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 937, Thu Sep 20 21:20:28 2001 UTC revision 1327, Fri May 16 16:59:31 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: Allen Leung
17    Date: 2003/05/16 13:05:00 CDT
18    Tag: leunga-20030516-cygwin-runtime
19    Description:
20    
21        Ported the runtime system to cygwin, which uses the unix
22    x86-unix bin files.  Missing/buggy features:
23    
24       o getnetbyname, getnetbyaddr: these functions seem to be missing in
25         the Cygwin library.
26       o Ctrl-C handling may be flaky.
27       o Windows system calls and Windows I/O are not supported.
28    
29       A new set of binfiles is located at:
30    
31         http://www.dorsai.org/~leunga/boot.x86-unix.tgz
32    
33       This is only needed for bootstrapping the cygwin version of smlnj.
34       Other x86 versions can use the existing binfiles.
35    
36    ----------------------------------------------------------------------
37    Name: Matthias Blume
38    Date: 2003/04/08 15:42:00 CDT
39    Tag: blume-20030408-listpair
40    Description:
41    
42    1. Added a target 'mlrisc' to installer.
43    
44    2. Added missing elements to structure ListPair.
45    
46    ----------------------------------------------------------------------
47    Name: Allen Leung
48    Date: 2003/01/07 10:40:00 EST
49    Tag: leunga-20030107-int-rem
50    Description:
51    
52    Fixed a bug in Int.rem(x,y) where y is a power of 2 on x86.
53    The arguments to the SUBL instruction were swapped.
54    
55    ----------------------------------------------------------------------
56    Name: Matthias Blume
57    Date: 2002/12/12 16:25:00 EST
58    Tag: blume-20021212-risc-ra
59    Description:
60    
61    Fixed a serious bug in the rewrite code for FP spilling/reloading that
62    sent the RA into an infinite loop when floating point registers get
63    spilled.  (Because of this bug, e.g., nucleic stopped compiling between
64    110.37 and 110.38.)
65    There was another set of potential problems related to the handling of
66    MLRISC annotations (but those did not yet cause real problems, apparently).
67    
68    ----------------------------------------------------------------------
69    Name: Matthias Blume
70    Date: 2002/12/06 22:40:00 EST
71    Tag: blume-20021206-cm-fileid
72    Description:
73    
74    Added a call of SrcPath.sync at the beginning of Parse.parse (in CM).
75    This fixes the problem of CM getting confused by files that suddenly
76    change their identity (e.g., by getting unlinked and recreated by some
77    text editor such as vi).  There might be a better/cheaper/cleaner way
78    of doing this, but for now this will have to do.
79    
80    ----------------------------------------------------------------------
81    Name: Matthias Blume
82    Date: 2002/10/28 09:50:00 EST
83    Tag: blume-20021028-typecheck
84    Description:
85    
86    Exported structure Typecheck from $smlnj/viscomp/core.cm.
87    
88    ----------------------------------------------------------------------
89    Name: Matthias Blume
90    Date: 2002/10/17 09:10:00 EDT
91    Tag: Release_110_42
92    Description:
93    
94    In good old tradition, there has been a slight hiccup so that we have
95    to patch 110.42 after the fact.   The old release tag has been replaced
96    (see below).
97    
98    The change solves a problem with two competing approaches the
99    configuration problem regarding MacOS 10.1 vs. MacOS 10.2 which got in
100    each other's way.
101    
102    This change only affects the runtime system code and the installer script.
103    (No new bootfiles.)
104    
105    ----------------------------------------------------------------------
106    Name: Matthias Blume
107    Date: 2002/10/16 12:00:00 EDT
108    Tag: Release_110_42_removed
109    Description:
110    
111    New working release.  New bootfiles.
112    
113    ----------------------------------------------------------------------
114    Name: Matthias Blume
115    Date: 2002/10/10 13:10:00 EDT
116    Tag: blume-20021010-ppc-divs
117    Description:
118    
119    The mltree operator DIVS must be implemented with an overflow check on
120    the PPC because the hardware indicates divide-by-zero using "overflow" as
121    well.
122    
123    ----------------------------------------------------------------------
124    Name: Matthias Blume
125    Date: 2002/07/23 11:45:00
126    Tag: blume-20020723-smlnj-home
127    Description:
128    
129    Sml now senses the SMLNJ_HOME environment variable.  If this is set,
130    then the bin dir is assumed to be in $SMLNJ_HOME/bin and (unless
131    CM_PATHCONFIG is also set), the path configuration file is assumed
132    to be in $SMLNJ_HOME/lib/pathconfig.  This way one can easily move
133    the entire tree to some other place and everything will "just work".
134    
135    (Companion commands such as ml-build and ml-makedepend also sense this
136    variable.)
137    
138    ----------------------------------------------------------------------
139    Name: Matthias Blume
140    Date: 2002/07/12 21:19:00 EDT
141    Tag: blume-20020712-liveness
142    Description:
143    
144    Exported two useful "step" functions from liveness module (MLRISC).
145    
146    ----------------------------------------------------------------------
147    Name: Matthias Blume
148    Date: 2002/07/05 16:00 EDT
149    Tag: Release_110_41
150    Description:
151    
152    New working release.  New bootfiles.
153    
154    ----------------------------------------------------------------------
155    Name: Matthias Blume
156    Date: 2002/07/05 10:25:00 EDT
157    Tag: blume-20020705-btimp
158    Description:
159    
160    Exported structure BTImp from $smlnj/viscomp/debugprof.cm so that
161    other clients can set up backtracing support.
162    
163    ----------------------------------------------------------------------
164    Name: Matthias Blume
165    Date: 2002/06/25 17:23:00 EDT
166    Tag: blume-20020625-fpmax
167    Description:
168    
169    Fixed a bug in translation of INLMAX (and INLMIN) for the floating-point
170    case.  (The sense of the isNaN test was reversed -- which made min and
171    max always return their first argument.)
172    
173    ----------------------------------------------------------------------
174    Name: Matthias Blume
175    Date: 2002/06/11
176    Tag: blume-20020611-unixpath
177    Description:
178    
179    Back-ported OS.Path.{from,to}UnixPath from idlbasis-devel branch.
180    
181    ----------------------------------------------------------------------
182    Name: Matthias Blume
183    Date: 2002/06/10 16:35:00 EDT
184    Tag: blume-20020610-ieeereal
185    Description:
186    
187    I back-ported my implementation of IEEEReal.fromString from the
188    idlbasis-devel branch so that we can test it.
189    
190    Another small change is that ppDec tries to give more information
191    than just "<sig>" in the case of functors.  However, this code is
192    broken in some mysterious way if the functor's body's signature
193    has not been declared by ascription but gets inferred from the
194    implementation.  This needs fixing...
195    
196    ----------------------------------------------------------------------
197    Name: Matthias Blume
198    Date: 2002/05/31
199    Tag: blume-20020531-btrace-mode
200    Description:
201    
202    Resurrected SMLofNJ.Internals.BTrace.mode.  (It accidentally fell by
203    the wayside when I switched over to using Controls everywhere.)
204    
205    ----------------------------------------------------------------------
206    Name: Lal George
207    Date: 2002/05/23 12:21:40 EDT
208    Tag: george-20020523-visual-labels
209    Description:
210    
211    Labels are now displayed in the graphical output to make
212    the fall-through and target blocks obvious.
213    
214    ----------------------------------------------------------------------
215    Name: Matthias Blume
216    Date: 2002/05/22 11:03:00 EDT
217    Tag: blume-20020522-shrink
218    Description:
219    
220    John tweaked yesterday's fix for 1131 to handle an out-of-memory
221    situation that comes up when allocating huge arrays.
222    
223    ----------------------------------------------------------------------
224    Name: Matthias Blume
225    Date: 2002/05/21 16:00:00 EDT
226    Tag: Release_110_40
227    Description:
228    
229    New working release (110.40).  New bootfiles.
230    
231    [Also: John Reppy fixed GC bug 1131.]
232    
233    ----------------------------------------------------------------------
234    Name: Matthias Blume
235    Date: 2002/05/21 12:35:00 EDT
236    Tag: blume-20020521-cmdoc
237    Description:
238    
239    CM documentation update.
240    
241    ----------------------------------------------------------------------
242    Name: Matthias Blume
243    Date: 2002/05/21 10:55:00 EDT
244    Tag: blume-20020521-misc
245    Description:
246    
247    - John tweaked runtime to be silent on heap export (except when
248      GC messages are on).
249    - I added a few more things (cross-compiling versions of CMB) to
250      config/preloads (as suggestions).
251    
252    ----------------------------------------------------------------------
253    Name: Matthias Blume
254    Date: 2002/05/20 22:25:00 EDT
255    Tag: blume-20020520-controls
256    Description:
257    
258    - Added ControlUtil structure to control-lib.cm.
259    - Use it throughout.
260    - Used Controls facility to define MLRISC controls (as opposed to
261      registering MLRISC control ref cells with Controls after the fact)
262    - Fixed messed-up controls priorities.
263    
264    * Removed again all the stuff from config/preloads that one wouldn't
265      be able to preload at the time the initial heap image is built.
266      (Many libraries, e.g., CML, do not exist yet at this time.  The
267       only libraries that can be preloaded via config/preloads are those
268       that come bundled with the bootfiles.)
269    
270    ----------------------------------------------------------------------
271    Name: Matthias Blume
272    Date: 2002/05/20 10:59:00 EDT
273    Tag: blume-20020520-preloads
274    Description:
275    
276    Added a lot of commented-out suggestions for things to be included
277    in config/preloads.
278    
279    ----------------------------------------------------------------------
280    Name: Allen Leung
281    Date: 2002/05/18 14:20:00 EDT
282    Tag: leunga-20020518-mdl
283    Description:
284    
285    o Made the mdl tool stuff compile and run again.
286    o I've disabled all the stuff that depends on RTL specifications; they
287      are all badly broken anyway.
288    
289    ----------------------------------------------------------------------
290    Name: Matthias Blume
291    Date: 2002/05/17 16:49:00 EDT
292    Tag: blume-20020517-controls
293    Description:
294    
295    0. John Reppy made several modifications to the SML/NJ library.
296       In particular, there is a shiny new controls-lib.cm.
297    
298    1. Pushed new controls interface through compiler so that everything
299       compiles again.
300    
301    2. Added FormatComb and FORMAT_COMB to the CML version of the
302       SML/NJ library (so that CML compiles again).
303    
304    3. Modified init scripts because XXX_DEFAULT environment variables
305       are no longer with us.  (Boot-time initialization is now done
306       using the same environment variables that are also used for
307       startup-time initialization of controls.)
308    
309    ----------------------------------------------------------------------
310    Name: Lal George
311    Date: 2002/05/15 09:20:10 EDT
312    Tag: george-20020515-pseudo-op-decls
313    Description:
314    
315            All pseudo-ops emitted before the first segment declaration
316    such as TEXT, DATA, and BSS directives are assumed to be global
317    declarations and are emitted first in the assembly file. This is
318    useful in a number of situations where one has pseudo-ops that are not
319    specific to any segment, and also works around the constraint that one
320    cannot have client pseudo-ops in the TEXT segment.
