last couple of days.
3416    
3417    Of practical relevance at this moment is a workaround for a pickling
3418    bug that Allen ran into the other day.  The cause of the bug itself is
3419    still unknown and it might be hard to fix it properly, but the
3420    workaround has some merits of its own (namely somewhat reducing pickling
3421    overhead for certain libraries).  Therefore, I think this solution should
3422    be satisfactory at this time.
3423    
3424    The rest of the changes (i.e., the vast majority) has to do with my
3425    ongoing efforts of providing direct support for C function calls from
3426    ML.  At the moment there is a new primop "RAW_CCALL", typing magic
3427    in types/cproto.sml (invoked from FLINT/trans/translate.sml), a new
3428    case in the FLINT CPS datatype (RCC), changes to cps/convert.sml to
3429    translate uses of RAW_CCALL into RCC, and changes to mlriscGen.sml to
3430    handle RCC.
3431    
3432    The last part (the changes to mlriscGen.sml) are still known to be
3433    wrong on the x86 and not implemented on all other architectures.  But
3434    the infrastructure is in place. I had to change a few functor
3435    signatures in the backend to be able to route the CCalls interface
3436    from MLRISC there, and I had to specialize the mltree type (on the
3437    x86) to include the necessary extensions. (The extensions themselves
3438    were already there and redy to go in MLRISC/x86).
3439    
3440    Everything should be very happy as soon as someone helps me with
3441    mlriscGen.sml...
3442    
3443    In any case, nothing of this should matter to anyone as long as the
3444    new primop is not being used (which is going to be the case unless you
3445    find it where I hid it :). The rest of the compiler is completely
3446    unaffected.
3447    
3448    ----------------------------------------------------------------------
3449    Name: Matthias Blume
3450    Date: 2001/01/05 00:30:00 JST
3451    Tag: blume-20010105-primops
3452    Description:
3453    
3454    Added some experimental support for work that I am doing right now.
3455    These changes mostly concern added primops, but there is also a new
3456    experimental C library in the runtime system (but currently not enabled
3457    anywhere except on Linux/X86).
3458    
3459    In the course of adding primops (and playing with them), I discovered that
3460    Zhong's INL_PRIM hack (no type info for certain primops) was, in fact, badly
3461    broken.  (Zhong was very right he labeled this stuff as "major gross hack".)
3462    To recover, I made type information in INL_PRIM mandatory and changed
3463    prim.sml as well as built-in.sml accordingly.  The InLine structure now
3464    has complete, correct type information (i.e., no bottom types).
3465    
3466    Since all these changes mean that we need new binfiles, I also bumped the
3467    version number to 110.32.1.
3468    
3469    ----------------------------------------------------------------------
3470    Name: Matthias Blume
3471    Date: 2000/12/30 22:10:00 JST
3472    Tag: blume-20001230-various
3473    Description:
3474    
3475    Added proxy libraries for MLRISC and let MLRISC libraries refer
3476    to each other using path anchors.  (See CM manual for explanation.)
3477    
3478    Updated CM documentation.
3479    
3480    Fixed some bugs in CM.
3481    
3482    Implemented "proxy" libraries (= syntactic sugar for CM).
3483    
3484    Added "-quiet" option to makeml and changed runtime system accordingly.
3485    
3486    Added cleanup handler for exportML to reset timers and compiler stats.
3487    
3488    ----------------------------------------------------------------------
3489    Name: Lal George
3490    Date: 2000/12/22 22:22:58 EST 2000
3491    Tag: Release_110_32
3492    Description:
3493    
3494            Infinite precision used throughout MLRISC.
3495            see MLRISC/mltree/machine-int.sig
3496    
3497    ----------------------------------------------------------------------
3498    Name: Matthias Blume
3499    Date: 2000/12/22 23:16:00 JST
3500    Tag: blume-20001222-warn
3501    Description:
3502    
3503    Corrected wording and formatting of some CM warning message which I
3504    broke in my previous patch.