321    
322            Because no segment is associated with these declarations it is
323    an error to allocate any space or objects before the first segment
324    directive and an exception will be raised. However, we cannot make
325    this check for client pseudo-ops.
326    
327            These top level declarations are a field in the CFG graph_info.
328    In theory you can continue to add to this field after the CFG has been
329    built -- provided you know what you are doing;-)
330    
331    ----------------------------------------------------------------------
332    Name: Matthias Blume
333    Date: 2002/05/13 16:40:00 EDT
334    Tag: blume-20020513-pp-etc
335    Description:
336    
337    A few minor bugfixes:
338    
339      - Stopgap measure for bug recently reported by Elsa Gunter (ppDec).
340        (Bogus printouts for redefined bindings still occur.  Compiler
341        bug should no longer occur now.  We need to redo the prettyprinter
342        from scratch.)
343    
344      - CM pathname printer now also adds escape sequences for ( and )
345    
346      - commend and docu fixes for ml-nlffi
347    
348    ----------------------------------------------------------------------
349    Name: Matthias Blume
350    Date: 2002/05/10 16:40:00 EDT
351    Tag: blume-20020510-erg-textio
352    Description:
353    
354    Applied the following bugfix provided by Emden Gansner:
355    
356        Output is corrupted when outputSubstr is used rather than output.
357    
358        The problem occurs when a substring
359    
360            ss = (s, dataStart, dataLen)
361    
362        where dataStart > 0, fills a stream buffer with avail bytes left.
363        avail bytes of s, starting at index dataStart, are copied into the
364        buffer, the buffer is flushed, and then the remaining dataLen-avail
365        bytes of ss are copied into the beginning of the buffer. Instead of
366        starting this copy at index dataStart+avail in s, the current code
367        starts the copy at index avail.
368    
369        Fix:
370        In text-io-fn.sml, change line 695 from
371             val needsFlush = copyVec(v, avail, dataLen-avail, buf, 0)
372        to
373             val needsFlush = copyVec(v, dataStart+avail, dataLen-avail, buf, 0)
374    
375    ----------------------------------------------------------------------
376    Name: Matthias Blume
377    Date: 2002/04/12 13:55:00 EDT
378    Tag: blume-20020412-assyntax
379    Description:
380    
381    1. Grabbed newer assyntax.h from the XFree86 project.
382    2. Fiddled with how to compile X86.prim.asm without warnings.
383    3. (Very) Minor cleanup in CM.
384    
385    ----------------------------------------------------------------------
386    Name: Matthias Blume
387    Date: 2002/04/01 (no joke!) 17:07:00 EST
388    Tag: blume-20020401-x86div
389    Description:
390    
391    Added full support for div/mod/rem/quot on the x86, using the machine
392    instruction's two results (without clumsily recomputing the remainder)
393    directly where appropriate.
394    
395    Some more extensive power-of-two support was added to the x86 instruction
396    selector (avoiding expensive divs, mods, and muls where they can be
397    replaced with cheaper shifts and masks).  However, this sort of thing
398    ought to be done earlier, e.g., within the CPS optimizer so that
399    all architectures benefit from it.
400    
401    The compiler compiles to a fixed point, but changes might be somewhat
402    fragile nevertheless.  Please, report any strange things that you might
403    see wrt. div/mod/quot/rem...
404    
405    ----------------------------------------------------------------------
406    Name: Matthias Blume
407    Date: 2002/03/29 17:22:00
408    Tag: blume-20020329-div
409    Description:
410    
411    Fixed my broken div/mod logic.  Unfortunately, this means that the
412    inline code for div/mod now has one more comparison than before.
413    Fast paths (quotient > 0 or remainder = 0) are not affected, though.
414    The problem was with quotient = 0, because that alone does not tell
415    us which way the rounding went.  One then has to look at whether
416    remainder and divisor have the same sign...  :(
417    
418    Anyway, I replaced the bootfiles with fresh ones...
419    
420    ----------------------------------------------------------------------
421    Name: Matthias Blume
422    Date: 2002/03/29 14:10:00 EST
423    Tag: blume-20020329-inlprims
424    Description:
425    
426    NEW BOOTFILES!!!    Version number bumped to 110.39.3.
427    
428    Primops have changed. This means that the bin/boot-file formats have
429    changed as well.
430    
431    To make sure that there is no confusion, I made a new version.
432    
433    
434    CHANGES:
435    
436    * removed REMT from mltree (remainder should never overflow).
437    
438    * added primops to deal with divisions of all flavors to the frontend
439    
440    * handled these primops all the way through so they map to their respective
441      MLRISC support
442    
443    * used these primops in the implementation of Int, Int32, Word, Word32
444    
445    * removed INLDIV, INLMOD, and INLREM as they are no longer necessary
446    
447    * parameterized INLMIN, INLMAX, and INLABS by a numkind
448    
449    * translate.sml now deals with all flavors of INL{MIN,MAX,ABS}, including
450      floating point
451    
452    * used INL{MIN,MAX,ABS} in the implementation of Int, Int32, Word, Word32,
453      and Real (but Real.abs maps to a separate floating-point-only primop)
454    
455    
456    TODO items:
457    
458    * Hacked Alpha32 instruction selection, disabling the selection of REMx
459      instructions because the machine instruction encoder cannot handle
460      them.  (Hppa, PPC, and Sparc instruction selection did not handle
461      REM in the first place, and REM is supported by the x86 machine coder.)
462    
463    * Handle DIV and MOD with DIV_TO_NEGINF directly in the x86 instruction
464      selection phase.  (The two can be streamlined because the hardware
465      delivers both quotient and remainder at the same time anyway.)
466    
467    * Think about what to do with "valOf(Int32.minInt) div ~1" and friends.
468      (Currently the behavior is inconsistent both across architectures and
469      wrt. the draft Basis spec.)
470    
471    * Word8 should eventually be handled natively, too.
472    
473    * There seems to be one serious bug in mltree-gen.sml.  It appears, though,
474      as if there currently is no execution path that could trigger it in
475      SML/NJ.  (The assumptions underlying functions arith and promotable do not
476      hold for things like multiplication and division.)
477    
478    ----------------------------------------------------------------------
479    Name: Matthias Blume
480    Date: 2002/03/27 16:27:00 EST
481    Tag: blume-20020327-mlrisc-divisions
482    Description:
483    
484    Added support for all four division operations (ML's div, mod, quot,
485    and rem) to MLRISC.  In the course of doing so, I also rationalized
486    the naming (no more annoying switch-around of DIV and QUOT), by
487    parameterizing the operation by div_rounding_mode (which can be either
488    DIV_TO_ZERO or DIV_TO_NEGINF).
489    
490    The generic MLTreeGen functor takes care of compiling all four
491    operations down to only round-to-zero div.
492    
493    Missing pieces:
494    
495      * Doing something smarter than relying on MLTreeGen on architectures
496        like, e.g., the x86 where hardware division delivers both quotient and
497        remainder at the same time.  With this, the implementation of the
498        round-to-neginf operations could be further streamlined.
499    
500      * Remove inlining support for div/mod/rem from the frontend and replace it
501        with primops that get carried through to the backend.  Do this for all
502        int and word types.
503    
504    ----------------------------------------------------------------------
505    Name: Matthias Blume
506    Date: 2002/03/25 17:25:00 EST
507    Tag: blume-20020325-divmod
508    Description:
509    
510    I improved (hopefully without breaking them) the implementation of Int.div,
511    Int.mod, and Int.rem.   For this, the code in translate.sml now takes
512    advantage of the following observations:
513    
514      Let  q = x quot y      r = x rem y
515           d = x div  y      m = x mod y
516    
517    where "quot" is the round-to-zero version of integer division that
518    hardware usually provides.  Then we have:
519    
520         r = x - q * y        where neither the * nor the - will overflow
521         d = if q >= 0 orelse x = q * y then q else q - 1
522                              where neither the * nor the - will overflow
523         m = if q >= 0 orelse r = 0 then r else r + y
524                              where the + will not overflow
525    
526    This results in substantial simplification of the generated code.
527    The following table shows the number of CFG nodes and edges generated
528    for
529            fun f (x, y) = x OPER y
530            (* with OPER \in div, mod, quot, rem *)
531    
532    
533        OPER | nodes(old) | edges(old) | nodes(new) | edges(new)
534        --------------------------------------------------------
535         div |         24 |         39 |         12 |         16
536         mod |         41 |         71 |         12 |         16
537        quot |          8 |         10 |          8 |         10
538         rem |         10 |         14 |          8 |         10
539    
540    
541    ----------------------------------------------------------------------
542    Name: Matthias Blume
543    Date: 2002/03/25 22:06:00 EST
544    Tag: blume-20020325-cprotobug
545    Description:
546    
547    Fixed a bug in cproto (c prototype decoder).
548    
549    ----------------------------------------------------------------------
550    Name: Matthias Blume
551    Date: 2002/03/25 16:00:00 EST
552    Tag: blume-20020325-raw-primops
553    Description:
554    
555    I did some cleanup to Allen's new primop code and
556    replaced yesterday's bootfiles with new ones.
557    (But they are stored in the same place.)
558    
559    ----------------------------------------------------------------------
560    Name: Matthias Blume
561    Date: 2002/03/24 22:40:00 EST
562    Tag: blume-20020324-bootfiles
563    Description:
564    
565    Made the bootfiles that Allen asked for.
566    
567    ----------------------------------------------------------------------
568    Name: Allen Leung
569    Date: 2002/03/23 15:50:00 EST
570    Tag: leunga-20020323-flint-cps-rcc-primops
571    Description:
572    
573      1. Changes to FLINT primops:
574    
575        (* make a call to a C-function;
576         * The primop carries C function prototype information and specifies
577         * which of its (ML-) arguments are floating point. C prototype
578         * information is for use by the backend, ML information is for
579         * use by the CPS converter. *)
580      | RAW_CCALL of { c_proto: CTypes.c_proto,
581                       ml_args: ccall_type list,
582                       ml_res_opt: ccall_type option,
583                       reentrant : bool
584                     } option
585       (* Allocate uninitialized storage on the heap.
586        * The record is meant to hold short-lived C objects, i.e., they
587        * are not ML pointers.  With the tag, the representation is
588        * the same as RECORD with tag tag_raw32 (sz=4), or tag_fblock (sz=8)
589        *)
590      | RAW_RECORD of {tag:bool,sz:int}
591      and ccall_type = CCALL_INT32 | CCALL_REAL64 | CCALL_ML_PTR
592    
593      2.  These CPS primops are now overloaded:
594    
595           rawload of {kind:numkind}
596           rawstore of {kind:numkind}
597    
598          The one argument form is:
599    
600             rawload {kind} address
601    
602          The two argument form is:
603    
604             rawload {kind} [ml object, byte-offset]
605    
606      3. RAW_CCALL/RCC now takes two extra arguments:
607    
608         a. The first is whether the C call is reentrant, i.e., whether
609            ML state should be saved and restored.
610         b. The second argument is a string argument specifying the name of
611            library and the C function.
612    
613         These things are currently not handled in the code generator, yet.
614    
615      4. In CProto,
616    
617         An encoding type of "bool" means "ml object" and is mapped into
618         C prototype of PTR.  Note that "bool" is different than "string",
619         even though "string" is also mapped into PTR, because "bool"
620         is assigned an CPS type of BOGt, while "string" is assigned INT32t.
621    
622      5. Pickler/unpicker
623    
624         Changed to handle RAW_RECORD and newest RAW_CCALL
625    
626      6. MLRiscGen,
627    
628         1. Changed to handle the new rawload/rawstore/rawrecord operators.
629         2. Code for handling C Calls has been moved to a new module CPSCCalls,
630            in the file CodeGen/cpscompile/cps-c-calls.sml
631    
632      7. Added the conditional move operator
633    
634             condmove of branch
635    
636         to cps.  Generation of this is still buggy so it is currently
637         disabled.
638    
639    ----------------------------------------------------------------------
640    Name: Lal George
641    Date: 2002/03/22 14:18:25 EST
642    Tag: george-20020322-cps-branch-prob
643    Description:
644    
645    Implemented the Ball-Larus branch prediction-heuristics, and
646    incorporated graphical viewers for control flow graphs.
647    
648    Ball-Larus Heuristics:
649    ---------------------
650    See the file compiler/CodeGen/cpscompile/cpsBranchProb.sml.
651    
652    By design it uses the Dempster-Shafer theory for combining
653    probabilities.  For example, in the function:
654    
655        fun f(n,acc) = if n = 0 then acc else f(n-1, n*acc)
656    
657    the ball-larus heuristics predicts that the n=0 is unlikely
658    (OH-heuristic), and the 'then' branch is unlikely because of the
659    RH-heuristic -- giving the 'then' branch an even lower combined
660    probability using the Dempster-Shafer theory.
661    
662    Finally, John Reppy's loop analysis in MLRISC, further lowers the
663    probability of the 'then' branch because of the loop in the else
664    branch.
665    
666    
667    Graphical Viewing:
668    ------------------
669    I merely plugged in Allen's graphical viewers into the compiler. The
670    additional code is not much. At the top level, saying:
671    
672            Control.MLRISC.getFlag "cfg-graphical-view" := true;
673    
674    will display the graphical view of the control flow graph just before
675    back-patching.  daVinci must be in your path for this to work. If
676    daVinci is not available, then the default viewer can be changed
677    using:
678    
679            Control.MLRISC.getString "viewer"
680    
681    which can be set to "dot" or "vcg" for the corresponding viewers. Of
682    course, these viewers must be in your path.
683    
684    The above will display the compilation unit at the level of clusters,
685    many of which are small, boring, and un-interesting. Also setting:
686    
687            Control.MLRISC.getInt "cfg-graphical-view_size"
688    
689    will display clusters that are larger than the value set by the above.
690    
691    
692    ----------------------------------------------------------------------
693    Name: Matthias Blume
694    Date: 2002/03/21 22:20:00 EST
695    Tag: blume-20020321-kmp-bugfix
696    Description:
697    
698    Changed the interface to the KMP routine in PreString and fixed
699    a minor bug in one place where it was used.
700    
701    ----------------------------------------------------------------------
702    Name: Allen Leung
703    Date: 2002/03/21 20:30:00 EST
704    Tag: leunga-20020321-cfg
705    Description:
706    
707      Fixed a potential problem in cfg edge splitting.
708    
709    ----------------------------------------------------------------------
710    Name: Allen Leung
711    Date: 2002/03/21 17:15:00 EST
712    Tag: leunga-20020321-x86-fp-cfg
713    Description:
714    
715      1. Recoded the buggy parts of x86-fp.
716    
717         a. All the block reordering code has been removed.
718            We now depend on the block placement phases to do this work.
719    
720         b. Critical edge splitting code has been simplified and moved into the
721            CFG modules, as where they belong.
722    
723         Both of these were quite buggy and complex.  The code is now much, much
724         simpler.
725    
726      2. X86 backend.
727    
728         a. Added instructions for 64-bit support.  Instruction selection for
729            64-bit has not been committed, however, since that
730            requires changes to MLTREE which haven't been approved by
731            Lal and John.
732    
733         b. Added support for FUCOMI and FUCOMIP when generating code for
734            PentiumPro and above.  We only generate these instructions in
735            the fast-fp mode.
736    
737         c. Added cases for JP and JNP in X86FreqProps.
738    
739      3. CFG
740    
741         CFG now has a bunch of methods for edge splitting and merging.
742    
743      4. Machine description.
744    
745         John's simplification of MLTREE_BASIS.fcond broke a few machine
746         description things:
747    
748         rtl-build.{sig,sml} and hppa.mdl fixed.
749    
750         NOTE: the machine description stuff in the repository is still broken.
751               Again, I can't put my fixes in because that involves
752               changes to MLTREE.
753    
754    ----------------------------------------------------------------------
755    Name: Matthias Blume
756    Date: 2002/03/20 15:55:00 EST
757    Tag: blume-20020320-kmp
758    Description:
759    
760    Implemented Knuth-Morris-Pratt string matching in PreString and used
761    it for String.isSubstring, Substring.isSubstring, and
762    Substring.position.
763    
764    (Might need some stress-testing.  Simple examples worked fine.)
765    
766    ----------------------------------------------------------------------
767    Name: Matthias Blume
768    Date: 2002/03/19 16:37:00 EST
769    Tag: blume-20020319-witnesses
770    Description:
771    
772    Added a structure C.W and functions convert/Ptr.convert to ml-nlffi-lib.
773    
774    This implements a generic mechanism for changing constness qualifiers
775    anywhere within big C types without resorting to outright "casts".
776    (So far, functions such as C.rw/C.ro or C.Ptr.rw/C.Ptr.ro only let you
777    modify the constness at the outermost level.)
778    The implementation of "convert" is based on the idea of "witness"
779    values -- values that are not used by the operation but whose types
780    "testify" to their applicability.  On the implementation side, "convert"
781    is simply a projection (returning its second curried argument).  With
782    cross-module inlining, it should not result in any machine code being
783    generated.
784    
785    ----------------------------------------------------------------------
786    Name: Matthias Blume
787    Date: 2002/03/15 16:40:00 EST
788    Tag: blume-20020315-basis
789    Description:
790    
791    Provided (preliminary?) implementations for
792    
793      {String,Substring}.{concatWith,isSuffix,isSubstring}
794    
795    and
796    
797      Substring.full
798    
799    Those are in the Basis spec but they were missing in SML/NJ.
800    
801    ----------------------------------------------------------------------
802    Name: Matthias Blume
803    Date: 2002/03/14 21:30:00 EST
804    Tag: blume-20020314-controls
805    Description:
806    
807    Controls:
808    ---------
809    
810    1. Factored out the recently-added Controls : CONTROLS stuff and put
811       it into its own library $/controls-lib.cm.  The source tree for
812       this is under src/smlnj-lib/Controls.
813    
814    2. Changed the names of types and functions in this interface, so they
815       make a bit more "sense":
816    
817          module -> registry
818          'a registry -> 'a group
819    
820    3. The interface now deals in ref cells only.  The getter/setter interface
821       is (mostly) gone.
822    
823    4. Added a function that lets one register an already-existing ref cell.
824    
825    5. Made the corresponding modifications to the rest of the code so that
826       everything compiles again.
827    
828    6. Changed the implementation of Controls.MLRISC back to something closer
829       to the original.  In particular, this module (and therefore MLRISC)
830       does not depend on Controls.  There now is some link-time code in
831       int-sys.sml that registers the MLRISC controls with the Controls
832       module.
833    
834    CM:
835    ---
836    
837      * One can now specify the lambda-split aggressiveness in init.cmi.
838    
839    ----------------------------------------------------------------------
840    Name: Allen Leung
841    Date: 2002/03/13 17:30:00 EST
842    Tag: leunga-20020313-x86-fp-unary
843    Description:
844    
845    Bug fix for:
846    
847    > leunga@weaselbane:~/Yale/tmp/sml-dist{21} bin/sml
848    > Standard ML of New Jersey v110.39.1 [FLINT v1.5], March 08, 2002
849    > - fun f(x,(y,z)) = Real.~ y;
850    > [autoloading]
851    > [autoloading done]
852    >       fchsl   (%eax), 184(%esp)
853    > Error: MLRisc bug: X86MCEmitter.emitInstr
854    >
855    > uncaught exception Error
856    >   raised at: ../MLRISC/control/mlriscErrormsg.sml:16.14-16.19
857    
858    The problem was that the code generator did not generate any fp registers
859    in this case, and the ra didn't know that it needed to run the X86FP phase to
860    translate the pseudo fp instruction.   This only happened with unary fp
861    operators in certain situations.
862    
863    ----------------------------------------------------------------------
864    Name: Matthias Blume
865    Date: 2002/03/13 14:00:00 EST
866    Tag: blume-20020313-overload-etc
867    Description:
868    
869    1. Added _overload as a synonym for overload for backward compatibility.
870       (Control.overloadKW must be true for either version to be accepted.)
871    
872    2. Fixed bug in install script that caused more things to be installed
873       than what was requested in config/targets.
874    
875    3. Made CM aware of the (_)overload construct so that autoloading
876       works.
877    
878    ----------------------------------------------------------------------
879    Name: Matthias Blume
880    Date: 2002/03/12 22:03:00 EST
881    Tag: blume-20020312-url
882    Description:
883    
884    Forgot to update BOOT and srcarchiveurl.
885    
886    ----------------------------------------------------------------------
887    Name: Matthias Blume
888    Date: 2002/03/12 17:30:00 EST
889    Tag: blume-20020312-version110392
890    Description:
891    
892    Yet another version number bump (because of small changes to the
893    binfile format).  Version number is now 110.39.2.  NEW BOOTFILES!
894    
895    Changes:
896    
897      The new pid generation scheme described a few weeks ago was overly
898      complicated.  I implemented a new mechanism that is simpler and
899      provides a bit more "stability":  Once CM has seen a compilation
900      unit, it keeps its identity constant (as long as you do not delete
901      those crucial CM/GUID/* files).  This means that when you change
902      an interface, compile, then go back to the old interface, and
903      compile again, you arrive at the original pid.
904    
905      There now also is a mechanism that instructs CM to use the plain
906      environment hash as a module's pid (effectively making its GUID
907      the empty string).  For this, "noguid" must be specified as an
908      option to the .sml file in question within its .cm file.