3505    
3506    ----------------------------------------------------------------------
3507    Name: Matthias Blume
3508    Date: 2000/12/22 21:20:00 JST
3509    Tag: blume-20001222-anchorenv
3510    Description:
3511    
3512    Fixed CM's handling of anchor environments in connection with CMB.make.
3513    
3514    ----------------------------------------------------------------------
3515    Name: Matthias Blume
3516    Date: 2000/12/22 13:15:00 JST
3517    Tag: blume-20001222-cleanup
3518    Description:
3519    
3520    Removed src/cm/ffi which does not (and did not) belong here.
3521    
3522    ----------------------------------------------------------------------
3523    Name: Matthias Blume
3524    Date: 2000/12/21 23:55:00 JST
3525    Tag: blume-20001221-exn
3526    Description:
3527    
3528    Probably most important: CM no longer silently swallows all exceptions
3529    in the compiler.
3530    Plus: some other minor CM changes.  For example, CM now reports some
3531    sizes for generated binfiles (code, data, envpickle, lambdapickle).
3532    
3533    ----------------------------------------------------------------------
3534    Name: Matthias Blume
3535    Date: 2000/12/15 00:01:05 JST
3536    Tag: blume-20001215-dirtool
3537    Description:
3538    
3539    - "dir" tool added.
3540    - improvements and cleanup to Tools structure
3541    - documentation updates
3542    
3543    ----------------------------------------------------------------------
3544    Name: Allen Leung
3545    Date: Thu Dec 14 03:45:24 EST 2000
3546    Description:
3547    Tag:  leunga-20001214-int-inf
3548    Description:
3549    
3550       In IntInf, added these standard functions, which are missing from our
3551    implementation:
3552    
3553        andb : int * int -> int
3554        xorb : int * int -> int
3555        orb  : int * int -> int
3556        notb : int -> int
3557         <<   : int * word -> int
3558        ~>>  : int * word -> int
3559    
3560       Not tested, I hope they are correct.
3561    
3562    ----------------------------------------------------------------------
3563    Name: Allen Leung
3564    Date: Fri Dec  8 19:23:26 EST 2000
3565    Description:
3566    Tag:  leunga-20001208-nowhere
3567    Description:
3568    
3569      Slight improvements to the 'nowhere' tool to handle OR-patterns,
3570    to generate better error messages etc.  Plus a brief manual.
3571    
3572    ----------------------------------------------------------------------
3573    Name: Lal George
3574    Date: 2000/12/08 09:54:02 EST 2000
3575    Tag: Release_110_31
3576    Description:
3577    
3578    - Version 110.31
3579    ----------------------------------------------------------------------
3580    Name: Allen Leung
3581    Date: Thu Dec  7 22:01:04 EST 2000
3582    Tag:  leunga-20001207-cell-monster-hack
3583    Description:
3584    
3585    Major MLRISC internal changes.  Affect all clients.
3586    Summary:
3587    
3588    1.  Type CELLS.cell = int is now replaced by a datatype.
3589        As a result, the old regmap is now gone.  Almost all interfaces
3590        in MLRISC change as a consequence.
3591    
3592    2.  A new brand version of machine description tool (v3.0) that generates
3593        modules expecting the new interface.  The old version is removed.
3594    
3595    3.  The RA interface has been further abstracted into two new functors.
3596        RISC_RA and X86RA.  These functors have much simpler interfaces.
3597        [See also directory MLRISC/demo.]
3598    
3599    4.  Some other new source->source code generation tools are available:
3600    
3601        a. MLRISC/Tools/RewriteGen -- generate rewriters from rules.
3602        b. MLRISC/Tools/WhereGen -- expands conditional pattern matching rules.
3603           I use this tool to generate the peephole optimizers---with the new
3604           cell type changes, peephole rules are becoming difficult to write
3605           without conditional pattern matching.
3606    
3607    5.  More Intmap -> IntHashTable change.  Previous changes by Matthias didn't
3608        cover the entire MLRISC source tree so many things broke.
3609    
3610    6.  CM files have been moved to the subdirectory MLRISC/cm.
3611        They are moved because there are a lot of them and they clutter up the
3612        root dir.
3613    
3614    7.  More detailed documentation to come...
3615    
3616        NOTE: To rebuild from 110.30 (ftp distribution), you'll have to do
3617        a makeml -rebuild first.  This is because of other other
3618        changes that Matthias has made (see below).
3619    
3620    
3621    ----------------------------------------------------------------------