909      This is most useful for code that is being generated by tools such
910      as ml-nlffigen (because during development programmers tend to
911      erase the tool's entire output directory tree including CM's cached
912      GUIDs).  "noguid" is somewhat dangerous (since it can be used to locally
913      revert to the old, broken behavior of SML/NJ, but in specific cases
914      where there is no danger of interface confusion, its use is ok
915      (I think).
916    
917      ml-nlffigen by default generates "noguid" annotations.  They can be
918      turned off by specifying -guid in its command line.
919    
920    ----------------------------------------------------------------------
921    Name: Lal George
922    Date: 2002/03/12 12 14:42:36 EST
923    Tag: george-20020312-frequency-computation
924    Description:
925    
926    Integrated jump chaining and static block frequency into the
927    compiler. More details and numbers later.
928    
929    ----------------------------------------------------------------------
930    Name: Lal George
931    Date: 2002/03/11 11 22:38:53 EST
932    Tag: george-20020311-jump-chain-elim
933    Description:
934    
935    Tested the jump chain elimination on all architectures (except the
936    hppa).  This is on by default right now and is profitable for the
937    alpha and x86, however, it may not be profitable for the sparc and ppc
938    when compiling the compiler.
939    
940    The gc test will typically jump to a label at the end of the cluster,
941    where there is another jump to an external cluster containing the actual
942    code to invoke gc. This is to allow factoring of common gc invocation
943    sequences. That is to say, we generate:
944    
945            f:
946               testgc
947               ja   L1      % jump if above to L1
948    
949            L1:
950               jmp L2
951    
952    
953    After jump chain elimination the 'ja L1' instructions is converted to
954    'ja L2'. On the sparc and ppc, many of the 'ja L2' instructions may end
955    up being implemented in their long form (if L2 is far away) using:
956    
957            jbe     L3      % jump if below or equal to L3
958            jmp     L2
959         L3:
960            ...
961    
962    
963    For large compilation units L2  may be far away.
964    
965    
966    ----------------------------------------------------------------------
967    Name: Matthias Blume
968    Date: 2002/03/11 13:30:00 EST
969    Tag: blume-20020311-mltreeeval
970    Description:
971    
972    A functor parameter was missing.
973    
974    ----------------------------------------------------------------------
975    Name: Allen Leung
976    Date: 2002/03/11 10:30:00 EST
977    Tag: leunga-20020311-runtime-string0
978    Description:
979    
980       The representation of the empty string now points to a
981    legal null terminated C string instead of unit.  It is now possible
982    to convert an ML string into C string with InlineT.CharVector.getData.
983    This compiles into one single machine instruction.
984    
985    ----------------------------------------------------------------------
986    Name: Allen Leung
987    Date: 2002/03/10 23:55:00 EST
988    Tag: leunga-20020310-x86-call
989    Description:
990    
991       Added machine generation for CALL instruction (relative displacement mode)
992    
993    ----------------------------------------------------------------------
994    Name: Matthias Blume
995    Date: 2002/03/08 16:05:00
996    Tag: blume-20020308-entrypoints
997    Description:
998    
999    Version number bumped to 110.39.1.  NEW BOOTFILES!
1000    
1001    Entrypoints: non-zero offset into a code object where execution should begin.
1002    
1003    - Added the notion of an entrypoint to CodeObj.
1004    - Added reading/writing of entrypoint info to Binfile.
1005    - Made runtime system bootloader aware of entrypoints.
1006    - Use the address of the label of the first function given to mlriscGen
1007      as the entrypoint.  This address is currently always 0, but it will
1008      not be 0 once we turn on block placement.
1009    - Removed the linkage cluster code (which was The Other Way(tm) of dealing
1010      with entry points) from mlriscGen.
1011    
1012    ----------------------------------------------------------------------
1013    Name: Allen Leung
1014    Date: 2002/03/07 20:45:00 EST
1015    Tag: leunga-20020307-x86-cmov
1016    Description:
1017    
1018       Bug fixes for CMOVcc on x86.
1019    
1020       1. Added machine code generation for CMOVcc
1021       2. CMOVcc is now generated in preference over SETcc on PentiumPro or above.
1022       3. CMOVcc cannot have an immediate operand as argument.
1023    
1024    ----------------------------------------------------------------------
1025    Name: Matthias Blume
1026    Date: 2002/03/07 16:15:00 EST
1027    Tag: blume-20020307-controls
1028    Description:
1029    
1030    This is a very large but mostly boring patch which makes (almost)
1031    every tuneable compiler knob (i.e., pretty much everything under
1032    Control.* plus a few other things) configurable via both the command
1033    line and environment variables in the style CM did its configuration
1034    until now.
1035    
1036    Try starting sml with '-h' (or, if you are brave, '-H')
1037    
1038    To this end, I added a structure Controls : CONTROLS to smlnj-lib.cm which
1039    implements the underlying generic mechanism.
1040    
1041    The interface to some of the existing such facilities has changed somewhat.
1042    For example, the MLRiscControl module now provides mkFoo instead of getFoo.
1043    (The getFoo interface is still there for backward-compatibility, but its
1044    use is deprecated.)
1045    
1046    The ml-build script passes -Cxxx=yyy command-line arguments through so
1047    that one can now twiddle the compiler settings when using this "batch"
1048    compiler.
1049    
1050    TODO items:
1051    
1052    We should go through and throw out all controls that are no longer
1053    connected to anything.  Moreover, we should go through and provide
1054    meaningful (and correct!) documentation strings for those controls
1055    that still are connected.
1056    
1057    Currently, multiple calls to Controls.new are accepted (only the first
1058    has any effect).  Eventually we should make sure that every control
1059    is being made (via Controls.new) exactly once.  Future access can then
1060    be done using Controls.acc.
1061    
1062    Finally, it would probably be a good idea to use the getter-setter
1063    interface to controls rather than ref cells.  For the time being, both
1064    styles are provided by the Controls module, but getter-setter pairs are
1065    better if thread-safety is of any concern because they can be wrapped.
1066    
1067    *****************************************
1068    
1069    One bug fix: The function blockPlacement in three of the MLRISC
1070    backpatch files used to be hard-wired to one of two possibilities at
1071    link time (according to the value of the placementFlag).  But (I
1072    think) it should rather sense the flag every time.
1073    
1074    *****************************************
1075    
1076    Other assorted changes (by other people who did not supply a HISTORY entry):
1077    
1078    1. the cross-module inliner now works much better (Monnier)
1079    2. representation of weights, frequencies, and probabilities in MLRISC
1080       changed in preparation of using those for weighted block placement
1081       (Reppy, George)
1082    
1083    ----------------------------------------------------------------------
1084    Name: Lal George
1085    Date: 2002/03/07 14:44:24 EST 2002
1086    Tag: george-20020307-weighted-block-placement
1087    
1088    Tested the weighted block placement optimization on all architectures
1089    (except the hppa) using AMPL to generate the block and edge frequencies.
1090    Changes were required in the machine properties to correctly
1091    categorize trap instructions. There is an MLRISC flag
1092    "weighted-block-placement" that can be used to enable weighted block
1093    placement, but this will be ineffective without block/edge
1094    frequencies (coming soon).
1095    
1096    
1097    ----------------------------------------------------------------------
1098    Name: Lal George
1099    Date: 2002/03/05 17:24:48 EST
1100    Tag: george-20020305-linkage-cluster
1101    
1102    In order to support the block placement optimization, a new cluster
1103    is generated as the very first cluster (called the linkage cluster).
1104    It contains a single jump to the 'real' entry point for the compilation
1105    unit. Block placement has no effect on the linkage cluster itself, but
1106    all the other clusters  have full freedom in the manner in which they
1107    reorder blocks or functions.
1108    
1109    On the x86 the typical linkage code that is generated is:
1110       ----------------------
1111            .align 2
1112       L0:
1113            addl    $L1-L0, 72(%esp)
1114            jmp     L1
1115    
1116    
1117            .align  2
1118       L1:
1119       ----------------------
1120    
1121    72(%esp) is the memory location for the stdlink register. This
1122    must contain the address of the CPS function being called. In the
1123    above example, it contains the address of  L0; before
1124    calling L1 (the real entry point for the compilation unit), it
1125    must contain the address for L1, and hence
1126    
1127            addl $L1-L0, 72(%esp)
1128    
1129    I have tested this on all architectures except the hppa.The increase
1130    in code size is of course negligible
1131    
1132    ----------------------------------------------------------------------
1133    Name: Allen Leung
1134    Date: 2002/03/03 13:20:00 EST
1135    Tag: leunga-20020303-mlrisc-tools
1136    
1137      Added #[ ... ] expressions to mlrisc tools
1138    
1139    ----------------------------------------------------------------------
1140    Name: Matthias Blume
1141    Date: 2002/02/27 12:29:00 EST
1142    Tag: blume-20020227-cdebug
1143    Description:
1144    
1145    - made types in structure C and C_Debug to be equal
1146    - got rid of code duplication (c-int.sml vs. c-int-debug.sml)
1147    - there no longer is a C_Int_Debug (C_Debug is directly derived from C)
1148    
1149    ----------------------------------------------------------------------
1150    Name: Matthias Blume
1151    Date: 2002/02/26 12:00:00 EST
1152    Tag: blume-20020226-ffi
1153    Description:
1154    
1155    1. Fixed a minor bug in CM's "noweb" tool:
1156       If numbering is turned off, then truly don't number (i.e., do not
1157       supply the -L option to noweb).  The previous behavior was to supply
1158       -L'' -- which caused noweb to use the "default" line numbering scheme.
1159       Thanks to Chris Richards for pointing this out (and supplying the fix).
1160    
1161    2. Once again, I reworked some aspects of the FFI:
1162    
1163       A. The incomplete/complete type business:
1164    
1165       - Signatures POINTER_TO_INCOMPLETE_TYPE and accompanying functors are
1166         gone!
1167       - ML types representing an incomplete type are now *equal* to
1168         ML types representing their corresponding complete types (just like
1169         in C).  This is still safe because ml-nlffigen will not generate
1170         RTTI for incomplete types, nor will it generate functions that
1171         require access to such RTTI.   But when ML code generated from both
1172         incomplete and complete versions of the C type meet, the ML types
1173         are trivially interoperable.
1174    
1175         NOTE:  These changes restore the full generality of the translation
1176         (which was previously lost when I eliminated functorization)!
1177    
1178       B. Enum types:
1179    
1180       - Structure C now has a type constructor "enum" that is similar to
1181         how the "su" constructor works.  However, "enum" is not a phantom
1182         type because each "T enum" has values (and is isomorphic to
1183         MLRep.Signed.int).
1184       - There are generic access operations for enum objects (using
1185         MLRep.Signed.int).
1186       - ml-nlffigen will generate a structure E_foo for each "enum foo".
1187         * The structure contains the definition of type "mlrep" (the ML-side
1188         representation type of the enum).  Normally, mlrep is the same
1189         as "MLRep.Signed.int", but if ml-nlffigen was invoked with "-ec",
1190         then mlrep will be defined as a datatype -- thus facilitating
1191         pattern matching on mlrep values.
1192         ("-ec" will be suppressed if there are duplicate values in an
1193          enumeration.)
1194         * Constructors ("-ec") or values (no "-ec") e_xxx of type mlrep
1195         will be generated for each C enum constant xxx.
1196         * Conversion functions m2i and i2m convert between mlrep and
1197         MLRep.Signed.int.  (Without "-ec", these functions are identities.)
1198         * Coversion functions c and ml convert between mlrep and "tag enum".
1199         * Access functions (get/set) fetch and store mlrep values.
1200       - By default (unless ml-nlffigen was invoked with "-nocollect"), unnamed
1201         enumerations are merged into one single enumeration represented by
1202         structure E_'.
1203    
1204    ----------------------------------------------------------------------
1205    Name: Allen Leung
1206    Date: 2002/02/25 04:45:00 EST
1207    Tag: leunga-20020225-cps-spill
1208    
1209    This is a new implementation of the CPS spill phase.
1210    The new phase is in the new file compiler/CodeGen/cpscompile/spill-new.sml
1211    In case of problems, replace it with the old file spill.sml
1212    
1213    The current compiler runs into some serious performance problems when
1214    constructing a large record.  This can happen when we try to compile a
1215    structure with many items.  Even a very simple structure like the following
1216    makes the compiler slow down.
1217    
1218        structure Foo = struct
1219           val x_1 = 0w1 : Word32.int
1220           val x_2 = 0w2 : Word32.int
1221           val x_3 = 0w3 : Word32.int
1222           ...
1223           val x_N = 0wN : Word32.int
1224        end
1225    
1226    The following table shows the compile time, from N=1000 to N=4000,
1227    with the old compiler:
1228    
1229    N
1230    1000   CPS 100 spill                           0.04u  0.00s  0.00g
1231           MLRISC ra                               0.06u  0.00s  0.05g
1232              (spills = 0 reloads = 0)
1233           TOTAL                                   0.63u  0.07s  0.21g
1234    
1235    1100   CPS 100 spill                           8.25u  0.32s  0.64g
1236           MLRISC ra                               5.68u  0.59s  3.93g
1237              (spills = 0 reloads = 0)
1238           TOTAL                                   14.71u  0.99s  4.81g
1239    
1240    1500   CPS 100 spill                           58.55u  2.34s  1.74g
1241           MLRISC ra                               5.54u  0.65s  3.91g
1242              (spills = 543 reloads = 1082)
1243           TOTAL                                   65.40u  3.13s  6.00g
1244    
1245    2000   CPS 100 spill                           126.69u  4.84s  3.08g
1246           MLRISC ra                               0.80u  0.10s  0.55g
1247              (spills = 42 reloads = 84)
1248           TOTAL                                   129.42u  5.10s  4.13g
1249    
1250    3000   CPS 100 spill                           675.59u  19.03s  11.64g
1251           MLRISC ra                               2.69u  0.27s  1.38g
1252              (spills = 62 reloads = 124)
1253           TOTAL                                   682.48u  19.61s  13.99g
1254    
1255    4000   CPS 100 spill                           2362.82u  56.28s  43.60g
1256           MLRISC ra                               4.96u  0.27s  2.72g
1257              (spills = 85 reloads = 170)
1258           TOTAL                                   2375.26u  57.21s  48.00g
1259    
1260    As you can see the old cps spill module suffers from some serious
1261    performance problem.  But since I cannot decipher the old code fully,
1262    instead of patching the problems up, I'm reimplementing it
1263    with a different algorithm.  The new code is more modular,
1264    smaller when compiled, and substantially faster
1265    (O(n log n) time and O(n) space).  Timing of the new spill module:
1266    
1267    4000  CPS 100 spill                           0.02u  0.00s  0.00g
1268          MLRISC ra                               0.25u  0.02s  0.15g
1269             (spills=1 reloads=3)
1270          TOTAL                                   7.74u  0.34s  1.62g
1271    
1272    Implementation details:
1273    
1274    As far as I can tell, the purpose of the CPS spill module is to make sure the
1275    number of live variables at any program point (the bandwidth)
1276    does not exceed a certain limit, which is determined by the
1277    size of the spill area.
1278    
1279    When the bandwidth is too large, we decrease the register pressure by
1280    packing live variables into spill records.  How we achieve this is
1281    completely different than what we did in the old code.
1282    
1283    First, there is something about the MLRiscGen code generator
1284    that we should be aware of:
1285    
1286    o MLRiscGen performs code motion!
1287    
1288       In particular, it will move floating point computations and
1289       address computations involving only the heap pointer to
1290       their use sites (if there is only a single use).
1291       What this means is that if we have a CPS record construction
1292       statement
1293    
1294           RECORD(k,vl,w,e)
1295    
1296       we should never count the new record address w as live if w
1297       has only one use (which is often the case).
1298    
1299       We should do something similar to floating point, but the transformation
1300       there is much more complex, so I won't deal with that.
1301    
1302    Secondly, there are now two new cps primops at our disposal:
1303    
1304     1. rawrecord of record_kind option
1305        This pure operator allocates some uninitialized storage from the heap.
1306        There are two forms:
1307    
1308         rawrecord NONE [INT n]  allocates a tagless record of length n
1309         rawrecord (SOME rk) [INT n] allocates a tagged record of length n
1310                                     and initializes the tag.
1311    
1312     2. rawupdate of cty
1313          rawupdate cty (v,i,x)
1314          Assigns to x to the ith component of record v.
1315          The storelist is not updated.
1316    
1317    We use these new primops for both spilling and increment record construction.
1318    
1319     1. Spilling.
1320    
1321        This is implemented with a linear scan algorithm (but generalized
1322        to trees).  The algorithm will create a single spill record at the
1323        beginning of the cps function and use rawupdate to spill to it,
1324        and SELECT or SELp to reload from it.  So both spills and reloads
1325        are fine-grain operations.  In contrast, in the old algorithm
1326        "spills" have to be bundled together in records.
1327    
1328        Ideally, we should sink the spill record construction to where
1329        it is needed.  We can even split the spill record into multiple ones
1330        at the places where they are needed.  But CPS is not a good
1331        representation for global code motion, so I'll keep it simple and
1332        am not attempting this.
1333    
1334     2. Incremental record construction (aka record splitting).
1335    
1336        Long records with many component values which are simulatenously live
1337        (recall that single use record addresses are not considered to
1338         be live) are constructed with rawrecord and rawupdate.
1339        We allocate space on the heap with rawrecord first, then gradually
1340        fill it in with rawupdate.  This is the technique suggested to me
1341        by Matthias.
1342    
1343        Some restrictions on when this is applicable:
1344        1. It is not a VECTOR record.  The code generator currently does not handle
1345           this case. VECTOR record uses double indirection like arrays.
1346        2. All the record component values are defined in the same "basic block"
1347           as the record constructor.  This is to prevent speculative
1348           record construction.
1349    
1350    ----------------------------------------------------------------------
1351    Name: Allen Leung
1352    Date: 2002/02/22 01:02:00 EST
1353    Tag: leunga-20020222-mlrisc-tools
1354    
1355    Minor bug fixes in the parser and rewriter
1356    
1357    ----------------------------------------------------------------------
1358    Name: Allen Leung
1359    Date: 2002/02/21 20:20:00 EST
1360    Tag: leunga-20020221-peephole
1361    
1362    Regenerated the peephole files.  Some contained typos in the specification
1363    and some didn't compile because of pretty printing bugs in the old version
1364    of 'nowhere'.
1365    
1366    ----------------------------------------------------------------------
1367    Name: Allen Leung
1368    Date: 2002/02/19 20:20:00 EST
1369    Tag: leunga-20020219-mlrisc-tools
1370    Description:
1371    
1372       Minor bug fixes to the mlrisc-tools library:
1373    
1374       1.  Fixed up parsing colon suffixed keywords
1375       2.  Added the ability to shut the error messages up
1376       3.  Reimplemented the pretty printer and fixed up/improved
1377           the pretty printing of handle and -> types.
1378       4.  Fixed up generation of literal symbols in the nowhere tool.
1379       5.  Added some SML keywords to to sml.sty
1380    
1381    ----------------------------------------------------------------------
1382    Name: Matthias Blume
1383    Date: 2002/02/19 16:20:00 EST
1384    Tag: blume-20020219-cmffi
1385    Description:
1386    
1387    A wild mix of changes, some minor, some major:
1388    
1389    * All C FFI-related libraries are now anchored under $c:
1390        $/c.cm      --> $c/c.cm
1391        $/c-int.cm  --> $c/internals/c-int.cm
1392        $/memory.cm --> $c/memory/memory.cm
1393    
1394    * "make" tool (in CM) now treats its argument pathname slightly
1395      differently:
1396        1. If the native expansion is an absolute name, then before invoking
1397           the "make" command on it, CM will apply OS.Path.mkRelative
1398           (with relativeTo = OS.FileSys.getDir()) to it.
1399        2. The argument will be passed through to subsequent phases of CM
1400           processing without "going native".  In particular, if the argument
1401           was an anchored path, then "make" will not lose track of that anchor.
1402    
1403    * Compiler backends now "know" their respective C calling conventions
1404      instead of having to be told about it by ml-nlffigen.  This relieves
1405      ml-nlffigen from one of its burdens.
1406    
1407    * The X86Backend has been split into X86CCallBackend and X86StdCallBackend.
1408    
1409    * Export C_DEBUG and C_Debug from $c/c.cm.
1410    
1411    * C type encoding in ml-nlffi-lib has been improved to model the conceptual
1412      subtyping relationship between incomplete pointers and their complete
1413      counterparts.  For this, ('t, 'c) ptr has been changed to 'o ptr --
1414      with the convention of instantiating 'o with ('t, 'c) obj whenever
1415      the pointer target type is complete.  In the incomplete case, 'o
1416      will be instantiated with some "'c iobj" -- a type obtained by
1417      using one of the functors PointerToIncompleteType or PointerToCompleteType.
1418    
1419      Operations that work on both incomplete and complete pointer types are
1420      typed as taking an 'o ptr while operations that require the target to
1421      be known are typed as taking some ('t, 'c) obj ptr.
1422    
1423      voidptr is now a bit "more concrete", namely "type voidptr = void ptr'"
1424      where void is an eqtype without any values.  This makes it possible
1425      to work on voidptr values using functions meant to operate on light
1426      incomplete pointers.
1427    
1428    * As a result of the above, signature POINTER_TO_INCOMPLETE_TYPE has
1429      been vastly simplified.
1430    
1431    ----------------------------------------------------------------------
1432    Name: Matthias Blume
1433    Date: 2002/02/19 10:48:00 EST
1434    Tag: blume-20020219-pqfix
1435    Description:
1436    
1437    Applied Chris Okasaki's bug fix for priority queues.
1438    
1439    ----------------------------------------------------------------------
1440    Name: Matthias Blume
1441    Date: 2002/02/15 17:05:00
1442    Tag: Release_110_39
1443    Description:
1444    
1445    Last-minute retagging is becoming a tradition... :-(
1446    
1447    This is the working release 110.39.
1448    
1449    ----------------------------------------------------------------------
1450    Name: Matthias Blume
1451    Date: 2002/02/15 16:00:00 EST
1452    Tag: Release_110_39-orig
1453    Description:
1454    
1455    Working release 110.39.  New bootfiles.
1456    
1457    (Update: There was a small bug in the installer so it wouldn't work
1458    with all shells.  So I retagged. -Matthias)
1459    
1460    ----------------------------------------------------------------------
1461    Name: Matthias Blume
1462    Date: 2002/02/15 14:17:00 EST
1463    Tag: blume-20020215-showbindings
1464    Description:
1465    
1466    Added EnvRef.listBoundSymbols and CM.State.showBindings.  Especially
1467    the latter can be useful for exploring what bindings are available at
1468    the interactive prompt.  (The first function returns only the list
1469    of symbols that are really bound, the second prints those but also the
1470    ones that CM's autoloading mechanism knows about.)
1471    
1472    ----------------------------------------------------------------------
1473    Name: Matthias Blume
1474    Date: 2002/02/15 12:08:00 EST
1475    Tag: blume-20020215-iptrs
1476    Description:
1477    
1478    Two improvements to ml-nlffigen:
1479    
1480      1. Write files only if they do not exist or if their current contents
1481         do not coincide with what's being written.  (That is, avoid messing
1482         with the time stamps unless absolutely necessary.)
1483    
1484      2. Implement a "repository" mechanism for generated files related
1485         to "incomplete pointer types".   See the README file for details.
1486    
1487    ----------------------------------------------------------------------
1488    Name: Matthias Blume
1489    Date: 2002/02/14 11:50:00 EST
1490    Tag: blume-20020214-quote
1491    Description:
1492    
1493    Added a type 't t_' to tag.sml (in ml-nlffi-lib.cm).  This is required
1494    because of the new and improved tag generation scheme.  (Thanks to Allen
1495    Leung for pointing it out.)
1496    
1497    ----------------------------------------------------------------------
1498    Name: Lal George
1499    Date: 2002/02/14 09:55:27 EST 2002
1500    Tag: george-20020214-isabelle-bug
1501    Description:
1502    
1503    Fixed the MLRISC bug sent by Markus Wenzel regarding the compilation
1504    of Isabelle on the x86.
1505    
1506    From Allen:
1507    -----------
1508     I've found the problem:
1509    
1510         in ra-core.sml, I use the counter "blocked" to keep track of the
1511         true number of elements in the freeze queue.  When the counter goes
1512         to zero, I skip examining the queue.  But I've messed up the
1513         bookkeeping in combine():
1514    
1515             else ();
1516             case !ucol of
1517               PSEUDO => (if !cntv > 0 then
1518                     (if !cntu > 0 then blocked := !blocked - 1 else ();
1519                                        ^^^^^^^^^^^^^^^^^^^^^^^
1520                      moveu := mergeMoveList(!movev, !moveu)
1521                     )
1522                  else ();
1523    
1524         combine() is called to coalesce two nodes u and v.
1525         I think I was thinking that if the move counts of u and v are both
1526         greater than zero then after they are coalesced then one node is
1527         removed from the freeze queue.  Apparently I was thinking that
1528         both u and v are of low degree, but that's clearly not necessarily true.
1529    
1530    
1531    02/12/2002:
1532        Here's the patch.  HOL now compiles.
1533    
1534        I don't know how this impact on performance (compile
1535        time or runtime).  This bug caused the RA (especially on the x86)
1536        to go thru the potential spill phase when there are still nodes on the
1537        freeze queue.
1538    
1539    
1540    
1541    
1542    ----------------------------------------------------------------------
1543    Name: Matthias Blume
1544    Date: 2002/02/13 22:40:00 EST
1545    Tag: blume-20020213-fptr-rtti
1546    Description:
1547    
1548    Fixed a bug in ml-nlffigen that was introduced with one of the previous
1549    updates.
1550    
1551    ----------------------------------------------------------------------
1552    Name: Matthias Blume
1553    Date: 2002/02/13 16:41:00 EST
1554    Tag: blume-20020213-cmlpq
1555    Description:
1556    
1557    Added new priority queue export symbols (which have just been added to
1558    smlnj-lib.cm) to CML's version of smlnj-lib.cm.  (Otherwise CML would
1559    not compile and the installer would choke.)
1560    
1561    ----------------------------------------------------------------------
1562    Name: Matthias Blume
1563    Date: 2002/02/13 16:15:00 EST
1564    Tag: blume-20020213-various
1565    Description:
1566    
1567    1. More tweaks to ml-nlffigen:
1568    
1569       - better internal datastructures (resulting in slight speedup)
1570       - "-match" option requires exact match
1571       - "localized" gensym counters (untagged structs/unions nested within
1572         other structs/unions or within typedefs get a fresh counter; their
1573         tag will be prefixed by a concatenation of their parents' tags)
1574       - bug fixes (related to calculation of transitive closure of types
1575         to be included in the output)
1576    
1577    2. Minor Basis updates:
1578    
1579       - added implementations for List.collate and Option.app
1580    
1581    ----------------------------------------------------------------------
1582    Name: Matthias Blume
1583    Date: 2002/02/11 15:55:00 EST
1584    Tag: blume-20020211-gensym
1585    Description:
1586    
1587    Added a "-gensym" option to command line of ml-nlffigen.  This can be
1588    used to specify a "stem" -- a string that is inserted in all "gensym'd"
1589    names (ML structure names that correspond to unnamed C structs, unions,
1590    and enums), so that separate runs of ml-nlffigen do not clash.
1591    
1592    ----------------------------------------------------------------------
1593    Name: Matthias Blume
1594    Date: 2002/02/11 12:05:00 EST
1595    Tag: blume-20020211-gensml
1596    Description:
1597    
1598    A quick fix for a problem with GenSML (in the pgraph-util library):
1599    Make generation of toplevel "local" optional.  (Strictly speaking,
1600    signature definitions within "local" are not legal SML.)
1601    
1602    Other than that: updates to INSTALL and cm/TODO.
1603    
1604    ----------------------------------------------------------------------
1605    Name: Matthias Blume
1606    Date: 2002/02/08 15:00:00 EST
1607    Tag: blume-20020208-uniquepid
1608    Description:
1609    
1610    0. Version number has been bumped to 110.38.1.  NEW BOOTFILES!!!
1611    
1612    1. The installer (config/install.sh) has gotten smarter:
1613    
1614         - Configuration options are a bit easier to specify now
1615           (in config/targets).
1616         - Bug in recognizing .tar.bz2 files fixed.
1617         - Installer automatically resolves dependencies between
1618           configuration options (e.g., if you ask for eXene, you will
1619           also get cml -- regardless whether you asked for it or not).
1620         - Installer can run in "quieter mode" by setting the environment
1621           variable INSTALL_QUIETLY to "true".  "Quieter" does not mean
1622           "completely silent", though.
1623         - Build HashCons library as part of smlnj-lib.
1624    
1625    2. A new scheme for assigning persistent identifiers to compilation
1626       units (and, by extension, to types etc.) has been put into place.
1627       This fixes a long-standing bug where types and even dynamic values
1628       can get internally confused, thereby compromising type safety
1629       (abstraction) and dynamic correctness.  See
1630    
1631         http://cm.bell-labs.com/cm/cs/who/blume/pid-confusion.tgz
1632    
1633       for an example of how things could go wrong until now.
1634    
1635       The downside of the new scheme is that pids are not quite as
1636       persistent as they used to be: CM will generate a fresh pid
1637       for every compilation unit that it thinks it sees for the first
1638       time.  That means that if you compile starting from a clean, fresh
1639       source tree at two different times, you end up with different
1640       binaries.
1641    
1642       Cutoff recompilation, however, has not been compromised because
1643       CM keeps pid information in special caches between runs.
1644    
1645    ----------------------------------------------------------------------
1646    Name: Lal George
1647    Date: 2002/02/07 15:34:13 EST 2002
1648    Tag: <none>
1649    Description:
1650    
1651    Compilers that generate assembly code may produce  global labels
1652    whose value is resolved at link time. The various peephole optimization
1653    modules did not take this in account.
1654    
1655    TODO. The Labels.addrOf function should really return an option
1656    type so that clients are forced to deal with this issue, rather
1657    than an exception being raised.
1658    
1659    ----------------------------------------------------------------------
1660    Name: Lal George
1661    Date: 2002/02/06 13:55:02 EST
1662    Tag: george-20020206-ra-breakup
1663    Description:
1664    
1665    1. A bug fix from Allen.
1666    
1667        A typo causes extra fstp %st(0)'s to be generated at compensation
1668        edges, which might cause stack underflow traps at runtime.  This
1669        occurs in fft where there are extraneous fstps right before the 'into'
1670        trap instruction (in this case they are harmless since none of the
1671        integers overflow.)
1672    
1673    2. Pulled out various utility modules that were embedded in the modules
1674       of the register allocator. I need these modules for other purposes, but
1675       they are not complete enough to put into a library (just yet).
1676    ----------------------------------------------------------------------
1677    Name: Matthias Blume
1678    Date: 2002/01/31 16:05:00 EST
1679    Tag: blume-20020131-sparc-ccalls
1680    Description:
1681    
1682    1. C-calls on Sparc needlessly allocated a huge chunk (96 bytes)
1683       of extra stack space by mistake.  Fixed.
1684    
1685    2. Bug in logic of handling of command-line options in ml-nlffigen fixed.
1686    
1687    ----------------------------------------------------------------------
1688    Name: Allen Leung
1689    Date: 2002/01/30
1690    Tag: leunga-20020130-nowhere-bug-fix
1691    Description:
1692    
1693       MLRISC bug fixes:
1694       1. Fixed a bindings computation bug in the 'nowhere' program generator tool.
1695       2. MachineInt.fromString was negating its value.
1696    
1697    ----------------------------------------------------------------------
1698    Name: Matthias Blume
1699    Date: 2002/01/29
1700    Tag: blume-20020129-INSTALL
1701    Description:
1702    
1703    - Added somewhat detailed installation instructions (file INSTALL).
1704    - Fixed curl-detection bug in config/install.sh.
1705    - It is now possible to select the URL getter using the URLGETTER
1706      environment variable:
1707    
1708          not set / "unknown"      --> automatic detection (script tries wget,
1709                                       curl, and lynx)
1710          "wget" / "curl" / "lynx" --> use the specified program (script "knows"
1711                                       how to properly invoke them)
1712          other                    --> use $URLGETTER directly, it must take
1713                                       precisely two command-line arguments
1714                                       (source URL and destination file name)
1715    
1716    ----------------------------------------------------------------------
1717    Name: Matthias Blume
1718    Date: 2002/01/28
1719    Tag: blume-20020128-sparc-ccalls
1720    Description:
1721    
1722    - Fixed problem with calculation of "used" registers in sparc-c-calls.
1723    - Make use of the allocParam argument in sparc-c-calls.
1724    
1725    ----------------------------------------------------------------------
1726    Name: Matthias Blume
1727    Date: 2002/01/28
1728    Tag: blume-20020128-allocParam
1729    Description:
1730    
1731    John Reppy:  Changes c-calls API to accept client-callback for
1732    allocating extra stack space.
1733    me: Corresponding changes to mlriscGen (using a dummy argument that
1734        does not change the current behavior).
1735    
1736    ----------------------------------------------------------------------
1737    Name: Matthias Blume
1738    Date: 2002/01/28 12:00:00
1739    Tag: Release_110_38
1740    Description:
1741    
1742    This time for real!!!
1743    
1744    ----------------------------------------------------------------------
1745    Name: Matthias Blume
1746    Date: 2002/01/28 10:56:00 EST
1747    Tag: blume-20020128-retraction
1748    Description:
1749    
1750    0. Retracted earlier 110.38.  (The Release_110_38 tag has been replaced
1751       with blume-Release_110_38-retracted.)
1752    
1753    1. Fixed a problem with incorrect rounding modes in real64.sml.
1754       (Thanks to Andrew Mccreight <andrew.mccreight@yale.edu>.)
1755    
1756    2. A bug in ml-nlffigen related to the handling of unnamed structs, unions,
1757       and enums fixed.  The naming of corresponding ML identifiers should
1758       now be consistent again.
1759    
1760    ----------------------------------------------------------------------
1761    Name: Allen Leung
1762    Date: 2002/01/27
1763    Tag: leunga-20020127-nowhere
1764    Description:
1765    
1766       Added a target called nowhere in the configuration scripts.
1767       Enabling this will build the MLRISC 'nowhere' tool (for translating
1768       programs with where-clauses into legal SML code) during installation.
1769    
1770    ----------------------------------------------------------------------
1771    Name: Matthias Blume
1772    Date: 2002/01/25 21:27:00 EST
1773    Tag: blume-Release_110_38-retracted
1774    Description:
1775    
1776    Call it a (working) release!  Version is 110.38. Bootfiles are ready.
1777    
1778    README will be added later.
1779    
1780    !!! NOTE:  Re-tagged as blume-Release_110_38-retracted. Original tag
1781    (Release_110_38) removed.  Reason: Last-minute bug fixes.
1782    
1783    ----------------------------------------------------------------------
1784    Name: Matthias Blume
1785    Date: 2002/01/25
1786    Tag: blume-20020125-ffi
1787    Description:
1788    
1789    A large number of tweaks and improvements to ml-nlffi-lib and
1790    ml-nlffigen:
1791    
1792       - ML represenation types have been streamlined
1793       - getter and setter functions work with concrete values, not abstract
1794         ones where possible
1795       - ml-nlffigen command line more flexible (see README file there)
1796       - some bugs have been fixed (hopefully)
1797    
1798    ----------------------------------------------------------------------
1799    Name: Lal George
1800    Date: 2002/01/24
1801    Tag: george-20020124-risc-ra-interface
1802    Description:
1803    
1804       There is a dramatic simplification in the interface to the
1805       register allocator for RISC architectures as a result of making
1806       parallel copy instructions explicit.
1807    
1808    ----------------------------------------------------------------------
1809    Name: Matthias Blume
1810    Date: 2002/01/22
1811    Tag: blume-20020122-x86-ccalls
1812    Description:
1813    
1814    Bug fix for c-calls on x86 (having to do with how char- and
1815    short-arguments are being handled).
1816    
1817    ----------------------------------------------------------------------
1818    Name: Matthias Blume
1819    Date: 2002/01/21
1820    Tag: blume-20020121-ff
1821    Description:
1822    
1823    Another day of fiddling with the FFI...
1824    
1825    1. Bug fix/workaround:  CKIT does not complain about negative array
1826       dimensions, so ml-nlffigen has to guard itself against this possibility.
1827       (Otherwise a negative dimension would send it into an infinite loop.)
1828    
1829    2. Some of the abstract types (light objects, light pointers, most "base"
1830       types) in structure C are now eqtypes.
1831    
1832    3. Added constructors and test functions for NULL function pointers.
1833    
1834    ----------------------------------------------------------------------
1835    Name: Matthias Blume
1836    Date: 2002/01/18
1837    Tag: blume-20020118-ready-for-new-release
1838    Description:
1839    
1840    Made config/srcarchiveurl point to a new place.  (Will provide boot
1841    files shortly.)
1842    
1843    Maybe we christen this to be 110.38?
1844    
1845    ----------------------------------------------------------------------
1846    Name: Matthias Blume
1847    Date: 2002/01/18
1848    Tag: blume-20020118-more-ffifiddle
1849    Description:
1850    
1851    Today's FFI fiddling:
1852    
1853      - Provided a structure CGetSet with "convenient" versions of C.Get.* and
1854        C.Set.* that use concrete (MLRep.*) arguments and results instead
1855        of abstract ones.
1856    
1857      - Provided word-style bit operations etc. for "int" representation
1858        types in MLRep.S<Foo>Bitops where <Foo> ranges over Char, Int, Short,
1859        and Long.
1860    
1861    ----------------------------------------------------------------------
1862    Name: Matthias Blume
1863    Date: 2002/01/18
1864    Tag: blume-20020118-use-x86-fp
1865    Description:
1866    
1867    Now that x86-fast-fp seems to be working, I turned it back on again
1868    by default.  (Seems to work fine now, even with the FFI.)
1869    
1870    Other than that, I added some documentation about the FFI to
1871    src/ml-nlffigen/README and updated the FFI test examples in
1872    src/ml-nlffi-lib/Tests/*.
1873    
1874    ----------------------------------------------------------------------
1875    Name: Allen Leung
1876    Date: 2002/01/17
1877    Tag: leunga-20020117-x86-fast-fp-call
1878    Description:
1879    
1880       1. Fixed a problem with handling return fp values when x86's fast fp
1881          mode is turned on.
1882    
1883       2. Minor pretty printing fix for cellset.  Print %st(0) as %st(0) instead
1884          of %f32.
1885    
1886       3. Added a constructor INT32lit to the ast of MLRISC tools.
1887    
1888    ----------------------------------------------------------------------
1889    Name: Matthias Blume
1890    Date: 2002/01/16
1891    Tag: blume-20020116-ffifiddle
1892    Description:
1893    
1894    More fiddling with the FFI interface:
1895    
1896     - Make constness 'c instead of rw wherever possible.  This eliminates
1897       the need for certain explicit coercions.  (However, due to ML's
1898       value polymorphism, there will still be many cases where explicit
1899       coercions are necessary.  Phantom types are not the whole answer
1900       to modeling a subtyping relationship in ML.)
1901    
1902     - ro/rw coersions for pointers added.  (Avoids the detour through */&.)
1903    
1904     - "printf" test example added to src/ml-nlffi-lib/Tests.  (Demonstrates
1905       clumsy workaround for varargs problem.)
1906    
1907    ----------------------------------------------------------------------
1908    Name: Lal George
1909    Date: 2002/01/15
1910    Tag: <none>
1911    Description:
1912    
1913    1. Since COPY instructions are no longer native to the architecture,
1914       a generic functor can be used to implement the expandCopies function.
1915    
1916    2. Allowed EXPORT and IMPORT pseudo-op declarations to appear inside a
1917       TEXT segment.
1918    
1919    ----------------------------------------------------------------------
1920    Name: Matthias Blume
1921    Date: 2002/01/15
1922    Tag: blume-20020115-ffiupdates
1923    Description:
1924    
1925    1. Fix for bug resulting in single-precision float values being returned
1926       incorrectly from FFI calls.
1927    
1928    2. Small modifications to C FFI API:
1929    
1930        - memory-allocation routines return straight objects (no options)
1931          and raise an exception in out-of-memory situations
1932        - unsafe extensions to cast between function pointers and pointers
1933          from/to ints
1934        - added structure C_Debug as an alternative to structure C where
1935          pointer-dereferencing (|*| and |*!) always check for null-pointers
1936        - added open_lib' to DynLinkage;  open_lib' works like open_lib
1937          but also takes a (possibly empty) list of existing library handles
1938          that the current library depends on
1939    
1940    ----------------------------------------------------------------------
1941    Name: Matthias Blume
1942    Date: 2002/01/10
1943    Tag: blume-20020110-newffigen
1944    Description:
1945    
1946    1. Updates to portable graph code.
1947    
1948    2. Major update to ml-nlffigen and ml-nlffi-lib.  Things are much
1949       more scalable now so that even huge interfaces such as the one
1950       for GTK compile in finite time and space. :-)
1951       See src/ml-nlffigen/README for details on what's new.
1952    
1953    ----------------------------------------------------------------------
1954    Name: Lal George
1955    Date: 2001/01/09 14:31:35 EST 2002
1956    Tag: george-20011206-rm-native-copy
1957    Description:
1958    
1959            Removed the native COPY and FCOPY instructions
1960            from all the architectures and replaced it with the
1961            explicit COPY instruction from the previous commit.
1962    
1963            It is now possible to simplify many of the optimizations
1964            modules that manipulate copies. This has not been
1965            done in this change.
1966    
1967    ----------------------------------------------------------------------
1968    Name: Lal George
1969    Date: 2001/12/06 16:50:13 EST 2001
1970    Tag: george-20011206-mlrisc-instruction
1971    Description:
1972    
1973    Changed the representation of instructions from being fully abstract
1974    to being partially concrete. That is to say:
1975    
1976      from
1977            type instruction
1978    
1979      to
1980            type instr                              (* machine instruction *)
1981    
1982            datatype instruction =
1983                LIVE of {regs: C.cellset, spilled: C.cellset}
1984              | KILL of {regs: C.cellset, spilled: C.cellset}
1985              | COPYXXX of {k: CB.cellkind, dst: CB.cell list, src: CB.cell list}
1986              | ANNOTATION of {i: instruction, a: Annotations.annotation}
1987              | INSTR of instr
1988    
1989    This makes the handling of certain special instructions that appear on
1990    all architectures easier and uniform.
1991    
1992    LIVE and KILL say that a list of registers are live or killed at the
1993    program point where they appear. No spill code is generated when an
1994    element of the 'regs' field is spilled, but the register is moved to
1995    the 'spilled' (which is present, more for debugging than anything else).
1996    
1997    LIVE replaces the (now deprecated) DEFFREG instruction on the alpha.
1998    We used to generate:
1999    
2000            DEFFREG f1
2001            f1 := f2 + f3
2002            trapb
2003    
2004    but now generate:
2005    
2006            f1 := f2 + f3
2007            trapb
2008            LIVE {regs=[f1,f2,f3], spilled=[]}
2009    
2010    Furthermore, the DEFFREG (hack) required that all floating point instruction
2011    use all registers mentioned in the instruction. Therefore f1 := f2 + f3,
2012    defines f1 and uses [f1,f2,f3]! This hack is no longer required resulting
2013    in a cleaner alpha implementation. (Hopefully, intel will not get rid of
2014    this architecture).
2015    
2016    COPYXXX is intended to replace the parallel COPY and FCOPY  available on
2017    all the architectures. This will result in further simplification of the
2018    register allocator that must be aware of them for coalescing purposes, and
2019    will also simplify certain aspects of the machine description that provides
2020    callbacks related to parallel copies.
2021    
2022    ANNOTATION should be obvious, and now INSTR represents the honest to God
2023    machine instruction set!
2024    
2025    The <arch>/instructions/<arch>Instr.sml files define certain utility
2026    functions for making porting easier -- essentially converting upper case
2027    to lower case. All machine instructions (of type instr) are in upper case,
2028    and the lower case form generates an MLRISC instruction. For example on
2029    the alpha we have:
2030    
2031      datatype instr =
2032         LDA of {r:cell, b:cell, d:operand}
2033       | ...
2034    
2035      val lda : {r:cell, b:cell, d:operand} -> instruction
2036        ...
2037    
2038    where lda is just (INSTR o LDA), etc.
2039    
2040    ----------------------------------------------------------------------
2041    Name: Matthias Blume
2042    Date: 2001/11/22 21:40:00 EST
2043    Tag: Release_110_37
2044    Description:
2045    
2046    Release 110.37.  This time for real.
2047    
2048    ----------------------------------------------------------------------
2049    Name: Matthias Blume
2050    Date: 2001/11/21 16:35:00 EST
2051    Tag: blume-20011121-foot-in-mouth
2052    Description:
2053    
2054    Removed the "Release_110_37" tag because of a serious bug.
2055    This will be re-tagged once the bug is fixed.
2056    
2057    ----------------------------------------------------------------------
2058    Name: Matthias Blume
2059    Date: 2001/11/21 16:14:00 EST
2060    Tag: blume-20011121-forgottenfile
2061    Description:
2062    
2063    Forgot to add a file.  (Just a .tex-file -- part of
2064    the CM manual source.)
2065    
2066    ----------------------------------------------------------------------
2067    Name: Matthias Blume
2068    Date: 2001/11/21 16:10:00 EST
2069    Tag: blume-20011121-invalid_110_37
2070    Description:
2071    
2072    Note: I removed the original tag "Release_110_37" from this commit
2073          because we found a serious bug in all non-x86 backends.
2074          - Matthias
2075    
2076    1. Modifications to the SML/NJ code generator and to the runtime system
2077       so that code object name strings are directly inserted into code
2078       objects at code generation time.  The only business the runtime system
2079       has with this is now to read the name strings on occasions.
2080       (The encoding of the name string has also changed somewhat.)
2081    
2082    2. CM now implements a simple "set calculus" for specifying export lists.
2083       In particular, it is now possible to refer to the export lists of
2084       other libraries/groups/sources and form unions as well as differences.
2085       See the latest CM manual for details.
2086    
2087    3. An separate notion of "proxy" libraries has again be eliminated from
2088       CM's model.  (Proxy libraries are now simply a special case of using
2089       the export list calculus.)
2090    
2091    4. Some of the existing libraries now take advantage of the new set
2092       calculus.
2093       (Notice that not all libraries have been converted because some
2094       of the existing .cm-files are supposed to be backward compatible
2095       with 110.0.x.)
2096    
2097    5. Some cleanup in stand-alone programs.  (Don't use "exnMessage" -- use
2098       "General.exnMessage"!  The former relies on a certain hook to be
2099       initialized, and that often does not happen in the stand-alone case.)
2100    
2101    ----------------------------------------------------------------------
2102    Name: Lal George
2103    Date: 2001/11/21  13:56:18 EST
2104    Tag: george-2001121-pseudo-ops
2105    Description:
2106    
2107      Implemented a complete redesign of MLRISC pseudo-ops. Now there
2108      ought to never be any question of incompatabilities with
2109      pseudo-op syntax expected by host assemblers.
2110    
2111      For now, only modules supporting GAS syntax are implemented
2112      but more should follow, such as MASM, and vendor assembler
2113      syntax, e.g. IBM as, Sun as, etc.
2114    
2115    ----------------------------------------------------------------------
2116    Name: Matthias Blume
2117    Date: 2001/11/14 11:52:00 EST
2118    Tag: blume-20011114-srcname
2119    Description:
2120    
2121    1. Routed the name of the current source file to mlriscgen where it
2122       should be directly emitted into the code object.  (This last part
2123       is yet to be done.)
2124    
2125    2. Some cleanup of the pgraph code to make it match the proposal that
2126       I put out the other day.  (The proposal notwithstanding, things are
2127       still in flux here.)
2128    
2129    ----------------------------------------------------------------------
2130    Name: Lal George
2131    Date: 2001/11/14 09:44:04 EST
2132    Tag:
2133    Description:
2134    
2135      Fix for a backpatching bug reported by Allen.
2136    
2137      Because the boundary between short and long span-dependent
2138      instructions is +/- 128, there are an astounding number of
2139      span-dependent instructions whose size is over estimated.
2140    
2141      Allen came up with the idea of letting the size of span
2142      dependent instructions be non-monotonic, for a maxIter
2143      number of times, after which the size must be monotonically
2144      increasing.
2145    
2146      This table shows the number of span-dependent instructions
2147      whose size was over-estimated as a function of maxIter, for the
2148      file Parse/parse/ml.grm.sml:
2149    
2150         maxIter            # of instructions:
2151            10                      687
2152            20                      438
2153            30                      198
2154            40                        0
2155    
2156      In compiling the compiler, there is no significant difference in
2157      compilation speed between maxIter=10 and maxIter=40. Actually,
2158      my measurements showed that maxIter=40 was a tad faster than
2159      maxIter=10! Also 96% of the  files in the compiler reach a fix
2160      point within 13 iterations, so fixing maxIter at 40, while high,
2161      is okay.
2162    
2163    ----------------------------------------------------------------------
2164    Name: Matthias Blume
2165    Date: 2001/10/31 15:25:00 EST
2166    Tag: blume-20011031-pgraph
2167    Description:
2168    
2169    CKIT:
2170    * Changed the "Function" constructor of type Ast.ctype to carry optional
2171      argument identifiers.
2172    * Changed the return type of TypeUtil.getFunction accordingly.
2173    * Type equality ignores the argument names.
2174    * TypeUtil.composite tries to preserve argument names but gives up quickly
2175      if there is a mismatch.
2176    
2177    installation script:
2178    * attempts to use "curl" if available (unless "wget" is available as well)
2179    
2180    CM:
2181    * has an experimental implementation of "portable graphs" which I will
2182      soon propose as an implementation-independent library format
2183    * there are also new libraries $/pgraph.cm and $/pgraph-util.cm
2184    
2185    NLFFI-LIB:
2186    * some cleanup (all cosmetic)
2187    
2188    NLFFIGEN:
2189    * temporarily disabled the mechanism that suppresses ML output for
2190      C definitions whose identifiers start with an underscore character
2191    * generate val bindings for enum constants
2192    * user can request that only one style (light or heavy) is being used;
2193      default is to use both (command-line arguments: -heavy and -light)
2194    * fixed bug in handling of function types involving incomplete pointers
2195    * generate ML entry points that take record arguments (i.e., using
2196      named arguments) for C functions that have a prototype with named
2197      arguments
2198      (see changes to CKIT)
2199    
2200    ----------------------------------------------------------------------
2201    Name: Allen Leung
2202    Date: 2001/10/27 20:34:00 EDT
2203    Tag: leunga-20011027-x86-fast-fp-call
2204    Description:
2205    
2206       Fixed the bug described in blume-20010920-slowfp.
2207    
2208       The fix involves
2209          1. generating FCOPYs in FSTP in ia32-svid
2210          2. marking a CALL with the appropriate annotation
2211    
2212    ----------------------------------------------------------------------
2213    Name: Matthias Blume
2214    Date: 2001/10/16 11:32:00 EDT
2215    Tag: blume-20011016-netbsd
2216    Description:
2217    
2218    Underscore patch from Chris Richards (fixing problem with compiling
2219    runtime system under recent NetBSD).
2220    
2221    ----------------------------------------------------------------------
2222    Name: Allen Leung
2223    Date: 2001/10/12 17:18:32 EDT 2001
2224    Tag: leung-20011012-x86-printflowgraph
2225    Description:
2226    
2227    X86RA now uses a valid (instead of dummy) PrintFlowgraph module.
2228    
2229    ----------------------------------------------------------------------
2230    Name: Lal George
2231    Date: 2001/10/11 23:51:34 EDT
2232    Tag: george-20011011-too-many-instrs
2233    Description:
2234    
2235    The representation of a program point never expected to see more
2236    than 65536 instructions in a basic block!
2237    
2238    ----------------------------------------------------------------------
2239    Name: Lal George
2240    Date: 2001/10/09 09:41:37 EDT
2241    Tag: george-20011008-mlrisc-labels
2242    Description:
2243    
2244    Changed the machine description files to support printing of
2245    local and global labels in assembly code, based on host assembler
2246    conventions.
2247    
2248    ----------------------------------------------------------------------
2249    Name: Matthias Blume
2250    Date: 2001/09/25 15:25:00 EDT
2251    Tag: blume-20010925-exninfo
2252    Description:
2253    
2254    I provided a non-hook implementation of exnName (at the toplevel) and
2255    made the "dummy" implementation of exnMessage (at the toplevel) more
2256    useful: if nothing gets "hooked in", then at least you are going to
2257    see the exception name and a message indicating why you don't see more.
2258    
2259    [For the time being, programs that need exnMessage and want to use
2260    ml-build should either use General.exnMessage (strongly recommended) or
2261    refer to structure General at some other point so that CM sees a
2262    static dependency.]
2263    
2264    [Similar remarks go for "print" and "use":  If you want to use their
2265    functionality in stand-alone programs generated by ml-build, then use
2266    TextIO.output and Backend.Interact.useFile (from $smlnj/compiler.cm).]
2267    
2268    ----------------------------------------------------------------------
2269    Name: Matthias Blume
2270    Date: 2001/09/20 17:28:00 EDT
2271    Tag: blume-20010920-slowfp
2272    Description:
2273    
2274    Allen says that x86-fast-fp is not safe yet, so I turned it off again...
2275    
2276    ----------------------------------------------------------------------
2277  Name: Matthias Blume  Name: Matthias Blume
2278  Date: 2001/09/20 17:20:00 EDT  Date: 2001/09/20 17:20:00 EDT
2279  Tag: blume-20010920-canonicalpaths  Tag: blume-20010920-canonicalpaths

Legend:
Removed from v.937  
changed lines
  Added in v.1327

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